Shifts and Patterns in Maltese 9783110496376, 9783110495638

Table of contents :
Contents
Preface
From Maltese phonology to morphogenesis: A tribute to David Cohen
Part I: Phonology
Minimalist representation of Maltese sounds
Phonological changes in Maltese: Evidence from onomastics
Lengthening as a discourse strategy in Maltese: Phonetic and phonological characteristics
Part II: Morphology & Syntax
How inflectional morphology meets subcategorization frame distinctions in Maltese
The origin of differential object marking in Maltese
Polar interrogative -š in Maltese: Developments and antecedents
On short and long forms of personal pronouns in Maltese
Connecting /t/ in Maltese numerals
Part III: Contact, Bilingualism & Technology
Languages in contact
Maltese loanword typology
Language profiling: The weaving of Maltese and English in Maltese children’s conversations
Digitizing the grammar and vocabulary of Maltese
Index of Authors
Index of Languages
Index of Subjects

Citation preview

Shifts and Patterns in Maltese

Studia Typologica

Beihefte / Supplements STUF – Sprachtypologie und Universalienforschung Language Typology and Universals Editors Thomas Stolz, François Jacquesson, Pieter C. Muysken Editorial Board Michael Cysouw (München), Ray Fabri (Malta), Steven Roger Fischer (Auckland), Bernhard Hurch (Graz), Bernd Kortmann (Freiburg), Nicole Nau (Poznán), Ignazio Putzu (Cagliari), Stavros Skopeteas (Bielefeld), Johan van der Auwera (Antwerpen), Elisabeth Verhoeven (Berlin), Ljuba Veselinova (Stockholm)

Volume 19

Shifts and Patterns in Maltese

Edited by Gilbert Puech, Benjamin Saade

ISBN 978-3-11-049563-8 e-ISBN (PDF) 978-3-11-049637-6 e-ISBN (EPUB) 978-3-11-049300-9 ISSN 1617-2957 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2016 Walter de Gruyter GmbH, Berlin/Boston Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Gilbert Puech and Benjamin Saade Preface | VII Martine Vanhove  From Maltese phonology to morphogenesis: A tribute to David Cohen | 1

Part I:

Phonology 

Gilbert Puech  Minimalist representation of Maltese sounds | 19 Andrei A. Avram  Phonological changes in Maltese: Evidence from onomastics | 49 Alexandra Vella, Flavia Chetcuti and Sarah Agius  Lengthening as a discourse strategy in Maltese: Phonetic and phonological characteristics | 91

Part II: Morphology & Syntax  Maris Camilleri  How inflectional morphology meets subcategorization frame distinctions in Maltese | 115 Hans-Jörg Döhla  The origin of differential object marking in Maltese | 149 David Wilmsen  Polar interrogative -š in Maltese: Developments and antecedents | 175 Thomas Stolz and Benjamin Saade  On short and long forms of personal pronouns in Maltese | 199 Christopher Lucas and Michael Spagnol  Connecting /t/ in Maltese numerals | 269

VI | Contents

Part III: Contact, Bilingualism & Technology  George Farrugia  Languages in contact. Is the gender assignment system in Maltese undergoing change? | 293 Bernard Comrie and Michael Spagnol  Maltese loanword typology | 315 Marie Azzopardi-Alexander  Language profiling: The weaving of Maltese and English in Maltese children’s conversations | 331 John J. Camilleri  Digitizing the grammar and vocabulary of Maltese | 359 Index of Authors | 387 Index of Languages | 391 Index of Subjects | 392

Preface The present volume Shifts and patterns in Maltese presents a selection of papers from the 4th International Conference on Maltese Linguistics held in Lyon in June 2014. The fourth edition marks the first time that the bi-annual conference was not held in Bremen or in Malta. Bremen as the seat of GĦILM (L-Għaqda Internazzjonali tal-Lingwistika Maltija/The International Association for Maltese Linguistics) and home of the Malta Centre hosted the conference in 2007 and 2009, while Malta was the conference venue in 2011. We take this change to yet another venue as a further sign that Maltese Linguistics as a subject in its own right will not be confined to an academic niche but has developed a considerable clout of its own. This volume shows two other encouraging ongoing developments: While the contributions to this volume reflect the diversity of topics Maltese Linguistics has to offer, the composition of the authorship also reflects the ongoing international interest in Maltese: the institutions of the authors are based in 8 different countries ranging from Lebanon to Sweden. The first contribution to the volume by Martine Vanhove presents a special tribute to the French linguist David Cohen who sadly passed away on March 9, 2013. His contributions to the description of Maltese especially in the realm of phonology and morphology set the foundations for many Maltese linguists of the new generation. The remainder of the book is divided broadly into three thematic sections:

1. Phonology In his paper, Gilbert Puech proposes a minimalist representation of sounds in Maltese drawing on prosodic, structural, and melodic components while also taking into account dialectal variation in Maltese. Andrei A. Avram sheds light on the nature and chronology of phonological changes in Maltese using a wealth of onomastic data including nicknames, surnames, and toponyms. Alexandra Vella, Flavia Chetcuti, and Sarah Agius investigate “hesitation lengthening” for Maltese, a discourse strategy related to (but not identical to) filled pauses and preboundary lengthening, on the basis of natural spoken language data.

2. Morphology & Syntax Maris Camilleri provides a new perspective on verbal inflection in Maltese, proposing a multidimensional model of paradigm, which accounts for subcategorization frames by taking into account the non-concatenative aspect of Mal-

VIII | Preface

tese verbal morphology. In his paper, Hans-Jörg Döhla discusses the origins of Differential Object Marking (DOM) in Maltese, putting the phenomenon in the context of other varieties of Arabic, and concluding that DOM in Maltese and other Arabic languages seems to be an exclusively contact-induced trait. The study by David Wilmsen concerns the use and development of polar interrogative -š in Maltese, tracing its likely diachronic pathway and drawing parallels to other varieties and historical stages of Arabic. Thomas Stolz and Benjamin Saade present a detailed corpus study of the variation between long and short independent pronouns in Maltese including phonological environment, person, text type, and copular use as influencing factors. The contribution by Christopher Lucas and Michael Spagnol sheds light on the use of the connecting /t/ in the Maltese numeral system, identifying the onset of counted nouns as the key determining factor in the variation, with syllable number and sound/broken plural distinction as only marginal influences.

3. Contact, Bilingualism & Technology In his paper George Farrugia describes the assignment of grammatical gender to loanwords in Maltese, drawing attention to the instability of gender assignment for English loans and the possible long-term influences of loan word gender on the Maltese gender system as a whole. Bernard Comrie and Michael Spagnol contribute to the better understanding of the composition of the Maltese lexicon by providing a detailed study on the proportion of loanwords by semantic field and parts of speech while using the results to put Maltese in the context of other languages in a comparative loanword typology. Marie Azzopardi-Alexander investigates the continuum of bilingual language practices in the speech of young children in Malta ranging from exclusive use of one language (Maltese or English) to various degrees of codes-witching and code-mixing which could also be better described with the term translanguaging. In the final contribution to the volume John J. Camilleri shows how the computational implementation of a Maltese grammar can lead to advances that are both theoretically relevant and practically oriented, as in the case with Ġabra, an online full-form lexicon of Maltese. We would like to thank all participants of the 4th International Conference on Maltese Linguistics for their interesting papers and stimulating discussions throughout the time in Lyon. A special thank you goes to Cornelia Stroh for her hard and patient work and support during all stages of the editing process. We are also grateful to Thomas Stolz for his support as the head of the Malta Centre in Bremen and as the president of GĦILM. The editors, Lyon and Bremen, May 2016

Martine Vanhove

From Maltese phonology to morphogenesis: A tribute to David Cohen Abstract: This paper is an overview of the pioneer contributions of David Cohen to Maltese synchronic and diachronic phonology. The focus is on four critical issues in this domain: the evolution and the phonological status of the pharyngeal fricative għ and the laryngeal fricative h; the influence of the “emphatic” and “back” consonants on the development of the current vocalic system; the raising and fronting of the back short and long vowels a and ā traditionally termed imaala; the historical phonology of the oldest written Maltese poem, the Cantilena. Cohen’s works are presented within the general background of Arabic dialectology in which he was completely at ease, and briefly compared with other scholar’s work on the same topics. Keywords: Maltese, phonology, synchrony, diachrony, comparative linguistics, Arabic dialectology

1 Introduction On 9 March 2013, David Cohen passed away at the age of 90. He was a great scholar and a man of perfect scientific integrity. He was also my supervisor when I was preparing my PhD thesis on Maltese in the late 1980’s. I owe him a lot scientifically and professionally. This is why I decided to take the opportunity of the fourth Għilm meeting in Lyon in 2013 to pay him a tribute, by recalling some of his work and his influence on Maltese linguistics, even though perhaps such a tribute is more modest than he deserves, and is already overshadowed by a monumental Festschrift volume (Lentin and Lonnet 2003) and several prestigious prizes. David Cohen dedicated a great deal of his research to historical and comparative linguistics in the domain of Semitic and Afroasiatic languages (which he persisted in calling “Hamito-Semitic”, because it was no less scientifically grounded than “Afroasiatic”). It is on this phylum that he based his theoretical

|| Martine Vanhove: French National Centre for Scientific Research, Langage, Langues et Cultures d’Afrique Noire (LLACAN, UMR 8135, CNRS, INALCO, Université Sorbonne Paris-Cité), 7, rue Guy Môquet, 94801 Villejuif cedex, France. E-mail: [email protected]

2 | Martine Vanhove

and general investigations about the human capacity for language and about how and why languages change and are what they are. He leaves his imposing Dictionnaire des racines sémitiques unfinished. Fortunately he managed to publish the first ten volumes – the first appeared in 1970 – and hopefully his collaborators from volume 3 onwards will complete his work. One of his most important contributions to research in the domain of historical and comparative linguistics concerns the evolution of aspectual systems and the renewal of verbal systems in general. His book, La phrase nominale et l’évolution du système verbal en sémitique. Etude de syntaxe historique, written as a Thèse d’Etat in the mid-1970’s and published in an abridged version in 1984, goes far beyond what the title suggests. It actually concerns three branches of the Afroasiatic languages, Semitic, Cushitic and Egyptian. Following this, he went on to expand his perspective cross-linguistically, developing what he was reluctant to call a “theory of aspect” in L’aspect verbal (1989). This book also includes his first published thoughts on the role of auxiliaries and verbal periphrastic constructions in general in the renewal of verbal systems (putting aside his annual Rapports sur les Conférences de l’Ecole Pratique des Hautes Etudes, IVe section, Sciences Historiques et philologiques, in particular that of the 1982–83 academic year and the section on “Les auxiliaires dans l’expression périphrastique du parfait”). He placed this study within the larger framework of “morphogenesis”, a term more adequate for his comparative research which was encompassing not only processes of reanalysis – now known as grammaticalization (see for example Heine 1993) – but also other processes of language change such as polarity, analogy and language contact. It is not the aim of this article to give a thorough account of Cohen’s contribution to linguistics since such an attempt would lead far beyond Maltese linguistics. Rather, I will recall very briefly his major contributions to what is traditionally termed “Arabic dialectology” and his volumes on the Arabic varieties of Tunis (Jewish variety, Cohen 1964 and 1975) and Mauritania (Cohen 1963), as well as his pioneer works on computational linguistics (Cohen 1970), language pathology and aphasia (Cohen et al. 1963 and Cohen and Gauthier 1965). A detailed bibliography was published in the Festschrift volume (Lentin and Lonnet 2003). Cohen contributed to the domain of Maltese linguistics both by his teaching and his research work. His main contributions concern phonology from a synchronic and diachronic viewpoint, and I will explore them in what follows in relation to a limited number of previous and subsequent works by other linguists.

From Maltese phonology to morphogenesis: A tribute to David Cohen | 3

2 Historical background After an early note on the vocalic system of Maltese (Cohen 1957–1960), Cohen wrote a fundamental paper on Maltese phonology entitled “Le système phonologique du maltais. Aspects synchroniques et diachroniques”. It is written within the structuralist theoretical framework of the time, comparing it with Arabic, from which Maltese stemmed. It was also a highly valuable input for sound research on Maltese phonology in the various theoretical frameworks that followed, although this is not always acknowledged in bibliographies, probably because it was written in French. It was first published almost half a century ago, in 1966, in the third issue of the newly founded Journal of Maltese Studies, and was reprinted in his book Etudes de linguistique sémitique et arabe in 1970. It is an extension of a paper he presented in 1965 at the Second International Congress of Dialectology, which was held in Marburg, and was published in the Proceedings of the conference in 1967 under the title “Contribution à la phonologie diachronique du maltais”. Cohen used two different sources for his phonology of Maltese. The first was Aquilina’s description of Maltese from the late 1930s, which was published in 1959 (The structure of Maltese. A study in mixed grammar and vocabulary), and Stumme’s Maltesische Studien. Eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache, published in 1904. He was only able to check his phonetic observations on a small sample of recordings that were kindly sent to him by Aquilina. Of course, for Stumme’s data he had to rely on the phonetic transcriptions of the various Maltese varieties he illustrated, since phonology was only beginning to emerge as a discipline at that time. Regarding Aquilina, his structural description, which he claimed was “never meant to fall within the specific category of any particular school of linguistics” (Aquilina 1973: iv) nevertheless acknowledges the influence of the theoretical framework developed by Firth (1957) at SOAS in London before World War II. His phonetic and phonological description is thus influenced by this theory. It is also heavily influenced by some etymological issues: the phonology section was subdivided into two sub-sections, Semitic and non-Semitic Maltese, a common practice among scholars of Maltese in the first half of the 20th century (cf. e.g., Sutcliffe 1936). Aquilina provided a thorough study of għajn regarding its vocalic and consonantal allophones. Still, in the “Semitic” part of Maltese, nothing was mentioned about the influence of the back and so-called “emphat-

4 | Martine Vanhove

ic” consonants (i.e. pharyngealized) on the vocalic system, which Cohen was the first to explain.1

3 The phonological status of għ and h I will not recall here the details of Cohen’s synchronic phonology, which are by now well established and uncontroversial, even if Maltese of course has evolved over the last 50 to 100 years. Instead, I will concentrate on issues which are particularly important in a comparative perspective. I will start with Cohen’s arguments concerning the phonological status of għajn and h. He borrowed the adjective “virtual” to describe these from Stumme (1904), the first scholar to discuss – albeit in a brief note – the unusual status of the various allophones in the different dialects of Maltese of what he recognized as developments of three Arabic phonemes, ʕ, ġ and h. Cohen states first that the treatment of għajn and h as phonemes in Maltese grammars are ultimately based on morphological criteria. In what follows, I will translate and summarize the main lines of his argument to make them accessible to those unfamiliar with French. 1. The Cv̆Ca pattern of Perfective verbs in the 3rd person singular masculine such as rema ‘he threw’ and sema’ ‘he heard’, which have two distinct paradigms for the 3rd person singular feminine and plural in the Perfective (Cohen 1970: 130):2 SG 1 rmeyt ‘I threw’ smayt ‘I heard’ 2 remeyt ‘you threw’ smayt ‘you heard’ M 3 rema ‘he threw’ sema. ‘he heard’ F rmiə̅t ‘she threw’ sem.et ‘she heard’ PL 1 rmeyna ‘we threw’ smayna ‘we heard’ 2 rmeytu ‘you threw’ smaytu ‘you heard’ 3 rmew ‘they threw’ sem.u ‘they heard’ and of the plural forms in the Imperfective: narmu, etc. nesm.ew, etc.

|| 1 Cowan (1966) made a similar, and probably independent, analysis of the emphatic consonants, which was published in the same issue of the Journal of Maltese Studies as Cohen’s (1966). 2 Cohen’s transcriptions are given as they appear in his article, minus the obvious typos.

From Maltese phonology to morphogenesis: A tribute to David Cohen | 5

2. The existence of a verb class whose paradigms go against the general principles of Maltese syllabic structures and stress rules: yoʔó.du ‘they will sit’, yelá.bu ‘they will play’ vs. the “normal” forms níktbu ‘we will write’ or nizílʔu ‘we will slide’. 3. The existence of forms with alternating allomorphs, namely the pharyngeal ħ in absolute final position or Ø (or a pharyngealization and/or lengthening of the adjacent vowel) before enclitics, e.g. biə̅ħ ‘he sold’, nbélek or nbé°lek ‘I will sell to you’. 4. The existence of other forms with the reverse allomorphic pattern, that is, a vowel in absolute final position, and the pharyngeal ħ before the enclitic pronouns of the 3rd person singular feminine and 3rd person plural. This is illustrated in verbs such as láʔa ‘he received’ vs. laʔáħħa ‘he received her’, laʔáħħom ‘he received them’, and in the genitive particle ta which alternates with taħ- with the same enclitic pronouns: taħħa ‘of her’, taħħom ‘of them’. 5. The allomorphs of these very same enclitic pronouns -ħa, -ħom, which in other phonetic contexts are not pronounced with an initial consonant, but are spelled with an h described as “quiescent” in grammars, These forms present special paradigms, e.g. verbs such as fé°em ‘he understood’, where the stem alternates between CeC before a vowel and CiC before a consonant. 6. The existence of a particular factitive pattern for this verb type with a long vowel instead of a geminate medial consonant: fiə̅em ‘he caused to understand’. These particular features were discussed at length and in great detail within the framework of generative linguistics by Brame a few years later, in an article entitled On the abstractness of phonology: Maltese ʕ (1972). Brame does not mention Cohen’s paper (nor Aquilina’s or Stumme’s for that matter). Nevertheless, he arrives at the very same conclusions as Cohen by proposing a set of rules typical of the generative approach of the time, some of them related to classes of consonants, in order to prove the necessity of postulating, in some instances, an underlying representation that deviates from and is not represented in the surface alternations. Brame posits the abstract existence of a fricative pharyngeal phoneme ʕ, but leaves aside one logical consequence of his approach, the possibility of a second abstract laryngeal phoneme h whose allophones he does not deal with in his paper, and which only partly overlap with those of għajn. In addition to the above mentioned morphological arguments, Cohen brings further arguments of a sociolinguistic and comparative nature. He explicitly mentions that morphology is not in itself a decisive argument that would make it possible to decide in favor of a “full” phoneme, but that the problem is crucial enough to be taken into account if one does not want to distort the language’s phonological system (which Schabert 1976 did to some extent in his description

6 | Martine Vanhove

of the diphthongs of two Maltese varieties, even though he mentions Cohen’s work). Cohen first refers to the fact that the non-articulation of għ and h is not a general phenomenon in all Maltese varieties. The speakers who did not pronounce them were in contact with others who, in the first half of the 20th century, did have various allophones of the virtual phoneme. These were ranging from a simple hiatus between two vowels to pharyngeal and laryngeal consonants. Thus the existence of the virtual phoneme was perceived by all speakers, at least in a good number of forms. This reminded Cohen of the Arabic variety of the Jewish population of Tunis (of which he was a native speaker) and which he had described in two volumes (Cohen 1964/1975). In this Arabic variety, the laryngeal phoneme was not pronounced as such, but existed in the contact variety spoken by Muslims, maintaining the awareness of a kind of “virtual” phoneme, with various contextual allophones, including vowel lengthening, a variant still found in today’s Maltese. From a comparative point of view, Cohen also mentions, as of particular importance, the partial assimilation of verbs with a pharyngeal as third root consonant to the paradigm of verbs ending in a semi-vowel. His argument, based on analogy, runs as follows: where the semi-vowel was represented by an autonomous phonetic quantity, the conjugation could align with that of verbs with a pharyngeal consonant, e.g. the 1st and 2nd persons singular of the Perfective, where *qata°t could be assimilated to ramayt in the form of ʔatayt. But when the form taken by the last semi-vowel of the stem was fused with the flectional morpheme, verbs with a final pharyngeal maintained a phonemic quantity realized as a vowel before the flectional morpheme, e.g. yaʔtew as opposed to yarmu. Cohen’s conclusion was that the various cognates of the Arabic pharyngeals and laryngeal were, in Maltese, in a transitory phase of an already advanced phonological shift. He considered that this evolution would most probably lead to their total disappearance eventually. This fact is well known by Maltese linguists, and is confirmed by teachers of Maltese who often point out the spelling mistakes made by students when they align their paradigms with that of other root types, in particular the mistakes that lead to consonant clusters similar to the affricate consonants of Maltese ċ and z. Cohen goes on to explain that this virtual phoneme, even if not realized in a stable way, keeps having an influence on contextual elements. In order to account for all the phenomena described, he favors the existence of a phoneme which he represents, unlike Brame, with the symbols “.” or “°”, which are intentionally abstract so that they cannot be associated with any particular phonetic realization. The complementary distribution of the various allophones Cohen proposes is not exhaustive because of the limited documentation he had access to, but is accurate for the Maltese language of the first half of the 20th century.

From Maltese phonology to morphogenesis: A tribute to David Cohen | 7

4 The influence of emphatic and back consonants on the vocalic system The second important point in Cohen’s articles (1967, 1970) concerns the effects of the partial loss and fusion of the back and emphatic consonants on the synchronic vocalic system. As is now well known, Maltese lost, probably at an early stage, all the emphatic consonants that are still preserved in most Arabic varieties when they were fused with the corresponding non-emphatic phonemes. One result of Cohen’s research on the vocalic system is the discovery that, like most Maghribi Arabic dialects, Maltese shared the innovation of an emphatic flap ṛ, which will be discussed further below. His reconstruction is based on the evidence of the vocalic system, whose phonology he first describes before arguing about diachronic evolution. Standard Maltese has four short phonological vowels, a, e, i, o, in a structuralist sense, i.e. when stressed. Actually, by now, there is probably a fifth vowel, u, because it is present in the massive and ongoing borrowings from Italian and English. Maltese also has six long vowels, which at the time of Cohen’s description were the five monophthongs ā́, ī́, ṓ, ḗ, ū́, and one diphthong iə̅́, which only occur in stressed position. The existence of the diphthong was not led by the orthography as Borg (1976: 200, n. 14) claims in his paper about the imaala in Maltese. It did exist, and still does, in some Maltese varieties, although not in Standard contemporary Maltese as described by Borg and Azzopardi-Alexander (1997) among others. Anyway, what Cohen clearly shows – and this more or less ended the speculations about a possible oriental origin of the Arabic ancestor of Maltese which were initiated by Stumme – is that this system is due to several historical, phonological and sociolinguistic factors. Cohen’s (partial) explanation of the current system can be partially summarized as follows: (i) the fusion, at an early stage of Maltese development, of the two short vowels i and a, as is the case in many Maghribi varieties, and their massive reintroduction under the pressure of intensive contact with Italian and Sicilian, which are genetically unrelated; (ii) the role of stress; (iii) the role of the syllabic structure; (iv) what Cohen calls a “shift of distinctivity” from consonantal features to vocalic quality and quantity; (v) borrowings from Italian and Sicilian. The last three factors are widely accepted and have been developed, in different theoretical frameworks. Puech (1979 and 1994) and Borg (1976, 1977 and 1996)

8 | Martine Vanhove

both elaborate on the various phonemic evolutions and diphthongizations of short and long vowels and on the imaala in various varieties of Maltese, not to mention my own co-authored work on the dialect of Mġarr (Camilleri and Vanhove 1994). The two first factors, on the other hand, seem to be overlooked sometimes when one reads, even today, arguments concerning the filiation of Maltese with one or other of the two major geographic divisions of Arabic dialects. Let us examine now in more detail Cohen’s diachronic argumentation. a. The short vowel ĕ, is a Maltese innovation and a cognate of both ă and ĭ; b. ă and ĭ both shifted to the three short vowels ă, ĕ, ĭ; c. ŭ only shifted to ŏ. Thus, the Maltese evolution bears witness to a tendency for ă and ĭ to fuse into one phoneme. Cohen establishes the following laws of phonological change, which I will present in here in reverse order starting with their Maltese outcomes, and which concern only stressed vowels: (1) The Maltese /ŏ/ is quite straightforward: it is cognate with Arabic *ŭ in almost all nominal and verbal forms; ŭ occurs only scarcely as an optional allophone of /ŏ/ (Cohen also notes that in non-accented positions, ŭ sometimes becomes /ă/ or /ĕ/). (2) Maltese /ĕ/ is cognate with:  Arabic *ĭ when preceded or followed by a current or former laryngeal, a pharyngeal, a velar or the uvular qâf (now pronounced as a glottal stop);  Arabic *ă when in contact with the etymological laryngeal h, in monosyllabic forms, in dissyllabic verbal forms in closed syllables, and in the C1VC2a verb pattern. (3) Maltese /ĭ/ is cognate with:  Arabic *ĭ in the context of non-back consonants, except in case of vowel harmony with /ŏ/;  Arabic *ă in dissyllabic forms in open syllables, in verbs whose basic pattern was *CV̆CV̆C-, in nouns derived from verbs such as bídi ‘beginning’ (as opposed to béda ‘he began’), in deverbal prefixes (tíbkir ‘early rising’, tíblit ‘ramming’, míbrūm ‘rolled’, etc.), and often, but not systematically in non-verbal forms in closed syllables (cf. bikri ‘early’). In other words, phonemic conditioning, morphology and analogy were the driving forces for the various phonemic shifts. But phonemic conditioning has often overridden morphological and analogical processes, and generally, the old *ă kept its vocalic quality when in contact with the former back consonants (the velars, pharyngeals and the uvular).

From Maltese phonology to morphogenesis: A tribute to David Cohen | 9

In connection with the vowel shifts that occurred in Maltese, Cohen also notices that in contact with the flap r, the vowel /ă/ has all three possible cognates: ĕ as in rézaħ [razaħa] ‘shiver’, ĭ as in rikeb [rakiba] ‘ride’, ă as in ǧarr ‘drag, take away’, just as with the other three Maltese phonemes, s, t and d, which are cognates with two phonemes, an emphatic and a non-emphatic one. Cohen concludes that, as in many Maghribi dialects but unlike classical Arabic, Maltese, at some stage in its development, had, in addition to the non-emphatic flap, an emphatic ṛ phoneme, as the case of ǧarr which retained the original ă bears witness. To sum up, in unstressed syllables, Arabic ă and ĭ have both evolved towards the vowel whose quality was imposed by the phonemic context, and with possible pharyngealization, vowel harmony, etc, and also by analogy with the morphological pattern of the form to which the vowel was assigned. But the predominance of phonemic conditioning did not apply alike to all grammatical categories. Cohen (1970) recalls that in a morphological context such as the verbal system, analogical and regularization factors are bound to play an important role, and these may go against phonological rules. One of them concerns the class of middle verbs in Arabic, usually an intransitive verb class. For instance the Maltese CVCC pattern in this verb category is characterized by a short vowel e, not an a, even when in contact with a back or a former emphatic consonant, e.g. ħeff ‘become light’, ħenn ‘be pathetic, pitiful’, qell ‘be rude’, seħħ ‘be valid’. Another example is linked to the former fourth derived verb form, the one with a prefix ʔa- in classical Arabic. In Maltese, the CeCa pattern includes verbs with a first back or emphatic consonant. Now, in many Maghribi varieties, there are a good number of pairs such as xla, yaxli ‘he devastated’: xla, yaxla ‘he was ruined’ which are differentiated by the vocalic opposition in final position after the second root consonant. This distinction goes back to the opposition between the causative fourth forms with the glottal stop prefix, and the middle form. Contemporary Maltese contains only fossilized remnants of this fourth form, but it is remarkable that the verbs which in Maghribi Arabic are of the non-middle type like xla, yaxli all belong to the CeCa type in Maltese, whatever the root consonants, whereas the verbs which are of the type xla, yaxla with a back or an emphatic consonant in their roots, all belong to the type CaCa. This can be illustrated by the opposition in Maltese between ġera ‘he ran’ and ġara ‘it took place’ which corresponds to the Maghribi opposition between ǧra, yaǧri and ǧṛa, yaǧṛa, the latter being pronounced with a pharyngealized ṛ. So, once a phoneme is lost, phonological conditionings tend to lose ground and eventually cease to operate; morphological analogy can become more and more predominant, as is the case with the CoCoC pattern for middle verbs,

10 | Martine Vanhove

which attracts other verbs of other patterns. But analogy and phonological conditioning did not occur simultaneously and in exactly the same way in all varieties of Maltese, and dialectal variation is still indicative of various stages in its phonological development. The vocalic cognate sets mentioned above are not the whole story and Cohen shows that the vowels of the Maltese disyllabic pattern with two short vowels CV̆CV̆C do not all go back to former vowels. Some short vowels were introduced in order to separate consonant clusters containing a sonorant (l, r, ṛ, m) as second element. Their quality is generally in harmony with the vowel characteristic of the pattern. Conversely, some of the former short vowels were eliminated: this was the case for disyllabic forms in the first syllable if the second syllable was long and in the second syllable if the initial consonant was a sonorant. In trisyllabic forms the second vowel was eliminated. Cohen concludes this part of his article cautiously. He admits that he did not have sufficient data to explain in detail all the losses of short vowels and the role of the various influences that took place when consonant clusters were separated. Numerous forms remain unexplained and there is still room for further research. However, as Cohen states, the mechanism that triggered the shift from the old vocalic system to the Maltese system is clear. The distinction between the Maltese phonemes /ĕ/ and /ĭ/ is an interrupted evolution freezing differences which were tending to be conditioned variants. The contrast between /ĕ/ and /ĭ/ thus partly represents the former contrast between /ă/ and /ĭ/. As for the existence of a short and a long a (/ă/ ~ /ā/), it has a straightforward explanation: the fusion of the emphatic consonant with the corresponding non-emphatic ones. The emphatic consonants had prevented the shift from /ă/ and /ā/ to /ĕ/ and /ē/, respectively, but once they were lost, /ă/ and /ā/ became contextually independent. Variants became phonemes by a transfer of distinctive features from the consonants to the vowels.

5 The imaala Cohen (1970) also suggests that a pre-Maltese stage contained a high frequency of what is traditionally termed imaala in Arabic grammars, that is, the raising and fronting of the back vowel a/ā towards i/ī through an intermediate stage e/ē, for both long and short vowels, except when the Maltese back vowel was in contact with the back and the emphatic consonants. Cohen recognizes that there are exceptions that need to be accounted for and that are triggered by different processes such as analogy. Borg (1976: 220) rises to the challenge by

From Maltese phonology to morphogenesis: A tribute to David Cohen | 11

showing that “both vocalic and consonantal conditioning are likely to have played their part in the present-day situation”, and by demonstrating analogy for some patterns where the imaala was “grammatically conditioned”. He also discusses the possible reintroduction of a new layer of Arabic variety at a later stage which counteracted the spread of the imaala even in conditioning contexts, but considers this highly speculative. It comes as a natural conclusion of all the previous arguments developed by Cohen that the imaala paved the way for the subsequent phonologization of the contrast between a and e, and that it necessarily preceded the loss of the emphatic consonants which occurred at a Maltese stage, just as, more recently, the general weakening of the back consonants triggered an opening of the adjacent closest vowels. But for /ŭ/ the opening to /ŏ/ occurred in all contexts.

6 The Cantilena The above reconstruction of the phonological system of Maltese (together with other arguments) was used for a subsequent joint research project that David Cohen and I undertook in 1986 on the famous Cantilena of Peter Caxaro, which was published in 1991. This paper proposes some new interpretations of the text based on comparative, phonological, morphological, syntactic, philological, and stylistic criteria. The analysis of the spelling system, even though the spelling is not transparent, shows that the degree of evolution of the Maltese phonological system in the 15th century was by no means as distant from its origins as it is today. Still, it already showed at an incipient stage the signs of further evolution. At that time, Maltese had preserved all the back consonants: the uvular q, the unvoiced fricative laryngeal h, as well as the distinctions between the voiced and unvoiced fricative velars x and ħ on the one hand, and between the pharyngeals and ɣ and ʕ on the other, even though the latter was already weak in final position. I was later able to confirm, in an article about the allophones of għajn in the idiolect of an elderly speaker of the village of Mtaħleb who had preserved the fricative pharyngeal articulation, that the weakest position was actually the final root position where għajn was never pronounced as such (Vanhove 1991–1992). There is no clear evidence in the spelling system of the Cantilena that the emphatic consonants were preserved at that time, but there are clear indications of their influence on the quality of the adjacent vowels. For example, it was possible to reconstruct for the word tale, a verb *tala ‘go up’ (< ṭalaʕ), instead of taalin ‘then’ (a meaning not attested in Arabic) or talahawn ‘come here’

12 | Martine Vanhove

as proposed by Wettinger (1978). The form *tala is semantically and stylistically more adequate than the two hypotheses he proposed. Cowan (1975) may be correct in seeing, in the word timayt – also written tumayt and reconstructed as *ṭamaʕ ‘to hope’ – an indication that the emphatic consonants were still preserved. But it is possible that this word may go back to *ihtamma ‘to worry about’, since neither the h nor the gemination are systematically noted in the Cantilena, and since the meaning is not incompatible with the context (‘where I hoped to find stones’ or ‘where I was worried about stones’). Our translation followed Cowan’s, but with caution. At the phonetic level, the analysis of the Cantilena shows (i) that there was as yet no neutralization between voiced and unvoiced consonants in final position (cf. ard); (ii) that the imaala was an already existing phenomenon, but not as strong as in today’s standard Maltese (cf. gireni ‘my neighbors’); (iii) that the short i had not disappeared in open syllables except in deverbal forms with prefixes; (iv) that vowel harmony with o was already pervasive. The text of the Cantilena keeps part of its mystery: we were not able to solve all the problems, neither were we fully satisfied with all our hypotheses. Still, the application of a rigorous comparative method, and concern about the semantic accuracy of the proposed etymons and their possible semantic shifts, as well as the syntax of the verses, were important to avoid the pitfalls of unlimited imagination. For instance in verse 5, hayran (< xayraan) does not mean ‘willing, wanting’ but ‘embarrassed, troubled, worried’, an adjective often found in Arabic love songs with the meaning of ‘desperately in love’. Furthermore it is not a deverbal form followed by its complement as Wettinger and Fsadni (1968) suggest, but an adjective with a stative meaning. Similarly minzeli does not mean ‘my fall’ but literally ‘the place where I go down, where I have to go’, and in this context metaphorically, ‘my destiny’. We thus propose a very different translation: ‘Where desperately in love, I go down the steps of my destiny in order to drown myself’.

7 Conclusion After these papers, David Cohen did not continue working on Maltese himself, but encouraged his students to do so. As a conclusion, I would like to acknowledge the fact that his influence on Maltese studies went far beyond the domain of phonology, at least as far as I am concerned, through his work on aspect, auxiliaries, and the morphogenesis of verbal systems. Regarding auxiliaries, Cohen never managed to express his thoughts fully in the form of a book.

From Maltese phonology to morphogenesis: A tribute to David Cohen | 13

But he started circulating his research on this topic in the early 1980’s during his lectures at the Ecole Pratique des Hautes Etudes in Paris. These lectures were summarized and published in the annual reports of that institution, and are now freely available online, thanks to the efforts of Gilbert Puech, who for several years was at the head of a vast enterprise digitalizing French academic research for the internet portal Persée.3 Later, part of Cohen’s approach to morphogenesis was included in his book L’aspect verbal (1989).4 It explains very clearly how, across a large variety of languages from different phyla, aspectual systems are renewed in cycles which split the two basic aspects, perfective and imperfective. These new divisions tend to disappear, allowing a new two-way distinction to reemerge, different from the initial one. I strongly encourage linguists working in this domain to read it, as well as his two last books, the first mainly dedicated to morphogenetic processes, the second to sociolinguistics: Essais sur l’exercice du langage et des langues. Vol. 1: Communication et langage (2006), and (with Aziza Boucherit) Le langage, les langues et les nécessités de la communication (2010).

References Aquilina, Joseph. 1973/1959. The structure of Maltese: A study in mixed grammar and vocabulary. Malta: The Royal University of Malta. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London: Routledge. Borg, Alexander. 1976. The Imaala in Maltese. Israel Oriental Studies 6. 191–223. Borg, Alexander. 1977. Reflexes of pausal forms in Maltese rural dialects? Israel Oriental Studies 7. 211–225. Borg, Alexander. 1996. On some Levantine linguistic traits in Maltese. Journal of Oriental Studies. Studies in Modern Semitic Languages 16. 133–152. Brame, Michael K. 1972. On the abstractness of phonology: Maltese ʕ. In Michael K. Brame (ed.), Contributions to generative phonology, 22–62. Austin & London: University of Texas Press. Camilleri, Antoinette & Martine Vanhove. 1994. A phonetic and phonological description of the Maltese dialect of Mġarr (Malta). Zeitschrift für arabische Linguistik 28. 87–110. Cohen, David. 1957–1960. Sur le vocalisme du maltais. Comptes Rendus du G.L.E.C.S. (Groupe Linguistique d’Etudes Chamito-Sémitiques) 8. 83–85. Cohen, David. 1963. Le dialecte arabe Hassānīya de Mauritanie. Paris: Klincksieck.

|| 3 http://www.persee.fr/web/ouvrages/home/prescript/fond/ephe. 4 This became the framework of my own research on the verbal system of Maltese (Vanhove 1991, 1993, 2003; Vanhove and Simeone-Senelle 1997).

14 | Martine Vanhove

Cohen, David. 1964–1975. Le parler arabe des Juifs de Tunis. Tome 1: Textes et documents linguistiques et ethnographiques. Tome 2: Etude linguistique. La Haye/Paris: Mouton. Cohen, David. 1967. Contribution à la phonologie diachronique du maltais. In Ludwig E. Schmitt (ed.), Verhandlungen des zweiten internationalen Dialektologenkongresses, 164–171. Wiesbaden: Steiner. Cohen, David. 1970. Études de linguistique sémitique et arabe. La Haye/Paris: Mouton. Cohen, David. 1970–99. Dictionnaire des racines sémitiques ou attestées dans les langues sémitiques comprenant un fichier comparatif de Jean Cantineau. 10 volumes. Paris/La Haye: Mouton and Paris: Maisonneuve et Larose. Cohen, David. 1973–74. Résumé des conférences. Annuaire de l’École Pratique des Hautes Études, IVe section. 175–191. Cohen, David. 1983. Rapport 1982–1983. Les auxiliaires dans l’expression périphrastique du parfait. Annuaire de l’École Pratique des Hautes Études, IVe section. 58–59. Cohen, David. 1984. La phrase nominale et l’évolution du système verbal en sémitique. Étude de syntaxe historique. Leuven/Paris: Peeters. Cohen, David. 1989. L’aspect verbal. Paris: Presses Universitaires de France. Cohen, David. 2006. Essais sur l’exercice du langage et des langues. Vol. 1: Communication et langage. Paris: Maisonneuve et Larose. Cohen, David & Aziza Boucherit. 2010. Le langage, les langues et les nécessités de la communication. Conversations à Censier. Limoges: Lambert Lucas. Cohen, David, Jean Dubois, Michel Gauthier, Henri Hecaen & R. Angelergues. 1963. Aspects du fonctionnement du code linguistique chez les aphasiques moteurs. Neuropsychologia 1. 165–177. Cohen, David & Michel Gauthier. 1965. Aspects linguistiques de l’aphasie. L'Homme. Revue française d'anthropologie 5. 5–31. Cohen, David & Martine Vanhove. 1991. La cantilène maltaise du 15ème siècle: Remarques linguistiques. Comptes rendus du G.L.E.C.S. (Groupe Linguistique d’Études Chamitosémitiques) 29/30, 1984–1986 (conférence of 19 March 1986). 177–220. Cowan, William. 1966. Loss of emphasis. Journal of Maltese Studies 3. 27–32. Cowan, William. 1975. Caxaro’s Cantilena: A checkpoint for change in Maltese. Journal of Maltese Studies 10. 4–10. Firth, John Rupert. 1957. Papers in linguistics 1934–1951. London: Oxford University Press. Heine, Bernd. 1993. Auxiliaries. Cognitive forces and grammaticalization. Oxford: Oxford University Press. Lentin, Jérôme & Antoine Lonnet (eds.). 2003. Mélanges David Cohen. Études sur le langage, les langues, les dialectes, les littératures, offertes par ses élèves, ses collègues, ses amis. Présentés à l’occasion de son quatre-vingtième anniversaire. Paris: Maisonneuve et Larose. Puech, Gilbert. 1979. Les parlers maltais. Essai de phonologie polylectale. Thèse d’Etat. Université de Lyon II. Puech, Gilbert. 1994. Ethnotextes maltais. (Studia Melitensia 1). Wiesbaden: Harrassowitz. Schabert, Peter. 1976. Laut- und Formenlehre des Maltesischen anhand zweier Mundarten. Erlangen: Palm & Enke. Stumme, Hans. 1904. Maltesische Studien. Eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache nebst Erläuterungen. Leipzig: J.C. Hinrich’sche Buchhandlung, coll. Leipziger Semitische Studien, I, 4. Sutcliffe, Edmund F. 1960/1936. A grammar of the Maltese language: With chrestomathy and vocabulary. La Valette: Progress Press.

From Maltese phonology to morphogenesis: A tribute to David Cohen | 15

Vanhove, Martine. 1991. L’expression du parfait en maltais. In Alan S. Kaye (ed.), Semitic studies in honor of Wolf Leslau on the occasion of his eighty-fifth birthday, November 14th, 1991. Vol. 2, 1601–1618. Wiesbaden: Harrassowitz. Vanhove, Martine. 1991–92. On the survival of [ʕ] in a Maltese idiolect at Mtaħleb in Malta. Journal of Afroasiatic Languages 3. 22–34. Vanhove, Martine. 1993. La langue maltaise. Études syntaxiques d'un dialecte arabe « périphérique ». Wiesbaden: Harrassowitz. Vanhove, Martine. 2003. Auxiliaires, grammaticalisation et transcatégorialité: Le cas des marqueurs de futur et d’injonction en maltais. In Stéphane Robert (ed.), Perspectives synchroniques sur la grammaticalisation. Polysémie, transcatégorialité et échelles syntaxiques, 149–162. Louvain/Paris: Peeters. Vanhove, Martine & Marie-Claude Simeone-Senelle. 1997. La formation et l’évolution d’auxiliaires et particules verbales dans des langues sémitiques: les langues sudarabiques modernes et le maltais. Mémoires de la Société de Linguistique de Paris. Grammaticalisation et Reconstruction. Nouvelle série 5. 85–102. Wettinger, Godfrey. 1978. Looking back on ‘The Cantilena’ of Peter Caxaro. Journal of Maltese Studies 12. 88–105. Wettinger, Godfrey & Michael Fsadni. 1968. Peter Caxaro’s Cantilena, a poem in Medieval Maltese. Malta.

| Part I: Phonology

Gilbert Puech

Minimalist representation of Maltese sounds Abstract: This paper develops an Element-based representation of Maltese sounds in a model derived from lateral CV-phonology framework. Morpholexical information is converted into linearized chains of phonological expressions (PE), articulated on three layers: prosodic, structural and melodic. On the structural layer, PEs are represented by monovalent elements {C} and {V} on three tiers. The melodic layer has elements {I} and {U} on the same tier, and {A} on another one. If two elements are present on the same layer, either one is the head of the expression. Categories which are defined by one element only may be subcategorized by a complementary element used as specifier. C- and Vpositions may be empty. The sound pattern of Maltese includes glides j and w, and ‘amodal’ glides, corresponding to graphemes h and għ. Vowels are simplex, or complex if they incorporate an empty position into a filled nucleus. Length is not a phonological prime. Stressed complex vowels are realized long. Pausal complex vowels may be diphthongized in dialectal Maltese. We conclude by highlighting major characteristics of a LPP (lateral prosodic phonology) model. Keywords: Maltese phonology, prosodic templates, lateral prosodic phonology (LPP), stem syllabic alternations, element theory, specifier elements, amodal glides, pausal diphthongs, word stress and vowel length

1 Introduction Major contributions have been devoted to Maltese phonology in the 1970s; in particular Cohen (1970) in a structural framework (see Vanhove, this volume) and Brame (1972), who focused on the phonological cycle and abstractness in generative phonology. In the 1990s, metrical theory (Angoujard 1990, Hayes 1995) and feature geometry (Hume 1994) brought new perspectives within nonlinear frameworks. In this paper, I propose a model of representation of sounds influenced by recent theories; in particular: distributed morphology in phases, cf. Halle and Marantz (1993), Wolf (2013, 2016); minimalist phonology, cf. Samuels (2011); strict CV Phonology, cf. Lowenstamm (1996, 2011); flat and lateral

|| Gilbert Puech: Professor emeritus, University of Lyon; UMR du CNRS ‘Dynamique du Langage’ (DDL). E-mail: [email protected]

20 | Gilbert Puech

phonology, cf. Scheer (2004); linearization of directed graphs, cf. Idsardi and Raimy (2013); Radical CV phonology, cf. Hulst (1995, 2005) and Element theory, cf. Harris & Lindsey (1995) and Backley (2011), among others. The paper is organized as follows: Section 2 analyzes the linearization of morpholexical information and prosodic requirements for stem nuclei to surface. Sections 3 and 4 account for the structural and melodic layers of consonants. Section 5 is devoted to the phonological expression of four Maltese glides: palatal, labiovelar, guttural and melody-empty, which respectively correspond to letters j, w, għ and h in the official orthographic code. Section 6 accounts for the phonological expressions of monophthong vowels and of dialectal pausal diphthongs. Section 7 deals with word stress assignment and vowel length, in a model in which ‘length’ is not a phonological prime. We conclude by highlighting that the present lateral prosodic model of phonology (LPP) is strictly representational. Finally, all PE’s of the Maltese sound pattern are recapitulated in appendix.

2 Root-and-pattern stems In root-and-pattern Maltese morphology stems are composed of two units conforming to a pattern: a root formed by a sequence of 3±1 consonants and embedded vowels. Canonical patterns result from the interdigitation of two nuclear positions between radical consonants; cf. Goldenberg (1998). Stem consonants are constant for all lexical entries sharing the same root. For Ussishkin and Twist (2009) “roots facilitate lexical access”, which contributes to legitimating their synchronic status. In surface forms, a stem vowel is present or absent, depending on the prosodic template and on inserted affixes. For strict CV-Phonology, Lowenstamm (1996: 24) claims that: syllable structure universally, i.e. regardless of whether the language is templatic or not, reduces to CV. Directly confronting phonological objects typically deviating from the alleged unique type, I will argue that for all languages closed syllables, geminate consonants, and long vowels must be reanalyzed in terms of sequences of light open syllables.

Polgárdi (2012: 111) proposed a ‘loose CV’ framework “where representations are made up of strictly alternating C and V positions, but where words can start with a V position and end in a C position”. I combine this approach with the use of the symbol # for the left word-edge and of % for the right-edge; cf. Samuels (2011). In this framework, morphological information is linearized into a chain whose edges are also phonological positions. Thus, the following information:

Minimalist representation of Maltese sounds | 21

ROOT: VERB MEASURE 1 TEMPLATE

k→t→b ‘write’ CVCVC default value

PERFECTIVE, VOC: 3 MASCULINE, SINGULAR is linearized into the chain: # → k → v → t → v → b → %

The insertion of an inflectional suffix in the perfective conjugation entails alternations in the stem syllabic profile.

2.1 Stem suffixation (perfective verb in measure 1) Verbs agree in person, gender and number in the conjugation. The uninflected root-stem conveys the value of the 3rd person, masculine, singular in the perfective. In other cases, a suffix is inserted after the last stem consonant: (1a) (1b)

V or VC: C or CV:

it ~ et (3F.SG) t (1/2M.SG)

u (3PL) nie ~ na (1PL)

tu (2PL)

The suffix is anchored to the final consonant of the stem. If the suffix begins in a vowel, it is directly attached to the stem; if it begins in a consonant, an empty nucleus separates the final radical consonant from the suffix C in ‘CV Phonology’. We represent empty nuclei by the symbol ‘·’ (median dot): (2a) (2b)

1/2 M.SG 3F.SG

# → C V1 C V C → C → % # → C V C V2 C → V C → %

: C · C V C · C : C V C · C V C

cf. ktibt cf. kitbet

Stem vowels require prosodic licensing to surface. An empty nucleus occupies the stem first nuclear position in (2a) and the second one in (2b).

2.2 A prosodic approach to stem nuclei alternation In Maltese root-and-pattern morphology, autonomous stems are subject to a prosodic constraint of minimality: (3)

PROSODIC MINIMALITY: The stem prosodic template is minimally a binuclear simplex trochee or a mononuclear complex foot Σ.

A trochee is a prosodic relation between two filled nuclei: a head (Hd) and a right-adjacent dependent (Dp). This is the case in zero-suffix verbal forms:

22 | Gilbert Puech

(4)

Hd Dp # C1 V C2 V C3 %

cf. 3M.SG kiteb

C1 and C2 function as onset consonants followed by a filled nucleus. C3 is not followed by a filled or empty nucleus but by the termination symbol. This consonant, which is part of the prosodic (underlined) stem domain, closes the final syllable. However, it does not confer any ‘weight’ of its own onto the preceding vowel, which, thus, may function as trochee’s dependent; cf. Puech (to appear). If the stem is followed by a suffix vowel, the second stem nucleus (in bold) in open syllable cannot be qualified as the stem domain prosodic head in Maltese: (5a)

# C V C V C V %

The nuclear position is unmarked for {V}. In other terms, the nuclear position is occupied by an empty nucleus: (5b)

# C V C

·

C V %

When a filled nucleus is left-adjacent to an empty nucleus, the latter is prosodically merged (henceforth: p-merged) with the filled nucleus to yield a complex mononuclear Σ-foot; cf. Puech (to appear): (5c)

Σ p-merge # C V C · C V %

cf. 3PL kitbu

The stem last radical is followed by an empty nucleus if followed by a suffix consonant: (6a)

# C V C V C

·

C %

This empty nucleus is p-merged to yield a complex mononuclear Σ-foot: (6b)

Σ p-merge # C V C V C · C %

The 1st stem nuclear position in open syllable does not surface: (6c) # C

·

Σ p-merge C V C · C %

cf. 1/2SG ktibt

To sum up, a stem simplex nucleus surfaces as a full vowel if it is the stem prosodic head, or in a closed syllable, else it is represented by an empty nucleus.

Minimalist representation of Maltese sounds | 23

2.3 Comparison with rule-based approach by vowel syncope Since Brame (1972), the standard approach to syllabic structure in Maltese has been based on the interaction of two major rules: (i) stress assignment, (ii) syncope of unstressed short vowels in open syllable. According to Brame’s rule, stress is cyclically assigned (i) on the ultima if it is superheavy (or the word is monosyllabic), else (ii) on the penult, if it is heavy (or the word is bisyllabic), else on the antepenult; cf. Wolf (2016), Borg (1997: 272). This rule is modeled on the Palestinian stress rule; cf. Brame (1974: 39). Hume keeps the same stress rule even if the organization of the derivations is adapted to the autosegmental framework for the interaction between stress and syncope; cf. Hume (1994: 194 and 201). Some forms, for instance kitbítu ‘she wrote it-M’ may not be correctly stressed in Maltese by the ‘Palestinian’ rule; see Section 7.1 below. I would like to emphasize what, in my opinion, is the main drawback of Brame’s rule-based approach: rules are formulated as if any short vowel might be syncopated, provided it has not been stressed in the first or second cycle. But in fact, only stem vowels are subject to syncope, never suffix vowels. This is why I propose that, in the stem domain, a nuclear position surfaces as a filled nucleus if the prosodic conditions are met, else as an empty nucleus. A benefit of this approach is also to eventually eliminate the need for syncope rules such as ‘X → Ø’ and the converse Epenthesis format ‘Ø → X’, which confer an unlimited and out of control generative power to rules. Alternatively, we have an overall representational view in projecting stem nuclei from the lexical line L0 to line L1 in accordance with prosodic constraints; cf. Ulfsbjorninn (2014): (7a)

(7b)

Hd Dp V V # k V t V b % Σ p-merge V · # k V t V b V %

L1 kiteb L0 (7c) Σ p-merge L1 · V · kitbu L0 # k V t V b · C % ktibt

3 Structure elements {C} and {V} for consonants In this section, we define the constituency of phonological expressions (PE) assigned to consonants. In our model of lateral CV-phonology, {C} and {V} are

24 | Gilbert Puech

monovalent elements which anchor alternating skeletal positions; cf. Živanović (2004). These elements have a phonetic interpretation (articulatory events and their acoustic effects); cf. Hulst (1995). The element {C} is referred to a categorical gesture of closure, while the element {V} involves the opposite and is associated to resonance, voice periodicity and/or continuous airflow.

3.1 Major categories All consonants imply a closure gesture, which is expressed by element {C} on the 1st tier, the vertex of a phonological expression. Major categories are expressed by a combination with element {V} on a second tier. If {C} and {V} are combined, the head of the expression is underlined. On this basis, we define three major categories for consonants: 1st tier Stops: 2nd tier

C

Fricatives:

C V

Sonorants: C V

In obstruents, the element {C} is the head. In fricatives, element {V} on the 2nd tier refers to turbulent airstream produced by a constriction. For sonorants the {V} head is associated to resonance. We still have to examine the case of affricates and to account for the voiced vs. unvoiced contrast among obstruents.

3.2 Affricates There is a long tradition of accounting for affricates as either ‘contoured stops’ or ‘non-contoured stops’. In the former case, an affricate’s phonological expression combines a stop component followed by a fricative component linked to the same skeletal position; cf. Clements (1999: 273). In Jakobson et al. (1952) affricates are [+strident] stops contrasting with simple stops. This analysis is particularly relevant for homorganic affricates in absence of “edge effects”; cf. Clements and Hume (1995: 256). In Maltese stops, affricates and fricatives share the same place of articulation: alveolar for /t ~ ʦ ~ s/, post-alveolar for /ʧ ~ ʤ ~ ʃ/; cf. Borg and Azzopardi-Alexander (1997: 301–302). For obstruents, {V} on the 2nd tier interprets turbulent airflow. In stops the position is vacant, while it is filled by {V} in fricatives. If affricates are a subcategory of stops, we propose that this position is occupied by element {V} used as specifier. We represent the specifier element as {V} (small capital): 1st tier Stops: 2nd tier

C

Fricatives:

C V

Affricates:

C V

Minimalist representation of Maltese sounds | 25

3.3 Voice in obstruents and voice agreement Except in positions of neutralization, voicing is contrastive in Maltese for obstruents. A third tier of structure is required to account for the voiced vs. unvoiced contrast in couples like /t ~ d/ or /s ~ z/. A new element {voice} may be introduced by the theory. Alternatively, element {V} may characterize voicing on the third tier. The position is left vacant for their unvoiced counterpart: 1st tier Stops: C ~ C 2nd tier V 3rd tier V

Fricatives: C ~ C V V V

Affricates:

C ~ C V

V

V

Some obstruents are always unvoiced, without voiced counterpart. This is the case of the glottal stop ʔ, or its dialectal variant q, and of the pharyngeal fricative ħ. This is also the case of affricates ʦ and ʧ, which may be considered as having a special status of “phonaesthemes”; cf. Borg (1997: 263). We represent these sounds with the element {C} on the 3rd tier: 1st tier Stop ʔ or q: 2nd tier 3rd tier

C

Fricative ħ:

C

C V C

Affricates ʦ and ʧ: C V

C

A word-final voiced obstruent is realized voiceless. Thus, the 3rd tier position is occupied by specifier element {V}, corresponding to IPA voiceless diacritic: (8)

kiteb gireż ħareġ

write:PERF.3M.SG lament:PERF.3M.SG go out:PERF.3M.SG

# kiteb % : # girez % : # ħareʤ % :

[kɩteb̥] cf. kitbu [gɩrez̥] cf. girżu [ħared̥ʒ] cf. ħarġu

In obstruent /CiCj/ clusters, there is regressive voice agreement: (9a) (9b)

ħobż ktibt

bread.NOUN:M.SG write:PERF.1/2SG

# ħob·z % : # k·tib·t % :

[ħob̥z̥] [ktib̥t]

cf. ħobza

In (9), underlyingly voiced Ci agrees with voiceless Cj: the 3rd tier position is occupied by specifier element {V} corresponding to IPA voiceless diacritic. In (10a), underlyingly voiceless Ci agrees with voiced Cj. If, however, Ci is marked for {C} on the 3rd tier, there is no agreement with voiced Cj. (10a) kitbu (10b) qbilt

write:PERF.3PL agree. PERF.1/2SG

# kɩt·bu % : # ʔ·bil·t % :

[kit̬bu] [ʔbilt]

cf. qabel

Table 1 recapitulates the representation of obstruents for the structure layer. It includes the fricative /ʒ/, which occurs in some loan-words independently of voicing agreement; cf. Borg & Azzopardi-Alexander (1997: 303).

26 | Gilbert Puech

Tab. 1: Phonological expressions of mainstream Maltese obstruents (structure layer).

Segment p

b

f

v

t

d

s

z

ʦ

ʣ

ʃ

ʒ

ʧ

ʤ

k

g

ħ

ʔ

PE

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

V

V

V

V

V

V

V

V

V

V

V

C

V

V

C

V

C

V

V

V

V V

C

C

3.4 Nasal and liquid sonorants Some versions of Element Theory refer to a specific element of nasality {N} to classify nasal consonants apart from liquids /l/ and /r/; cf. Backley (2011: 145). Since sonorants, being {V}-headed, are intrinsically voiced, the contrast between interrupted or free oral airflow may be supported by the element {C} on the third tier for nasals. l and r are voiced coronal sonorants; l is a lateral approximant, r is realized as a trill, a flap or a retracted alveolar approximant; cf. Borg (1997: 256), Borg & Azzopardi-Alexander (1997: 302), Puech (1994). In Maltese, like in many Arabic dialects, if the lateral sonorant stands for the definite article, it assimilates to a coronal stem-initial consonant (except /ʤ/); cf. Sutcliffe (1936: 18). This militates for leaving the 3rd tier position vacant for l. The contextual realization of Maltese r depends on individual speech habits and dialectal variation. According to Ladefoged & Maddieson (1996: 216), We may say that in languages with consonant clusters, rhotics tend to occur close to the syllable nucleus … Frequently they share this privileged position with lateral approximants, and/or nasals. The affinity between rhotics and vowels is apparent in a number of other ways as well.

On these grounds, I am proposing that for r the position is occupied by specifier element {V}, while it is occupied by {C} in nasals and vacant for l: Tab. 2: Phonological expressions of mainstream Maltese nasals and liquids (structure layer).

Segment

m

n

r

l

PE

C

C

C

C

V

V

V

V

C

C

V

Minimalist representation of Maltese sounds | 27

3.5 Lost or dying guttural sonorants Maltese inherited from Arabic a voiced postvelar approximant ɣ and a voiced pharyngeal approximant ʕ (għajn represented by the digraph għ). The fate of these two sounds from postmedieval to modern standard and dialectal Maltese is a topic on its own and will not be dealt with in this paper. Suffice it to say that ɣ and ʕ were still attested in the 20th century; cf. Stumme (1904), Borg (1978 and 1997), Vanhove (1991), Puech (1994). In mainstream Maltese, ɣ and ʕ have merged and, according to Cohen (1970), became a ‘virtual’ segment; cf. Vanhove (this volume). Brame (1972: 51) “pinpoints the abstract segment” as being a “voiced pharyngeal spirant”. Contrary to Brame (1972: 60) I do not think that “the child coming to the language-learning situation is capable of inducing ʕ on the basis of Maltese phonetics alone”. In Section 5, I analyze the reflex of ɣ/ʕ as a guttural glide. What is beyond any doubt, however, is that these consonants, when they were integrated as sonorants in the sound pattern, behaved as liquids and nasals with respect to stem syllabification constraints. The modern spelling system still attests this stage. Compare, for instance, stem orthographic alternations in the following forms: (11a)

‘to write’ kiteb nikteb niktbu

PERF.3SG IMPF.1SG IMPF.1PL

(11b) ‘to lose’ tilef nitlef nitilfu

(11c)

‘to play’ lagħab nilgħab nilagħbu

A medial CCC cluster is allowed in (11a) but broken in (11b) and (11c). Orthographic għajn in (11c) behaves as the liquid l in (11b).

3.6 Structure elements for nasal, oral and guttural sonorants The 3rd tier position is {C} for nasals, specifier element {V} for r, else is vacant: Tab. 3: Phonological expressions of Maltese sonorants (structure layer).

Segment

m

n

r

l

*ɣ

*ʕ

PE

C

C

C

C

C

C

V

V

V

V

V

V

C

C

V

28 | Gilbert Puech

4 Melodic elements {I} ~ {U} ~ {A} for consonants 4.1 The melodic layer for obstruents Maltese has five major areas of articulation for consonants: labial, alveolar, alveopalatal, (advanced) velar, guttural; see Aquilina (1959), Borg (1997), Borg & Azzopardi-Alexander (1997). The ‘labial’ area covers bilabial articulations, labio-dental articulations and the labio-velar articulation of the approximant w. The ‘coronal’ area covers articulations from the dento-alveolar to the postalveolar place. The (advanced) ‘velar’ area extends from post-palatal to velar articulations for k and g, depending on the context; see Borg and Azzopardi-Alexander (1997: 300–303). The ‘guttural’ area encompasses places of articulation from the oropharynx to the larynx inclusively. Three different active articulators have access to this area: the tongue body for retracted velar or uvular sounds, the tongue root or the epiglottis for pharyngeal sounds and the larynx for the glottal stop; see Laufer (1988), McCarthy (1991), and, for different realizations of /ħ/ in Maltese, Borg (1997: 260). Tab. 4: Place of articulation of Maltese consonants (q in italics is attested dialectally).

Category

Stop

Fricative

Labial

p/b

f/v

t/d

s/z

ʦ/ʣ

ʃ/ʒ

ʧ/ʤ

Affricate

Sonorant

Oral glides

m

w

Coronal Alveolar Alveopalatalal Velar (advanced)

n/l/r j

k/g

Guttural Postvelar

q

[x]

ʔ

[h]

ħ

Pharyngeal Glottal

The founding principle of Element Theory is that consonants’ melody is interpreted by the same three monovalent features: {I U A} as for vowels; cf. Harris and Lindsey (1995), Backley (2011), among others. Yet, we want to characterize four areas: labial, coronal, velar and guttural. Furthermore, the coronal area encompasses too many contrastive sounds to be referred to one melodic element only. First, let us observe that in Maltese there is no argument to combine elements {I}

Minimalist representation of Maltese sounds | 29

and {U} in the same phonological expression, while the element {A} may be combined with either {I} or {U}. This is the reason why {I} and {U} are seated on the same tier, while {A} is seated on a distinct second melodic tier. In some versions of the theory, a ‘neutral’ element {@} is thought of as “a blank canvas to which the colors represented by [A], [I] and [U] can be applied”; cf. Harris and Lindsey (1995). I propose an alternative approach. On the {A} melodic tier, the position may be vacant, marked for {A}, or occupied but the specifier element represented as {A}. In the latter case, the specifier element plays the role of the terminal node [-anterior] for coronals in Hume’s analysis; cf. Hume (1994: 225). In Jakobson et al. (1952: 30), “gravity characterizes labial consonants as against dentals, as well as velars vs. palatals”. According to Backley (2011: 79) “the same groupings can be expressed by employing a geometry such as the one shown [below], where a peripheral node dominates [labial] in labial consonants and [dorsal] in velar consonants”: place peripheral [labial]

coronal

[dorsal]

In Jakobson et al. (1952: 28), labials and (dento)alveolars are [diffuse], while (alveo)palatals and velars are [compact]. In Table 5, coronals are marked for {I} and peripherals for {U}. In ‘compact’ postalveolars [ʃ, ʒ, ʧ, ʤ] the melodic 2nd tier position is occupied by element {A} used as specifier, while it is vacant for (dento)alveolars. Labials and velars are marked for {U}. Compared to labials, velars are rather ‘compact’. This position is similarly occupied by melodic specifier element, while it is vacant for labials. The melodic 2nd tier specifier is represented by {A} (small capital). Tab. 5: Phonological expressions of Maltese obstruents (including the dialectal variant q for ʔ).

Segment p b

f

v

t

d

s

z

ʦ

ʣ

ʃ

ʒ

ʧ

ʤ

k

g

*q

ħ

ʔ

Structure C C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

V

V

V

V

V

V

V

V

V

V

V

C

V

V

C

V

I

I

I

V Melody

U U

V U

U

V I

I

I

V V

C

I

I

I

I

U

U

U

A

A

A

A

A

A

A

C

C

A

A

30 | Gilbert Puech

4.2 The melodic layer for sonorants Labial and coronal nasals are characterized by the melodic element {U} and {I} respectively, and similarly the labiovelar and palatal glides. The rhotic liquid is realized as an apical trill, a flap or, particularly in some villages, a retracted alveolar approximant. The lateral liquid is, in general, an apico-alveolar approximant. If we characterize these two coronal sounds with the element {I}, we need a secondary distinction between them. I do not adopt the views according to which coronal sounds are characterized by either the element {I} or {A}; cf. contra Broadbent (1997, 1999) and Backley (2011: 3.6.2). On another hand, in Maltese and in some Arabic dialects, r partially behaves as a guttural sound. In Maltese, this is the case, for instance, in bi-radical verbs whose stem pattern in the perfect is CC. The vocalism is in general ie, as in bies ‘he kissed’, ried ‘he wanted’; in a few cases ie freely alternates with ā, as in sāb ~ sieb ‘he found’. If, however, the second radical is /r/, the inclination (imaala) of ā to ie has been prevented, as in dār ‘he turned’, fār ‘he grew angry’, mār ‘he went’, sār ‘he cooked’, tār ‘he flew’, żār ‘he visited’. There are also alternations like resaq/ jersaq ‘to approach’ in which the timbre e of the first vowel implies a ‘guttural’ behavior of r; cf. Puech (2014: 81). On these grounds, I propose for the sound r that the 2nd tier is occupied by the melodic specifier {A}. Thus, sounds [n, l, r] are still classified as coronal sonorants, but the particular behavior of r is accounted for; on r' PE, cf. Scheer (1998). Table 6 recapitulates the phonological expressions of Maltese sonorant consonants. Tab. 6: Phonological expressions of Maltese sonorants (including dying ɣ and ʕ).

Segment

m

n

r

l

*ɣ

*ʕ

Structure

C

C

C

C

C

C

V

V

V

V

V

V

C

C

V

U

I

I

U

Melody

I A

A

A

Minimalist representation of Maltese sounds | 31

5 Glides In phonology, glides traditionally refer to a category of sounds which show up as a voiced approximant in some contexts and as vowel in others. This is clearly the case for j/i and w/u alternations in Maltese. These glides are realized as approximant sonorants in C-position and as vowels in nuclear position. Thus, we analyze the vertex (1st tier position) as empty in the morpholexical representation, and occupied by {C} or {V} in actual utterances. The palatal glide is characterized by element {I} and the labiovelar glide by {U}.

5.1 {I} / {U}-melody glides The 3M.SG person prefix in the imperfect is realized j before a vowel and i before a consonant: (12a) jikser (12b) ikisser

break:IMPF.3M.SG smash:IMPF.3M.SG

(MEASURE 1) (MEASURE 2: INTENSIVE)

Some stems have the labiovelar glide as 1st radical. This glide is realized as w if the initial stem vowel is present, else as u: (13a) wera (13b) uriet

show:PERF.3M.SG show:PERF.3F.SG

In (12) and (13), the vertex position is filled by the element {C} before a filled nucleus and by the element {V} before a consonant. Thus, glides may be underlyingly represented with an empty vertex on the structural layer and {I} or {U} on the melodic tier. Once the form is linearized, the vertex position is filled by {C} or {V}, according to the context.

5.2 {A}-melody amodal glide In contemporary Maltese, we analyze għajn and h as ‘amodal’ glides whose morpholexical vertex is similarly empty. A glide is amodal1 if it has no target sound but is realized in distinct modalities. Let us first analyze the case of għajn.

|| 1 I borrow the term from Ladd (2014: 23), where it is applied to graphemes. Needless to say, it is intended to capture the same phonological complexity as Cohen’s “segment virtuel” and Brame’s “abstract segment”.

32 | Gilbert Puech

Some stems seem to be built after a ‘CVC’ pattern rather than ‘CVCVC’: (14)

għamel

do:PERF.3M.SG

:

[ā́mel]

Some speakers keep the initial vowel long even if it is unstressed, others not: (15)

għamilt

do:PERF.1/2SG

:

[āmílt] / [amílt]

Initial vowel length results from the fusion of the glide with the first stem nucleus: (16a)

għ a m e l # · V C V C A A [ ā m e l

(16b) % ]

għ a m i l # · V C V C A A [ ā m í l

·

t C % t

]

For speakers who no longer keep the initial vowel long if it is unstressed, there is no independent evidence that an analysis of the template as ‘·VCVC’ is to be preferred over the pattern ‘CVC’ for these stems (or rather ‘V̇CVC’, see below 6.2). Concerning verbs whose 3M.SG form in the perfect is ‘CC’, there are two quite distinct paradigms; cf. Sutcliffe (1936), Camilleri (2014: 60): (17a) sâb sâbet sibt sibna sîbu

[sāb̥] find PERF [sābet] [sib̥t] [sibna] [sību]

3M.SG 3F.SG 1/2SG 1PL 3PL

(17b) lagħab lagħbet lgħabt lgħabna lagħbu

[lāb̥] play PERF [lābet] [lāb̥t/lab̥t] [lābna] [lābu]

Orthographic forms for the verb ‘to play’ match the phonological forms when the sound system encompassed the voiced pharyngeal approximant ʕ; see above. In contemporary Maltese, evidence for postulating the persistence of ʕ in the sound pattern no longer exists. Yet, the two templates in (17) have to be kept distinct. The template in (17a) is characterized as ‘C /V C’ (or rather ‘CV̇C’, see Section 6.2 below), with an ablaut alternation ā/i. The template in (17b) may be analyzed as ‘CVʕVC’, where (small capital) ʕ represents the guttural glide.

5.3 Melody-empty amodal glide In his Lexicon, Vassalli (1796: 232) writes about the letter h: “Questa lettera nel Maltese à suono aspirato”. For Agius de Soldanis (1750: 75) h is “gentilmente aspirato”. The sound, inherited from Arabic, is appropriately defined as a glottal fricative; cf. Laufer (1991). As in other Semitic languages, /h/ has [h], [ɦ] and zero as allophones. Agius de Soldanis (1750: 129), for instance, records the word for ‘gold’

Minimalist representation of Maltese sounds | 33

as deep (deheb in Vassalli 1796 & 1827). In the 19th century, /h/ is lost as a fullfledged fricative. In contemporary Maltese, graphic h has been kept where it is expected on etymological grounds, but has no target sound on its own. Radical h may be found in a limited number of stems such as: (i) First radical: hemeż ‘to attach’ [ēmes] (ii) Second radical: fehem ‘to understand’ [fēm] cf. fhimt PERF.1/2SG [fimt] xehed ‘to testify’ [ʃēt] cf. xhedt PERF.1/2SG [ʃēt] (iii) Third radical: koroh ‘ugly’ ADJ.PL [koroħ] cf. M ikreh, F kerha Observing that there is an e-to-i ablaut in the paradigm of fehem, but not for xehed, Camilleri (2014: 66) concludes that “nothing synchronically conditions or is able to account for the observed distinct paradigmatic behavior across these CV:C [h-medial] verbs”. Her remark may be extended to other cases: alternations in stems involving an etymological h are scarce and unstable. Quite to the opposite, alternations associated to the presence of orthographic h in the following object pronouns are regular and productive: -ha or hie OBJECT PRONOUN.3F.SG -hom OBJECT PRONOUN.3PL -hu OBJECT PRONOUN.3 M.SG in internal position (u in final position) Other object pronouns are: -na or nie OBJECT PRONOUN.1PL -kom OBJECT PRONOUN.2PL -ek, ik or ok OBJECT PRONOUN.2SG With respect to syllabic alternations in stems, h-suffixes behave as C-suffixes although the h is silent: (18a) ħelisna (18b) ħelisha

deliver:PERF.3M.SG:OBJ.1PL deliver:PERF.3M.SG:OBJ.3F.SG

: :

[ħelisna] [ħelisa]

In intervocalic position, h shows up as a glide homorganic to the close vowel: (19a) bnieha (19b) bluha

build: PERF.3M.SG:OBJ.3F.SG stupidity: NOUN:F.SG

: :

[bnɩ̄ja] [blūwa]

After a low vowel in (20), h does not show up either; the hiatus is avoided by fusion of the stem vowel with the suffix vowel; cf. Borg (1997: 274–276): (20)

rahom

see:PERF.3M.SG:OBJ.3PL

:

[rɔ̄m]

Orthographic h in modern Maltese is interpreted by an empty vertex.

34 | Gilbert Puech

In (21), the initial nucleus (in bold) is the stem trochee’s head in phase 1. In phase 2 (concatenation of an object pronoun), V2 is ‘protected’ by the prosodic merge with the empty position (standing for orthographic h), cf. above 2.2. (21a)

Σ p-merge # C V1 C V2 C · V % [ ħ e l í s a ] ħ e l i s h a ‘he set her free’

(21b)

Σ p-merge # C V1 C V2 C · C V % [ ħ e l í s n a ] ħ e l i s n a ‘he set us free’

In our model of CV-phonology, the prosodic merge between the empty position and V2 preserves the well-formedness of (21a) with respect to C/V alternations. In synchrony, some alternations have to be taken care of by morphophonological rules, which reflect historical developments. Thus, h is realized as [ħ] in word final position after a long vowel in (22) and after radical /ħ/ in (23): (22a) kitbuh (22b) koroh

write:PERF.3PL:OBJ.3M.SG ugly:ADJ.PL

: :

[kɩtbūħ] [koroħ]

(23)

open:PERF.3SG:OBJ.3F.SG

:

[fetaħħa]

fetaħha

cf. F kerha

These cases illustrate phonology-conditioned suppletive allomorphy in Maltese; cf. Mascaró (1996), Paster (2015), Wolf (2015).

5.4 Phonological expressions for glides In morpholexical representation, a glide’s vertex is left empty. The position, underlyingly occupied by ‘·’, is marked for {C} or {V} in linearized chains for oral glides. For guttural glides, the empty position is prosodically merged. Tab. 7: Morpholexical representation of glides.

Segment

j

w

għ

h

Structure

·

·

·

·

V

V

V

V

U

I

Melody

A

Minimalist representation of Maltese sounds | 35

6 Vowels and diphthongs Vowels are distributed between simplex and complex nuclei. Complex nuclei surface as long vowels in stressed position only: vowel-length requires prosodic licensing. Standard Maltese chiefly differs from dialects spoken in the villages of Malta and Gozo by specific constraints on vowels. In particular, complex nuclei in village dialects may be diphthongized in word-final pausal position.

6.1 Standard Maltese vowels Standard Maltese has five short vowels and six long vowels, as shown in Table 6 (notice that, as Borg 1997, I use the symbol ‘a’ with the value of a non-fronted low vowel). Readers are referred to Borg (1997: 264 and 268) for a list of minimal pairs between short vowels, on the one hand, and between long vowels on the other. Minimal pairs below are quoted after Azzopardi-Alexander (2003) for short lax ɩ, long ɩ̄, and tense (long) ī (stress is on the penultimate syllable): (24a) qalilha (24b) qalhielha (24c) qalila

ɩ ɩ̄ ī

ʔalɩla ʔalɩ̄la ʔalīla

say:PERF.3M.SG:IND:OBJ.3F.SG say: PERF.3M.SG:OBJ.3F.SG:IND:OBJ.3F.SG troublesome:ADJ.F.SG

Tab. 8: Short and long vowels in Standard Maltese.

Short vowels

Long vowels

tense Close Half-close Open

lax

ī ɩ

ʊ ɛ

ɔ a

ū

ɩ̄ ɛ̄

ɔ̄ ā

6.2 Phonological expressions of vowels As we said before, there is a distinction between simplex and complex vowels. Both have the element {V} as vertex on the first structure tier. Complex vowels incorporate an empty nuclear position. Thus, we will represent them as {V̇}. In this analysis, length is not a phonological prime. In Maltese, a complex vowel surfaces as long if and only if it bears the word stress.

36 | Gilbert Puech

On the melodic layer, in absence of front rounded vowels, {I} and {U} do not combine. Element {A} combines with {I} and {U}; the degree of openness depends on which element is the head of the expression: Tab. 9: Phonological expressions of Standard Maltese vowels.

Segment

ī

ɩ / ɩ̄

ɛ / ɛ̄

a/ā

ɔ / ɔ̄

Structure

V̇

V/ V̇

V / V̇

V / V̇

V / V̇

V

V̇

Melody

I

I

I

U

U

U

A

A

A

A

A

ʊ

ū

6.3 Monophthong vowels in dialectal Maltese Maltese village dialects have a ‘quadrangular’ system of monophthong vowels; the allophonic status of mid-vowels is dialect-dependent. Readers are referred to my contribution in Puech (2014), although a different architecture organizes the ‘minimalist’ representations that I propose here. Tab. 10: Monophthong vowels in dialectal Maltese (Malta and Gozo). Short vowels tense lax

Long vowels

ɩ

ʊ

ī̄ ɩ̄

Half-close

[ɛ]

[ɔ]

[ɛ̄]

[ɔ̄]

Open

æ

ɑ

ǣ

ɑ̄

Close

ū ʊ̅

Tab. 11: Phonological expressions of dialectal Maltese monophthongs (V̇’s are long if stressed).

Segment

ī

ɩ / ɩ̄

[ ɛ / ɛ̄ ]

[æ/ ǣ]

ɑ / ɑ̄

[ ɔ / ɔ̄ ]

ʊ / ʊ̅

ū

PE

V̇

V / V̇

V / V̇

V / V̇

V / V̇

V / V̇

V / V̇

V̇

I

I

I

I

U

U

U

U

A

A

A

A

A

A

Minimalist representation of Maltese sounds | 37

In her study of Naduri (a Gozitan village dialect), Azzopardi-Alexander (2011: 246) adds to this inventory the rounded close front vowel /y/, for instance in bût ‘pocket’ transcribed as /byt/. This representation, however, does not satisfy the minimality prosodic constraint on templatic stems; cf. 2.2. In fact, I consider the vowel in [byt] as a reduced realization of a pausal diphthong /eu/; see below.

6.4 Pausal diphthongs in dialectal Maltese Borg (1976) is the first linguist, as far as I know, to analyze the ‘pausal’ context in which word-final long vowels become diphthongized. An in-depth diachronic and synchronic analysis of specificities occurring in village dialects would go beyond the limits of the present contribution; see Aquilina and Isserlin (1981), Puech (1994). Suffice it to say that some Gozitan dialects exhibit all the potentialities of the system. Let us first compare some monosyllables in Standard and Gozitan Maltese: (25)

dnub dnub

‘sin’ ‘tails’

Standard: dnūb dnūb

Gozitan: dneub dnoub

(26)

tqil fqir

‘heavy’ ‘poor’

tʔı̄l fʔı̄r

tʔeil fʔoir

(27)

ktieb

‘book’

ktɩ̄b ~ ktɩəb

ktɩɛb

(28)

ħmar

‘donkey’

ħmār

ħmʊɔr ~ ħmɔ̄r

Pausal diphthongs are to be classified in two types: i. closing diphthongs: tense /i/ or /u/ target from lower and more central quality ii. opening diphthongs: from lax /ɩ/ or /ʊ/ to lower and more central target Targets aimed at by closing diphthongs are represented by elements {I} and {U} respectively; the starting point is situated between {I} and {A}, for prototypical realizations ei and eu, or between {U} and {A} for prototypical realizations oi and ou. The starting point for opening diphthongs is {I} for the prototypical realization of ɩɛ, and {U}, for prototypical realization ʊɔ. Trajectories are shown in Figure 1: I

U

A Fig. 1: Melodic trajectories of pausal diphthongs.

38 | Gilbert Puech

In our model, a complex vowel incorporates a prosodic empty nuclear position into a filled nucleus (on this mechanism, see Ulfsbjorninn 2014). In monophthongs, the targeted nucleus is immediately aimed at. In closing diphthongs, a phonologized onset delays the realization of the target {I} or {U}. The diphthong trajectory starts from a lower value towards the targeted head of the expression. We represent the phonological expression in underlining the diphthong’s head. The starting point is on a scale bounded by an upper value, {I} or {U}, and a lower value {A}. The outcome depends on several factors, such as the context, individual speech habits and village discriminant phonological and phonetic characteristics. In Gozitan Maltese the dependent part of the diphthong is subject to vowel harmony with the stem thematic vowel; cf. Puech (1978). In opening diphthongs, the trajectory is defined from the head {I} or {U} towards an undershot target (A}. There is no vowel harmony. For dialectal shades of realization, cf. Borg (1976, 1978, 1997), Puech (1978, 1994), Aquilina & Isserlin (1981), among others. Tab. 12: Phonological expressions of pausal diphthongs in dialectal Maltese.

Segment

ei

oi

eu

ou

ɩɛ

ʊɔ

PE

V̇

V̇

V̇

V̇

V̇

V̇

I,I

U,I

I,U

U,U

I,I

U,U

A

A

A

A

A

A

7 Word stress and vowel length In Maltese, word stress is assigned either to the ultimate or to the penultimate nucleus. Complex vowels are realized by a long allophone if they are stressed, else their allophone is short. In phonemic terms, vowel length is contrastive in Maltese. Actual vowel length depends on several factors, among which the persistent influence of a guttural glide, the word syllabic profile and, of course, intonation; cf. Hume et al. (2009), Vella (2009). Some loanwords keep an etymological stress on the word-final vowel, like in libertá, or an antepenultimate stress, like in kámera. In many cases, there is a ‘maltesized’ variant, e.g. kâmra; cf. Mifsud (1995) & MLRS (these entries).

Minimalist representation of Maltese sounds | 39

7.1 Word stress Stress is either on the penultimate or the final filled nucleus. Again, except for loanwords which have kept their etymological prosodic profile, stress is never antepenultimate in Maltese. The stressed vowel is the head of the word-final foot. If the ultimate vowel may not be the foot head stress is penultimate: (29a) (29b) (29c) (29d) (29e)

ħobża kiteb kitbu kitibhom kitibhomlu

bread:NOUN.F.SG write:PERF.3M.SG write:PERF.3PL write:PERF.3M.SG:OBJ.3PL write:PERF.3M.SG:OBJ.3PL:IND.OBJ.M.SG

[ħɔ́bza] [kíteb] [kítbu] [kitíbom] [kitibómlu]

The case of the 3rd person feminine is particularly interesting: (30a) (30b) (30c)

kitbet kitbitu kitbitek

write:PERF.3F.SG write:PERF.3F.SG:OBJ.3M.SG write:PERF.3F.SG:OBJ.2M.SG

[kítbet] [kitbítu] [kitbítek]

Brame’s stress rule assigns stress to the antepenultimate vowel; then, the syncope rule deletes the short medial vowel. In (30b, c) Brame’s rules yield the wrong outputs *kitibtu and *kitibtek if the syncope rule is applied right-to-left or *kitibtu and *kitibtek if syncope applies simultaneously to both the second stem vowel and the suffix vowel. In our approach, V2 in phase 1 is reduced to an empty nucleus in open syllable. Suffix vowels are never subject to reduction. The nucleus in the suffix it is stressed, being the head of the word-final foot. If the ultimate vowel is the word-final foot head, it is stressed: (31a) (31b) (31a)

berikt (ma) kitibx miktub

bless:PERF.1/2SG write:PERF.3M.SG:NEG write:PAST P.3M.SG

[beríkt] [kitíbʃ] [mɩktū́b]

7.2 Short and long allophones for complex nuclei A morpholexical complex nucleus is realized as a long vowel if it is stressed, else as a short one. In the following example, the plural suffix is the complex vowel /u̇/. This complex vowel is realized short it is not stressed; if it is the head of the word-final foot, it is stressed and realized long: (32a) (32b) (32c)

kitbu (ma) kitbux kitbuli

write:PERF.3PL write:PERF.3PL:NEG write:PERF.3PL:IND.OBJ.1SG

[kítbu] [kitbū́ʃ] [kitbū́li]

40 | Gilbert Puech

(32d)

kitbuhomli

write:PERF.3PL:OBJ.3PL:IND.OBJ.1SG

[kitbuwómli]

Notice that vowel length is not analyzed as a phonological prime. The phonological representation corresponding to (32) are: (33a)

kítbu

:

(33b)

kitbū́ʃ

:

(33c)

kitbū́li

:

(33d)

w

Σ # C V C

·

# C V C

·

# C V C

·

# C V C

·

kitbu ómli :

C V̇ % Σ C V̇ C % Σ C V̇ C V̇ % Σ C V̇ · V C ·

C V̇ %

The stressed nucleus, shown in bold in (33), is penultimate or ultimate, depending on the prosodic position of the word-final foot head.

8 Conclusion In this paper, I have proposed a model which draws inspiration from different frameworks: the minimalist program in phonology, ‘CV phonology’ models, Element Theory, among others. For the representations of sounds, the model articulates three layers. The prosodic layer determines the stem foot and the word-final foot. The stem syllabic profile is determined in the 1st phase and is not modified in the 2nd phase (expansion by object pronouns and enclitic negation). Stress is assigned to the nuclear head of the word-final foot. Complex nuclei are realized as long vowels if they are stressed, else as short. The structure layer represents CV positions with elements {C} and {V} on three tiers. Major categories of consonants are determined by the combination of {C} and {V} on the two first tiers. A vertex (1st tier position) may be left vacant. This is the case for empty nuclei and underlying glides. In surface utterances, the position may be occupied by {V} or {C}, prosodically merged with a preceding full-fledged position, or filled by epenthesis. On the 3rd tier, obstruent clusters agree in voicing. The melodic layer represents the ‘color’ of vowels and consonants with prime elements {I, U} on a first tier and {A} on a distinct tier. Elements may be inserted or copied in two types of contexts: adjacency and head-dependent lateral relation. For instance, the former case accounts for spreading {A} from a guttural consonant to an adjacent nucleus. Evidence for the latter case is confirmed by vowel harmony (VH) linking a radical nuclear position to the word-

Minimalist representation of Maltese sounds | 41

final nuclear position of a suffix vowel (progressive VH) or the word-initial nuclear position (regressive VH). A model of representation in which the number of ‘elements’ is maximally reduced needs nevertheless to account for all relevant distinctions. There is a trade-off between less elements and more structure, or more elements and less structure. In our analysis, primitive elements are {C} and {V} for the structure layer, and {I/U} and {A} for the melodic layer. We also define two specifier elements: {V} and {A}. The value attributed to {V} depends on which tier it occupies a position, but in all cases is related to element {V}’s basic properties. Similarly, {A} is postulated in cases where some attenuated effects correlated to {A} are ‘visible’. For instance, the timbre of lax [ɩ] and [ʊ] is kept distinct from that of their tense counterpart [i] and [u] in combining {I} and {U} respectively with {A}. Finally, pre-modern Maltese (18th century) had a full set of guttural sounds, including two sonorants and one glottal fricative. In modern Maltese, we borrowed the term ‘amodal’ to qualify the phonological expressions of guttural glides. In contemporary Maltese, however, due to the loss of phonetic contrasts, conditioned allomorphy and paradigm leveling, it is doubtful that “children coming to the language-learning situation” may still induce autonomous PEs for these sounds. Borrowing the term from Spagnol (2011), I would rather say that it is a tale of another morphology. In contributing to an analysis of phonological representations in Maltese, I tried to lay out some foundations for a model of Lateral Prosodic Phonology.

Abbreviations 1, 2, 3 ADJ

Dp F

Hd IMPF IND M NEG OBJ

PE PERF PL

p-merge SG

1st, 2nd, 3rd person adjective dependent feminine head imperfect indirection masculine negator object phonological expression perfect plural prosodic merge singular

42 | Gilbert Puech

Appendix 1: The Maltese alphabet The letters are listed in the order adopted in Aquilina’s dictionary (1987–1990) with their IPA correspondence and definition. See Borg (1997), Borg & Azzopardi-Alexander (1997: Section 3), Azzopardi-Alexander (2003) and Aquilina & Isserlin (1981) for local dialects. Letters a, A b, B ċ, Ċ d, D e, E f, F ġ, Ġ g, G h, H ħ, Ħ i, I i ie, Ie j, J k, K l, L m, M n, n għ, Għ o, O p, P q, Q r, R s, S t, T u, U v, V w, W x, X ż, Ż z, Z

IPA symbol [ɐ] or [ɐ̄] [b] [ʧ] [d] [ɛ] or [ɛ̄] [f] [ʤ] [g] [x, ħ or h] [i, ī] [ɩ] [ɩ̄, ɩə] [j] [k] [l] [m] [n] [ɔ] [p] [ʔ] [r, ɾ] [s] [t] [ū] [ʊ] [v] [w] [ʃ] [z] [ʦ, ʣ]

Standard Maltese Dialectal sounds Low central vowel, short or long Fronted [æ], backed [ɑ] Voiced bilabial stop Voiceless alveopalatal affricate Voiced dento-alveolar stop Mid front vowel, short or long Voiceless labiodental fricative Voiced alveopalatal affricate Voiced palatovelar stop Amodal* melody-empty glide Voiceless guttural fricative Tense high front vowel, long or short Pausal diphthongs ei/oi Lax high front vowel (short) Lax high front vowel (long) Pausal diphthong [ɩɛ] Palatal glide Voiceless palatovelar stop Alveolar lateral sonorant Bilabial nasal sonorant Alveolar nasal sonorant Amodal* {A}-melody glide Dying sonorants: [ɣ], [ʕ] Mid back rounded vowel (short or long) Voiceless bilabial stop Voiceless glottal stop Postvelar stop [q] Voiced alveolar trill or flap Approximant [ɹ, ɻ] Voiceless alveolar fricative Voiceless dentoalveolar stop Tense rounded high back vowel Pausal diphthongs ou/eu Lax high back round short vowel [ʊ̄] and pausal dipht. [ʊɔ] Voiced labiodental fricative Voiced labiovelar glide Voiceless alveopalatal fricative Voiced alveolar fricative Voiceless or voiced alveolar affricate

* A letter is amodal if it has no target sound but is realized in distinct modalities; cf. Ladd (2014).

Minimalist representation of Maltese sounds | 43

Appendix 2: Phonological expressions (PE) Empty position (vertex) and specifier elements {V} (structure layer) & {A} (melodic layer):

1st tier:

STRUCTURE

MELODIC

‘·’

in nuclear position:

empty nucleus

in consonant position:

glide’s underlying vertex

2nd tier: {V}

affricates (stop subcategory): delayed fricative release

3rd tier: {V}

for obstruents: for sonorants:

regressive voicing assimilation r-manner of articulation

A- tier:

subcategorization: for obstruents

relative compactness: – (alveo)palatals vs. (dento)alveolars – (advanced) velars vs. labials – r vs. l and n – lax ɩ vs. tense i, lax ʊ vs. tense u

{A}

for sonorants: for vowels:

Phonological expressions of obstruents

Segment p

b

f

v

t

d

s

z

ʦ

ʣ

ʃ

ʧ

ʤ

k

g

*q

ħ

ʔ

*h

Structure C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

C

V

V

V

V

V

V

V

V

V

V

C

V

C

V

I

I

I

I

I

I

A

A

A

V Melody

U

U

V U

U

V I

I

I

Phonological expressions of sonorants

Segment

m

n

r

l

*ɣ

*ʕ

Structure

C

C

C

C

C

C

V

V

V

V

V

V

C

C

V

U

I

I

U

Melody

I A

A

A

V V

C

U

U

U

A

A

A

V

C

C

A

A

V

44 | Gilbert Puech

Phonological expressions of glides

Segment

j

w

għ

h

Structure

·

·

·

·

V

V

V

V

U

I

Melody

A

Phonological expressions of Standard Maltese vowels

Segment

ī

ɩ/ɩ̄

ɛ/ɛ̄

a/ā

ɔ/ɔ̄

Structure

V̇

V/V̇

V/V̇

V/V̇

V/V̇

V

V̇

Melody

I

I

I

U

U

U

A

A

A

A

A

ʊ

ū

Phonological expressions of dialectal Maltese monophthongs

Segment

ī

ɩ/ɩ̄

[ ɛ/ɛ̄ ]

[ æ/ǣ ]

ɑ/ɑ̄

[ ɔ/ɔ ]

ʊ/ʊ̄

ū

PE

V̇

V/V̇

V/V̇

V/V̇

V/V̇

V/V̇

V/V̇

V̇

I

I

I

I

U

U

U

U

A

A

A

A

A

A

Phonological expressions of pausal diphthongs in dialectal Maltese

Segment

ei

oi

eu

ou

ɩɛ

ʊɔ

PE

V̇

V̇

V̇

V̇

V̇

V̇

I,I

U,I

I,U

U,U

I,I

U,U

A

A

A

A

A

A

Minimalist representation of Maltese sounds | 45

References Agius de Soldanis, Giovanni Pietro F. 1750. Della Lingua Punica presentemente usata da Maltesi & c. overro Nuovi documenti, li quali possono servire di lume all'antica lingua etrusca; stesi in due dissertazioni &c. Roma. Published by Joe Zammit Ciantar, Facsimile edition: Malta 2007. Angoujard, Jean-Pierre. 1990. Metrical structure of Arabic. Dordrecht: Foris Publications. Aquilina, Joseph. 1959. The structure of Maltese: A study in mixed grammar and vocabulary. Malta: Royal University of Malta. Aquilina, Joseph. 1987 & 1990. Maltese-English dictionary. (2 volumes). Malta: Midsea Books. Aquilina, Joseph & Benedikt S. J. Isserlin. 1981. A Survey of contemporary dialectal Maltese. University of Leeds. Leeds: B.S.J. Isserlin. Azzopardi-Alexander, Marie. 2003. The vowel system of Maltese: from production to perception. AIDA 5th Conference Proceedings, Cádiz. Azzopardi-Alexander, Marie. 2011. The vowel system of Xlukkajr and Naduri. In Sandro Caruana, Fay Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 235–255. Berlin: Akademie Verlag. Backley, Phillip. 2011. An Introduction to element theory. Edinburgh: Edinburgh University Press. Borg, Albert J. & Marie Azzopardi-Alexander. 1997. Maltese: Descriptive grammars. London: Routledge. Borg, Alexander. 1976. Reflexes of pausal forms in Maltese rural dialects. Israel Oriental Studies 7. 211–225. Borg, Alexander. 1978. A historical and comparative phonology and morphology of Maltese. Israel: The Hebrew University. Borg, Alexander. 1997. Maltese phonology. In Alan S. Kaye (ed.), Phonologies of Asia and Africa. Winona Lake, Indiana: Eisenbrauns. Brame, Michael K. 1972. On the abstractness of phonology: Maltese ʕ. In Michael K. Brame (ed.), Contributions to generative phonology, 22–61. Austin: University of Texas Press. Brame, Michael K. 1974. The cycle in phonology: Stress in Palestinian, Maltese, and Spanish. Linguistic Inquiry 5(1). 39–60. Broadbent, Judith M. 1997. A review of Hume’s account of consonant vowel interactions in Maltese Arabic. In Pier Marco Bertinetto, Livio Gaeta, Georgi Jetchev & David Michaels (eds.), Certamen Phonologicum III. Papers from the Third Cortona Phonology Meeting, April 1996, 27–44. Torino: Rosenberg & Seller. Broadbent, Judith M. 1999. A new approach to the representation of coronal segments. In S. J. Hannahs & Mike Davenport (eds.), Issues in phonological structure, Papers from an international workshop, 1–25. Amsterdam: Benjamins. Camilleri, Maris. 2014. The stem in inflectional verbal paradigms in Maltese. Ph.D. dissertation. University of Surrey. Clements, George Nick. 1999. Affricates as noncontoured stops. In Osamu Fujimura, Brian D. Joseph & Bohumil Palek (eds.), Proceedings of LP '98: Item order in language and speech, 271–299. Prague: The Karolinum Press. Clements, George Nick & Elizabeth V. Hume. 1995. The internal organization of speech sounds. In John A. Goldsmith (ed.), The handbook of phonological theory, 245–306. Cambridge, MA: Blackwell. Cohen, David. 1970. Études de linguistique sémitique et arabe. La Haye: Mouton.

46 | Gilbert Puech

Goldenberg, Gideon. 1998. Studies in Semitic linguistics: Selected writings. Jerusalem: Magnes Press, Hebrew University. Halle, Morris & Alec Marantz. 1993. Distributed morphology and the pieces of inflection. In Kenneth Hale & Samuel J. Keyser (eds.), The view from building 20: Essays in linguistics in honor of Sylvain Bromberger, 111–176. Cambridge, Massachusetts: MIT Press. Harris, John & Geoff Lindsey. 1995. The elements of phonological representation. In Jacques Durand & Francis Katamba (eds.), Frontiers of phonology. Atoms, structures, derivations, 34–79. London: Longman. Hayes, Bruce. 1995. Metrical stress theory principles and case studies. Chicago, Ill./London: The University of Chicago Press. Hulst, Harry van der. 1995. Radical CV phonology: The categorial gesture. In Jacques Durand & Francis Katamba (eds.), Frontiers of phonology. Atoms, structures, derivations, 80–116. London: Longman. Hulst, Harry van der. 2005. The molecular structure of phonological segments. In Philip Carr, Jacques Durand & Colin J. Ewen (eds.), Headhood, elements, specification and contrastivity. Phonological papers in honor of John Anderson, 193–234. Amsterdam: Benjamins. Hume, Elizabeth V. 1994. Front vowels, coronal consonants, and their interaction in nonlinear phonology. (Outstanding dissertations in linguistics). New York: Garland Pub. Hume, Elizabeth, Jennifer Venditti, Alexandra Vella & Samantha Gett. 2009. Vowel duration and Maltese ‘għ’. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 15–46. Amsterdam: Benjamins. Idsardi, William & Eric Raimy. 2013. Three types of linearization and the temporal aspects of speech. In Theresa Biberauer & Ian Roberts (eds.), Challenges to linearization, 31–56. Berlin: De Gruyter Mouton. Jakobson, Roman, C. Gunnar M. Fant & Morris Halle. 1952. Preliminaries to speech analysis; the distinctive features and their correlates. Cambridge, Massachusetts: MIT Press. Ladd, Robert D. 2014. Simultaneous structure in phonology. Oxford: Oxford University Press. Ladefoged, Peter & Ian Maddieson. 1996. The sounds of the world’s languages. Oxford Cambridge, MA.: Blackwell. Laufer, Asher. 1988. The emphatic and pharyngeal sounds in Hebrew and in Arabic. Language and Speech 31(2). 181–205. Laufer, Asher. 1991. The ‘glottal fricatives’. Journal of International Phonetic Association 21(2). 91–93. Lowenstamm, Jean. 1996. CV as the only syllable type. In Jacques Durand & Bernard Laks (eds.), Current trends in phonology: Models and methods, 419–443. Manchester UK: University of Salford. Lowenstamm, Jean. 2011. The phonological pattern of phi-features in the perfective paradigm of Moroccan Arabic. Brill’s Annual of Afroasiatic Languages and Linguistics 3. 140–201. Maltese Language Resource Server (MLRS) – http://mlrs.research.um.edu.mt/resources/gabra/roots – http://mlrs.research.um.edu.mt/CQPweb/malti03/ Mascaró, Joan. 1996. External allomorphy as emergence of the unmarked. In Jacques Durand & Bernard Laks (eds). Current trends in phonology: Models and methods, 473–483. Salford: ESRI.

Minimalist representation of Maltese sounds | 47

McCarthy, John J. 1991. Semitic gutturals and distinctive feature theory. In Bernard Comrie & Mushira Eid (eds.), Perspectives on Arabic linguistics III. Papers from the third annual symposium on Arabic linguistics, 63–91. Amsterdam: Benjamins. Mifsud, Manwel. 1995. Loan verbs in Maltese: A descriptive and comparative study. (Studies in Semitic languages and linguistics). Leiden/New York: Brill. Paster, Mary. 2015. Phonologically conditioned suppletive allomorphy: Cross-linguistic results and theoretical consequences. In Eulàlia Bonet, Maria Rosa Lloret & Joan Mascaró (eds.), Understanding allomorphy: Perspectives from Optimality Theory, 218–253. London: Equinox. Polgárdi, Krisztina. 2012. The distribution of vowels in English and trochaic proper government. In Bert Boma & Roland Noske (eds.), Phonological explorations: Empirical, theoretical and diachronic issues, 111–134. Berlin: Mouton de Gruyter. Puech, Gilbert. 1978. A cross-dialectal study of vowel harmony in Maltese. Papers from the fourteenth regional meeting. Chicago Linguistic Society. 377–389. Puech, Gilbert. 1994. Ethnotextes maltais. (Studia Melitensia). Wiesbaden: Harrassowitz. Puech, Gilbert. 2014. Prime constituents of Maltese sounds. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics,61–88. Berlin: Akademie Verlag. Puech, Gilbert. to appear. Maltese kiteb vs. Tunisian (Sūsa) ktib. A prosodic CV approach. Samuels, Bridget D. 2011. Phonological architecture: A biolinguistic perspective. Oxford: Oxford University Press. Scheer, Tobias. 1998. La structure interne des consonnes. Langues et grammaire II & III: Phonologie, 141–172. Scheer, Tobias. 2004. A lateral theory of phonology: What is CVCV, and why should it be? Berlin: Mouton de Gruyter. Spagnol, Michael. 2011. A tale of two morphologies. Verb structure and argument alternations in Maltese. Dissertation. Universität Konstanz. Stumme, Hans. 1904. Maltesische Studien: eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache, nebst Erläuterungen. Leipzig: J. C. Hinrichs. Sutcliffe, Edmund F. 1936. A grammar of the Maltese language. London: Oxford University Press. Ulfsbjorninn, Shanti J. 2014. A field theory of stress. Ph. D. Thesis, University of London, S.O.A.S. Ussishkin, Adam & Alina Twist. 2009. Auditory and visual lexical decision in Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics. Selected papers from the 1st International Conference on Maltese Linguistics, Bremen, 18–20 October, 2007, 233–249. Amsterdam: Benjamins. Vanhove, Martine. 1991. On the survival of [ʕ] in a Maltese idiolect at Mtaħleb in Malta. Journal of Afroasiatic Languages 3. 22–334. Vassalli, Michael A. 1796. Ktŷb yl klŷm Mâlti, ’mfysser byl-Latǐn u byt-Taljânm, sive liber dictionum Melitensium. Roma. Reedited by Frans Sammut. 2002. Lexicon ta’ Mikiel Anton Vassalli. Malta: SKS. Vassalli, Michelantonio. 1827. Grammatica della Lingua Maltese. 2nd edition. Malta. Vella, Alexandra. 2009. On Maltese prosody. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics. Selected papers from the 1st International Conference on Maltese Linguistics, Bremen, 18–20 October, 2007, 47–68. Amsterdam/Philadelphia: Benjamins. Wolf, Matthew. 2013. Candidate chains, unfaithful spell-out, and outwards-looking phonologically-conditioned allomorphy. Morphology 23. 145–178.

48 | Gilbert Puech

Wolf, Matthew. 2015. Lexical insertion occurs in the phonological component. In Eulàlia Bonet, Maria Rosa Lloret & Joan Mascaró (eds.), Understanding allomorphy: Perspectives from Optimality Theory, 361–407. London: Equinox. Wolf, Matthew. 2016. Cyclicity and non-cyclicity in Maltese: Local ordering of phonology and morphology in OT-CC. In John McCarty & Joe Pater (eds.), Harmonic serialism and harmonic grammar. London:Equinox. Živanović, Peter. 2004. C and V as elements, Yet Another Krazy Attempt (YAKA). Web.

Andrei A. Avram

Phonological changes in Maltese: Evidence from onomastics Abstract: The present paper analyzes the onomastic evidence for some of the phonological changes in the history of Maltese. The analysis is based on a rich corpus of surnames, nicknames and place-names, recorded between the 14th and the 18th centuries. Onomastics provides valuable insights into the evolution of Maltese consonants and vowels. It also sheds light on the emergence of some of the phonological rules typical of Modern Maltese, such as obstruent devoicing in word-final position and regressive voicing assimilation. Finally, it is shown that evidence from onomastics appears to disconfirm some previous claims regarding the nature and chronology of some phonological changes in Maltese. Keywords: surnames, nicknames, place-names, historical phonology

1 Introduction Evidence from onomastics has only rarely been considered in previous work on the historical phonology of Maltese (see, however, Borg 1976 and 1978, Avram 2012 and 2014). The present paper is an attempt at analyzing the relevance of onomastica data for the study of the diachrony of Maltese consonants and vowels as well as for the emergence of some of the phonological rules specific to Modern Maltese. The corpus of data consists of Maltese surnames (Wettinger 1968, 1969 and 1999, Cassar 2005), nicknames (Wettinger 1971, Camenzuli 2002), and placenames (Abela 1647, Dessoulavy 1957, Aquilina 1961, Wettinger 1983 and 2000), recorded between the 14th and the 18th centuries. All entries include the year of the attestation and the source. When an exact year could not be established a hyphen is used to indicate approximate dates: a year preceded by a hyphen reads ‘in or before’, if followed ‘in or after’. Relevant portions are highlighted in boldface. Some of the original translations into English have been slightly amended; the translations into Italian are reproduced from Abela (1647).

|| Andrei A. Avram: University of Bucharest, Department of English, 7-13 Pitar Mos Str., 010451 Bucharest, Romania. E-mail: [email protected]

50 | Andrei A. Avram

The paper is organized as follows. Section 2 focuses on consonants. The topics covered are: neutralization of voicing in word-final position; regressive voicing assimilation; the interdental fricatives ṯ and ḏ; the voiceless velar fricative ḫ and the voiceless pharyngeal fricative ḥ; the voiced velar fricative ġ and the voiced pharyngeal fricative ʕ; the voiceless laryngeal fricative h. Section 3, on vowels, is concerned with: word-final imāla; word-medial imāla; deletion of short unstressed vowels in word-initial position; “secondary/auxiliary” vowels; vowel harmony. The findings and their implications are discussed in Section 4.

2 Consonants 2.1 Neutralization of voicing in word-final position Modern Maltese exhibits devoicing of obstruents in word-final position (Cohen 1966: 13, Cohen 1967: 165, Borg 1975: 19–20, Borg 1997: 250, Cardona 1997: 103– 104). Evidence from surnames and from nicknames of Arabic origin shows that the rule of word-final devoicing was not part of the phonology of 15th century Maltese: (1a) (1b) (1c) (1d) (1e)

busalib 1419 haiuz 1419 Cathaldus Muhumud 1467 mifsud 1480L-Chiccu labiad 1483 DEF-Chiccu DEF white ‘Chiccu the white one’

[Wettinger 1968: 33] [Wettinger 1968: 30] [Wettinger 1983: 39] [Wettinger 1968: 42] [Wettinger 1971: 40]

A devoiced word-final obstruent occurs in only one form: (2)

aius 1480-

[Wettinger 1968: 30]

The spelling of place-names also points to the continuous occurrence of voiced obstruents in word-final position until much later, towards the end of the 16th century: (3a)

(3b)

mita Jlchaded 1461 DEF:iron of ‘of the iron’ bita muezeb 1486 of gutter ‘of the gutter’

[Wettinger 1983: 53]

[Wettinger 1983: 32]

Phonological changes in Maltese. Evidence from onomastics | 51

(3c)

(3d)

(3e)

ta michammed 1500 of Muḥammad ‘of Muḥammad’ ta misged 1537 of synagogue ‘of the synagogue’ il mueyed 1590 DEF tablelands ‘tablelands’

[Wettinger 1983: 40]

[Wettinger 1983: 33]

[Wettinger 1983: 37]

The following are exceptions: (4a)

(4b)

gebel labiath 1501 stone/hill DEF:white ‘the white rock, the white hill-side’ merhelet ilmohos 1523 pen DEF:goats ‘the goats’ pen’

[Wettinger 1983: 64]

[Wettinger 1983: 34]

However, the digraph in labiath ‘white’ appears to be simply an idiosyncratic spelling. In addition, the same place-name is also recorded with final (5)

ta gebel labiod 1538 of stone/hill DEF:white ‘of the white rock, of the white hill-side’

[Wettinger 1983: 64]

Clear instances of word-final devoiced obstruents are attested in surnames, nicknames and place-names from the 17th century onwards: (6a) (6b)

(6c)

(6d)

Mihammet 1632 [Wettinger 1983: 40] Mitahlep 1647 [Abela 1647: 65] place where cows are milked ‘luogo, oue si mungeua il latte’ (= place where cows were milked) Andar l’Jsuet 1773 [Wettinger 1983: 65] threshing.floor DEF:black ‘Iswed’s (= the dark-skinned’s) threshing floor’ ta’ Lisuet -1798 [Camenzuli 2002: 322] of DEF:black ‘of black complexion, dark-skinned’

52 | Andrei A. Avram

(6e)

ta’ Supperf1 -1798 of proud ‘proud, arrogant’

[Camenzuli 2002: 323]

As shown by example (6e), word-final obstruent devoicing is also attested in nicknames of Romance origin. To sum up, evidence from surnames, nicknames and place-names is suggestive of the failure of obstruents to undergo devoicing in word-final position in the 15th century. This is confirmed by textual evidence such as Pietru Caxaru’s Cantilena2 (c. 1450), which contains no instance of devoiced word-final obstruents.3 On the other hand, the occurrence in surnames, nicknames and place-names beginning with the 17th century of devoiced word-final obstruents accords well with the fact that these are already attested in 1588, in Hieronymus Megiser’s word-list4, where devoiced word-final obstruents occur in 10 forms out of 13. Voiced word-final obstruents still outnumber devoiced ones in Philip Skippon’s word-list5, published in 1664, and in Giovan Francesco Buonamico’s Sonnet6 (c. 1675), out of five potential occurrences, only two forms exhibit a devoiced word-final obstruent.

2.2 Regressive voicing assimilation In Modern Maltese, regressive voicing assimilation occurs when voiced and voiceless obstruents occur in adjacent position; consequently, adjacent obstruents agree in voicing, i.e. they are all either voiced or voiceless (Borg 1975: 15– 16, Borg 1997: 250, Cardona 1997: 104). A 16th century place-name exhibits what may be the first occurrence in Maltese of regressive voicing assimilation. Consider the forms in (7a) and (7b), which correspond to the Modern Maltese plural miżbliet and the singular miżbla ‘dunghill’ (Schembri 1998: 177) respectively: (7a)

te miczbelet 1541 of dunghill ‘of the rubbish dumps or town middens’

[Wettinger 1983: 35]

|| 1 Spelled supperv in Modern Maltese. 2 The text is reproduced in Wettinger and Fsadni (1968: 36) and in Wettinger and Fsadni (1983: 47). 3 See also Cohen and Vanhove (1991: 181). 4 First published in 1601; see Cowan (1964) and Cassola (1987–1988). 5 Reproduced in Cachia (2000: 34–40). 6 See the text in Cachia (2000: 18).

Phonological changes in Maltese. Evidence from onomastics | 53

(7b)

misbile dela saccaya 1545 dunghill of:DEF Saqqajja ‘the rubbish dump or town midden at Saqqajja’

[Wettinger 1983: 35]

In (7a) the digraph appears to be an attempt at rendering a [z], which has emerged via regressive voicing assimilation.7 However, in (7b) the use of suggests that the sound at issue is [s]. These examples may reflect variation in the application of regressive voicing assimilation. Regressive voicing assimilation is also attested in 1588 in Megiser’s wordlist. However, textual evidence from the 17th century shows that either the rule had not run its full course or perhaps it was still subject to variation. Thus, only a minority of the relevant forms recorded by Skippon in 1664 exhibit regressive voicing assimilation, while in Buonamico’s Sonnett (c. 1675), obstruents undergo devoicing via regressive assimilation in all three relevant forms.

2.3 The interdental fricatives ṯ and ḏ The place-names recorded by Giovanni Francesco Abela (1647), whose work is widely regarded as indispensable to the study of Maltese place-names, shed light on the fate of the interdental fricatives (see also Avram 2012: 102, Vella 2013, Avram 2014: 25–26). All the relevant forms identified in Abela (1647) are reproduced below. The reflex of the Arabic voiceless interdental fricative t is spelled with the digraph , and is found in three forms: (8a)

(8b)

(8c)

Ghar Buthomna8 1647 cave big:measure.for.corn ‘The Cave of the Measure of Corn’ Ben Varrath 1647 son heir ‘figlio dell’ Herede’ (= ‘son of the heir’) Ghar el MethKub 1647 cave DEF perforated ‘grotta pertugiata, ò forata’ (= ‘perforated cave’)

[Abela 1647: 73]

[Abela 1647: 71]

[Abela 1647: 22]

|| 7 Cf. Arabic sabla ‘manure, dung’ (Wehr 1976: 396). 8 Cf. tomna ‘a kind of measure for corn and pulse’ (Busuttil 1900: 484), Modern Maltese tomna ‘measure (of corn)’, cf. also Arabic tumna ‘a dry measure’ (Wehr 1976: 107).

54 | Andrei A. Avram

Consider next the two examples in which the digraph transcribes the reflex of the Arabic voiced interdental fricative ḏ: (9a)

(9b)

Dhoccara9 1647 wild.fig ‘Wild Fig’ Bir el dheeb 1647 well DEF gold ‘pozzo d’oro’ (= ‘gold well’)

[Abela 1647: 16]

[Abela 1647: 107]

Finally, in five place-names the digraph corresponds to Arabic ḍ: (10a) Blata el baydha 1647 [Abela 1647: 20] rock DEF white;F ‘rocca bianca’ (= ‘white rock’) (10b) SieKia il baidha 1647 [Abela 1647: 25] aqueduct DEF white ‘Condotto bianco’ (= ‘white aqueduct’) (10c) Hal Dheeif 1647 [Abela 1647: 84] village weak ‘Casale del macilente, o debole’ (= ‘village of the weak one’) (10d) El Baiadhat 1647 [Abela 1647: 97] DEF white fields ‘terreni biancacci’ (= ‘whitish fields’) (10e) Blata el baidh 1647 [Abela 1647: 122] rock DEF white ‘pietra bianca’ (= ‘white rock’’ There is independent evidence which suggests that the forms in (8)–(10) contain indeed the fricative interdentals ṯ and ḏ. The voiceless interdental t occurs in only one form, spelled with in the Cantilena (c. 1450). This single instance is dismissed by Cohen and Vanhove (1991: 1979) as being “sans doute déjà une exception à l’époque”. However, t is also found in eleven reflexes of *ṯ spelled with (six occurrences), (one occurrence), (one occurrence) and (three occurrences), in Megiser’s word-list, recorded in 1588. As put by Dessoulavy (1937: 1182), “the words seem to show that the soft th sound was still heard in Malta in the sixteenth century”. The same word-list also includes an instance of a reflex of *ḏ, spelled with . This evidence points to the continu-

|| 9 Cf. duccar ‘caprificating fig’ (Busuttil 1900: 53), Modern Maltese dukkar ‘wild fig’ (Schembri 1998: 44).

Phonological changes in Maltese. Evidence from onomastics | 55

ous existence in (some varieties of) Maltese of the fricative interdentals ṯ and ḏ before the 17th century, when Abela (1647) recorded the place-names of Malta. Furthermore, there is independent evidence for their occurrence in the 17th century, and even in the 18th century. As shown in Vella (2013) and Avram (2014: 24 and 26–29), the interdental fricatives ṯ and ḏ are found in the dictionary of Maltese (c. 1640) attributed to the French knight Thezan10, and also in two works by Giovanni Pietro Francesco Agius de Soldanis (1750 and 1750–1767). Importantly, both Thezan and de Soldanis list forms in which, as in Abela (1647), ḏ also occurs as a reflex of Arabic ḍ.

2.4 The voiceless uvular stop q The Modern Maltese reflex of *q is the voiceless glottal stop ’ (Cohen 1966: 15; Cohen 1967: 166). The various spellings in surnames, nicknames and placenames show that earlier Maltese still had the voiceless uvular stop q. One frequently occurring spelling is . Consider first 15th and 16th century surnames and nicknames: (11a) Gullielmu Staferagi cuchede 1420Gullielmu Staferagi nit ‘Gullielmu Staferagi a louse just hatched’ (11b) Lucas Baldakin alias miseuhac 1512 Lucas Baldakin alias variegated ‘Lucas Baldakin the variegated’ (11c) Joannes Vella alias Carchille 1531 Joannes Vella alias kind.of.pastry

[Wettinger 1971: 45]

[Wettinger 1971: 44]

[Wettinger 1971: 44]

Note in example (11c) the use of in early Romance loanwords, in which Maltese q is a reflex of an etymological /k/.11 The use of and occasionally of for q is also attested in 15th and 16th century place-names. These include the examples below: (12a) mahanuc 1419 strangled ‘the strangled’

|| 10 Edited by Cassola (1992). 11 Cf. Modern Maltese qarċilla (Wettinger 1971: 44).

[Borg 1976: 22]

56 | Andrei A. Avram

(12b) calet il habid 1487 fort DEF slaves ‘the slaves’ fort’ (12c) il chubejleth ricac 1530 DEF field-strips narrow ‘the narrow field-strips’ (12d) misbile dela saccaya 1545 Saqqajja dunghill of:DEF ‘the rubbish dump or town midden at Saqqajja’ (12e) il hofra fucanie 1587 DEF hole upper ‘the upper depression’

[Wettinger 1983: 56]

[Wettinger 1983: 63]

[Wettinger 1983: 35]

[Wettinger 1983: 61]

The same is also true of a 17th century place-names: (13a) Ghar Barca 1647 [Abela 1647: 65] cave blessing ‘Grotta della Benedittione’ (= ‘the Cave of the Blessing’) (13b) ta sicayac 1659 [Wettinger 1983: 47] of small.alley ‘of the small alley’ Another spelling is , which occurs in the following nicknames: (14a) Luca Lukis beraq 1525 Luca Lukis lightning ‘Luca Lukis the lightning’ (14b) Lumejna bita saqora 1483 Lumejna of sack ‘Lumejna of the sack’

[Wettinger 1971: 40]

[Wettinger 1971: 46]

Also attested is , in 16th century place-names: (15a) Jl

harke il chamra 1523 field DEF red ‘the red field’ (15b) gued kleya 1533 valley Qleigħa ‘the valley at Qleigħa’ (15c) habel Jtikil 1538 field-strip DEF:heavy ‘the field-strip of the heavy man’ (nickname)

[Wettinger 1983: 65]

DEF

[Wettinger 1983: 55]

[Wettinger 1983: 64]

Phonological changes in Maltese. Evidence from onomastics | 57

With very few exceptions, Abela (1647) uses for reflexes of *q. Here are some of his examples of place-names: (16a) Blat el Kamar 1647 rock DEF moon ‘Roccami della luna’ (= ‘the Rocks of the Moon’) (16b) Ghar BaKar 1647 cave cows ‘grotta delle vacche’ (= ‘the Cave of the Cows’) (16c) MaKluba 1647 overturned;F ‘Riuoltata’ (= ‘overturned’) (16d) Tal eenieK 1647 of:DEF young.she-goats ‘of the young she-goats’

[Abela 1647: 88]

[Abela 1647: 73]

[Abela 1647: 100]

[Abela 1647: 99]

Yet another spelling is , attested in the following nickname: (17)

Joanna filia Blasii Ketcuti alias manja Joanna daughter Blasius Ketcuti alias eats bachira 1532 cow [Wettinger 1971: 44] ‘Joanna daughter of Blasius Ketcuti alias ate a cow’.

The digraph also occurs, more frequently, in place-names: (18a) machalube 1547 overturned ‘overturned’ (18b) Jl charcha Jisighire 1574 DEF field small ‘the small field’ (18c) fuch Jl gued Jl chibir 1581 on DEF valley DEF big ‘above the big valley’ (18d) choleja 1585 small.fort ‘small fort’

[Borg 1978: 22]

[Wettinger 1983: 62]

[Wettinger 1983: 66]

[Wettinger 1983: 44]

58 | Andrei A. Avram

Note, in (18c), that is also used for [k], in chibir ‘big’.12 This is further proof that the Maltese reflex of *q was still a voiceless uvular stop, perceived by the transcriber as similar/identical to [k]. The digraph is also found in early 17th century place-names: (19a) tal machnuch 1610 of:DEF hoarse ‘of the hoarse person’ (19b) el ghilechi sighar 1611 DEF fields small:PL ‘the small fields’

[Wettinger 1983: 39]

[Wettinger 1983: 62]

The single exception in which a reflex of *q is not rendered by any letter must be a scribal error: (20)

gued il huasa13 1498 valley DEF wide ‘the wide valley’

[Wettinger 1983: 62]

Similar various orthographic choices to render reflexes of *q are attested in other early Maltese texts. Thus, around 1450, Caxaru uses (four occurrences), (one occurrence) and (two occurrences) in his Cantilena. In 1588, Megiser transcribes reflexes of *q with (one occurrence), (three occurrences) and (one occurrence). In 17th century texts the situations is as follows. Skippon uses almost exclusively (34 occurrences), (one occurrence) and (one occurrence) in 1664, in his word-list; Buonamico resorts to (three occurrences) and to (one occurrence) in his Sonnet (c. 1675). Of these spellings, and are not attested in 15th and 16th century Maltese surnames, nicknames and place-names. Later records, of late 18th century nicknames exhibit less variation, the only spellings attested being and . Consider the examples under (21) and (22) of nicknames recorded between 1771 and 1798: (21a) il

Mactula -1798 murdered ‘the murdered’ (21b) ta’ Misruca -1798 of stolen ‘stolen’

[Camenzuli 2002: 323]

DEF

|| 12 Cf. the nickname il-Chibir ‘the huge one’ (Camenzuli 2002: 320). 13 Cf. Modern Maltese wiesa’ [wiəsa’].

[Camenzuli 2002: 323]

Phonological changes in Maltese. Evidence from onomastics | 59

(22a) ta’ Busach14 -1798 of big:belly ‘pot-bellied’ (22b) talmhallach -1798 of:DEF hanged ‘the hanged’

[Camenzuli 2002: 321]

[Camenzuli 2002: 322]

Again, these spellings differ from those in other contemporary sources, such as late 18th sermons, whose authors resort to and (Bonelli 1897).

2.5 The voiceless velar fricative ḫ and the voiceless pharyngeal fricative ḥ Modern Maltese has lost the distinction between the voiceless velar fricative ḫ and the voiceless pharyngeal fricative ḥ. As shown below, reflexes of *ḫ are mostly spelled with or in 15th century surnames and nicknames. These competing spellings are found even in variants of the same form, compare (23a) to (23b): (23a) haxixe 1419 (23b) chaxixe 1480-

[Wettinger 1968: 31] [Wettinger 1968: 31]

The same hold true for 16th century place-names. Consider the pairs of variants under (24) and (25), in which the same noun is spelled either with or with : (24a) il

chireybeth 1509 small.ruins ‘the small ruins’ (24b) ilhireybet 1527 DEF:small.ruins ‘the small ruins’

[Wettinger 1983: 50]

(25a) habel ta mneihor 1537 field of little.nose ‘the field of the little nose’ (25b) mineychor 1543 little.nose ‘little nose’

[Wettinger 1983: 48]

DEF

|| 14 Cf. Modern Maltese żaqq ‘belly’ (Schembri 1998: 305).

[Wettinger 1983: 42]

[Wettinger 1983: 48]

60 | Andrei A. Avram

The following example illustrates a rare spelling, namely, the use of : (26)

ta xueyac 1539 of little.old.man ‘of the little old man’

[Wettinger 1983: 43]

The reflexes of *ḫ in 15th and 17th century surnames and nicknames display the same alternation in the use of or . This is illustrated with orthographic variants of several forms corresponding to the Modern Maltese surnames Buhagiar (27), Hakem (28), Melliehi15 (29), Micallef (30), and Muhammed (31) respectively: (27a) buhaiar 1419 (27b) buchaiar 1480-

[Wettinger 1968: 33] [Wettinger 1968: 33]

(28a) (28b) (28c) (28d)

[Wettinger 1968: 40] [Wettinger 1968: 40] [Wettinger 1968: 40] [Wettinger 1968: 40]

hakem 1419 hakim 1419 chakim 1419 hakim 1480-

(29a) mellahi 1419 (29b) mellachi 1480-

[Wettinger 1968: 42] [Wettinger 1968: 42]

(30a) (30b) (30c) (30d)

mihallef 1419 mihallif 1419 Mahallif 1420michallif 1480-

[Wettinger 1968: 42] [Wettinger 1968: 42] [Wettinger 1971: 43] [Wettinger 1968: 42]

(31a) (31b) (31b) (31c) (31d) (31e)

muhamud 1419 muhumud 1419 muhumudi 1419 Muhumud 1467 michammed 1500 Mihammet 1632

[Wettinger 1968: 42] [Wettinger 1968: 42] [Wettinger 1968: 42] [Wettinger 1983: 39] [Wettinger 1983: 40] [Wettinger 1983: 40]

The pairs of orthographic variants under (32)–(35) attest to the alternation of and in 15th and 16th century place-names as well: (32a) mirachil 1467 [ animal.yards ‘the animal yards’

|| 15 Literally ‘from Mellieħa’ [melliəħa], a town in Malta.

Wettinger 1983: 34]

Phonological changes in Maltese. Evidence from onomastics | 61

(32b) mirahel 1487 animal.yards ‘the animal yards’

[Wettinger 1983: 34]

(33a) ta bir il-chamem 1508 of well DEF:bathhouse ‘of the well of the bathhouse’ (33b) ta bir il hammem 1558 of well DEF bathhouse ‘of the well of the bathhouse’

[Wettinger 1983: 58]

(34a) ta chumayra 1530 of small.donkey ‘of the small donkey’ (34b) ta humayra 1543 of small.donkey ‘of the small donkey’

[Wettinger 1983: 45]

(35a) il

[Wettinger 1983: 61]

chofra fucanie 1548 DEF hole upper ‘the upper depression’ (35b) Jl hofra Jl fucanie 1587 DEF hole DEF upper ‘the upper depression’

[Wettinger 1983: 58]

[Wettinger 1983: 45]

[Wettinger 1983: 61]

The digraph is also found in 17th century place-names, in which it represents a reflex of either *ḫ in (36), or of *ḥ, in (37): (36)

tal machnuch 1610 of:DEF hoarse ‘of the hoarse person’

[Wettinger 1983: 39]

(37)

tal moftech 1611 of:DEF key ‘of the key’

[Wettinger 1983: 38]

In addition, the reflex of *ḥ is also spelled with , as in the following example: (38)

il

miseyrac 1611 small.open.pace ‘the small open place’

[Wettinger 1983: 38]

DEF

The spellings of the surnames, nicknames and place-names considered so far suggest that the two fricatives *ḫ and *ḥ had already merged as early as the 15th century. Contemporary textual evidence also seems to point to the same conclu-

62 | Andrei A. Avram

sion. Thus, in the Cantilena (c. 1450) reflexes of *ḫ are represented by (one occurrence), (two occurrences) and (two occurrences), while reflexes of *ḥ are spelled with (one occurrence) and (three occurrences). In the words recorded in 1588 by Megiser reflexes of *ḫ are rendered with (one occurrence) and (eight occurrences), whereas reflexes of *ḥ are represented by (one occurrence), (two occurrences), (one occurrence), (four occurrences) and no letter (once). Skippon uses only two spellings in his 1664 word-list: (one occurrence) and (16 occurrences) for reflexes of *ḫ, and (13 occurrences) for reflexes of *ḥ. In Buonamico’s Sonnet (c. 1675) the following spellings are found: (three occurrences) for reflexes of *ḫ, (four occurrences) and (five occurrences) for reflexes of *ḥ. However, in Abela (1647) with one single exception all reflexes of *ḫ are spelled with , whereas all reflexes of *ḥ are spelled with . Consider some of the placenames recorded by Abela (1647), containing reflexes of *ḫ, in the examples under (39), and of *ḥ, in the forms under (40) respectively: (39a) Vyed l’Aherief 16 1647 valley DEF:lambs ‘valle d’agnelli’ (= ‘the Valley of the Lambs’) (39b) Dachlet 1647 entrance ‘entrata’ (= ‘entrance’) (39c) Ras chanzir 1647 head pig ‘the cape of the pig’ (39d) Tal Cherba 1647 of:DEF ruin ‘del edificio rouinato’ (= ‘of the ruined building’) (40a) Bir el Hut 1647 well DEF fish ‘the water tank of the fish’ (40b) Bir Miftuh 1647 well open ‘the open water tank’ (40c) Taht el gebel 1647 under DEF stone ‘sotto la rocca’ (= ‘under the rock’)

|| 16 This is the only exception in which a reflex of *ḫ is spelled with .

[Abela 1647: 73]

[Abela 1647: 25]

[Abela 1647: 16]

[Abela 1647: 98]

[Abela 1647: 85]

[Abela 1647: 103]

[Abela 1647: 28]

Phonological changes in Maltese. Evidence from onomastics | 63

Finally, in late 18th century nicknames an identical spelling is used for reflexes of both *ḫ and *ḥ as in the examples under (41) and (42) respectively: (41a) il-Haddiela -1798 DEF:paralyzed ‘the paralyzed’ (41b) Mohhu Bilma -1798 brain:3SG.POSS with:water ‘ignorant’ (lit. ‘his brains are with water’) (41c) talMishut -1798 of:DEF cursed ‘the cursed’ (42a) Mahrus -1798 guarded ‘kept in custody’ (42b) ta’ Harrasci -1798 of harsh ‘the harsh one’ (42c) talHobla -1798 of:DEF pregnant ‘pregnant’

[Camenzuli 2002: 322]

[Camenzuli 2002: 322]

[Camenzuli 2002: 323]

[Camenzuli 2002: 322]

[Camenzuli 2002: 323]

[Camenzuli 2002: 322]

The use of for reflexes of both fricatives at issue is attested in late 18th century texts (Bonelli 1897) as well. To sum up, evidence from 17th century place-names (in Abela 1647) suggests that the voiceless velar fricative ḫ and the voiceless pharyngeal fricative ḥ had not as yet merged in the 17th century. Late 18th century spellings of nicknames already point to the tendency towards the merger of these two fricatives.

2.6 The voiced velar fricative ġ and the voiced pharyngeal fricative ʕ As is well known, Modern Maltese no longer has the voiced velar fricative ġ and the voiced pharyngeal fricative ʕ (Borg 1997: 246; Cardona 1997: 22–23). As mentioned by Borg (1997: 246), Modern Standard Maltese exhibits the outcome of “the reinterpretation of the OA [= Old Arabic] velar and pharyngeal fricative pair [] and [ɣ], chiefly as vocalic length”. The two consonants still occur in 15th and 16th century place-names. In the case of ġ, the most frequent spelling is, by far, , as in the examples below:

64 | Andrei A. Avram

(43a) gar Jlkebir 1467 cave DEF:big ‘the big cave’ (43b) tal magalac 1500 of:DEF enclosure ‘of the enclosure’ (43c) gulejca 1514 small.field ‘small field’ (43d) tal gureife 1537 of:DEF small.room ‘of the small upper floor room’ (43e) gadiret Jxorraf 1564 lake DEF:Xorraf ‘Xorraf’s lake’ (43f) il galca Jseude 1584 DEF field DEF:black ‘the black field’

[Wettinger 1983: 62]

[Wettinger 1983: 37]

[Wettinger 1983: 43]

[Wettinger 1983: 44]

[Wettinger 1983: 56]

[Wettinger 1983: 65]

In a number of place-names ġ is rendered by : (44a) xaret il hadiri 1487 scrubland DEF lake ‘the scrubland at the lake’ (44b) misirach sihjr 1521 small.open.place small ‘the small open place’ (44c) il harke il chamra 1523 DEF field DEF red ‘the red field’

[Wettinger 1983: 56]

[Wettinger 1983: 62]

[Wettinger 1983: 65]

The use of digraphs is also attested. One is , in the following 16th century place-names: (45a) il

galca sighire 1508 field small ‘the small field’ (45b) el ghuleica 1542 DEF small.field ‘the small field’ (45c) il charcha Jissighire 1574 DEF field DEF:small ‘the small field’

[Wettinger 1983: 62]

DEF

[Wettinger 1983: 43]

[Wettinger 1983: 62]

Phonological changes in Maltese. Evidence from onomastics | 65

Moreover, the same spelling is found in 17th century place-names and it is the only one used by Abela (1647): (46a) el

ghilechi sighar 1611 [Wettinger 1983: 62] small.fields small;PL ‘the small fields’ (46b) Ghar Ghliem 1647 [Abela 1647: 72] cave slave ‘Grotta del Seruo’ (= ‘the Slave’s Cave’) (46c) Hal Seyegh 1647 [Abela 1647: 101] village silversmith ‘Casale dell’ Argentiere’ (= ‘the Silversmith’s village’) DEF

Another digraph is , which occurs in the 16th century place-name below: (47)

il

charcha Jisighire 1574 DEF field DEF:small ‘the small field’

[Wettinger 1983: 62]

The same spelling is found in one 17th century place-name, in word-final position: (48)

Sebbiech 1647 painter ‘Tintore’ (= ‘painter’)

[Abela 1647: 69]

Given Abela’s consistency in transcribing reflexes of *ġ with , in this form stands for the voiceless velar fricative [x]. This is an allophone of ġ, a result of obstruent devoicing in word-final position. Finally, in the corpus there is just one exception, in which a reflex *ġ is not spelled with a consonant letter: (49)

zayra 1419 small ‘small’

[Borg 1976: 194]

In all likelihood, this apparent exception can be dismissed as a misspelling, where should read . As shown above, all reflexes of *ġ are rendered with a consonant letter or a digraph, and there is no case of the use of double vowel letters suggestive of the reinterpretation of *ġ as vocalic length. A comparison with other early texts yields the following results. Caxaru uses (one occurrence> in his Cantilena (c. 1450). Megiser writes ġ with (one occurrence) in 1588. Buonamico uses (two occurrences) in his Sonnet (c. 1675). Finally, in 1664 Skippon is remarkably consistent in the almost exclusive use of (12 occurrences), with one exception when he uses , in word-final position. As in the place-names considered above, in all these early sources, *ġ

66 | Andrei A. Avram

is always transcribed by a consonant letter. On the other hand, and are used to render ġ only in place-names, in which the use of is not attested In late 18th century nicknames, while no longer appears, continues to be used: (50a) il

Bahal -1798 DEF mule ‘the bastard’17 (50b) ta’ Berhuda -1798 of flea ‘with fleas’

[Camenzuli 2002: 324]

[(Camenzuli 2002: 322]

Of the two digraphs formerly used, and , only the latter is found: (51)

ta’ Misbuch -1798 of painted ‘the outstripped18

[Camenzuli 2002: 322]

Given that occurs in word-final position, it reflects a phonetic realization as a voiceless velar fricative [x], explicitly noted and commented upon by contemporary writers.19 On the other hand, continues to occur in other late 18th century sources, such as sermons (Bonelli 1897). Consider next the fate of the voiced pharyngeal fricative ʕ. In 15th century surnames and 16th century place-names, the most frequently found spelling is , both word-initially and word-medially, as illustrated by the forms under (52), corresponding to the Modern Maltese surnames Abdilla, Agius and Harabi respectively and the place-names under (53): (52a) habdille 1419 (52b) haius 1419 (52c) harabi 1419

[Wettinger 1968: 30] [Wettinger 1968: 30] [Wettinger 1968: 40]

(53a) calet il habid 1487 fort DEF slaves ‘the slaves’ fort’ (53b) tal husayfar 1507 of:DEF small bird ‘of the small bird’

[Wettinger 1983: 56]

|| 17 Cf. Modern Maltese bagħal ‘mule’. 18 Cf. Modern Maltese miżbugħ ‘painted’. 19 See Vassalli (1796: 314).

[Wettinger 1983: 47]

Phonological changes in Maltese. Evidence from onomastics | 67

(53c) merhelet ilmohos 1523 pen DEF:goats ‘the goats’ pen’

[Wettinger 1983: 34]

The digraph appears in one 16th century place-name: (54)

bital charab 1529 of:DEF Arab ‘of the Arabs’

[Wettinger 1983: 52]

The reflex of word-initial *ʕ is not transcribed with any consonant letter in several surnames and nicknames recorded in the 15th century: (55a) abdille 1480(55b) aius 1419

[Wettinger 1968: 30] [Wettinger 1968: 30]

Similarly, in the following 16th century place-name word-initial ʕ is not rendered in the orthography by any consonant letter: (56)

te gued bir abdalla 1542 of valley well Abdilla ‘of the valley of Abdilla’s well’

[Wettinger 1983: 57]

Word-medial ʕ, if transcribed, is spelled also spelled with , as in the nickname in (57) and in the place-names under (58): (57)

Jacobo Chirmel alias mehze 1494 Jacobo Chirnel alias goat ‘Jacobo Chirmel alias the goat’

(58a) macahad ilme 1504 seat water ‘the place where water collects’ (58b) merhelet il mohos 1523 DEF goats pen ‘the goats’ pen’ (58c) ta lahueyn 1549 of DEF:small.spring ‘of the small spring’

[Wettinger 1971: 44]

[Wettinger 1983: 32]

[Wettinger 1983: 34]

[Wettinger 1983: 43]

68 | Andrei A. Avram

On the other hand, intervocalic ʕ is not represented by any consonant letter in variants of the place-name Qleigħa:20 (59a) gued il culeja 1488 valley DEF Qleigħa ‘the valley at Qleigħa’ (59b) gued il culaya 1500 valley DEF Qleigħa ‘the valley at Qleigħa’ (59c) gued kleya 1533 valley Qleigħa ‘the valley at Qleigħa’

[Wettinger 1983: 55]

[Wettinger 1983: 55]

[Wettinger 1983: 55]

Mention should also be made of word-final *ʕ in the various reflexes of Ar. mitā‘. In 15th and 16th century nicknames and place-names three such reflexes occur: bita and, more rarely, mita, and ta/tha/te, the latter replacing the first two before the end of the 16th century. The spelling of these reflexes – i.e. the absence of any consonant letter – suggests that the voiced pharyngeal fricative ʕ no longer occurred in word-final position, if preceded by /a/. Consider the nicknames under (60) and the place-names under (61) respectively: (60a) Lumejna bita saqora 1483 Lumejna of sack ‘Lumejna of the sack’ (60b) Palma et Catharina bita gilidi 1485 Palma and Catharina of skin ‘Palma and Catherina of the skin’ (60c) Luchia bita lispital 1498Luchia of DEF:hospital ‘Luchia from the hospital’

[Wettinger 1971: 46]

(61a) mita Jlchaded 1461 of DEF:iron ‘of the iron’ (61b) bita muezeb 1486 of gutter ‘of the gutter’

[Wettinger 1983: 53]

[Wettinger 1971: 43]

[Wettinger 1971: 46]

[Wettinger 1983: 54]

|| 20 In the Modern Maltese spelling the digraph represents here an etymological ʕ, cf. Arabic quley‘a ‘small fort’.

Phonological changes in Maltese. Evidence from onomastics | 69

(61c) ta ilculeyat 1496 of DEF:small.forts ‘of the small forts’ (61d) tha Jlmreychilet 1542 of DEF:small.animal.yards ‘of the small animal yards’ (61e) te gued bir abdalla 1542 of valley well Abdilla ‘of the valley of Abdilla’s well’

[Wettinger 1983: 50]

[Wettinger 1983: 34]

[Wettinger 1983: 57]

Reflexes of word-initial and word-medial *ʕ are almost always spelled with a double vowel letter by Abela (1647): (62a) Hal dheeif 1647 [Abela 1647: 84] village weak ‘Casale del macilente, ò debole’ (= ‘the village of the weak’) (62b) Kalaa tàl Miggiär 1647 [Abela 1647: 68] fortress of:DEF Miġar ‘the fortress of Miġar’ (62c) Oosc el Hamiema 1647 [Abela 1647: 27] nest DEF dove ‘nido della Colomba’ (= ‘the Dove’s nest’) (62d) Tal eenieK 1647 [Abela 1647: 99] of:DEF young.she-goats ‘of the young she-goats’ (62e) Vyed el Aasel 1647 [Abela 1647: 71] valley DEF honey ‘torrente di Mele’ (= ‘the river of Honey’) There are several exceptions. These include a form spelled with a single vowel letter in word-initial position, but which also has a variant with a double vowel letter. Compare the spellings of ‘slaves’ and ‘slave’ respectively in the following place-names: (63a) Kalaa tal Abid 1647 [Abela 1647: 71] fort of:DEF slaves ‘forte schiaui (Aabid detti in Arabo)’ (= ‘the slaves’ fort’) (63b) Vyed tà Bir Aabd Alla 1647 [Abela 1647: 100] valley of well slave God ‘valle ou’ è il pozzo del seruo di Dio’ (= ‘valley where the well of God’s slave is’]

70 | Andrei A. Avram

Abela’s spellings of the reflexes of Ar. mitā‘ suggest that ʕ no longer occurred in this position if preceded by the vowel /a/: (64a) tà Bunachla 1647 of big:palm.tree ‘of the big palm tree’ (64b) ta Lahgiar 1647 of DEF:stones ‘delle pietre’ (= ‘of the stones’)

[Abela 1647: 68]

[Abela 1647: 106]

It appears, then, that ġ and ʕ had not as yet merged in Abela’s time. This is further confirmed by the following place-name which contains reflexes of both *ġ and *ʕ, in the same, word-initial position: (65)

Aayn Ghrab 1647 spring raven ‘Fontana del coruo’ (= ‘the Fountain of the raven’)

[Abela 1647: 66]

While Abela’s spellings of reflexes of word-initial and word-medial *ʕ with a double vowel letter are suggestive of vocalic length, reflexes of *ʕ continue to be rendered with consonant letters as well. For example, the reflex of *ʕ is spelled with in the following 17th century place-name: (66)

gudia el halia 1654 hillock DEF high ‘the high hillock’

[Wettinger 1983: 61]

Other 15th, 16th and 17th century sources also display a great variety in the spelling of reflexes of *ʕ. Thus, in the Cantilena (c. 1450), stands for ʕ wordinitially (four occurrences) and word-medially (four occurrences); in wordmedial position ʕ is also transcribed with (one occurrence); ʕ is not represented by any letter in intervocalic (five occurrences) and word-final position (two occurrences). In 1588, Megiser uses in his word-list (one occurrence) and (one occurrence) for ʕ in word-medial position; on the other hand, ʕ is not represented by a letter: word-initially (four times), word-medially (once) and word-finally (three times). By far the widest range of orthographic options is found in Skippon’s word-list in 1664: no letter (10 times), a double vowel letter (twice) or a vowel letter (once) in word-initial position; (one occurrence), a double vowel letter (three occurrences) or no letter (once) in intervocalic position; a vowel (twice) or no letter (nine times) in other word-medial contexts; (three occurrences), (two occurrences) or no letter (three times) in word-

Phonological changes in Maltese. Evidence from onomastics | 71

final position.21 Finally, Buonamico does not use any consonant letter in his Sonnet (c. 1675): reflexes of *ʕ are transcribed with a double vowel letter both word-initially (two occurrences) and word-medially (two occurrences); one word-initial ʕ is not represented by any letter. The occasional use of a double vowel letter by both Skippon and Buonamico might suggest that the gradual reinterpretation of *ʕ as vocalic length had already started, not only in wordinitial, but also in word-medial position. The voiced pharyngeal fricative ʕ is still attested in word-medial position in a late 18th century nickname: (67)

talmhallach -1798 of:DEF hanged ‘the hanged’

[Camenzuli 2002: 322]

The use of is rather surprising since in other contemporary texts, if transcribed, ʕ is spelled with both word-initially and word-medially, e.g. in late 18th century sermons (Bonelli 1897). As in earlier nicknames and place-names, ʕ does not occur word-finally, if preceded by the vowel /a/. In this case, *ʕ appears as , as in the reflex of Ar. mitā‘ (68a) ta’ Coleriti -1798 of angry ‘angry, enraged’ (68b) ta’ Fitta -1798 of tedious ‘importunate person’ (68c) ta’ Gamiema -1798 of turtle-dove ‘grumbling person’

[Camenzuli 2002: 322]

[Camenzuli 2002: 323]

[Camenzuli 1798: 324]

To conclude, the spelling of surnames, nicknames and place-names is further confirmation of the fact that both the voiced velar fricative ġ and the voiced pharyngreal fricative ʕ survive to some extent in most dialects of Maltese, i.e. they are still two different phonemes, at least until the end of the 18th century.22 The use of for both consonants may be indicative of the tendency towards

|| 21 This variety of spellings clearly disconfirms Cachia’s (2000: 41) claim that Skippon spelled “words with ‘ayn […] with double vowels”. 22 And possibly in early 19th century as well.

72 | Andrei A. Avram

their merger, which also accords with explicit descriptions and comments in contemporary sources.23

2.7 The voiceless laryngeal fricative h The inventory of consonantal phonemes of Modern Maltese does not include the laryngeal fricative h (Cohen 1967: 164, Borg 1973: 8, Borg 1975: 11, Cardona 1997: 23). The following 15th and 16th century place-names show that h still occurred in Maltese, including in intervocalic position: (69a) zahara 1480blossom ‘blossom’ (69b) siheym 1514 small.allotment ‘small allotment’ (69c) siheym 1530 small.allotment ‘small allotment’ (69d) ysiheyem 1558 DEF:small.allotment ‘the small allotment’

[Borg 1976: 194]

[Wettinger 1983: 42]

[Wettinger 1983: 42]

[Wettinger 1983: 42]

This accords with the situation reflected by 15th and 16th century texts. In the Cantilena (c. 1450), 11 out of 12 possible occurrences of h are reflected in the spelling, invariably : all eight instances of word-initial h, three out of four cases of word-medial h, two of which in intervocalic position. Similarly, in Megiser’s word-list (1588), h is rendered with in three out of four possible cases: two word-initially and one word-medially in intervocalic position; the only word-final *h is not reflected in the spelling. Relevant forms in 17th century place-names show that h continued to occur, including in intervocalic position. Consider the following examples: (70a) Kibur elihut 1647 graves DEF:Jew ‘Sepulchri de’ Giudei’ [= ‘graves of the Jews’]

|| 23 Such as Vassalli (1796: 239 and 314).

[Abela 1647: 82]

Phonological changes in Maltese. Evidence from onomastics | 73

(70b) ras e Raheb 1647 [Abela 1647: 24] head DEF monk ‘Capo, ò Promontorio del monaco’ (= ‘the cape of the monk’) (70c) Ta Buha 1647 [Abela 1647: 97] of father:3SG.POSS ‘di suo padre’ (= ‘of her father’) Corroborating evidence is provided by other samples of 17th century Maltese. Thus, seven forms in Skippon’s word-list (1664) testify to the occurrence of h: two word-initially and five word-medially, of which four in intervocalic position; in two cases, h is not represented in intervocalic position, but, in one of these forms it is also spelled with . Finally, in Buonamico’s Sonnet (c. 1675), h is represented in the spelling with in all four possible instances: twice word-initially and twice word-medially in intervocalic position. Word-initial h is also found in late 18th century nicknames: (71)

ta’ Hafif -1798 of light ‘silly, mentally defective’

[Camenzuli 2002: 322]

The occurrence of word-initial h is corroborated by evidence provided by late 18th century texts (Bonelli 1897), in which h, rendered by , which is well attested not only word-initially, but also word-medially, and which apparently still occurs even in word-final position.

3 Vowels 3.1 Word-final imāla Borg (1976: 193) writes that “such scanty records of early Maltese as have come down to us suggest that final imaala (both of /*-aa/ and of the feminine ending /*-a/) was the rule in certain 15th and 16th century Maltese dialects”. Borg (1976: 194) concludes “that we are here dealing with a phonologically conditioned sound change” and that this “is made clear by the fact that /a/ is preserved in the environment of historically back or emphatic consonants”. This conclusion is illustrated with forms in which the reflexes of the feminine ending */-a/ and of */-a:/ are /a/ in a historically back or emphatic environment, e.g. garca ‘drowning’, cf. Ar. ġarqa, sura ‘appearance’, cf. Ar. ṣūṛa, but /e/ elsewhere, e.g. simine ‘obesity’, cf. Ar. simna, me ‘no’, cf. Ar. mā (Borg 1976: 193-

74 | Andrei A. Avram

194). In other words, the two reflexes of the feminine ending */-a/ and of */-a:/ are in complementary distribution. Consider, however, further evidence from various types of proper names. On the one hand, /a/ also turns up where it would not be expected to appear, given the pattern of complementary distribution posited by Borg (1976). In several 15th century surnames and nicknames /a/ occurs in a non-back or nonemphatic environment: (72a) Caruana 24 1419 (72b) haneja25 1420arc:F ‘arc’ (72c) chagira gebisa 1467 stone:F hard:F ‘hard rock’ (72d) xelluxa26 1499 root:F ‘root’

[Wettinger 1968: 35] [Wettinger 1971: 43]

[Wettinger 1971: 43]

[Wettinger 1971: 46]

Also, 15th and 16th century surnames, in (73), and nicknames, as in (74) and (75), exhibit variants with /e/ or /a/ in a non-back or non-emphatic environment: (73a) saliba 1419 (73b) salibe 1480s (73c) saliba 1480s

[Wettinger 1971: 44] [Borg 1976: 194] [Wettinger 1971: 44]

(74a) zurafe 1419 giraffe:F ‘giraffe’ (74b) zurafa 1419 giraffe:F ‘giraffe’ (74c) zurafe 1480s giraffe:F ‘giraffe’

[Borg 1976: 194]

[Wettinger 1968: 47]

[Wettinger 1968: 47]

|| 24 Cf. Arabic Kayrawān, the etymon suggested by Aquilina (1964: 154). 25 Wettinger (1971: 43) does not provide any translation. The Modern Maltese form is ħnejja ‘an arch, a vault’ (Busuttil 1900: 112), ‘arch, vault, arc’ (Schembri 1998: 91), cf. Arabic ḥanīya. 26 Cf. Arabic šilša ‘root’, the etymon proposed by Wettinger (1971: 38).

Phonological changes in Maltese. Evidence from onomastics | 75

(75a) farruge 1420s hen:F ‘hen’ (75b) farrugia 1495 hen:F ‘hen’

[Borg 1976: 194]

[Wettinger 1971: 40]

Finally, a number of 15th and 16thcentury place-names have variants with /e/ or /a/ in a non-back or non-emphatic environment: (76a) tafalie 1487 clayey:F ‘the clayey ground’ (76b) tafalia 1499 clayey:F ‘the clayey ground’ (77a) tal guistanie 1508 of:DEF central:F ‘of the central (portion)’ (77b] ta hustanie 1532 of central:F ‘of the central (portion)’ (77c) tal uistanie 1542 of central:F ‘of the central (portion)’ (77d) tal guystania 1548 of:DEF central:F ‘of the central (portion)’ (78a) mixta 1456 wintering.place:F ‘wintering place’ (78b) mixta 1572 wintering.place:F ‘wintering place’ (78c) mixte 1577 wintering.place:F ‘wintering place’

[Wettinger 1983: 63]

[Wettinger 1983: 63]

[Wettinger 1983: 61]

[Wettinger 1983: 61)

[Wettinger 1983: 61]

[Wettinger 1983: 61]

[Wettinger 1983: 35]

[Wettinger 1983: 35]

[Wettinger 1983: 35]

On the other hand, */a/ is not always preserved in a back environment. In the following place-name it is /e/ which occurs in this phonological environment:

76 | Andrei A. Avram

(79)

ta hubeyce 1556 of little.basil:F ‘of the little basil’

[Wettinger 1983: 44]

Also, /e/ alternates with /a/ in a historically emphatic environment. Consider the 15th century variants of the Modern Maltese surname Abdilla: (80a) habdille 1419 (80b) abdille 1480(80c) abdalla 1542

[Wettinger 1968: 30] [Wettinger 1968: 30] [Wettinger 1983: 57]

Place-names also have variants with /e/ and /a/ in a back environment: (81a) Il

harke il chamra 1523 field:F DEF red:F ‘the red field’ (81b) Il charcha Jssighire 1574 DEF field:F DEF:small:F ‘the small field’

[Wettinger 1983: 65]

DEF

[Wettinger 1983: 62]

Finally, compare the form in (82a)27, which corresponds to the Modern Maltese surname Zaira, to those occurring in the place-names in (82b, c): (82a) zayra 1419 small:F ‘small’ (82b) il galca sihire 1508 DEF field:F small:F (82c) Il charcha Jisighire 1574 DEF field:F DEF:small:F ‘the small field’

[Borg 1976: 194]

[Wettinger 1983: 62] [Wettinger 1983: 62]

3.2 Word-medial imāla According to Borg (1976: 200), “basing oneself on such early sources of Maltese as are available, one is inclined o conclude that historically, the monophthongal and diphthongal renderings of the high imaala vowel in Maltese are both very old”. The earliest examples provided by Borg (1976: 200) are from

|| 27 For which the “suggested derivation” in Borg (1978: 194) is żgiiṛa. This is also the etymon of the forms in (82b, c).

Phonological changes in Maltese. Evidence from onomastics | 77

Caxaru’s Cantilena (c. 1450) and from Megiser’s wordlist, collected in 1588. The former contain the monophthongal reflex /e:/ of */a:/, e.g. giren ‘neighbors’, cf. Ar. ǧīrān; miken ‘place’, cf. Ar. makān; zimen ‘time’, cf. Ar. zamān. As for the latter, these illustrate the diphthongal reflex /ie/ of */a:/, e.g. chtieb ‘book’, cf. Ar. kitāb; mnieher ‘nose’, cf. Ar. manāḫir; siech ‘thigh’, cf. Ar. sāq. The following place-name contains the first attestation of the reflex /e:/ of */a:/: (83)

motalchadedin 1372 of:DEF:blacksmiths ‘of the blacksmiths’

[Wettinger 1983: 53]

The reflex /e:/ of */a:/ is also found in the 15th century, in forms corresponding to the Modern Maltese surname Butigieg: (84a) butigegi 1419 (84b) butugegi 1419 (84c) butigeg 1480-

[Wettinger 1968: 33] [Wettinger 1968: 33] [Wettinger 1968: 33]

The same reflex is attested in 15th century surnames: (85a) fiteni28 1419 (85b) radene29 1494

[Wettinger 1968: 38] [Wettinger 1971: 45]

Further proof of the fact that the passage of */a:/ to /e:/ took place at a very early stage is provided by Romance loanwords which are found in Maltese surnames. Consider the word for ‘hospital’; the relevant entry in Vella (1848: 124) is “spedal (volg. sptar) Spedale. An hospital”.30 The 15th century Maltese forms occurring in surnames exhibit either /a:/ or /e:/. Consider the variants below: (86a) (86b) (86c) (86d)

spitali 1419 spiteri 1419 spital 1480 spiteri 1480

[Wettinger 1968: 45] [Wettinger 1968: 45] [Wettinger 1968: 45] [Wettinger 1968: 45]

|| 28 Aquilina (1964: 155) suggests the following etymon: “Ar. fātin ‘misleading, disturber, tempter, seducer, alluring, captivating’”. However, fiteni should be compared with Arabic fattān ‘fascinating, captivating, enchanting; charming; tempter, seducer; denunciator, informer, slanderer’ (Wehr 1976: 696); the fact that the reflex of Arabic /tt/ is spelled with a single is not necessarily surprising: cf. radene in (85b), with a single , etymologically derived from Arabic raddāna. 29 Cf. Arabic raddāna ‘spinning-wheel’, etymon suggested by Borg (1976: 194). 30 The Modern Maltese form is sptar (Busuttil 1900: 350, Schembri 1998: 253). However, the older form spedal is still recognizable in spedalier ‘hospitaller’ (Schembri 1998: 254).

78 | Andrei A. Avram

Note, incidentally, that the forms with /e:/ outnumber those with /a:/. Thus, in 1419 there are 12 occurrences of spiteri vs. only 3 of spitali (Wettinger 1968: 45). Similarly, in 1480 there are 7 attestations of spiteri in 1480, whereas spital is only recorded once (Wettinger 1968: 45). This suggests that the variant of pronunciation with [e:] was rather well entrenched. The reflex /e:/ of */a:/ is also found in several 15th century place-names: (87a) mita Jlchaded 1461 of DEF:blacksmith ‘of the blacksmith’ (87b) mita sigeyret 1477 of small.trees ‘of the small trees’ (87c) bir tahteni 1487 well lower ‘the lower well’ (87d) gibeylet 1490 small.stones ‘small boulders’ (87e) il hubeylet 1496 DEF field-strips ‘small field-strips’ (87f) rahal sifileni 1497 village lower ‘the lower village’

[Wettinger 1983: 53]

[Wettinger 1983: 54]

[Wettinger 1983: 61]

[Wettinger 1983: 49]

[Wettinger 1983: 50]

[Wettinger 1983: 61]

The reflex /ie/ of */a:/ is first recorded in a 15th century surname: (88)

fitien 1419

[Wettinger 1968: 38]

This diphthongal reflex also occurs in 16th century place-names: (89a) il

cubeyliet 1508 field-strips ‘small field-strips’ (89b) migarriet 1543 small.watercourses ‘small watercourses’

[Wettinger 1983: 50]

DEF

[Wettinger 1983: 32]

Phonological changes in Maltese. Evidence from onomastics | 79

(89c) ichineysiet31 1544 DEF:small.leases ‘small census paying lands’

[Wettinger 1983: 48]

It is also in a 16th century place-name that the reflex /i:/ of */a:/ is first attested: (90)

ilhureybit 1528 DEF:small.ruins ‘the small ruins’

[Wettinger 1983: 50]

3.3 Deletion of short unstressed vowels in word-initial open syllables On the basis of evidence from Caxaru’s Cantilena, Cowan (1975: 8) shows that unlike unstressed */i/ and */u/, which were elided directly when occurring in a word-initial open syllable, unstressed */a/ was first raised to /i/ and then deleted. Relevant examples from the Cantilena include miken ‘place’, cf. Ar. makān and zimen ‘time’, cf. Ar. zamān. This intermediate stage of word-medial imāla is also illustrated by 15th century place-names, such as those corresponding to Modern Maltese Mrieḥel (91) and Msida (92) respectively: (91a) mirachil 1467 animal.yard ‘animal yard, sheep-fold’ (91b) mirahel 1487 animal.yard ‘animal yard, sheep-fold’ (92a) misidae 1494 fishing.place ‘fishing place’ (92b) miside 1495 fishing.place ‘fishing place’

[Wettinger 1983: 34]

[Wettinger 1983: 34]

[Wettinger 1983: 36]

[Wettinger 1983: 36]

|| 31 Cf. diminutive of chens ‘lease’, which occurs e.g. in the place-name Ichens ta fuch Jl gued Jl chibir ‘the land on long lease above the big valley’ (Wettinger 1983: 66). The Modern Maltese form is spelled ċens (Schembri 1998: 29).

80 | Andrei A. Avram

Consider next the following place-names recorded in the 15th and the 16th centuries, which include forms corresponding to Modern Maltese maḥsel ‘laundry; washing-house, lavatory’ (Schembri 1998: 160): (93a) lu

mahasel 1487 washing.place ‘the washing place’ (93b) il mahisel 1495 DEF washing.place ‘the washing place’ (93c) tal mihasil 1504 of:DEF washing.place ‘of the washing place’

[Wettinger 1983: 33]

DEF

[Wettinger 1983: 33]

[Wettinger 1983: 33]

Compare also the form in (94a)32, which corresponds to the Modern Maltese surname Zaira, to those found in the place-names in (94b, c): (94a) zayra 1419 small ‘small’ (94b) il galca sihire 1508 DEF field small ‘the small field’ (94c) Jl charcha Jisighire 1574 DEF field DEF:small ‘the small field’

[Borg 1976: 194]

[Wettinger 1983: 62]

[Wettinger 1983: 62]

The forms in (93a, b) illustrate the initial stage, in which the reflex /a/ of */a/ is still preserved, whereas in the third one (93c) this undergoes imāla to /i/. Similarly, /a/ still occurs in (94a), but it is raised to /i/ in (94b, c). As shown by Cowan (1975: 9), the deletion of a short unstressed vowel in word-initial open syllables is first attested as early as c. 1450, in Caxaru’s Cantilena: tred ‘you want’, cf. Ar. turīd. However, evidence from 16th and 17th century place-names suggests that forms preserving this vowel and those in which this has been deleted must have co-existed for a rather long period. Consider the following variants of 16th and 17th century place-names, with or without the /i/ in a word-initial open syllable:

|| 32 For which the “suggested derivation” in Borg (1976: 194) is żgiiṛa. This is also the etymon of the forms in (94b, c).

Phonological changes in Maltese. Evidence from onomastics | 81

(95a) ilchreibe 1514 DEF:little.ruin ‘the little ruin’ (95b) tal chireibe 1533 of:DEF little.ruin ‘of the little ruin’ (96a) ta scayac 1548 of small.alley ‘of a small alley’ (96b) ta sicayac 1659 of small.alley ‘of a small alley’ (97a) Ghar SKalli 1647 cave Sicilian ‘grotta Siciliana’ (= ‘Sicilian cave’) (97b) Mars SiKalli 1647 harbour Sicilian ‘Porto de’ Siciliani’ (= ‘the Sicilians’ Harbour’)

[Wettinger 1983: 44]

[Wettinger 1983: 44]

[Wettinger 1983: 47]

[Wettinger 1983: 47]

[Abela 1647: 73]

[Abela 1647: 106]

The form in (95a) is chronologically the second attestation of the deletion of a short unstressed vowel in a word-initial open syllable, after tred ‘you want’, cf. Ar. turīd, found in Caxaru (c. 1450). Note that the short vowel /i/ is still found in a form (95b) attested 19 years later. Similarly, although the form in (96a) illustrates the deletion of the short vowel /i/, this still occurs in a form (96b) recorded 111 years later. Finally, variants with or without /i/ are attested in the same year, as in (97a, b). On the other hand, some 17th century place-names are only recorded in forms which illustrate the deletion of a short unstressed /i/ in a word-initial open syllable. Consider the following examples: (98a) Aayn Clieb 1647 spring dogs ‘fonte de’ cani’ (= ‘the dogs’ spring’) (98b) El Mchasel 1647 DEF washing.place ‘lauator’ (98c) Ras el Cneyes 1647 head DEF churches ‘capo […] delle Chiese’ (= ‘the cape of the Churches’)

[Abela 1647: 80]

[Abela 1647: 28]

[Abela 1647: 67]

82 | Andrei A. Avram

Finally, the examples under (95)–(97), and possibly those under (98), suggest that Maltese may have had at the time what Cantineau (1960: 110 and 113–114) calls “voyelles ultra-brèves”.33 This would accord both with the variation in the spelling of the place-names in (95)–(97) and with the absence of the short etymological vowel in the forms under (98), which are recorded in the same source with those under (97).

3.4 “Secondary/auxiliary” vowels As shown by Borg (1978: 22), “notarial documents from the 15th and 16th centuries containing transcriptions of Maltese place-names display the secondary vowels a or i in medial consonant clusters”. According to Borg (1978: 21), the quality of the so-called “secondary/auxiliary” vowel is phonologically conditioned: “the Maltese helping vowel [is] a intervening between C2 and C3 when either of these has the backness feature”. There is some evidence form place-names that a “secondary/auxiliary” vowel was still inserted in the 17th century, as illustrated by the following example: (99)

L’Aharasc 1647

[Abela 1647: 25]

DEF:harsh

‘punta così detta, dalla qualità di quell terreno aspro, e ruuido’ [= ‘peak so called given the quality of that harsh and rough ground’] Given that Borg (1978) only looks at place-names, it is worth examining surnames as well. Generally, the complementary distribution of the “secondary/auxiliary” vowels, i.e. [a] in a back environment and [i] elsewhere, is also attested in 15th century surnames and nicknames. Consider the following examples, which correspond to the Modern Maltese surnames Chetcuti (Wettinger 1968: 35), Mahduf (Wettinger 1968: 41), and Maħnuq (Wettinger 1969: 104) respectively: (100a) (100b) (100c) (100d)

kiticuti 1419 mahaduf 1419 Mahanuc 1419 keticuti 1480s

[Wettinger 1968: 35] [Wettinger 1968: 41] [Wettinger 1969: 104] [Wettinger 1968: 35]

|| 33 Such vowels are attested in North-African dialects of Arabic, to which Maltese historically belongs. The existence of ultra short vowels may have been characteristic of earlier Maltese as well.

Phonological changes in Maltese. Evidence from onomastics | 83

However, there are some exceptions, not mentioned by Borg (1978). Note in the following form the occurrence of [i] instead of the expected [a] in a back environment: (101)

il

mahisel 1495 washing.place ‘the washing place’

[Wettinger 1983: 33]

DEF

Also, the insertion of the “secondary/auxiliary” vowel appears to have been subject to variation, that is, the same surname is recorded both with and without the “secondary/auxiliary” vowel. Consider the forms below corresponding to the Modern Maltese surname Dingli: (102a) (102b) (102c)

dinkille 1419 dinkili 1419 dingli 1480s

[Wettinger 1968: 37] [Wettinger 1968: 37] [Wettinger 1968: 37]

Finally, as shown in the next section, the quality of the “secondary/auxiliary” vowel could also be determined by vowel harmony.

3.5 Vowel harmony Surnames also testify to the occurrence of vowel harmony34 in 15th and 16th century Maltese. According to Wettinger (1968: 33), the following forms correspond to the Modern Maltese surname Buġibba: (103a) (103b) (103c)

buiubbe 1419 buiubbe 1480s Bu Jubbe 1540

[Wettinger 1968: 33] [Wettinger 1968: 33] [Wettinger 1983: 58]

As can be seen, in the examples under (103) the vowel in the second syllable harmonizes with that in the first one. However, there is variation in the operation of vowel harmony. Consider the following 15th century variants with or without vowel harmony of the Modern Maltese surnames Butigieg (104), Muhammed (105) and Scicluna (106) respectively:

|| 34 As will be seen, in all examples a vowel is copied, regressively or progressively. In other words, vowel harmony is of the vowel copying type.

84 | Andrei A. Avram

(104a) butugegi 1419 (104b) butigegi 1419 (104c) butigeg 1480s (105a) (105b) (105c) (105d)

muhumud 1419 muhumudi 1419 muhumudo 1419 muhamud 1419

(106a) (106b) (106c)

xikulune 1419 xekilune 1419 xikilune 1419

[Wettinger 1968: 33] [Wettinger 1968: 33] [Wettinger 1968: 33] [Wettinger 1968: 42] [Wettinger 1968: 42] [Wettinger 1969: 100] [Wettinger 1968: 42] [Wettinger 1968: 45] [Wettinger 1969: 97] [Wettinger 1968: 45]

Note that the surname in (106) contains a “secondary/auxiliary” vowel. Its nature is determined by vowel harmony in (106a), whereas it is the non-back environment which enforces the choice of the vowel, i.e. [i], in (106b, c).

4 Conclusions In many cases, evidence from onomastics confirms the absolute and relative chronologies suggested in previous work on the historical phonology of Maltese. Consider first consonants. Both onomastics and other textual evidence suggest that the rules of word-final devoicing of obstruents and of regressive voicing assimilation, discussed in Sections 2.1 and 2.2 respectively, start operating towards the end the 16th century. As shown in Section 2.4, for a long period of time the reflex of *q was still a uvular voiceless stop, in confirmation of Cohen and Vanhove’s (1991: 5) conclusion that “l’articulation actuelle en occlusive glottale n’était pas encore, pour le moins, généralisée” at the time of Caxaru’s Cantilena (c. 1450) and that “elle ne l’était pas d’ailleurs à l’époque de Vassalli”. Consider next vowels. The onomastic evidence discussed in Section 3.3 confirms the analysis by Cowan (1975: 8), according to whom, short unstressed */a/ in open word-initial syllables was first raised to /i/ and subsequently underwent deletion. Also, as seen in Section 3.4, the quality of the “secondary/auxiliary” vowel generally conforms to the pattern of complementary distribution posited by Borg (1978: 21). Finally, the data discussed in Section 3.5 show that vowel harmony was pervasive in 15th century Maltese. This accords with Cohen and Vanhove (1991: 6), who conclude their comments on vowel harmony in Caxaru’s Cantilena (c. 1450) by stating that “[l’harmonisation vocalique] était très forte à cette époque”.

Phonological changes in Maltese. Evidence from onomastics | 85

In other cases, however, evidence provided by the transcriptions of proper names points to rather different conclusions or, at the very least, sheds light on controversial issues. The data presented in Section 2.3 disconfirm recurrent claims in the literature regarding the dating of the loss of fricative interdentals. This widely believed to have occurred before the separation of Maltese from the Arab world. Cohen (1966: 13), for instance, states that “la confusion des deux articulations [dentales et interdentales] peut […] remonter à un stade pré-maltais”. The same claim is repeated in Cohen (1967: 168): “on peut attribuer à un stade pré-maltais la confusion des dentales et des interdentales”. Similarly, according to Vanhove (1994: 170), “l’absence des interdentales remonte à la période pré-historique du maltais”. With respect to one of the interdental fricatives, Agius (1996: 272) puts forth the hypothesis according to which “Maltese does not have the interdental fricative /d/ and historically may have never experienced this articulation”. More recently, Vanhove (2000: 189) also concludes that “l’absence des interdentales […] semble bien être un héritage direct de la variété d’arabe maghrébin dont provient le maltais”. In light of the data examined it is clear that the demise of t occurred much more recently than hitherto assumed. The data examined in Section 2.3 also suggest a different evolution from Arabic to Maltese of the voiced interdentals and dentals (see Avram 2014: 30). The generally accepted relative chronology of the merger of the voiced interdentals with the voiced dentals (see e.g. Cohen 1966: 12, Cohen 1967: 169) can be schematically summarized as follows: stage (i) d ḏ ḍ > d ḍ; stage (ii) d ḍ > d. However, the data show that the voiced interdental fricative is also the reflex of Arabic ḍ, and attest to the existence of a stage in which ḏ ḍ > ḏ. Consequently, this suggests a rather different sequence of events: stage (i) d ḏ ḍ > d ḏ ḍ; stage (ii) d ḏ ḍ > d ḏ; stage (iii) d ḏ > d. As is well known, the current phonetic realization of the reflex of *q as a voiceless glottal stop has been taken by some linguists as proof of the influence of an alleged Punic substrate (Aquilina 1981) or of the Oriental origin of Maltese (Stumme 1904). The spellings of 18th century nicknames discussed in Section 2.4 demonstrate that the reflex of *q was still realized phonetically as a voiceless uvular stop. This clearly rules out a Punic substratal effect, which would have manifested itself much earlier in the history of Maltese. Moreover, as put by Cohen and Vanhove (1991: 181), “le passage de /q/ à /’/ n’est […] pas contemporain de l’arabisation de l’île”. Vanhove (1994: 171) also concludes that “la quasigénéralisation de l’occlusive uvulaire sourde vers l’occlusive laryngale est un phénomène récent”. Finally, the fact that Maltese did have the voiceless uvular stop q also disconfirms Krier’s (1976: 35) claim that in the adaptation of Ro-

86 | Andrei A. Avram

mance loanwords “the strangest case of mutation is represented by ‘the glottal stop’ which replaces voiceless and voiced velars”. Krier (1976: 35) writes that this distinguishes Maltese form the Arabic dialects spoken in the Maghreb, in which “Romance /k/ is borrowed under the form of /q/”. In fact, exactly as in the Maghrebian dialects, /k/ in early Romance loanwords is first rendered by [q]. It is only at a later stage that the phonetic realization of the reflex of *q shifted to [’], the voiceless glottal stop of Modern Maltese. According to Borg (1997: 259), “the unconditional merger of *ḥ and *x had probably already occurred in Medieval M”, given “the use of the digraph ch for the reflex of both OA sounds in the late 15th century poem, Peter Caxaru’s Cantilena”. However, as concluded by Cohen and Vanhove (1991: 4), in Caxaru (c. 1450) “nous avons bien là une preuve supplémentaire de l’existence ancienne de la vélaire sourde en maltais”. Also, the spelling of place-names in Abela (1647) indicates that ḫ and ḥ were still distinct phonemes (Section 2.5). Similarly, the different transcriptions of reflexes of *ġ and *ʕ (Section 2.6) suggests that at least until the 17th century the two fricatives had not merged yet and had not been reinterpreted as vocalic length. Consequently, Cantineau’s (1960: 72) claim that “en maltais, sans doute sous l’influence du substrat punique, les ḫ anciens sont passés à ḥ et les ġ anciens à ε; ḥadem, yaḥdem « travailler », ḥames « cinq », ḥuk « ton frère », etc.; baεal « mulet », εada « demain », εâni « riche », etc.” cannot be upheld. Turning to vowels, as shown in Section 3.1, 15th and 16th century onomastic evidence includes numerous exceptions to the pattern of complementary distribution of the reflexes /e/ and /a/ of the feminine ending */-a/ and of */-a:/ posited by Borg (1976: 194). The phonetic realization and phonological status – both diachronically and synchronically – of the vowel orthographically rendered with the digraph is a matter of some debate in the literature.35 Essentially, the controversy focuses on its nature: monophthongal, i.e. /i:/, or rather diphthongal, i.e. /ie/.36 As shown in Section 3.2, the monophthongal reflex /e:/ of */a:/ first occurs in a place-name recorded as early as 1372, i.e. some 80 years earlier than Caxaru’s Cantilena (c. 1450), hitherto believed to contain the first such attestations (e.g. Cowan 1975: 7, Borg 1976: 200, and Pace 2009: 26). As for the diphthongal reflex of */a:/, Cowan (1975: 7) writes that “if we examine documentary material that

|| 35 For a recent overview see Pace (2009). 36 See e.g. Cohen (1966: 10). For word-final */a:/ in Maltese see also Avram (1994). For the reflexes of */a:/ in the Maltese dialects see Borg (1978: 196). Synchronic imāla is discussed by Borg (1997: 271–273).

Phonological changes in Maltese. Evidence from onomastics | 87

comes after Caxaru, we find that by the time of Megiser in 1588 […] the plain non-final e had become diphthongized to /ie/”. Borg (1978: 200) also states that the first attestations of this diphthongal reflex are those in Megiser’s wordlist. Actually, as shown in Section 3.2, the first occurrence of /ie/ in a surname attested in 1419 precedes those in Megiser by 169 years, and three other instances occur in place-names between 80 to 44 years before Megiser. The forms recorded by Megiser in 1588 and spelled with – see Section 3.2 – are open to two interpretations: while they might stand for a diphthong, presumably [ie] or [iə], they might equally well represent the long vowel [i:], given the German orthographic conventions. However, the occurrence of in notarial documents written by Sicilians, who were unlikely to use it in order to render the long vowel [i:], suggests that at the time the vowel thus transcribed was a diphthong in (some varieties of) Maltese. Moreover, the monophthongal reflex /i:/ of */a:/ is first attested in 1528, that is, 60 years before Megiser. The question arises whether these monophthongal and diphthongal reflexes of */a:/ really “need not be taken as representing different historical stages of the same vowel shift”, as put by Borg (1976: 200). The data in Section 3.2 suggest the following evolution: */a:/ first undergoes imāla to /e:/, and then /e:/ either develops into either diphthong /ie/ or is raised to /i:/. As seen in Section 3.5, regressive or progressive vowel harmony also plays a role in selecting the so-called “secondary/auxiliary” vowel, overriding the complementary distribution posited by Borg (1978: 21). To sum up, the data examined in this paper show that onomastics provides valuable insights into the development of the phonology of Maltese.

References Abela, Giovanni Francesco. 1647. Della descrittione di Malta isola nel mare siciliano. Con le sue antichita, ed altre notitie. Malta: Paolo Bonacota. Agius, Dionisius A. 1996. Siculo Arabic. London and New York: Kegan Paul International. Aquilina, Joseph. 1961. Influenze arabe sulla toponomastica maltese. In Atti del VII Congresso Internazionale di Scienze Onomastiche, 131–146. Florence-Pisa. Aquilina, Joseph. 1964. A comparative study in lexical material relating to nicknames and surnames. Journal of Maltese Studies 2. 147–176. Aquilina, Joseph. 1981. Papers in Maltese Linguistics. La Valetta: The University of Malta. Avram, Andrei A. 1994. On final */aa/ in Maltese. Revue roumaine de linguistique 39(1). 65–73. Avram, Andrei A. 2012. Some phonological changes in Maltese reflected in onomastics. Bucharest Working Papers in Linguistics 14(2). 99–119. Avram, Andrei A. 2014. The fate of the interdental fricatives in Maltese. Romano-Arabica 14. 19–32.

88 | Andrei A. Avram

Bonelli, Luigi. 1897. Il dialetto maltese. Archivio Glottologico Italiano 6. 37–70. Borg, Alexander. 1973. The segmental phonemes of Maltese. Linguistics 109. 5–11. Borg, Alexander. 1975. Maltese morphophonemics. Journal of Maltese Studies 10. 11–28. Borg, Alexander. 1976. The imaala in Maltese. Israel Oriental Studies 6. 191–223. Borg, Alexander. 1978. Historical aspects of auxiliary vowels in Maltese. Wiener Zeitschrift für die Kunde des Morgenlandes 70. 15–34. Borg, Alexander. 1997. Maltese phonology. In Alan S. Kaye (ed.), Phonologies of Asia and Africa,vol. 1, 245–285. Winona Lake: Eisenbrauns. Busuttil, V. 1900. Diziunariu mill Malti ghall Inglis. Giabra ta dwar 30,000 chelma mfissrin mill Malti ghall Inglis bil fraseologia. Valetta – Sliema: N. C. Cortis & Sons. Cachia, Lawrenz. 2000. Habbew l-ilsien Malti. Malta: Veritas Press. Camenzuli, Anthony. 2002. Defamatory nicknames and insults in late eighteenth century Malta: 1771–1798. Melita Historica 13(3). 319–327. Cantineau, Jean. 1960. Cours de phonétique arabe. Paris: Librairie C. Klincksieck. Cardona, Tony. 1997. Introduzzjoni għal-lingwistika Maltija. L-Imsida: Mireva Publications. Cassar, Mario. 2005. Vestiges of Arabic nomenclature in Maltese surnames. Paper presented at XXII Congresso Internazionale di Scienze Onomastiche/22nd International Congress of Onomastic Sciences, 28.8–4.9.2005, Pisa. Cassola, Arnold. 1987–1988. Una edizione diversa della lista di voci maltesi del seicento di Hieronymus Megiser. Journal of Maltese Studies 17/18. 72–86. Cassola, Arnold. 1992. The Biblioteca Vallicelliana “Regole per la Lingua Maltese”. Malta: Said International. Cohen, David. 1966. Le système phonologique du maltais. Aspects synchroniques et diachroniques. Journal of Maltese Studies 3. 1–26. Cohen, David 1967. Contribution à la phonologie diachronique du maltais. In Verhandlungen des zweiten internationalen Dialektologenkongresses, 164–171. Wiesbaden: Steiner. Cohen, David & Martine Vanhove. 1991. La Cantilène maltaise du 15ème siècle: remarques linguistiques. Comptes-rendus du Groupe linguistique d’études chamito-sémitiques 29/30. 177–200. Cowan, William. 1964. An early Maltese word-list. Journal of Maltese Studies 2. 217–225. Cowan, William. 1975. Caxaru’s Cantilena: A checkpoint for change in Maltese. Journal of Maltese Studies 10. 4–10. Dessoulavy, C. L. 1937. Review of Edmund Sutcliffe, A grammar of the Maltese language with chrestomathy and vocabulary. Bulletin of the School of Oriental Studies 8(4). 1175–1188. Dessoulavy, C. L. 1957. Quelques noms propres maltais. Journal of the Faculty of Arts 1(1). 40–47. Krier, Fernande. 1976. Le maltais au contact de l’italien. Hamburg: Helmut Buske Verlag. Pace, Joe Felice. 2009. The tortuous itinerary of the Maltese vowel ie. Ilsienna 1. 21–50. Schembri, Ludovik. 1998. Dictionary Dizzjunarju. Maltese-English. English-Maltese. Mġarr: Colour Image. Soldanis, Giovanni Pietro Francesco Agius de. 1750. Nuova scuola di grammatica per agevolmente apprendere la lingua punica maltese, aperta agli studenti maltesi, e forestieri abitanti in Malta. Rome: Generoso Salomoni. Soldanis, Giovanni Pietro Francesco Agius de . 1750–1767. Ḋamma tal Kliem Kartaginis mscerred fel fom tal Maltin u Ghaucin. Ms 143/I–IV, National Library of Malta. Stumme, Hans. 1904. Maltesische Studien. Eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache nebst Erläuterungen. Leipzig: Hinrichs’sche Buchhandlung.

Phonological changes in Maltese. Evidence from onomastics | 89

Vanhove, Martine. 1994. La langue maltaise un carrefour linguistique. Revue du monde musulman et de la Méditerranée 71(1). 167–183. Vanhove, Martine. 2000. Le maltais et les interférences linguistiques. In Sonia Cristofaro & Ignazio Putzu (eds.), Language in the Mediterranean area. Typology and convergence. (Il progetto MEDTYP: Studio dell’area linguistica mediterranea), 187–199. Milan: Franco Angeli. Vassalli, Mikiel Anton. 1796. Ktŷb yl klŷm Mâlti ’mfysser byl-latin u byt-talyân. Lexicon Melitense-Latino-Italum. Rome: Antonio Fulgone. Vella, Francis. 1848. Dizionario portatile delle lingue maltese, italiana, inglese, parte prima. Livorno: Stamperia degli Artisti Tipografi. Vella, Olvin. 2013. A Gozitan dialect for Standard Maltese in the eighteenth century. Paper presented at the 4th International Conference on Maltese Linguistics, 17–19 June 2013, “Lumière” University, Lyon. Wehr, Hans. 1976. A dictionary of Modern written Arabic, edited by J. M. Cowan, Ithaca, NY, Spoken Language Services. Wettinger, Godfrey. 1968. The distribution of surnames in Malta in 1419 and in the 1480s. Journal of Maltese Studies 5. 25–48. Wettinger, Godfrey. 1969. The Militia list of 1419–20: A new starting point for the study of Malta’s population. Melita Historica 5(2). 80–106. Wettinger, Godfrey. 1971. Late medieval Maltese nicknames. Journal of Maltese Studies 6. 34–46. Wettinger, Godfrey. 1983. Some grammatical characteristics of the place-names of Malta and Gozo in early modern times. Journal of Maltese Studies 15. 31–68. Wettinger, Godfrey. 1999. The origin of Maltese surnames. Melita Historica 12(4). 333–344. Wettinger, Godfrey. 2000. Place-names of the Maltese islands: Ca. 1300–2000. M’Scala: PEG Publications. Wettinger, Godfrey & Michael Fsadni. 1968. Peter Caxaru’s Cantilena. A poem in medieval Maltese. Malta. Wettinger, Godfrey & Michael Fsadni. 1983. L-għanja ta’ Pietru Caxaru. Poeżija bil-Malti medjevali. Malta.

Alexandra Vella, Flavia Chetcuti and Sarah Agius

Lengthening as a discourse strategy in Maltese: Phonetic and phonological characteristics Abstract: Pausing and lengthening in speech seem to be inextricably linked. This is evidenced by the phenomenon of pre-boundary lengthening (Vaissière 1983). Analysis of new spontaneous (but, significantly, non-Map Task) data from Maltese has brought to light a distinct form of lengthening, noted also for other languages (e.g. Campione and Véronis’s 2006 “hesitation lengthening”) which involves a longer than long duration, does more than demarcate the edges of intonation phrases, and functions as more than simply as a sign of hesitation. This paper describes the distributional and the phonetic and phonological, particularly prosodic characteristics, of this phenomenon in Maltese. It examines its functions in discourse and shows how it is similar to, but also different from both filled and unfilled pauses. Keywords: pausing, lengthening, hesitation lengthening, discourse strategies, Maltese phonetics and phonology

1 Introduction This paper deals with lengthening in Maltese, of the type which has been referred to in the literature (Campione and Véronis 2006) as “hesitation lengthening”. Such lengthening is often associated with boundaries, and therefore will be examined alongside other related phenomena which occur in speech, namely, pauses of both types: the silent or unfilled type, and filled pauses (Cruttenden 1997). In this paper we attempt to show that “hesitation lengthening” is a phenomenon distinct from other related phenomena such as that of pre-boundary lengthening often noted to occur in the context of boundaries that is before

|| Alexandra Vella: University of Malta, Institute of Linguistics, Msida MSD 2080, Malta. E-mail: [email protected] Flavia Chetcuti: University of Malta, Institute of Linguistics, Msida MSD 2080, Malta. E-mail: [email protected] Sarah Agius: University of Malta, Institute of Linguistics, Msida MSD 2080, Malta.

92 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

unfilled pauses. It is also distinct from filled pauses although the latter too can involve “enhanced” lengthening of a similar sort to the “hesitation lengthening” that is examined in this paper. Our goal is therefore to show that whilst, at face value, “hesitation lengthening”, like, to a certain extent, both unfilled but even more so, filled pauses, may be perceived as a reflex of “hesitation”, these phenomena should rather be seen as elements of spontaneous speech which contribute in an important way to managing and maintaining discourse flow. The goal of the present paper is therefore twofold: on the one hand, it provides a description of the phonetic and phonological, particularly prosodic, characteristics of the phenomenon of “hesitation lengthening” in Maltese, as distinct from other related phenomena; on the other, it attempts to show that this phenomenon, alongside other related phenomena, is a discourse strategy used by speakers to manage and maintain the speech flow. More specifically, this paper seeks to identify the characteristics of this phenomenon as used in the Map and Conversation Tasks from the MalToBI corpus (Vella and Farrugia 2006), as well as in data from the MAMCO corpus (e.g. Vella and Paggio 2013). The paper is organized as follows. Section 2 provides a brief overview of research on the phenomena of pausing and lengthening, including on various discourse functions they perform, noted in the literature. Section 3 introduces the data used as the basis for the analysis, briefly explaining also how the phenomenon in question came to light, and what decisions were taken with respect to its annotation. Section 4 provides a description of the distributional, as well as the phonetic and phonological, particularly intonational, characteristics of this lengthening phenomenon. An attempt is also made to compare the behavior of this phenomenon to that of other features of spontaneous speech, collectively often referred to as “normal disfluencies” (Shriberg 1994), in so far as its discourse functions go. Finally, Section 5 provides some preliminary conclusions and outlines possible avenues for further research.

2 Hesitation: its manifestations and use as a discourse strategy Pausing often involves what Cruttenden (1997) refers to as silent or unfilled pauses, and is necessary, amongst other things, to allow a speaker to take breath. It is also used as a means to organize information into chunks which can be decoded, often occurring at major or minor constituent boundaries and having a demarcation function. Whilst pausing could be seen to reflect hesitation, research has shown that silent pauses are much more than simply markers of

Lengthening as a discourse strategy in Maltese | 93

hesitation and in fact fulfill a wide variety of functions. For example, a pause can allow the speaker time to think about what s/he wants to say in monologue contexts; in dialogue contexts it can signal to the interlocutor that the speaker is thinking about what to say next. Pausing can therefore be seen as an important means enabling the speaker to organize and manage the speech flow. Pauses can also be filled. Filled pauses (FPs), unlike unfilled ones, do not serve a demarcation function. Like unfilled pauses, however, they serve a variety of purposes, and again, although such pauses might appear, at face value, to be a reflex of hesitation, they can, like silent pauses, also be seen to function as a means to organize and manage the speech flow. Briefly therefore, pauses, unfilled as well as filled, are acknowledged to be more than simply markers of hesitation. As Couper-Kuhlen (1986: 75) notes, they not only have “a performance-related origin – a pause for breath”, but they are often used “to search for a word or to plan”. This role in planning has been noted extensively in the literature for both types of pauses across different texttypes, dialogue as well as monologue, and different languages (e.g. Cutler and Pearson 1986, Grosz and Hirschberg 1992, Swerts 1998, Clark and Fox Tree 2002), Maltese included (Vella et al. 2011 and 2014). Research has also shown that pausing often works hand-in-hand with lengthening, the well-studied phenomenon involved often being referred to as pre-boundary lengthening (e.g. Vaissière 1983, Hayes 1997). Pre-boundary lengthening (PBL) involves lengthening of the element before the boundary. A tendency for lengthening seems to apply also to FPs which, although not necessarily lengthened themselves, often seem to introduce an element of hesitation and/or additional length simply by virtue of their presence, which essentially, results in an “additional” element being added to the utterance. In the course of work on the annotation of spoken Maltese data, a phenomenon which is different to the phenomena mentioned above, but in some ways similar, came to light. This phenomenon consists of some elements in the speech flow itself, rather than intervening pauses, unfilled or filled, being characterized by a noticeably longer than long duration. This lengthening phenomenon was noted to occur in a much wider variety of contexts than the contexts involving either unfilled pauses, and consequently PBL, or FPs. A similar phenomenon, referred to, for example, by Campione and Véronis (2006) as “hesitation lengthening”, has been noted in the literature, although various other terms have been used (e.g. “vocal hesitations” Guaïtella 1993; “stretched out sounds”, Eklund and Shriberg 1998; “prolongations” Moniz et al. 2007). In this paper, we will adopt the term “hesitation lengthening” to refer to this phenomenon. As mentioned above, one important element in speech is the need for the speaker to organize and manage both the discourse content and its flow. As already mentioned above, Cruttenden (1987) notes, for example, that silent or

94 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

unfilled pauses are often, though not exclusively, used at major or minor constitutent boundaries. The location of silent pauses and the boundaries which accompany them aligns quite well with constituent boundaries at levels of structure other than the phonological one, giving structure to the phonology in so doing, but also organizing the content. FPs do not share this role at the structural level with unfilled ones. Both types of pauses however be seen to be similar in that both serve a planning function, giving the speaker time for lexical retrieval and planning (e.g. Goldman-Eisler 1968, Rose 1998). Other functions have also been suggested for both unfilled and filled pauses, and these include a role in managing turn-taking, pauses of both types being used as floor-holders (e.g. Eklund and Shriberg 1998) as well as a means of coordinating with another speaker during communication (e.g. Brennan and Williams 1995). Interestingly, the literature also suggests roles for hesitation lengthening, as a planning device (e.g. Maclay and Osgood 1959, Butterworth 1975, Rose 1998), as a floor-holder (Eklund and Shriberg 1998), and as an aid to turn-taking (e.g. Guaïtella 1993). In the words of the latter (Guaïtella 1993: 131), the role of vocal hesitations of different types in communication “is probably to create a ‘time of rest’ during which both the speaker and listener can review what has been said and extrapolate what is going to be said”, essentially “the same psycholinguistic purpose (stalling during processing)” (Rose 1998: 68). Because the data reported on here is not a monologue, the speech flow is managed by more than one person and therefore the use of pauses, unfilled as well as filled, as well as that of hesitation lengthening, in maintaining speech flow and moving the discourse forward becomes even more significant. This paper seeks to examine the phenomenon of hesitation lengthening in Maltese data, first, with a view to confirming its existence as a phenomenon in its own right, similar in nature to standard PBL, although the lengthening seems more pronounced. Hesitation lengthening however does seem to differ from PBL in that this type of lengthening is possible at a much wider range of locations, and certainly not always or only at the structural ones at which one would expect a phonological boundary to be placed (this may or may not coincide with syntactic boundaries). Second, the paper attempts to describe the phonetic and phonological, as well as functional characteristics of hesitation lengthening. To summarize, the paper attempts to establish whether there are any similarities between the phenomenon of hesitation lengthening, distributionally, but also intonationally and functionally, as compared to PBL and FPs. In the following sections, we examine data from Maltese in order to confirm the existence of the phenomenon we are referring to as hesitation lengthening, to describe both its distributional, as well as its phonetic and phonological characteristics, and to examine the similarities and/or differences in function between this phenomenon and PBL, on the one hand, and FPs on the other.

Lengthening as a discourse strategy in Maltese | 95

3 Methodology and annotation of data The data analyzed forms part of the MalToBI corpus, reported on in Vella and Farrugia (2006) and the MAMCO corpus, reported on, in, amongst others, Vella and Paggio (2013) and Paggio and Vella (in press). The MalToBI corpus includes recordings involving 8 Map Tasks and 8 Conversation Tasks designed to be representative of spoken Standard Maltese. The MAMCO corpus consists of 12 first encounter conversations, also involving speakers of spoken Standard Maltese. The MalToBI corpus includes data from a balanced group of female and male speakers from two age groups, 18 to 29, and over 30 and having different levels of education (up to secondary, post-secondary or tertiary). The MAMCO corpus includes data from 6 female and 6 male University students all aged 19 to 25. Both corpora consist of data from speakers of Standard Maltese. The Map Task involves quasi-conversational, rather than fully spontaneous speech. The Maltese Map Task was designed on lines similar to those of the task used to collect the data contained in the HCRC Map Task Corpus (Carletta et al. 1995). The spoken data collected is spontaneous in nature whilst at the same time also comprising an element of controlled elicitation. The Maltese Map Task included locations – “target items” – consisting of all-sonorant material. These target items represent different syllable structure and accentual possibilities in Maltese. The Conversation Task gave speakers the opportunity to speak freely within a given scenario. One of the speakers was asked to pretend that s/he knew someone who was seeking to fill a vacant post in the company s/he worked for. The task involved this speaker talking to the other speaker and trying to gather information which could be relayed on, regarding the suitability or otherwise of this person as a possible candidate for the job. Speakers changed roles when they felt they had nothing more to say. For the MAMCO recordings, participants were asked to simulate first encounter conversations, and an important prerequisite for this was that the two participants had not met before. The participants were instructed to get to pretend that they were at a party, and were introducing themselves to a previously unknown partner. Introduction over, they were free to decide what to talk about with their previously unknown interlocutor. Annotation of the data from the above-mentioned corpora was carried out using PRAAT (Boersma and Weenink 2008) and took place in two stages. First, the orthographic annotation of the eight Map Tasks from the MalToBI was carried out in the context of the project SPAN following a set of conventions which were developed for this purpose in the process of this same project (Vella et al. 2010 and the forthcoming website for a summary of the annotation conventions used). After this, the Conversation Task data and the MAMCO corpus data were transcribed in parallel, following the same standards and conventions as the Map Tasks.

96 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

It was in the course of the annotation of these new, relatively more spontaneous, and significantly non-Map Task data, that a “new” feature was noted. This involves certain elements in the discourse which intuitively appear to have a noticeably longer than long duration. This feature appeared to occur relatively frequently in these data in comparison to in the Map Task data and a decision was therefore taken to mark up the phenomenon in the SP1 and SP2 tiers and to update the SPAN guidelines accordingly. An important reason for annotating this “new” phenomenon is that doing so enhances searchability. Speaker tiers in the SPAN annotations contain word-by-word annotations of the contributions of each speaker. In the examples shown in this paper, a semiliteral translation is provided underneath each speaker tier. The annotation makes use of standard orthography except in cases where it is considered important to record deviation from the norm of any sort such as elisions (e.g. naq’a ‘a little bit’), insertions (e.g. hekk[e] ‘so[eh]’) or variations (e.g. &laqqas for lanqas ‘not even’). It was decided to mark this new phenomenon of lengthening using a plus sign, +, to be placed following the element intuitively perceived as lengthened by the annotator (e.g. imma+ ‘but’, lej+n ‘towards’). Two examples involving hesitation lengthening are shown in Figure 1 and foregrounded by means of bold black outline boxes. In some of the examples shown in this paper, an additional “syllable” tier has been added with the purpose of highlighting more precisely which element or elements of the phonetic/phonological string are subject to lengthening.

Fig. 1: Two examples involving hesitation lengthening – the function word il+‘the’ and the content word il-famuża+ ‘the famous [one=(event)]’.

Lengthening as a discourse strategy in Maltese | 97

An important reason for annotating this “new” phenomenon is that doing so enhances searchability. In the rest of this paper, an attempt will be made to describe both distributional patterns, and phonetic and phonological, particularly prosodic characteristics, of this phenomenon, as well as functions which seem to be associated with its occurrence in data from the corpora described above.

4 Results and discussion This section consists of five main sections. The first presents concrete evidence to support the existence of the hesitation lengthening noted by transcribers in the data as a distinct phenomenon (Section 4.1). This is followed (Section 4.2) by an examination of the locations in which hesitation lengthening can occur, at the local level (Subsection 4.2.1) and in the more extended domain (Subsection 4.2.2). Some observations on the types of lexical items which seem more prone to lengthening will also be provided (Subsection 4.2.3). A preliminary description of some of the phonetic and phonological, particularly prosodic, characteristics of this phenomenon is then provided (Section 4.3). This is followed by an examination of some of the functions which seem to be associated with its use (Section 4.4). Lastly, a comparison of the characteristics of this phenomenon as compared to the related phenomena of PBL and FPs is provided (Section 4.5).

4.1 Evidence for the phenomenon of “hesitation lengthening” This section is intended to provide evidence in support of the existence of a distinct phenomenon of “hesitation lengthening” noted to occur in the Maltese data described in Section 3 above. All instances of imma ‘but’, those annotated as non-lengthened, as well as those annotated as lengthened (imma+), in 2 recordings from the MAMCO data analyzed, were divided into syllables and durations measured. Figure 2 below presents the results of this preliminary investigation. These instances occurred in a variety of contexts, with silence on both sides as well as on neither side, as well as with speech only to the left or only to the right. Given the limited data, it was impossible to work out whether or not duration is influenced by context. What is clear from this preliminary examination of the data, however, is that instances of this word intuitively identified by the transcribers as being lengthened and annotated as imma+, are in fact substantially longer in duration than their counterparts annotated as imma.

98 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

Fig. 2: Duration of imma vs. imma+.

These, admittedly limited, data, also suggested that whereas the two syllables in the non-lengthened instances of imma have similar durations, this is not the case for the two syllables in the lengthened instances of imma+, where the second syllable is, in all cases, at least twice as long, or even longer, as the first syllable. These preliminary results suggest that the choice transcribers made to mark up an element as noticeably lengthened or not was not just a random response to an impression of “lengthening”. More detailed work on this aspect is worth carrying out.

4.2 Locations in which hesitation lengthening occurs This section examines hesitation lengthening more closely with a view to establishing which locations it tends to occur in, starting with the local level (Subsection 4.2.1), moving on to the larger domain (Subsection 4.2.2), and ending with a brief examination of the types of lexical items which appear to be more prone to being lengthened (Subsection 4.2.3).

Lengthening as a discourse strategy in Maltese | 99

4.2.1 The local level The data analyzed show that hesitation lengthening is not a feature of the stressed syllable (which often, though certainly not always, falls on the penultimate syllable in Maltese, Vella 2009), unless this happens to be the final syllable. As can be seen in Figure 3, lengthening occurs on the final open syllable of the disyllabic word għadni+ ‘I’ve yet to’. (Please note that, where necessary, lexically stressed syllables in the examples used in this paper are indicated in bold in the text.) Lengthening of the final syllable of a lexical item, rather than of a penultimate stressed syllable, occurs frequently in the data analyzed. In this example, as also in all the examples of imma ‘but’, discussed in Section 4.1, the final unstressed syllable is an open syllable. In such cases, it is easy to “perceive” the lengthening as affecting mainly the final vowel.

Fig. 3: Lengthening on the final unstressed syllable of a disyllabic word.

The example shown in Figure 4 involves lengthening on the monosyllabic word, tfa+l ‘children’, which, in this case, involves a closed syllable. The annotator’s perception here was also that the lengthening affects mainly the vowel, rather than the whole of the rhyme, which explains the annotator’s placement of the + immediately following the vowel.

100 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

Fig. 4: Lengthening on a closed monosyllabic lexical item.

Whilst the example of lengthening of tfal ‘children’, shown in Figure 4 above, suggests that what gets lengthened could be the vowel rather than the whole syllable, it would certainly be premature to reach such a conclusion without further study. Thus, for example, Figure 5 below shows an example of lengthening on il- ‘the’, whose rhyme is similar in structure to that of tfal: the annotator’s intuition was that lengthening affected the whole rhyme, rather than simply the vowel in this case.

Fig. 5: Lengthening on il- ‘the’, comprising a closed monosyllable.

Lengthening as a discourse strategy in Maltese | 101

Thus, although lengthening seems to occur mostly on the final syllable (see also discussion in Subsection 4.2.1 and also Figure 1 above), this preliminary analysis suggests that it might be worth investigating whether the element that is being lengthened is in fact the syllable, or whether it could be some other domain such as the rhyme, the vowel nucleus or the coda, or potentially, also the onset.

4.2.2 The extended domain Analysis of the locations in which hesitation lengthening occurs shows that this phenomenon, unlike PBL, is not restricted to occurring at boundaries, although it must be kept in mind that this type of lengthening, arising as it does in the context of hesitation, renders the notion of what counts as a boundary tricky at best. The data analyzed include various examples of hesitation lengthening occurring at points where one would not expect a boundary to occur. It is worth recalling that many authors (e.g. Cruttenden 1997) have often noted the tendency for pauses between phonological units to occur at constituent boundaries rather than within constituents. Two such examples consisting of lengthening which effectively breaks up a constituent are shown in Figures 6 and 7 below.

Fig. 6: Example of lengthening on the definite article il- ‘the’.

102 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

Fig. 7: Example of lengthening on the quantifier ħafna+ ‘many’.

In the first of these two examples shown in Figure 6, the definite article in il+ħames waħda ‘the fifth one’, il- is lengthened and distanced, as a result, from the rest of the noun phrase ħames waħda ‘fifth one’. Similarly, in the second example shown in Figure 7, the quantifier ħafna+ ‘many’ is virtually cut off from what comes next (not seen in this Figure but which in fact involves a full repair), by the lengthening and by the two false starts which follow. In neither case would it be usual for lexical elements of this sort to be separated from the following noun in contexts devoid of hesitation.

4.2.3 Which lexical items can be lengthened? As should already be apparent, hesitation lengthening does not seem to be limited to happening only to certain categories of words. Preliminary examination of the lexical items more prone to occur in contexts of hesitation lengthening suggests that there is a slightly greater preponderance of function words rather than content words amongst them, particularly of connecting items of different sorts. Table 1 below lists lexical items noted in the data analyzed to be lengthened more frequently than others:

Lengthening as a discourse strategy in Maltese | 103

Tab. 1: Examples of frequently lengthened lexical items in the data analyzed.

allura

‘so’

hawn/haw’

‘I mean’

il-

‘the’

imma

‘but’

iva/le

‘yes/no’

jiena

I’

tipo

‘sort of’

u

‘and’

xi

‘about’ or ‘some’

It must be noted that some of the elements in the list above have a kind of “pivotal” nature in coordinating elements within a discourse. Although this role will not be discussed further here, it is clearly another aspect which might be worth investigating further.

4.3 Intonational characteristics of hesitation lengthening Analysis of the instances involving hesitation lengthening shows that these elements have a mid-level F0. Given the analysis to date, however, it seems to be unclear whether this mid-level F0 contour starts before the lengthened syllable or not. In fact, it looks like the mid-level intonation which seems to be associated with elements which are lengthened could possibly extend the whole of the element whose part is being lengthened, rather than simply characterizing the part of the element which gets lengthened. For example, in the example shown in Figure 8 below involving the conjunction imma ‘but’, the mid-level F0 on imma extends from the start of imma through to the end. In this case, imma is both preceded and followed by a pause, but other instances of this phenomenon with pause on either one side or not at all, also seem to be characterized by this level intonation. This mid-level F0 is strikingly similar to that found in FPs, which are also characterized by a mid-level F0 throughout their realization (Vella et al. 2011). Figure 9 below shows an example of the FP Ee, preceded and followed by silence, with its characteristic mid-level F0. This is followed by Figure 10 which illustrates a sequence involving hesitation lengthening on the definite article il-+ ‘the’ which is followed by the FP ee. A slight lowering in F0 can be seen at the start of il-+, which settles into the mid-level F0 noted here to be characteristic of hesitation lengthening. This same mid-level F0 can be seen to extend throughout the FP ee.

104 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

Fig. 8: Intonation of Imma+.

Fig. 9: Mid-level F0 characteristic of FPs.

Fig. 10: Mid-level F0 on hesitation lengthened il- and following FP ee.

Lengthening as a discourse strategy in Maltese | 105

To conclude this section, Figure 11 can serve to illustrate the distinctive intonation which is characteristic of the phenomenon under investigation. The example illustrated contains two instances of use of the content word naqra ‘a little’. The first naqra does not involve hesitation lengthening, occurring instead in post-nuclear position to the verb Tbeżżagħni ‘It scares me’. Here one can see that there is a fall from the stressed syllable of Tbeżżagħni, followed by the phrase accent slight rise described in Vella (2003) on naqra. The second instance of naqra, by contrast, is an example of the phenomenon we are describing here and is characterized both by the lengthening under discussion as well as by the mid-level F0 characteristic of this phenomenon. Once again it can be noted that in this case, the mid-level F0 seems to extend through the whole of naqra rather than only on the final syllable, the slight fall in this case possibly being the result of microprosodic effects, including the presence of the “qr” /ʔr/ medial sequence in this lexical item.

Fig. 11: Intonation of naqra vs. naqra+.

As can be noted from the examples given above, the data also shows that this mid-level F0 characterizes the lengthened elements irrespectively of whether this lengthening occurs in function words or content words.

4.4 The functions of hesitation lengthening The picture which is beginning to emerge is in fact that hesitation lengthening is often used to buy time for lexical retrieval and to help plan what to say next, as in the example in Figure 6. It is also used in relation to repairing false starts,

106 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

as in the example in Figure 7. Moreover, as we will see later in this section, sometimes these two functions combine. A further example involving the function of buying time for lexical retrieval and planning can be seen in Figure 12 below. As in the case of the example illustrated in Figure 6 where the speaker stumbles on the first syllable of ħames, ħa... ‘fifth’ which is then retrieved, in this example, there is a stumble on the first syllable tie+ of tielgħa ‘going up’. There are two differences between these two examples. In the former case, the speaker backtracks slightly, introduces the definite article il-+, and lengthening this before recovering; in the latter case, lengthening tie+ serves to get the speaker back on track.

Fig. 12: Lengthening related to lexical retrieval of the word tielgħa ‘going up’.

An example similar to that in Figure 7 involving hesitation lengthening functioning in a context requiring repair can be seen in Figure 13.

Fig. 13: Repair-related lengthening of a preposition+definite article before noun.

Lengthening as a discourse strategy in Maltese | 107

The example shown in Figure 7 is an example in which hesitation lengthening results in a complete change of tack. Hesitation lengthening can however also serve as a successful means of repair as can be seen in the example in Figure 13. In this case, ma+l- falls short of the required assimilation of the /l/ to [ʃ] required before the noun xogħol. By stalling on ma+l, the speaker succeeds in buying time for repair, although reinforcement of the repair results in a second lengthening of the first part of the repaired ma+x. This preliminary analysis therefore suggests that there is a link between use of this phenomenon for lexical access and its use in contexts of repair, some of which in fact result from difficulties of lexical access. Two further illustrative examples are shown in Figures 14 and 15. In the example in Figure 14, the speaker lengthens the definite article il-+ before inserting a break and the discourse marker haw’ followed by another break during which the speaker continues to search for a way to express the intended meaning, which in the end, is described in very broad terms as ilfamuża+ ‘the famous (one=event)’.

Fig. 14: Lengthening in the context of lexical retrieval.

In Figure 15 below, the speaker lengthens the definite article il-+ before the lexical item “confectioner”, then goes on to use the same discourse marker used in the example shown in Figure 14 above, namely haw’, as a prelude to introducing the intended lexical item “grocer”. The above are a few examples of the way that hesitation lengthening is used by speakers as a way of managing the speech flow, particularly as a means to

108 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

lexical retrieval and repair rather than as simple hesitation. It is interesting to note also that hesitation lengthening seems to combine with: a) pauses, so that elements before a boundary, which might also be subject to PBL could be subject to further “enhanced” lengthening, and/or additional pauses which might be introduced; b) with FPs, which sometimes also occur in the vicinity of elements which have been lengthened, and/or which themselves can sometimes be subject to further “enhanced” lengthening.

Fig. 15: Lengthening in the context of repair related to lexical retrieval.

A function of hesitation lengthening which has not been explored here is its possible use as a floor-holder and hence as a means to successful turn-taking. This is certainly an avenue worth exploring. Even at this preliminary stage in the study of this phenomenon, however, it is safe to say that hesitation lengthening contributes to maintaining speech flow, a role which it shares with other elements of discourse.

4.5 Hesitation lengthening vs. PBL and FPs As mentioned in Section 4.2.1 above, hesitation lengthening involves lengthening of an existing element, often but not always all or part of the final syllable. Such lengthening of the final syllable, regardless of whether this syllable is stressed or unstressed, has been suggested for example by Cruttenden (1997) as one of the characteristics of PBL. In this respect, the phenomenon in question therefore

Lengthening as a discourse strategy in Maltese | 109

seems to pattern with PBL. The phenomenon under study is however different from PBL since this type of lengthening does not seem to be restricted to particular structural location as in the case of PBL which occurs pre-boundary. More specifically, PBL, unlike hesitation lengthening, seems to be more of a phonological phenomenon which maps on directly to phonological structure and particularly prosodic organization, and does not seem to play a role at the discourse level. At face value, hesitation lengthening differs from FPs in that whilst the latter involve an “additional” element being introduced into the speech flow, hesitation lengthening is, as it were, superimposed on “existing” material. Nevertheless, the data analyzed has shown that elements involving hesitation lengthening may be similar to FPs, certainly in their intonation, which typically has the same mid-level F0 contour as do FPs (see Section 4.3. above), but also in that there seem to be few restrictions on what elements can be affected by hesitation lengthening. Also, elements marked by hesitation lengthening seem to have a similar role in planning and managing speech flow as do FPs. Table 2 below provides a comparison between the phenomenon of hesitation lengthening and PBL on the one hand and FPs on the other. Tab. 2: Characteristics of PBL and FPs compared to those of hesitation lengthening.

Similarities and Differences PBL

Hesitation Lengthening

FPs

Existing element

Existing element

Additional element

Element before boundary

Often but not always final syllable Whole additional element

Restricted to pre-boundary Unrestricted

Pause in vicinity (before and/ or after)

Long

Longer

Long or Longer

Typical edge “tones”

Mid-level F0

Mid-level F0

The above clearly shows that hesitation lengthening shares some of the characteristics of PBL but differs from it in some important ways, for example with respect to the important grammatical role of the latter, and the related possibilities for where the two types of phenomena can occur, PBL being much more restricted in this respect. By contrast, hesitation lengthening shares a number of characteristics with FPs, not only at the level of its intonation, but also in so far as the ways in which it contributes to maintaining speech flow.

110 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

5 Conclusion This paper identifies and describes the phenomenon of hesitation lengthening in Maltese, establishing its existence by means of a preliminary durational investigation of a small subset involving “distinct” data sets (Section 4.1). It goes on to describe (Section 4.2) where hesitation lengthening occurs, both locally (final syllable? sub-part of the syllable?) and in the extended domain (at which points within the utterance?), as well as which types of lexical items. It then provides a first description of phonetic and phonological, particularly intonational characteristics, associated with this phenomenon (Section 4.3), as well as of aspects of the discourse role of this phenomenon (Section 4.4). Whilst further analysis is certainly needed, this preliminary investigation suggests that elements involving hesitation lengthening are indeed much longer than similar non-lengthened elements (Section 4.1). The preliminary investigation also suggests a tendency for the lengthened element to be the final syllable, even when this is unstressed (Section 4.2.1). It is not yet clear however whether hesitation lengthening affects the whole syllable or whether it can affect specific sub-part/s of syllables (Section 4.2.1). It is also not yet clear whether hesitation lengthening substitutes for pauses rather than co-occurring with them (Section 4.2.2), although it does appear from this preliminary investigation that lengthening tends to occur more frequently, though not exclusively, on function words, particularly coordinating elements, rather than on content words (Section 4.2.3). Lastly, this investigation provides some firm evidence that, as in other languages, hesitation lengthening is more than simply a marker of hesitation (Section 4.4). Rather it can be seen to constitute a specific strategy that speakers have at their disposal to keep conversation flowing by providing time for lexical access and allowing repairs to be made, for instance. This is in line with what other researchers have found to be the case for other languages. Other functions such as the role in turn-taking and in coordinating with other speakers in the course of communication, reported e.g. by Brennan and Williams (1995) and Eklund and Shriberg (1998), are also likely to be involved in the Maltese data, the nature and extent of such functions being other possible avenues for further research in this area. To summarize, hesitation lengthening is similar to but distinct from PBL and overall seems to be more similar, in both form and function, to FPs. Further research into this phenomenon is necessary as this would provide useful insights into the nature of conversational processes in Maltese and more generally.

Lengthening as a discourse strategy in Maltese | 111

References Boersma, Paul & David Weenink. 2008. PRAAT: Doing phonetics by computer. (Version 5.0.08). http://www.praat.org visited 11-Feb-08. Brennan, Susan E. & Maurice Williams. 1995. The feeling of another’s knowing: Prosody and filled pauses as cues to listeners about the metacognitive states of speakers. Journal of Memory and Language 34. 383–398. Butterworth, Brian. 1975. Hesitation and semantic planning in speech. Journal of Psycholinguistic Research 4. 75–87. Campione, Estelle & Jean Véronis. 2006. Pauses and hesitations in French spontaneous speech. Proceedings of disfluency in spontaneous speech. 43–46. Carletta, Jean, Amy Isard, Jacqueline Kowtko, Gwyneth Doherty-Sneddon & Anne H. Anderson. 1995. The coding of dialogue structure in a corpus. Proceedings of the twentieth workshop on language technology: Corpus-based approaches to dialogue modelling. 25–34. Clark, Herbert & Jean E. Fox Tree. 2002. Using uh and um in spontaneous speech. Cognition 84. 73–111. Couper-Kuhlen, Elizabeth. 1986. An introduction to English prosody. London: Edward Arnold. Cruttenden, Alan. 1997. Intonation. Cambridge: Cambridge University Press. Cutler, Ann & Mark Pearson. 1986. On the analysis of prosodic turn-taking cues. In Catherine Johns-Lewis (ed.), Intonation and discourse, 139–155. London: Croom Helm. Eklund, Robert & Elizabeth E. Shriberg. 1998. Crosslinguistic disfluency modelling: A comparative analysis of Swedish and American English human-human and human-machine dialogues. Proceedings of the 3rd international conference on spoken language processing. 2631–2634. Goldman-Eisler, Frieda. 1968. Psycholinguistics: Experiments in spontaneous speech. London: Academic Press. Grosz, Barbara & Julia Hirschberg. 1992. Some intentional characteristics of discourse structure. Proceedings of the International Conference on Spoken Language Processing, Banff, Canada. 429–432. Guaïtella, Isabelle. 1993. Functional, acoustic and perceptual analysis of vocal hesitations in spontaneous speech. Proceedings of ESCA workshop on prosody, Lund, Sweden. 128–131. Hayes, Bruce. 1997. Metrical stress theory: Principles and case studies. Chicago and London: Chicago University Press. Maclay, Howard & Charles E. Osgood. 1959. Hesitation phenomena in spontaneous English speech. Word 15. 19–44. Moniz, Helena, Ana Isabel Mata & M. Céu Viana. 2007. On filled-pauses and prolongations in European Portuguese. Proceedings of interspeech, Antwerp, Belgium. 2645–2648. Paggio, Patrizia & Alexandra Vella. in press. Overlaps in Maltese conversational and taskoriented dialogues. Post-proceedings of the 1st European symposium on multimodal communication, Malta. Rose, Ralph L. 1998. The communicative value of filled pauses in spontaneous speech. Master’s dissertation, University of Birmingham, Birmingham, UK. Shriberg, Elizabeth E. 1994. Preliminaries to a theory of speech disfluencies. Ph.D. dissertation, University of California at Berkeley, USA. Swerts, Marc. 1998. Filled pauses as markers of discourse structure. Journal of Pragmatics 30. 485–496.

112 | Alexandra Vella, Flavia Chetcuti and Sarah Agius

Vaissière, Jacqueline. 1983. Language-independent prosodic features. In Anne Cutler & D. Robert Ladd (eds.), Prosody: Models and measurements, 53–66. Hamburg: Springer. Vella, Alexandra. 2003. Phrase accents in Maltese: Distribution and realisation. Proceedings of the 15th international congress of phonetic sciences, Barcelona, Spain. 1775–1778. Vella, Alexandra. 2009. Maltese intonation and focus structure. In Ray Fabri (ed.), Maltese linguistics: A snapshot. In memory of Joseph A. Cremona. (Il-Lingwa Tagħna Vol. 1), 63–92. Bochum: Brockmeyer. Vella, Alexandra & Paulseph-John Farrugia. 2006. MalToBI – building an annotated corpus of spoken Maltese. In Rüdiger Hoffmann & Hansjörg Mixdorff (eds.), Proceedings of Speech Prosody 2006, Dresden, Germany. Dresden: TUD Press. Vella, Alexandra, Flavia Chetcuti, Sarah Grech & Michael Spagnol. 2010. Integrating annotated spoken Maltese data into corpora of written Maltese. Proceedings of the 7th international conference on language resources and evaluation, Workshop on language resources and human language technologies for Semitic languages, Valletta, Malta. 83–90. Vella, Alexandra & Patrizia Paggio. 2013. Overlaps in Maltese: A comparison between Map Task dialogues and multimodal conversational data. NEALT Proceedings. Northern European Association for Language and Technology, 4th Inordic symposium on multimodal communication, November 15–16, Gothenburg, Sweden. 21–29. Vella, Alexandra, Michael Spagnol, Sarah Grech & Flavia Chetcuti. 2011. A preliminary investigation of filled pauses in Maltese. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 255–276. Berlin: Akademie Verlag. Vella, Alexandra, Michael Spagnol, Sarah Grech & Flavia Chetcuti. 2014. The role of silence in spoken Maltese: Durational and distributional characteristics of breaks and pauses. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 89–113. Berlin: Akademie Verlag.

| Part II: Morphology & Syntax

Maris Camilleri

How inflectional morphology meets subcategorization frame distinctions in Maltese Abstract: In this paper we provide a detailed description of the inflectional verbal paradigm in Maltese. As we do so, it will be illustrated how there is another dimension to the Maltese inflectional verbal paradigm, apart from the concatenative/affixal dimension. We will observe how the paradigm also displays a non-concatenative dimension. It is on this dimension that we will be concentrating upon, illustrating how a number of correlations are present in the paradigm, particularly relations between the stem vowel and the suffixes in the Imperfective sub-paradigms, and where in certain contexts and under specific constraints, the choice of either of the dependencies employed come to reflect and correlate with the subcategorization frame associated with the verb. We will see how it is precisely due to the observation of these morphological behaviors internal to the paradigm that eventually led us to the identification of equipollent verbs in Maltese. Keywords: non-concatenative morphology, inflectional paradigms, equipollent verbs

1 Introduction In this paper I aim to provide an account of an intricate relation that exists between inflectional morphology and the realization of the verb’s valency. I do this by providing a new synchronic multidimensional model of inflectional verbal morphology in Maltese, which is surface-driven, stem-based and paradigm-oriented, where the locus of the analysis lies within what can be observed from the paradigm. This model will eventually also enable us to consider the binyanim system of related verbal forms from a new angle. Through this study it will be shown that there is another dimension to Maltese inflectional verbal morphology, whereby contra Spagnol (2011: 37), who

|| Maris Camilleri: University of Vienna, Department of Linguistics, Sensengasse, 3a, Vienna, Austria. E-mail: [email protected]

116 | Maris Camilleri

claims that “[v]erb inflection is concatenative, with prefixation and suffixation to a stem-base”, we here illustrate with evidence that suggests that inflectional morphology in Maltese cannot be solely affixal in nature. Rather, Maltese verbal paradigms across the board involve stem-alternations which, as we will see, translate into morphosyntactic- and morphosemantic-feature realization, which in turn yield significant competing multiple exponence interactions with both the affixal component as well as the templatic/stem-shape-conditioning (Camilleri 2014a). For the purpose of the point which we want to consider in this paper, it is not any random stem-alternation that we will be interested in, although the different patterns yielded, which can be equated to morphological features (Corbett and Baerman 2006: 232) are interesting in their own right, we are here interested in exploiting a set of dependencies previously unaccounted for, and the choice of dependencies employed internal to the paradigm.1 These dependencies appear in a constrained set of contexts, and the choice of the dependency employed appears to correlate with subcategorization frame alternations. The observation of such a paradigmatic dependency as well as the consequential correlation with a-structure, meant an a priori understanding of the paradigm as a morphological entity in its own right, such that it is a dynamic system of intricate networks/relations and dependencies which contribute to increased interpredictability that in turn facilitates learning and considerably minimizes the idiosyncratic information necessary within the lexicon (Ackerman et al. 2009, Ackerman and Malouf 2013). Such networks and dependencies are in a sense unnecessary, i.e. arbitrary, given that the paradigmatic structure in itself still fulfils its function in realizing morphosyntactic and morphosemantic features independent of the presence/absence of such networks. Nevertheless, this aspect of paradigmatic structure is systematic, and thus, unraveling this internal paradigmatic system and what it tells us in terms of a possible function, constitutes understanding better the paradigm as a morphological entity. Camilleri (2014b) identifies a number of such paradigmatic dependencies internal to the Maltese verbal paradigms. For what interests us in this paper, we shall be focusing upon those that occur internal to the paradigm of vowel-final (V-final) verbs. It is not any V-final verb that we will be concentrating on, however. It is only non-għ-final verbs that are able to exploit the correlation of the dependency relations with valence alternations. On the basis of our surfacebased model, whereby a verb like qata’ ‘cut’ and qara ‘read’ essentially share a

|| 1 Refer to Camilleri (2014b) for more detail, particularly for the discussions on the presence of forty-four patterns of stem-alternations across the different Maltese verbal paradigms, which are then classified more broadly in eighteen classes.

How inflectional morphology meets subcategorization frame distinctions | 117

stem-shape, i.e. CVCV and a vocalic pattern, i.e. the set of two stem-vowels in the Perfective 3SG.M citation form, which is a-a across both these verbs, the only way in which these verbs can be disambiguated is by idiosyncratically identifying these verbs by something that is yet again salient at the surface-form of these verbs. Essentially this identifier is the 3SG.F-realizing exponent, such that /ʔat-et/ (qatgħet ‘she cut’) takes an -et suffix and is [-et selecting], whereas /ʔra:t/ (qrat ‘she read’) takes a -t suffix and is hence [-t selecting]. The requirement to specifically identify these two sets of V-final verbs, i.e. diachronically għ-final vs. non-għ-final verbs, via the use of such diacritics, itself follows from the idiosyncratic complexities yielded by the very għ-final loss throughout the course of diachrony. The ability to differentiate these pairs of verbs at the 3SG.F exponent level and the inability of the għ-final set of verbs to parallel with the non-għfinal counterparts will be made clearer throughout this paper. The paper is divided as follows: In the following section I provide an account of the affixal dimension to Maltese inflectional morphology, while in Section 3 I proceed to provide a detailed account of the paradigmatic dependencies that exist, while Section 4 and Section 5 discuss the constraints which condition the choice of one paradigmatic dependency over another in the Semitic Maltese (SM) verbal paradigms and what this implies. In Section 6 the intricate behaviors across the non-SM lexicon are focused upon. Section 7 then concludes this paper.

2 The affixal inflectional inventory In what follows I provide the key analytical background necessary to better understand a number of assumptions being upheld in this paper. Apart from an analysis whereby I assume, following Baerman and Corbett (2012) that stems realize morphosyntactic and morphosemantic features, which may not form a natural class and be morphemic in nature, what is rather important is how I arrive to the stem-form and how the dependency relations are formed, particularly given the fact that the dependency we shall concentrate upon entails one between a stem-vowel and a suffixal exponent. If we are to define the stem as that morphological form which remains from a word-form after all affixes have been removed, and of which a lexical item may have more than one (Aronoff 1992: 14, Stump 2001: 33) (also refer to Anderson’s 1992 lexical stem set hypothesis), then to get to the stem, we have to know what constitutes the affixal inventory in the language. There have been three main affixal inventories postulated in Maltese, prior to the one being assumed here. These analyses may be split into what one can refer to or interpret as positing a multiple inflectional class (Mifsud 1995, Spagnol 2011),

118 | Maris Camilleri

vs. a single inflectional class (Fabri 2009). The segmentation here parallels closer to that of Fabri (2009), albeit with a distinct end analysis. The single inflectional affix inventory being posited here cuts across all verbs in Maltese, irrespective of their origin. What allomorphy exists is all phonologically-conditioned, based upon the stem as well as the quality of the stem-vowel, in the case of the 3SGF exponent, when this takes a VC shape. This unification at the affixal level forms part of the larger output of the multidimensional model employed, such that no distinction at the inflectional morphology level is motivated across the SM and non-SM verbs.2 To arrive at the inflectional inventory, Spencer’s (2012: 98) Stem Maximisation Principle towards segmentation is employed, where stems are maximized over inflection, yielding an increase in the number of stems but essentially a ‘simple[r] statement of exponence’ (Spencer 2012: 89). The resulting segmentation has been also enhanced by a number of discussions on the segmentation analysis present in the literature on Arabic vernaculars. Table 1 below represents the affixal inventory. Tab. 1: The inflectional affixal inventory employed in this study.

Morphosyntactic values

Perfective

Imperfective

1SG

-t

n- ~ m- ~ r-

2SG

-t

/t- ~ s- ~ ʃ- ~ ʤ- ~ ʣ- ~ ʦ- ~ ʧ- ~ z-/

3SG.M

ø

j- ~ i-

3SG.F

-(V)t

/t- ~ s- ~ ʃ- ~ ʤ- ~ ʣ- ~ ʦ- ~ ʧ- ~ z-/

1PL

-na3

n- ~ m-

2PL

-t-u

/t- ~ s- ~ ʃ- ~ ʤ- ~ ʣ- ~ ʦ- ~ ʧ- ~ z-/ -u ~ -w

3PL

-u ~ -w

j- ~ i-

-u ~ -w

-u ~-w

|| 2 Having said that only one inflectional class is being postulated across the Maltese verbal paradigms, one should mention that this is really what we can refer to as the canonical inflectional affixal inventory for Maltese. The reason for this is due to the availability of ACCand DAT-realized SUBJ exponents across impersonal and pseudo-verbs (Camilleri 2014b). 3 An anonymous reviewer asks why I am providing the list of allomorphs available but did not do so for the Perfective 1PL-realized suffix, which according to the reviewer should be: -na ~ -nie. I do not do so because the form -nie is only present as a morphophonological effect with the addition of the NEG exponent -x or the suffixation of ACC and DAT pronominal forms. I am here only concentrating on changes in the stem and/or affixal exponents in verbal paradigms that do not involve any additional affixed pronouns and/or the NEG-realizing exponent.

How inflectional morphology meets subcategorization frame distinctions | 119

A number of interesting behaviors obtain once we provide such an inventory, e.g. the identification of the presence of a non-autonomous (Zaliznjak 1973, Baerman et al. 2005: 42–44) 2SG set of values, given that the 2SG-realizing exponent is syncretic with exponent in the 1^2 SG PERS cells in the Perfect subparadigm and with the 3SG.F exponent in the Imperfect sub-paradigm when this is a -t (Camilleri 2014a, b). Additionally, I consider the initial vowel internal to the Imperfect sub-paradigmatic word-forms following the prefix, at least when this vowel is present, to not be part of the morphological stem. Rather, following Puech (2011) I consider this vowel as a formative vowel that is present for phonological reasons, and is neither part of the stem nor the prefix, and functions as a phonological extension of the morphological stem.

3 Dependencies internal to and across the paradigms of V-final verbs Given that the correlation between the employment of a paradigmatic dependency and the nature of the verb’s subcategorization frame only targets V-final verbs, we will here only be discussing the array of dependencies internal to the paradigms of V-final verbs. Before delving straight into these, it should be mentioned rather clearly from the outset that claims made here are only relevant for Standard Maltese, and evidence from non-Standard dialects, including Naxxari and Gozitan dialects, for example, is indicative of the application of independent systems and mechanisms. Additionally, the networks and dependencies to be discussed here are solely those that figure in V-final paradigms when these do not involve the attachment of additional non-SUBJ pronominal forms, and/or the NEG-realizing suffix -x. If we focus on non-għ-final V-final verbs first, we observe two dependencies internal to each sub-paradigm. The nature of the dependencies across the subparadigms is not of the same sort, however. Internal to the Perfective subparadigm, the dependency involves an implicative relation between the stemvowels across the Perfective 1^2 stem, the Perfective 3SG.F stem and the Perfective 3PL stem. Internal to the Imperfective sub-paradigm, on the other hand, the dependency involves an implicational relation between the stem-vowel in the stem across the SG cells and the suffix employed across the PL cells. Before looking at the details, through this generalized overview of the dependencies available, we can observe a broader overarching behavior, such that these dependencies involve a major ASP-based split: Perfective – Imperfective as well as a NUM-based

120 | Maris Camilleri

split internal to the Imperfective sub-paradigm. The two implicational dependencies internal to each sub-paradigm are presented in (1). Perfective:

a. b. Imperfective: a. b.

(1)

aj[1^2] ej[1^2] a[SG] i[SG]

→ → → →

ā[3SG.F] ie[3SG.F] -w[PL] -u[PL]

→ →

a[3PL] e[3PL]

The two distinct dependencies available internal to the Perfective and Imperfective sub-paradigms of non-għ-final V-final verbs

The choice of one or the other of the dependencies is constrained, as we will see further below, and interdependencies between the two sub-paradigmatic dependencies, although constrained, also exist. The reason for not including the diachronically għ-final set of verbs is because if we consider the surface paradigm of qata’ ‘cut, guess’ below in Table 2, we observe that while the a[SG] -w[PL] dependency in the Imperfective sub-paradigm is employed, none of the Perfective sub-paradigmatic dependencies are employed, given the [-et selecting] identification associated with these verbs. Tab. 2: The paradigm of qata’ ‘cut, guess’, representing diachronically għ-final verbs.

Morphosyntactic values

qata’ ‘cut, guess’ [-et selecting]

Perfective 1^2

qtaj-t

3SG.M

qata’

3SG.F

qatgħ-et /ʔat-et/

3PL

qatgħ-u /ʔata-w/

Imperfective SG

j-a-qta’

PL

j-a-qtgħ-u /j-a-ʔta-w/

Having now understood why diachronically għ-final verbs do not quite fit the entire picture being established here, we now move on to consider in more detail the paradigmatic networks themselves. The two sets of dependencies across the two sub-paradigms are able to multiply out and yield the four logical combinatory patterns, as in (2) below. What is important to mention is that these patterns are not themselves equally distributed across the paradigms of non-għfinal V-final Maltese verbs, and the representation of these patterns below are in descending order of frequency, with illustrative SM and non-SM verb examples that exhibit the combinatorial pattern of dependencies within its paradigm.

How inflectional morphology meets subcategorization frame distinctions | 121

a. b. c. d.

(2)

Perfective aj[1^2] -ā[3SG.M] - a[3PL] Imperfective a[SG] -w[PL] (e.g. qara ‘read’, avża ‘warn’, pparkja ‘park’) Perfective ej[1^2] -ie[3SG.M] - e[3PL] Imperfective i[SG] -u[PL] (e.g. mexa ‘walk’, spella ‘spell’, rrikonoxxa ‘acknowledge’) Perfective ej[1^2] -ie[3SG.M] - e[3PL] Imperfective a[SG] -w[PL] (e.g. nbeda ‘be started’, tbaxxa ‘be lowered, lower’ tgawda ‘be enjoyed’) Perfective aj[1^2] -ā[3SG.M] - a[3PL] Imperfective i[SG] -u[PL] (e.g. ta ‘give’, xtara ‘buy’, (staħa) ‘be shy’, (sparixxa) ‘disappear’ (the latter two verbs only optionally employ this cross-paradigm dependency)) The four logical combinatorial patterns of cross-paradigm dependencies in descending order of use

The combination of paradigmatic dependencies in (2a) may be analyzed as some sort of default, especially when one considers that the overwhelming majority of non-SM loans, including newer integrations in the system, do in fact exhibit this pattern. Table 3 provides the paradigms of qara ‘read’, eduka ‘educate’, and pparkja ‘park’ to illustrate this pattern. V-final -are-derived Romance loans and English loans, excluding spella ‘spell’, all take this pattern of stemalternations.4 Tab. 3: Representing the: Perfective aj[1^2] - ā[3SG.M] - a[3PL] Imperfective a[SG] -w[PL] combinatorial pattern of dependencies in the paradigms of qara ‘read’, eduka ‘educate’, and pparkja ‘park’.

Morphosyntactic values

Perfective

Imperfective

1^2

3SG.M

3SG.F

3PL

3SG

3PL

qraj-t

qara

qrā-t

qra-w

j-a-qra

j-a-qra-w

eduka ‘educate’

edukaj-t

eduka

edukā-t

eduka-w

j-eduka

j-eduka-w

pparkja ‘park’

pparkjaj-t pparkja

pparkjā-t

pparkja-w

j-i-pparkja

j-i-pparkja-w5

qara ‘read’

|| 4 When it comes to the stem-alternation patterns associated with non-Semitic verbs which employ this combination of paradigmatic dependencies, the pattern displayed is that which we also find in the paradigm of non-Ist binyan għ-final verbs. This observation made in Camilleri (2014b) could only be possible through the surface and stem-based approach that has been adopted. While the details fall outside of the scope of this research study, nevertheless, it is important to keep in mind that għ-final and non-għ-final verbs should not necessarily be conceived of as being quite distinct, when considering other paradigmatic behaviors. 5 Note that the dashes (-) around the first vowel in these verb forms is simply to show that under my account, following Puech (2011), in particular, these vowels are neither part of the

122 | Maris Camilleri

The pattern in (2b) is manifest, amongst others, in the paradigms of SM verbs such as mexa ‘walk, all non-prefixed V-final quadriliteral verbs, all IInd binyan V-final verbs, all IIIrd binyan verbs with an ie-e vocalic pattern, and Romance verbs patterning on such forms, irrespective of their thematic origin, as well as other V-final primary loans of an -ere/-ire origin, and English spella ‘spell’. The third available pattern consists of a far more restricted set and essentially involves a lexical set of SM verbs including nesa ‘forget’ and a constrained set of SM verbs, and all secondary generation loans of -ere/-ire origins, therefore including verbs such as tbata ‘be suffered’, and tgawda ‘be enjoyed’. The remaining fourth combinatory pattern is only obligatorily exhibited by the SM paradigms of xtara ‘buy’ and ta ‘give’, and optionally by the paradigm of staħa ‘be shy’, whose paradigm involves either: Perfective ej[1^2] - ie[3SG.M] - e[3PL] Imperfective i[SG] -u[PL] and Perfective aj[1^2] - ā[3SGM] - a[3PL] Imperfective i[SG] -u[PL], and sparixxa ‘disappear’, whose paradigm involves either: Perfective ej[1^2] - ie[3SGM] - e[3PL] Imperfective i[SG] -u[PL] and Perfective ej[1^2] - ie[3SG.M] - e[3PL] Imperfective a[SG] -w[PL]. The paradigm of such verbs is represented in Table 4. Tab. 4: Representing the: Perfective aj[1^2] – ā[3SG.M] – a[3PL] Imperfective i[SG] –u[PL] combinatorial pattern of dependencies in the paradigms of xtara ‘buy’, ta ‘give’, staħa ‘be shy’ and sparixxa ‘disappear’.

Morphosyntactic values

Perfective

Imperfective

1^2

3SG.M

3SG.F

3PL

3SG

3PL

xtara ‘buy’

xtraj-t

xtara

xtrā-t

xtra-w

j-i-xtri

j-i-xtr-u

ta ‘give’

taj-t

ta

tā-t

ta-w

j-agħti

j-agħt-u

staħa ‘park’

stħaj-t

staħa

stħā-t

stħa-w

j-i-stħi

j-i-stħ-u

sparixxa ‘disappear’

sparixxaj-t

sparixxa

sparixxā-t sparrixa-w

j-i-sparixxi

j-i-sparixx-u

At this point in our discussion, we can compare the scheme of combinatorial dependencies in (2) with Tables 5 and 6 below, which represents Hoberman and Aronoff’s (2003: 78) take on the combinations available across the SM and non-SM data set. They solely concentrate on the Perfective 1^2 stem-vowel and the Imper-

|| stem nor part of the prefix. Their function as epenthetic vowels (as the i in jipparkja), or as formative vowels (as the a in jaqra) is constrained by various factors which we do not need to delve into here.

How inflectional morphology meets subcategorization frame distinctions | 123

fect SG stem-vowel, however, without illustrating the dependencies available. Through the representation in their tables, we can illustrate what we can rephrase as the impossibility of SM verbs to exhibit the combinatorial pattern in (2d), i.e. one where the stem in the Perfective 1^2 stem involves an aj diphthong and the stem-vowel in the Imperfective sub-paradigm is an i, and the impossibility of the combinations in (2a) and (2d) across the non-SM set of verbs. As exhibited in (2), however, all verbs, be they SM or non-SM, exhibit any of dependencies available, even if in a rather limited fashion. Their use of (+) in Table 5 refers to the fact that it is a small class of verbs that allow for the combination of dependencies (2a) across the SM set of verbs. They motivate the alleged restrictions upon the combinatorial patterns by saying that a non-SM verb “… has either front-vowel endings in the Perfect[ive] and Imperfect[ive], or back-vowel endings in the Perfect[ive] and Imperfect[ive]” (Hoberman and Aronoff 2003: 72). While this behavior does in fact account for the overwhelming majority of non-SM verbs, this does not do justice to the full array of non-SM verbs, as we will see in more detail in Section 6. Tab. 5: The dependencies claimed to exist across Semitic V-final verbs according to Hoberman and Aronoff (2003: 78).

Imperfective -i

-a

-ejt

+

+

-ajt

–

(+)

Perfective

Tab. 6: The dependencies claimed to exist across non-Semitic V-final verbs according to Hoberman and Aronoff (2003: 78).

Imperfective -i

-a

-ejt

+

–

-ajt

–

+

Perfective

The above description of which verbs exhibit the patterns available was meant as a taxonomy. This taxonomy is however not random, and can be arrived at via a combination of properties, which together trigger the choice of paradigmatic dependencies employed. While the main focus here is on the Imperfective sub-

124 | Maris Camilleri

paradigmatic dependencies, given that such dependencies are being discussed for the first time following Camilleri (2014b), it is an apt opportunity to consider the constraints that govern such selection choices in more detail.

4 The constraints governing the taxonomy across SM verbs If we first consider the Perfective sub-paradigmatic dependency, which is only partially correlated with valence alternations and differences, we observe that the split in the use of the two different dependencies across non-għ triliteral Ist binyan verbs, is cross-classified with the vocalic pattern of the Perfective 3SG.M form. E-a vowel patterned verbs such as nesa ‘forget’, mexa ‘walk’ and wera ‘show’ realize the ej-ie-e dependency in the Perfective sub-paradigm, while a-a counterparts such as qara ‘read’ and dara ‘get used to’ display the aj-ā-a dependency.6 All quadriliteral V-final verbs, on the other hand, irrespective of their vocalic pattern, take an ej-ie-e dependency, as is the case across the paradigms of fisqa ‘nourish a child’, saqsa ‘ask’ and staqsa ‘say’.7 While such a vocalic pattern – dependency choice across the triliteral subset is maintained across the different binyanim as well, the only exception to this is what takes place across the IInd and Vth binyanim, where irrespective of the vocalic pattern of the verb, the ej-ie-e paradigmatic dependency is employed

|| 6 It should be mentioned that here għ- and h-initial diachronially CVCV Ist binyan verbs such as għama ‘blind’, għala ‘boil’, għola ‘rise’ and heda ‘pass time’, should not be considered, when discussing the correlation involved between the vocalic pattern and the Perfective subparadigmatic dependency employed. This is because as discussed in Camilleri (2014b), on the basis of the synchronic and surface analysis that has been formulated, such verbs take a completely different stem-shape, i.e. V:CV, and consequently are not meant to necessarily display the same set of paradigmatic behaviors. In fact, both għama ‘blind’ and għala ‘boil’, although diachronically taking an a-a vocalic pattern, which is in fact synchronically ā-a, exhibit an ej-ie-e Perfective sub-paradigmatic dependency, instead of the aj-ā-a dependency which a-a non-għ-final CVCV verbs employ. 7 Note that given the surface and stem-based analysis in Camilleri (2014b), instead of referring to triliteral vs. quadriliteral distinctions, reference to the stem-shape was made, such that SM a-a CVCV verbs were immediately associated with an aj-ā-a dependency while e-a CVCV verbs were immediately associated with an ej-ie-e dependency. On the other hand, SM C()VCCV verbs were immediately identified with an ej-ie-e dependency. Refer to Camilleri (2014b) for more details on how a full integration of both SM and non-SM CVCiCjV verbs is captured within the lexicon.

How inflectional morphology meets subcategorization frame distinctions | 125

in the Perfective sub-paradigm. Table 7 below summarizes the interesting dependency contrasts/non-contrasts across the Perfective sub-paradigms of related verbs. Particularly observe the contrast employed across the paradigms of Ist binyan dara ‘get used to’, which takes an aj-ā-a Perfective sub-paradigmatic dependency, vs. IInd and Vth binyanim darra ‘cause to get used to’ and ddarra ‘be made to get used to’, respectively, where irrespective of the a-a vocalic pattern, an ej-ie-e dependency is employed. Tab. 7: Representing the Perfective sub-paradigmatic dependencies across related verbs, whilst bringing out the interaction between the vocalic pattern conditioning and the binyan conditioning.

Morphosyntactic values

Perfective 1^2

3SG.M

3SG.F

3PL

wera ‘show’ – I binyan

wrej-t

nt-wera ‘be shown’ – VIIth binyan

nt-wrej-t

wera

wrie-t

wre-w

nt-wera

nt-wrie-t

nt-wre-w

mela ‘fill’ – Ist binyan

mlej-t

mela

mlie-t

mle-w

mlej-t

mela

mlie-t

mle-w

qara ‘read’- Ist binyan

qraj-t

qara

qrā-t

qra-w

n-qara ‘be read’ - VIIth binyan

n-qraj-t

n-qara

n-qrā-t

n-qra-w

mexa ‘walk’ – Ist binyan

mxej-t

mexa

mxie-t

mxe-w

mexxa ‘cause to walk, lead’ – IInd binyan

mexxej-t

mexxa

mexxie-t

mexxe-w

t-mexxej-t

t-mexxa

t-mexxie-t

t-mexxe-w

st

th

mela ‘become full’ – VIII binyan

th

t-mexxa ‘be walked, led’ – V binyan dara ‘get used to’ – Ist binyan

draj-t

dara

drā-t

dra-w

darra ‘cause to get used to’ – IInd binyan

darrej-t

darra

darrie-t

darre-w

d-darrej-t

d-darra

d-darrie-t

d-darre-w

th

ddara ‘be made to get used to’ – V binyan

Table 7 was meant to also illustrate the reason for stating that the Perfective sub-paradigm is only partially correlated with valency alternations. This is because the IInd and Vth binyanim do not form a natural class of verbs, when it comes to the valency associated with these verbs. There is additionally a clear difference in the subcategorization frame employed by the related forms dara ‘get used to’ – darra ‘make get used to’ – ddarra ‘be made get used to’: dara

126 | Maris Camilleri

, darra and ddarra . Note that even if dara and ddarra are both optionally monovalent, when they only display a SUBJ argument, the thematic role associated with the SUBJ grammatical function across the two verbs at the argument-structure level is not the same. That in the Ist binyan is an Agent, while that in the Vth binyan verb is a Patient. It should be made clear here that the Xth binyan verbs such as staħba ‘hide’, staħa ‘be shy’ and stagħna ‘become prosperous’, although seemingly taking an a-a pattern, are not the same kettle of fish and the a-a pattern they display does not constitute the vocalic pattern in parallel to the a-a of qara ‘read’. Rather, the actual morphological stem is in fact -ħba and so on (Darden 1992). Consequently, there is thus no expected correlation between the stem-vowel and the dependency employed. In fact, evidence that there is no such correlation is illustrated through the Perfective sub-paradigms of staħba ‘hide’, stagħna ‘become prosperous’, staħa ‘be shy’ and stenna ‘wait’, where the former three, alhough displaying an a-a set of vowels in the Perfective 3SG.M form, may optionally exhibit both of the dependencies available in the Perfective sub-paradigm, as in Table 8. Tab. 8: Representing the distinct behavior employed across Xth binyan verbs.

Morphosyntactic values

staħba ‘hide’

Perfective

1^2

3SG.M

3SG.F

3PL

staħbaj-t ~ staħbej-t

staħba

staħbā-t ~ staħbie-t staħba-w ~ staħbe-w

stagħna ‘become stagħnaj-t ~ stagħnej-t stagħna stagħnā-t ~ stagħnie-t stagħna-w ~ stagħne-w prosperous’ staħa ‘be shy’

stħaj-t ~ stħej-t

staħa

stħā-t

stħa-w ~ stħe-w

stenna ‘wait’

stennej-t

stenna

stennie-t

stenne-w

When we then consider the Imperfective sub-paradigmatic dependencies, we observe a more direct correlation between the morphological paradigmatic dependency and the verbs’ valency. In both the Ist and Xth binyanim, however, there is no such correlation employed. If we for example focus upon Ist binyan verbs, since these are the ones that do actually take a proper vocalic pattern, all such a-a non-għ-final V-final verbs are immediately correlated with an a -w dependency in the Imperfective sub-paradigm. On the other hand, e-a verbs displaying an ej-ie-e dependency in the Perfective sub-paradigm are then idiosyncratically split between a -w vs. i -u dependency taking verbs. Beda ‘start’

How inflectional morphology meets subcategorization frame distinctions | 127

and nesa ‘forget;, for example, display the former dependency, while verbs such as mexa ‘walk’ and wera ‘show’ display the latter dependency. Across the nonIst/Xth binyanim, on the other hand, we observe a IInd/IIIrd vs. Vth/VIth/VIIth/ VIIIth binyanim split. IInd and IIIrd binyanim verbs display an i -u Perfective subparadigmatic dependency in their paradigms, while verbs of binyanim V till VIII display an a -w dependency. The same follows for prefixed quadriliteral verbs. Illustrative paradigms are provided in Table 9 below. There are three major exceptions to this otherwise systematic behavior. These are the IInd binyan għejja ‘tire’, VIth binyan tnieda ‘be launched’, and VIIIth binyan xtara ‘buy’. We will be considering these exceptions to the rule in more detail below. Tab. 9: Illustrating the binyan-conditioned choice of Imperfect sub-paradigmatic dependency.

Morphosyntactic values

Imperfective 3SG.M

3PL

j-mexxi

j-mexx-u

nieda ‘lauch’ – III binyan

j-niedi

j-nied-u

t-mexxa ‘be walked, led’ – Vth binyan

j-i-t-mexxa

j-i-t-mexxe-w

t-tama ‘hope’ – VIth binyan

j-i-t-tama

j-i-t-tama-w

n-bena ‘be built’ – VII binyan

j-i-n-bena

j-i-bne-w

nt-rema ‘be thrown away’ – VIIth binyan

j-i-nt-rema

j-i-nt-reme-w

nd

mexxa ‘cause to walk, lead’ – II binyan rd

th

th

n-xara ‘be brought’ – VII binyan

j-i-n-xara

j-i-n-xra-w

mela ‘be filled’ – VIIIth binyan

j-i-mela

j-i-mle-w

s-saqsa ‘be asked’ – Prefixed quadriliteral

j-i-s-saqsa

j-i-s-saqse-w

What follows below in (3) is a visual representation of the above described constraints, including the vocalic pattern and binyan conditionings, and the relations they impose with respect to the paradigmatic dependencies employed. These account for the taxonomy of the paradigmatic dependencies employed across the Maltese inflectional verbal paradigms just presented in the previous section.

128 | Maris Camilleri

Choice of paradigmatic dependency in the Perfective sub-paradigm

(a) Ist

Vocalic pattern

Choice of paradigmatic dependency in the Imperfective sub-paradigm Choice of paradigmatic dependency in the Perfective sub-paradigm Vocalic pattern

Choice of paradigmatic dependency in the Imperfective sub-paradigm

ej

aj

e-a

a-a

i –u

a- w

(c) Vth

Choice of paradigmatic dependency in the Imperfective sub-paradigm (3)

a-a

i -u

(d) VIth/VIIth ej

a-a

e-a

a -w

aj

ie/e-a a-a

a -w

Choice of paradigmatic (e) VIIIth ej dependency in the Perfective sub-paradigm Vocalic pattern

ej

e-a

a -w

ej

e-a

(b) IInd

a -w

aj[xtara]

a-a

i -u

(f) IIIrd

ej

aj

ie-a

a-a

i -u

The paradigmatic relations that obtain in a propos the vocalic pattern across non-għ-final V-final verbs in the SM8

|| 8 Note that Xth binyan verbs are here not represented as it was shown that these do not involve a vocalic pattern in the same way as other V-final verbs across the binyanim do. IXth binyan verbs are not considered as none of these involve V-final verbs.

How inflectional morphology meets subcategorization frame distinctions | 129

5 What do we make out of these conditionings? Having provided a description of what is going on vis-à-vis V-final SM verbs in the language, and the dependencies employed internal to the paradigm, and the changes taking place, one should mention that while we will here motivate a correlation with valency, it should nevertheless be highlighted that the alternation in the dependencies across the paradigms of related verb-forms, when this takes place, are arbitrary, in the sense that there is nothing phonological or morphophonological that allows for: drajt ‘I got used to’, but not *darrajt or *ddarrajt in the Standard variety, for example. Additionally, nothing prohibits the form *nimxa or *nimxew internal to the paradigm of mexa ‘walk’. The ablautchanges as well as the correlated suffixal exponent changes in the Imperfect sub-paradigm are uncalled for. Incidentally, all such dependencies simply increase the number of alternating stems in the paradigm, yielding increasingly non-canonical behaviors, on a view that internal to the canonical paradigm, stems are to be invariant (Corbett 2007, Baerman and Corbett 2012). In this section we thus aim to assess what constitutes the underpinnings of this morphologically complex behavior, by first closely considering what we get out of this morphologically-triggered behavior (Section 5.1), and what parallels (Section 5.2) and idiosyncrasies (Section 5.3) we observe in the broader system.

5.1 The constraints and what they tell us From the data and the distribution of the dependencies presented so far, one observes that across the IInd–VIIIth binyanim, there evidently exists a correlation between the choice of paradigmatic dependency and the subcategorization frame. All verbs across the Vth–VIIIth binyanim, as exemplified in the previous section, illustrate an a -w dependency and are verbs that express reflexive, anticausative, passive and reciprocal semantic interpretations (Borg 1988, Spagnol 2011), which we can here collapse as interpretations that associate with intransitive valence structures, where by intransitive, I specifically mean non-OBJ or non-DAT OBJ recipient/goal argument-taking verbs, with the latter including verbs such as għajjat ‘shout to someone’, ċempel ‘phone someone’, sellem ‘greet’, tbissem ‘smile’, (dialectal) sieħ ‘call out to someone’, (Standard) sejjaħ ‘call out to someone’ given that it is possible to have intransitive forms taking DAT morphology to realize non-recipient/goal thematic-roles, including affected and benefactive DAT-realized thematic roles (refer to Camilleri and Sadler 2012 for more detail). This dependency is thus present across the paradigm of Vth–

130 | Maris Camilleri

VIIIth binyanim verbs, with the exception of a special use of VIth binyan tnieda ‘be launched’ and VIIth binyan xtara ‘buy’. On the contrary, all IInd and IIIrd binyan verbs, at least the ones for which there is no equivalent in other binyanim, are transitive, and illustrate an i -u dependency.9 By specifically focusing on non-għ-final V-final verbs, however, we see that the i -u dependency simply enhances the prototypical transitive a-structure realization expressed by the IInd binyan. In addition to this, one may here want to add that with respect to other data, which we will come to in the following section, Saydon (1958) has already identified the presence of the i stem-vowel across the Imperfect SG cells as a key identifier of transitive subcategorization frames in particular, and is reminiscent of the use of such a stem-vowel across the diachronic IVth binyan, which was an inherently causative binyan. In fact, through such an account of the inflectional morphology, other diachronically IVth binyan verbs were identified, apart from wera ‘show’, and these were ta ‘give’ and tema’ ‘feed’, where while synchronically also taking an inherent causative predication, the non-għ-final V-final ones synchronically also preserve and fossilize the morphology associated with this binyan (Camilleri 2014b, Camilleri et al. 2014). To conclude, we can therefore here take the correlation further to say that this dependency at the inflectional level interplays with the binyanim alternation such that it results in multiple exponent relations, where the inflectional morphology appears to be also participating in signaling changes in a-structure. I am specifically shying away from referring to a possible inflectional – derivational correlation, particularly because it remains debatable whether what we have is truly derivation across the binyanim, given the arguments in Camilleri (2012). But then again refer to Borg (1988), Borg and Mifsud (1999) and Spagnol (2011).

5.2 Additional SM data Our claim that inflectional morphology appears to be realizing information related to the subcategorization frame in some sort of multiple exponence relation with the binyanim-based templatic distinctions, can be further enhanced by

|| 9 What I mean with this statement here is that verbs such as IInd binyan ħammar ‘cause to become red’, may function as a labile predicate that may be used inchoatively with the meaning of ‘become red’, which is the meaning expressed by the IXth binyan counterpart ħmar, or for example the use of saddad ‘rust s.th’ with the meaning of ‘become rusty’, which may be otherwise realized by the Vth binyan form ssaddad. Refer to Spagnol (2011) for more detail on the matter.

How inflectional morphology meets subcategorization frame distinctions | 131

the identification of two equipollent verbs in the language (Camilleri 2014a,b). These include the Ist binyan verb għama ‘blind’ and IInd binyan għejja ‘tire’. According to Spagnol (2011: 113) għama is a labile verb, i.e. a verb that without changing its lexical form is both the inchoative ‘be blind’ and the causative ‘blind someone’. Here, on the contrary, I want to illustrate that għama is an equipollent verb, since it has been shown in Camilleri (2014a,b) to be associated with two distinct paradigms, hence pointing at an equipollent behavior (Haspelmath 1993, Comrie 2006), and that the two subcategorization frames associated with għama are clearly disambiguated from within the paradigm through the very Imperfect sub-paradigmatic dependency employed. The paradigm associated with the intransitive/inchoative għama ‘be blind’ takes an a -w dependency, while transitive għama ‘blind someone’ takes an i -u dependency, as illustrated in Table 10. Tab. 10: The distinct paradigms related with għama ‘be/become blind, blind someone’.

għama ‘be/become blind’

għama ‘blind someone’

1^2

għamej-t /a:mej-t/

għamej-t/a:mej-t/

3SG.M

għama /a:ma/

għama /a:ma/

3SG.F

għamie-t /a:mie-t/

għamie-t /a:mie-t/

3PL

għame-w /a:me-w/

għame-w /a:me-w/

SG

j-a-għma /j-a:ma/

n-a-għmi /j-a:mi/

PL

j-a-għame-w /j-a:me-w/

n-a-għm-u /j-a:m-u/

Morphosyntactic values Perfective

Imperfective

It is therefore rather clear that it is the very inflectional morphology, particularly the paradigmatic dependencies in the Imperfect sub-paradigm functioning as morphological features internal to the paradigm, that are realizing this difference at the subcategorization frame. This behavior thus further corroborates our hypothesis of interdependence between inflectional morphology and subcategorization frame differences, independent of any interaction with the binyanim distinctions. The behavior of għama is not an isolated case, IInd binyan għejja is yet another equipollent verb, which when used inchoatively means ‘be tired’, and when used transitively means ‘tire someone’. The interesting complication which għejja adds to the system is that it is the only IInd binyan verb that allows for an a -w dependency, at least in the paradigm associated with the inchoative mean-

132 | Maris Camilleri

ing, as illustrated in Table 11. The function of this dependency in the Imperfect sub-paradigm appears to substitute and fulfill the role of the Ist binyan *għeja ‘tire s.one’, which is obsolete in the author’s idiolect. The end effect, at least for speakers who still retain the Ist binyan *għeja, is the presence of verb pairs such as: Ist binyan bidel – IInd binyan biddel ‘change’, Ist binyan silef – IInd binyan sellef ‘lend’, dialectal Ist binyan sieħ – IInd binyan sejjaħ ‘call s.one’, Ist binyan *fitel – Impersonal IInd binyan fettel- ‘decide abruptly’, IInd binyan saddad – Vth binyan ssaddad ‘become rusty’, IInd binyan ħammar – IXth binyan ħmar ‘became red’, and others. Tab. 11: The emergence of IInd binyan għejja ‘be/become tired, tire someone’ as an equipollent verb.

għejja ‘be/become tired’

għejja ‘tire someone’

1^2

għejjej-t /ej:ej-t/

għejjej-t /ej:ej-t/

Morphosyntactic features Perfective

3SG.M

għejja /e:j:a/

għejja /e:j:a/

3SG.F

għejjie-t /ej:ie-t/

għejjie-t /ej:ie-t/

3PL

għejje-w /ej:e-w/

għejje-w /ej:e-w/

SG

j-għejja /j-e:j:a/

j-għejji /j-ej:i/

PL

j-għejje-w /j-ej:e-w/

j-għejj-u /j-ej:-u/

Imperfective

While both these two verbs provide us with valuable data that brings out further the correlation between the inflectional morphology and changes in the subcategorization frame, the paradigms associated with għejja also provides us with an important set of data whereby a IInd binyan V-final verb, at least in one of its associated paradigms, goes against the robust correlation between the IInd binyan and the choice of an i -u dependency in the Imperfect sub-paradigm. In the section which follows, four other SM verbs are discussed, which this time seems to mar the neat i -u – transitive correlation and the binyan-conditioned paradigmatic dependencies employed.

5.3 Idiosyncrasies There are essentially four verbs that we will be concerned with in this section, and whose paradigmatic behavior, whilst overlapping in some properties, dif-

How inflectional morphology meets subcategorization frame distinctions | 133

fers in other properties. These verbs are: salla ‘(euphemistic for) blasphemize, Muslim act of prayer’ and sella ‘send greetings’, xtara ‘buy’, and the use of an emergent paradigm that is synchronically being associated with the verb tnieda ‘be launched’. If we first start by considering the verbs salla ‘(euphemistic for) blasphemize, Muslim act of prayer’ and sella ‘send greetings’, these are essentially IInd binyan verbs, which while not synchronically associated with other forms in any other binyanim, are intransitive in meaning. Nevertheless, as illustrated through the paradigms in Table 12 below, they formally still exhibit the paradigmatic dependency choice exhibited by non-equipollent IInd binyan verbs in the Imperfective sub-paradigm, i.e. i -u. While the formal correlation between the binyan and the choice of Imperfective sub-paradigmatic dependency is maintained, i.e. whereby the IInd binyan implies the presence of an i -u dependency in the Imperfective sub-paradigm, the correlation between this dependency and the subcategorization frame that is expressed is not maintained. What this suggests, as will also be enhanced by the non-SM data presented in Section 6, is that a CVCiCiV stem-shape is strongly conditioning the choice of the i -u dependency in the Imperfective sub-paradigm. Tab. 12: Representing the: Perfective ej[1^2] - ie[3SGM] - e[3PL] Imperfective i[SG] -u[PL] combinatorial pattern of dependencies in the paradigms of salla ‘blasphemize, Muslim prayer’ and sella ‘send greetings’.

Morphosyntactic features

Perfective

Imperfective

1^2

3SG.M

3SG.F

3PL

3SG

3PL

salla ‘(euphemistic for) blasphemize, Muslim prayer’

sallej-t

salla

sallie-t

salle-w

j-salli

j-sall-u

sella ‘send greetings’

sellej-t

sella

sellie-t

selle-w

j-selli

j-sell-u

The correlation between the stem-shape (and binyan) and the choice of the dependency is on the other hand not as strong across the synchronic paradigm of VIIIth binyan xtara ‘buy’, which still employs an i -u dependency, (as illustrated in Table 4), even if all other Vth–VIIIth binyanim verbs illustrate an a -w dependency. What’s interesting, on the other hand, is that the dependency employed in the Imperfective sub-paradigm is nevertheless mapped and cross-classified with what is expected from the transitive subcategorization frame associated with this verb, at least if we exclude the above exceptions of an otherwise rather robust generalization as exemplified by salla ‘(euphemistic for) blasphemize, Muslim prayer’ and sella ‘send greetings’ of an otherwise rather robust generalization.

134 | Maris Camilleri

The remaining idiosyncratic verb displays the most intricate paradigmatic behavior. VIth binyan tnieda ‘be launched’, in relation to IIIrd binyan nieda ‘launch’, illustrate the choice of an a -w vs. i -u dependency. Once again, the alternation in the paradigmatic dependency employed relates with the intransitive – transitive alternation. Up to this point, this is the paradigmatic output we expect. Yet, in the Imperfective sub-paradigm, the additional employment of the i -u dependency is emerging, as illustrated in Table 13. Tab. 13: The paradigm nieda ‘launch’ and the two paradigms associated with tnieda ‘be launched’.

Morphosyntactic values

nieda ~ neda ‘launch’

tnieda ~ tneda ‘be launched’

1^2

nidej-t ~ nedejt

t-nidej-t ~ tnedej-t

3SG.M

nieda ~ ne:da

t-nieda ~ t-neda

3SG.F

nidie-t ~ nedie-t

t-nedie-t ~ t-nidie-t

3PL

nide-w ~ nede-w

t-nede-w ~ t-niede-w

SG

j-niedi ~ j-ne:di

j-i-t-nieda ~ j-i-t-ne:da ~ j-i-t-niedi ~ j-i-t-ne:di

PL

j-nied-u ~ j-ne:d-u j-i-t-niede-w ~ j-i-t-nede-w ~ j-i-t-nied-u ~ j-i-t-ne:d-u

Perfective

Imperfective

Evidence that this is what is happening synchronically comes from data from the MLRS Corpus. Compare (4a/b and 4c/d): (4a)

(4b)

(4c)

(4d)

Fil-ħames snin li ġejj-in in.DEF-five year.PL COMP coming.PASS.PTCP-PL j-i-t-nede-w ... 3-EPENT.VWL-PASS-launch.IMPV-PL ‘In the five years to come X will be launched’ (23 hits) Qed j-i-t-nied-u anke ... PROG 3-EPENT.VWL-PASS-launch.IMPV-PL as well ‘X are also being launched’ (38 hits) Ser j-i-t-nieda ktieb ... FUT 3-EPENT.VWL-PASS-launch.IMPV.SG.M book ‘A book will be launched’ (50 hits) Il-proċess li qed j-i-t-niedi ... DEF-process COMP PROG 3-EPENT.VWL-PASS-launch.IMPV.SG.M ‘The process that is being launched’ (5 hits)

How inflectional morphology meets subcategorization frame distinctions | 135

From the number of hits representing the uses of the two distinct forms in the Imperfective 3SG.M and 3PL cells, it is the PL cell which is showing the most frequent use of the -u suffixed verb-form. This choice of suffix is even surpassing the expected -w suffixed counterpart. Compare (4a)–(4b). The opposite scenario is the case, however, in the 3SG.M cell. One might want to hypothesize, therefore, that it is the PL stem that is motivating the activation of the i -u dependency in the Imperfective sub-paradigm. What could be yielding this overabundant behavior may be either heteroclisis, such that the distinct stems in the Imperfective sub-paradigm of nieda ‘launch’ are adopted as the stems in the paradigm for tnieda ‘be launched’ as well. Alternatively, it could be an influence of the frequent use of the PL form of the archaic verb tniegħed ‘cry for mercy’, somewhat fossilized in a common prayer in the Imperfective 1PL form: nitnegħdu /nitnie:du/ ~ /nitne:du/. Irrespective of what triggers this overabundance in the paradigm, it appears that the PL cells are the driving force here. If such a trend is to get going, the i -u vs. a -w choice of dependency and their correlation with the subcategorization frame, which is being motivated here, may not be retained with such neatness. Rather, just like salla ‘(euphemistic for) blasphemize, Muslim prayer’ and sella ‘send regards’, in the case of tnieda ‘be launched’, we have an intransitive taking an i -u Imperfective paradigmatic dependency, at least optionally, and just like xtara ‘buy’, we have a verb across the Vth–VIIIth binyanim which is overriding the binyan-conditioning and allows for the presence of the i -u dependency in one of its paradigms.

6 The semi-parallel behavior across non-SM V-final verbs Non-Semitic verbs also display the dependencies employed in both subparadigms, yet, the contexts as to where and when the dependency is in a correlation with valency. The correlation is in fact only systematically employed with second generation loans, as will be illustrated in this section. Broadly, non-Semitic V-final loans, which constitute the largest set of the lexicon, have integrated on the basis of either the pattern of paradigmatic dependencies employed in the paradigm of mexa ‘walk’, i.e. an ej-ie-e Perfective sub-paradigmatic dependency and the i -u dependency in the Imperfective subparadigm, or the pattern displayed by that of the paradigm of qara ‘read’, i.e. an aj-ā-a Perfective sub-paradigmatic dependency and an a -w Imperfective subparadigm. In the case of Romance loans (Mifsud 1995) the basis of this choice of

136 | Maris Camilleri

integration is the thematic origin of these verbs. Are-themed verbs in Italian/Sicilian pattern on the paradigm of qara ‘read’, while ere/ire-themed verbs pattern on the mexa ‘walk’ paradigm. English V-final loans all pattern on the qara ‘read’ paradigm, with spella ‘spell’ being the exceptional English verb whose paradigm patterns on that of mexa ‘walk’. Notwithstanding this rather neat overall behavior, Mifsud (1995) identifies a number of verbs, particularly those with a CVCiCiV stem-shape, as well as other lexical idiosyncrasies such as kanġa ‘change hue’, bimba ‘(child’s word for) drink’ and a particular use of sparixxa ‘disappear’, which do not follow the pattern dictated by their thematic origin. From within the synchronic-based account being upheld here, thematic origin is not treated as a motivation for the current state of affairs. Rather, the thematic-based analysis constitutes a diachronic explanation of a synchronically fossilized complex morphological behavior. It is not the place here to discuss how thematic-class based explanations are transferred into a synchronic and surface-based account, however refer to Camilleri (2014b) for a detailed analysis and how a child learning the language, with potentially no awareness of a divide between SM and non-SM, or any knowledge of the thematic-based patterns in Romance, is still able to deduce patterns from the synchronic surface-based information available in the lexicon. What is crucial for us here are primarily two sets of non-SM data. The first is the behavior of CVCiCiV verbs, which are the key set of data Mifsud discusses to demonstrate how a diachronically-based thematic has to be overridden, and how a paradigmatic pattern distribution cannot be employed. The second set of verbs we will consider here is not discussed in Mifsud. This includes a sub-set of second-generation loans, particularly the ere/ire-derived ones. Such loans include prefixed verbs that are related with non-prefixed verbal counterparts, including: t-CVCiCiV verbs related with their primary CVCiCiV counterparts, e.g. baxxa ‘lower’ – tbaxxa ‘be lowered’, keċċa ‘expell’ – tkeċċa ‘be expelled’ and poġġa ‘place, cohabit’ – tpoġġa ‘be placed’; t-CVCiCjV verbs related with their primary CVCiCjV counterparts, e.g. gawda ‘enjoy’ – tgawda ‘be enjoyed’, solva ‘solve’ – ssolva ‘be solved’, and serva ‘serve’ – sserva ‘be served’; and t-CV:C(C)V verbs related with CV:C(C)V verbs: bata ‘suffer’ – tbata ‘be suffered’, fera ‘wound’ – tfera ‘be wounded’, segwa ‘follow’ – ssegwa ‘be followed’ and rieġa ‘direct’ – trieġa ‘be followed’. Additionally, apart from the two major sets of non-SM verbs, another class of verbs to be discussed is the set of what we here refer to as emergent equipollent nonSM verbs, which in turn illustrate parallel non-conformant behaviors, when one considers the diachronically-based explanation to the observed paradigmatic behaviors these verbs illustrate. These three sets of data will be discussed separately through Section 6.1–Section 6.3.

How inflectional morphology meets subcategorization frame distinctions | 137

6.1 CVCiCiV non-SM verbs Non-SM CVCiCiV verbs across the board, just like their SM counterparts, illustrate the same stem-shape conditioning, where an ej-ie-e i -u combinatorial pattern of cross-paradigm dependencies is exhibited in their paradigm. This pattern cuts across vocalic patterns, just as was the case across the SM data set discussed in Section 4, and in this case also cuts across the different thematic origins of the verbs. Are-derived verbs such as baxxa ‘lower’, patta ‘pay back’ and pixxa ‘urinate’, all display the same pattern of paradigmatic dependencies as ere/ire-derived verbs such as falla ‘fail’, fotta ‘ruin’ and ninna ‘(child’s word for) sleep’, as shown in Table 1. Mifsud (1995: 122) refers to the behavior exhibited by such a sub-set of are-derived verbs as involving an ‘oscillation between conjugations’, as they come to pattern with ere/ire-derived verbs. Tab. 14: The Perfective ej[1^2] – ie[3SG.M] – e[3PL] Imperfective i[SG] –u[PL] combinatorial pattern of dependencies in the paradigms of non-SM CVCiCiV verbs irrespective of their thematic origin.

Morphosyntactic values

Perfective

Imperfective

1^2

3SG.M

3SG.F

3PL

3SG

3PL

baxxa ‘lower’

baxxej-t

baxxa

baxxie-t

baxxe-w

j-baxxi

j-baxx-u

patta ‘pay back’

pattej-t

patta

pattie-t

patte-w

j-patti

j-patt-u

pixxa ‘urinate’

pixxej-t

pixxa

pixxie-t

pixxe-w

j-pixxi

j-pixx-u

falla ‘fail’

fallej-t

falla

fallie-t

falle-w

j-falli

j-fall-u

fotta ‘ruin’

fottej-t

fotta

fottie-t

fotte-w

j-fotti

j-fott-u

ninna ‘(child’s word for) sleep’

ninnej-t

ninna

ninnie-t

ninne-w

j-ninni

j-ninn-u

are-derived

ere/ire-derived

What differs from the behavior across the non-għ-final IInd binyan SM counterparts, is the fact that while the choice of the Imperfective sub-paradigmatic dependency reflected a correlation with a transitive valence, with the exception of salla ‘(euphemistic for) blasphemize, Muslim prayer’ and sella ‘send greetings’, the employment of this pattern here across the non-SM counterparts is solely formal in the sense that it is purely morphologically-conditioned, and not correlated with any robust valency/subcategorization frame realizations. This is because of number of such verbs are in fact intransitive, e.g. pixxa ‘urinate’, patta ‘pay back’, poġġa ‘(in the sense of) cohabit’, and certain uses of baxxa ‘lower’. The presence of an i -u dependency in the Imperfective sub-paradigm,

138 | Maris Camilleri

therefore, is simply a formal one, i.e. based upon and triggered by an analogy with Semitic CVCiCiV verbs, and is conditioned by the very stem-shape, and not by any valency considerations, as was additionally the case across the SM counterparts, excluding the lexical idiosyncrasies discussed.

6.2 ere/ire-derived second generation loans If we now consider primary – second generation loan pairs as the ones illustrated in the beginning of this section, and contrast the behavior between arederived and ere/ire-derived ones, a nesa ‘forget’ type of paradigmatic analogy is employed across the ere/ire-derived verbs, although not discussed in Mifsud (1995), and hence the combinatorial cross-paradigmatic dependency employed is Perfective ej-ie-e and Imperfect a -w. A qara ‘read’ paradigmatic analogy is retained across both primary and secondary generation loans and the Perfective aj-ā-a and Imperfective a -w combinatorial cross-paradigm dependencies are employed. The paradigms in Table 15 are meant to contrast the distinct behaviors firstly between are- and ere/ire-derived verbs, and secondly the distinct Imperfective paradigmatic dependencies employed across primary and secondary generation ere/ire-derived loans. Tab. 15: The contrasting cross-paradigm dependency changes or lack thereof across arederived and ere/ire-derived primary – second generation pairs.

Morphosyntactic values

Perfective

Imperfective

1^2

3SG.M

3SG.F

3PL

3SG

3PL

vara ‘launch’

varaj-t

vara

varā-t

vara-w

j-vara

j-vara-w

tvara ‘be launched’

t-varaj-t

t-vara

t-varā-t

t-vara-w

j-i-t-vara

j-i-t-vara-w

kanta ‘sing’

kantaj-t

kanta

kantā-t

kanta-w

j-kanta

j-kanta-w

tkanta ‘be sung’

t-kantaj-t

t-kanta

t-kantā-t

t-kanta-w

j-i-t-kanta

j-i-t-kanta-w

baxxa ‘lower’

baxxej-t

baxxa

baxxie-t

baxxe-w

j-baxxi

j-baxx-u

t-baxxa ‘be lowered’

t-baxxej-t

t-baxxa

t-baxxie-t

t-baxxe-w

j-i-t-baxxa

j-i-t-baxxe-w

fera ‘wound’

ferej-t

fera

ferie-t

fere-w

j-feri

j-fer-u

tfera ‘be wounded’

tferej-t

tfera

tferie-t

tfere-w

j-i-t-fera

j-i-t-fere-w

gawda ‘enjoy’

gawdej-t

gawda

gawdie-t

gawde-w

j-gawdi

j-gawd-u

are-derived

ere/ire-derived

How inflectional morphology meets subcategorization frame distinctions | 139

Morphosyntactic values

Perfective

1^2

3SG.M

3SG.F

Imperfective

3SG

3PL

tgawda ‘be enjoyed’

t-gawdej-t t-gawda

t-gawdie-t t-gawde-w

3PL

j-i-t-gawda

j-i-t-gawde-w

fotta ‘ruin’

fottej-t

fotta

fottie-t

fotte-w

j-fotti

j-fott-u

tfotta ‘be ruined’

t-fottej-t

tfotta

t-fottie-t

t-fotte-w

j-i-t-fotta

j-i-t-fotte-w

As one can see, no changes occur in the paradigms of are-derived primary and secondary generation loans that do not involve a CVCiCiV stem-shape, such that both verbs in the pair employ a Perfective aj-ā-a and Imperfective a -w combinatorial cross-paradigm dependency. On the other hand, ere/ire-derived verbs, whilst all showing an ej-ie-e dependency in the Perfective sub-paradigm, which is thematically-driven, however, at least on the basis of a diachronic account, in the Imperfective sub-paradigm, primary generation loans demonstrate a thematically-consistent dependency, i.e. i -u, and their second generation counterparts take an a -w dependency, such that the combinatorial cross-paradigm dependency is: Perfective ej-ie-e and Imperfective i -u, which is the dependency employed by Vth binyan SM verbs, and e-a voweled VIIth–VIIIth binyan verbs. What is crucial is that second generation across the board, unlike their primary counterparts, are all intransitive in nature, and consequently, the employment of the a -w dependency is highly correlated with the intransitive valency. What we conclude out of this behavior, with additional insight from the previous section, is that the observed paradigmatic effects corroborate Vanhove’s (1993: 27) claim that loan integration in Maltese first took place formally, i.e. by a simple analogy based on form, with semantic integration following after. Across primary loans and CVCiCiV verbs we clearly see that this is in fact the case. Diachronically, non-CVCiCiV primary loans simply integrated upon the qara ‘read’ or mexa ‘walk’ type of paradigmatic dependencies on the basis of their thematic origin. CVCiCiV verbs, on the other hand, display a stem-shape conditioned cross-paradigmatic dependency without any consideration of the thematic origin, or the acidity of the verb. Second generation loans on the other hand are all intransitive, displaying passive, anticausative, inchoative, and reflexive interpretations, thereby clearly exhibiting the correlation between the choice of the Imperfective sub-paradigmatic a -w stem-vowel – suffix dependency and the intransitive valency. The concluding remark here is that in fact, we can now fill in the ejt – a relation in Hoberman and Aronoff (2003: 78) table, represented in Table 6 in Section 3, with a +, given that, as has been just shown, such a dependency does in fact exist across non-SM verbs, just as it exists across SM verbs.

140 | Maris Camilleri

6.3 Emerging non-SM equipollent verbs In this section I wish to discuss a set of changes which seem to be taking place in the paradigms of a number of type C Romance loans (Mifsud 1995). From an MLRS search one finds that the verbs skada ‘expire’, nvada ‘invade’ and sploda ‘explode’, with one hit each for the first two and two hits for sploda ‘explode’ appear to be synchronically associated with two distinct paradigms. The behavior displayed, however, does not appear as systematic as the paradigm associations present across the SM set discussed in Section 5.1–Section 5.2. The conclusion derived out of this data set, as we will see, should only be considered as a deduction out of the data which is currently available. What is unfortunate is that as illustrated in the data in (5) below, one only finds Imperfective 3SG.M or 3PL forms. A real deduction of what is in fact taking place cannot be made in a conclusive manner. The need to know what all the SG and PL word-forms look like is necessary in order to test the robustness of the Imperfective subparadigm stem-vowel – suffixal exponent dependency, which is what we have been focusing upon in this paper. It may well turn out to be the case that these verbs pose as exceptions to the otherwise very robust generalization across Vfinal[-t selecting] Standard verbs. If the speaker of the utterance in (5b), which is incidentally ungrammatical to me and for other speakers who I have consulted, (just as jisploda in (5c)), allows for the stem -invadi in the Imperfective SG cells and not the expected *-invada, in accordance with the established Imperfective sub-paradigmatic dependencies, then this suggests that an i -w dependency that we have not yet met across the Maltese data set so far, is emerging. (5a)

(5b)

(5c)

Iċ-ċens tal-MDC t-ħalla DEF-groundrent of.DEF.MDC PASS-leave.PFV.3SG.M j-i-skada meta seta’ 3-EPENT.VWL-expire.IMPV.SG.M when can.PFV.3SG.M j-i-ġi m-ġedded 3-FRM.VWL-come.IMPV.SG.M PASS.PTCP-renew.SG.M ‘The MDC’s groundrent was left to expire when it could have been renewed (MLRS) …irwejjaħ xejn pjaċevoli li j-kun-u qegħd-in ...smell.PL nothing nice.PL COMP 3-be.IMPV-PL stay.PROG.PTCP-PL j-i-nvade-w dar-ek. 3-EPENT.VWL-invade.IMPV-PL house-2SG.GEN …smells that are not as nice that are invading your house. (MLRS) … j-i-sploda l-bomba, u …3-EPENT.VWL-explode.IMPV.SG.M DEF-bomb and

How inflectional morphology meets subcategorization frame distinctions | 141

(5d)

j-a-ħrab minn fuq il-post. 3-FRM.VWL-flee.IMPV.SG.M from on DEF-place he explodes the bomb and flees from the place. (MLRS) …id-dejn pubblik-u tagħ-ha t-ħalla …DEF-debt.SG.M public-SG.M of-3SG.F.ACC PASS-leave.PFV.3SG.M j-i-sploda. 3-EPENT.VWL-expire.IMPV.SG.M …its public debt was left to explode.’ (MLRS)

If we focus on (5a) and (5d), which I regard as grammatical, even if I would personally still use the forms jiskadi and jisplodi respectively, it is striking to observe that the use of the a-taking Imperfective SG stems instead of the expected i, at least in accordance with their ere thematic origin, are both embedded under the permissive predicate tħalla ‘be left, allowed, permitted’. In such an embedded context, skada ‘expire’ loses its usual fi- ‘in’ headed collocational OBL argument, yet remains intransitive. It appears that the change in form observed is also correlated with a thematicargument change, where the SUBJect grammatical function is more likely to map onto a patient/experiencer than to a theme, in these embedded constructions. If we are on the right track with respect to our interpretation of the observed behavior, then in the case of sploda ‘explode’, which can be used both transitively and intransitively, a change in acidity results in a change in the paradigmatic dependency at least on the basis of the data fragment available. Having said this, while these verbs appear to be still in an optional stage of change, the fact that the presence of the a stem and the -w suffixed stem do in fact constrain the subcategorization frame possibilities in a particular construction implies that the very change in the inflectional morphology functions roughly in the same way as the t- prefix (and its allomorphs) in the case of second generation loans e.g. tvara ‘be launched’, tkanta ‘be sung’ and tfotta ‘be ruined’, for example, as discussed in the previous section. The interesting difference is that the type C verbs being discussed in this section do not take second generation loan counterparts, and are themselves primary derived loans. Thus, while no additional affixal morphology reflects the change in valency, the inflectional morphology, in this case the morphological feature internal to it, i.e. the paradigmatic dependency between the vowel and the suffix, appears to have become the source from where such changes are starting to be observed. A change in the Imperfective SG stem-vowel in particular, with the implicational effect that conditions the choice of the PL suffixal allomorph, yields the same effect as though they had a related second generation loan. Since we find the same verb form with distinct associated paradigms, then essentially we can conclude that these verbs are emerging as equipollent verbs

142 | Maris Camilleri

that provide themselves as a non-concatenative alternative to a second generation loan verb.10 What follows below is a discussion of the paradigms associated with sparixxa ‘disappear’, which while having proved problematic in Mifsud’s theme-based diachronic account, it for us on the other hand an important piece of data. For Mifsud, sparixxa ‘disappear’ represents yet another verb that does not conform to the diachronic theme-based analysis. While it is diachronically an ire-derived verb, it also involves paradigmatic forms that pattern as though the verb was a primary are-derived loan, and which for our purposes here, involves a Perfective aj-ā-a and Imperfective a -w cross-paradigm dependency, as illustrated in Table 16. This data set is attributed to Schabert’s (1976: 108) observation, who records the use of sparixxajt ‘I/you disappeared’ instead of sparixxejt (and the rest of the other forms) in the coastal Eastern town San Ġiljan. Tab. 16: The presence of both mexa-type and qara-type modeled paradigms associated with sparixxa ‘disappear’ .

Morphosyntactic values

sparixxa ‘disappear’ (expected paradigm)

sparixxa ‘disappear’ (another alternative paradigm)

1^2

sparixxej-t

sparixxaj-t

3SG.M

sparixxa

sparixxa

3SG.F

sparixxie-t

sparixxa:-t

3PL

sparixxe-w

sparixxa-w ~ sparixxe-w

3SG

j-i-sparixxi

j-i-sparixxa

3PL

j-i-sparixx-u

j-i-sparixxa-w ~ j-i-sparixxe-w

Perfective

Imperfective

This fact prompts Mifsud (1995: 126) to assume that the motivation for the oscillating paradigmatic behavior is the influence of a frequent use of English in this area, saying that the locals are ‘highly exposed to the interference of English in everyday Maltese speech’. The immediate question which arises out of this possible explanation would then be why it is not the case that other verbs show such a similar behavior and why should it be special. In the dialectal varieties I

|| 10 It remains an open question whether there are any other such emerging equipollent nonSM verbs in the Maltese system.

How inflectional morphology meets subcategorization frame distinctions | 143

am familiar with (Naxxari and Mosti), both paradigms associated with sparixxa ‘disappear’ are available. Yet, their choice is not in free variation however. Rather, this verb’s distinct paradigmatic behaviors seem to illustrate the same equipollent behavior we have observed across SM għama ‘be/become blind, blind someone’ and għejja ‘be/become tired, tire someone’ in Section 5.2. Once again, the paradigm illustrating the Perfective aj-ā-a and Imperfective a -w cross-paradigm dependencies yields an inchoative meaning: ‘vanish oneself, become thin’, while on the other hand, the paradigm exhibiting the Perfective ej-ie-e and Imperfective i -u cross-paradigm dependencies refers to the transitive/causative meaning of ‘vanish/disappear something’. In this respect, therefore, the oscillation does not remain ‘anomalous’ (Mifsud 1995: 126). Having argued for such an analysis, I have also found the following Facebook comment on Google (speaker’s location not disclosed) presented in (6). While grammatical, I would have used the Imperfect 3PL jisparixxaw ‘they disappear’. (6)

...il-problem-i ta’ Malta kollha j-i-sparixx-u ...DEF-problem-PL of Malta all.PL 3-EPENT.VWL-disappear.IMPV-PL bħalma sparixxa-w meta kien hemm ... like.COMP disappeared.PFV.3-PL when be.PFV.3SG.M there ‘...all of Malta’s problems will disappear just as they disappeared when there was Google’

What the data set in (6) represents, at least from the paradigmatic fragment available, is that while the Imperfective sub-paradigm, under the assumption that the established SG and PL dependencies are robust enough across non-għ-final V-final verbs, employs i -u dependency, the Perfective sub-paradigm employs the aj-ā-a dependency. If we assume that the speaker is consistent in the use of the verb forms, then the cross-paradigm combinatorial pattern of dependencies is Perfective aj-ā-a and Imperfective i -u. Note that as the i -u Imperfective dependency is here being associated with an intransitive use of this verb, it is not the case that the choice of paradigmatic dependency necessarily reflects in any way the valency of the verb sparixxa in given syntagmatic contexts. Additionally, if we are to accept the paradigmatic fragment in (6), and what can be deduced out of it, on condition that this is in fact representative of the speaker’s competence, then the combinatorial pattern of paradigmatic dependencies we end up with requires that we substitute the – with a (+) in the cell with the ajt i combinatorial pattern across the non-SM verbs in Hoberman and Aronoff’s (2003: 78) tableau represented in Table 6. If this is correct, then we can conclude that every single combinatorial pattern of paradigmatic dependencies in Maltese is present across both the SM and non-SM data set, even if with different frequencies and restrictions.

144 | Maris Camilleri

With this data set we conclude our account of the non-SM verbs. Essentially, we observe that inflectional morphology and valency correlations are only partially employed in the non-SM verbal paradigms. They are mainly present across second generation loans, which do obey the strict correlation between the a -w dependency in the Imperfective sub-paradigm and an intransitive valency. Emergent equipollent data also seems to be pointing at this direction, although the formal dependency employed, particularly the presence of i -w types of dependencies may be akin to the very intermediate phase these verbs are still part of. Additionally, dialectal sparixxa ‘disappear’ was shown to also display a behavior that parallel SM equipollent verbs.

7 Conclusion The primary concern is this paper was to bring to light the inflectional verbal paradigm in Maltese as a system of interrelated networks from where we are then able to better understand how inflectional morphology interacts with other areas that interface with the morphological system. This was done through a multi-dimensional model that was essentially synchronic, surface-based, stemoriented and paradigm-driven. Following the inventory of inflectional affixes, it was shown how, non-għ-final V-final Maltese inflectional paradigms make use of two different dependencies internal to each of the two sub-paradigms, such that all four combinatorial patterns of cross-paradigm dependencies are attested across the different verbal paradigms in the lexicon, even if with different frequencies and distributions across the SM and non-SM lexicon. The unraveling of such dependencies were in fact also used to then fill in a number of lacunae in Mifsud’s (1995) and Hoberman and Aronoff’s (2003) descriptions. The correlation observed with valency/acidity in particular, specifically required us to concentrate on the network internal to the Imperfective sub-paradigm, which dependency essentially exploits the NUM-based split available internal to this sub-paradigm, and involves a relation between the Imperfective stem-vowel in the SG cells and the suffix in the PL cells. As we looked at the paradigm as an entity in itself, it became clearer that Maltese inflectional morphology is not simply affixal in nature, but also non-concatenative, mainly through the very presence of stem-alternations. Across the data set presented here, stem-alternations mainly came about via morphophonological conditions and arbitrary ablaut-changes. Additionally, apart from realizing grammatical information, particularly morphosyntactic and morphosemantic features, we here observe how once the choice of the Imperfective sub-paradigmatic dependency corre-

How inflectional morphology meets subcategorization frame distinctions | 145

lates with valency differences, an analysis was provided whereby this inflectional morphology behavior itself comes to realize such valency-related information as well. Identifying the conditions and constraints across the SM and non-SM data sets when the realization of this sort of information is available, constituted a major part of our discussion here. Such a realization through inflectional morphology was either the sole exponent, as was the case when identifying a number of equipollent verbs in Maltese, or else, the inflectional morphology realizing this information was in a multiple exponence relation with other affixal and templatic morphology, including stem-shape conditioning, i.e. templatic morphology. The identification of the possibility for inflectional morphology in Maltese, particularly the dependencies within the paradigm, which were analyzed as morphological features in themselves, as an exponent of valence information, allowed us to better distinguish equipollent from labile verbs. The nature of our analysis, which formed part of a broader project aiming to unify the SM and non-SM lexicon, allowed us to shed a new light upon the paradigms of V-final non-SM verbs in the language. Apart from the discovery of emergent equipollent verbs in this part of the lexicon, which made up for these verbs’ lack of secondary generation loans, important claims where made with respect to the paradigmatic behaviors and contrasts observed when focusing in detail upon secondary generation loans. It was shown how the robust correlation between intransitivity and the choice of the a -w Imperfective sub-paradigmatic dependency goes against the diachronic theme-driven analysis. Additionally we have also seen how in-depth considerations of the paradigms of non-SM loans suggest that form precedes function in Maltese. It was shown how non-SM verbs first integrated formally and analogically upon other existing patterns of paradigmatic behaviors. This thus implies that once second generation loans were formed upon these, valency distinctions (if the case), are reflected internal to the inflectional morphology through the systematic use of patterns of paradigmatic dependencies that already exist across other similar SM verbs, which as we also discussed here, express the same subcategorization frame. Acknowledgment: The ERC-2008-AdG-230268 MORPHOLOGY grant, which allowed this research to be possible, is hereby acknowledged.

146 | Maris Camilleri

Abbreviations 1, 2, 3 ACC COMP DAT DEF EPENT F FRM FUT GEN IMPV M

1st, 2nd, 3rd person accusative complementizer/complement dative definite epenthetic feminine formative future genitive imperfective masculine

OBJ OBL PASS PFV PL PROG PTCP SG

SM SUBJ VWL XCOMP

object oblique passive perfective plural progressive participle singular Semitic Maltese subject vowel open complement

References Ackerman, Farrell, James B. Blevins & Robert Malouf. 2009. Parts and wholes: Patterns of relatedness in complex morphological systems and why they matter. In James B. Blevins & Juliette Blevins (eds.), Analogy and grammar, 54–81. Oxford: Oxford University Press. Ackerman, Farrell & Robert Malouf 2013. Morphological organization: The low conditional entropy conjecture. Language 89(3). 429–464. Anderson, Stephen R. 1992. A-morphous morphology. Cambridge: Cambridge University Press. Aronoff, Mark. 1992. Stems in Latin verbal morphology. In Mark Aronoff (ed.), Morphology now, 5–32. Albany: SUNY Press Baerman, Matthew & Greville G. Corbett. 2012. Stem alternations and multiple exponence. Word Structure 5(1). 52–68. Baerman, Matthew, Dunstan Brown & Greville G. Corbett. 2005. The syntax-morphology interface: A study of syncretism. Cambridge: Cambridge University Press. Borg, Albert J. 1988. Ilsienna. Malta: Ħas Sajjied. Borg, Albert J. & Manwel Mifsud. 1999. Il-forom (binyanim) in Maltese: Inflection or derivation? Paper presented at the MMM2, Malta. Camilleri, Maris. 2012. Morphological complexity in Maltese: A divergence from canonicity. Proceedings of the MMM8 . September, 2011, Cagliari, Sardegna. 91–112. Camilleri, Maris. 2014a. Stems as multiple exponents in Maltese and morphological complexity. Lingua e Linguaggio 1. 43–63. Camilleri, Maris. 2014b. The stem in inflectional verbal paradigms in Maltese. PhD Diss. University of Surrey. Camilleri, Maris & Louisa Sadler. 2012. On the analysis of non-selected datives in Maltese. In Miriam Butt & Tracy Holloway King (eds.), Proceedings of the LFG12 Conference, 118–138. Stanford: CSLI Publications. Camilleri, Maris, Shahimaa ElSadek & Louisa Sadler. 2014. A cross dialectal view of the Arabic dative alternation. Acta Linguistica Hungarica 61(1). 3–44.

How inflectional morphology meets subcategorization frame distinctions | 147

Comrie, Bernard. 2006. Transitivity pairs, markedness, and diachronic stability. Linguistics 44. 303–318. Corbett, Greville G. 2007. Canonical typology, suppletion and possible words. Language 83. 8–42. Corbett, Greville G. & Matthew Baerman. 2006. Prolegomena to a typology of morphological features. Morphology 16.231–246. Darden, Bill. 1992. The Cairene Arabic verb without form classes. In Diane Brentari, Gary N. Larson & Lynn A. MacLeod (eds.), The joy of grammar: A Festschrift in honor of James D. McCawley, 11–24. Amsterdam/Philadelphia: Benjamins. Fabri, Ray. 2009. Stem allomorphy in the Maltese verb. Ilsienna – Our Language 1. 1–20. Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie & Maria Polinsky (eds.), Causatives and transitives, 87–120. Amsterdam/Philadelphia: Benjamins Hoberman, Robert D. & Mark Aronoff. 2003. The verbal morphology of Maltese: From Semitic to Romance. In Joseph Shimron (ed.), Language processing and language acquisition in languages of Semitic, root-based, morphology, 61–78. Amsterdam/Philadelphia: Benjamins. Maltese Language Resource server (MLRS): http://mlrs.research.um.edu.mt/#MLRS Mifsud, Manwel. 1995. Loan verbs in Maltese: A descriptive and comparative study. Leiden: Brill. Puech, Gilbert. 2011. Syllable structure and stress in Maltese. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 277–316. Berlin: Akademie Verlag. Saydon, Pietru P. 1958. The vocalization of the verb in Maltese. Orbis 7. 168–182. Schabert, Peter. 1976. Laut- und Formenlehre des Maltesischen anhand zweier Mundarten. Erlangen: Palm & Enke. Spagnol, Michael. 2011. A tale of two morphologies: Verb structure and alternations in Maltese. Ph.D dissertation, University of Konstanz. Spencer, Andrew. 2012. Identifying stems. Word Structure 5. 88–108. Stump, Gregory T. 2001. Inflectional morphology: A theory of paradigm structure. Cambridge: Cambridge University Press. Vanhove, Martine. 1993. La langue maltaise: Études syntaxiques d’un dialecte arabe «périphérique». Wiesbaden: Harrassowitz. Zaliznjak, Andrej A. 1973. O ponimanii termina ‘padež’ v lingvističeskix opisanijax. In Andrej A. Zaliznjak (ed.), Problemy grammatičeskogo modelirovanija, 53–87. Moscow: Nauka. [Reprinted in: Andrej A. Zaliznjak. 2002. Russkoe imennoe slovoizmenenie: s priloženiem izbrannyx rabot po sovremennomu russkomu jazyku i obščemu jazykoznaniju, 613–647. Moscow: Jazyki slavjanskoj kul ́tury.]

Hans-Jörg Döhla

The origin of differential object marking in Maltese Abstract: Differential object marking (DOM) is a prominent feature of Maltese grammar. If we take a closer look at the possibility of internal development we come to the conclusion that DOM is not a genuine feature of pan-Arabic grammar. The source for its emergence in some more or less isolated, i.e. not interconnected Arabic vernaculars can be found in contact situations where the contact language serves as a source language for the induced replication of the pattern. In the case of Maltese, an Old Sicilian superstrate (and later adstrate) is assumed as the origin of the phenomenon. Keywords: Maltese, Arabic, Sicilian, object marking, language contact, language change

1 Introduction Generally, differential object marking (DOM) is a morphosyntactic phenomenon which can be found in many languages of the world (cf. Iemmolo 2010/11), a part of them as internal creations, and in other cases through contact-induced replication. In this paper, we will scrutinize the origin of DOM in Maltese departing from the fact that DOM is not a genuinely Arabic feature. Therefore, after introducing into the structure and its function, we will show that, despite other claims, there is no prepositional marking of the direct object in Classical Arabic. Furthermore, we will advance the view that the appearance of DOM in some historical and present Arabic vernaculars can all be explained as contactinduced replication. Finally, we will defend the Old Sicilian superstrate hypothesis by providing linguistic and historical evidence.

|| Hans-Jörg Döhla: Ruhr-University Bochum, Institute of Romance Studies, Universitätsstr. 150, 44801 Bochum, Germany. E-mail: [email protected]

150 | Hans-Jörg Döhla

2 DOM – generalities DOM is a term coined by Bossong (1982 and 1985) which designates variation in morphological marking of the direct object in a two argument transitive construction. This variation is conditioned mainly by the inherent, referential and discourse pragmatic properties of the object referent.1 Thus Iemmolo (2010/11: 268–269), in one of the latest studies on DOM in general2, states that [..] overt coding serves to signal that a lexical NP fulfilling the role of direct object is highly topical, while indexation is naturally associated with topical NPs and signals the discourse prominence of such direct objects. [..] it is the lower frequency of topical/topicworthy referents in direct object position that explains why these direct objects are less easy to process. [..] agents are human, definite and topical and direct objects are less human, [less] definite and [less] topical than agents, the reversal of this situation would be likely to be harder to process.

This approach has its roots in Comrie’s (1989: 128) markedness hypothesis: […] the most natural kind of transitive construction is one where the A is high in animacy and definiteness, and the P is lower in animacy and definiteness; and any deviation from this pattern leads to a more marked construction. [..] the construction which is more marked in terms of the direction of information flow should also be more marked formally, [..]

(Cf. also Aissen 2003, García García 2010 and Weber 2010). In other words, if the two core arguments of a transitive construction do not display their respective prototypical properties, there is a higher chance that one of the two arguments is marked formally, i.e. in our case the direct object.3 Following Chafe (1976), Keenan (1976), Givón (1976), Borg and Comrie (1984), Dixon (1989) and Comrie (1989) the prototypical characteristics of the two arguments of a transitive construction can be determined in the following way (see Table 1):

|| 1 Because of the limited space available we cannot present a full discussion of DOM. Therefore, other triggering but marginal aspects, such as verbal semantics (cf. Heusinger 2008), are left aside in the following study. 2 Dalrymple and Nikolaeva (2011: 220) come to a similar conclusion when they write that “[o]n our analysis, ‘true’ optionality does not arise: objects with the same semantic features are either grammatically marked or unmarked, depending either on their information-structure role or their grammatical function”. 3 There is also the opposite phenomenon where the subject is marked formally. Analogically, it is called differential subject marking (cf. Hoop and Swart 2009).

The origin of differential object marking in Maltese | 151

Tab. 1: Prototypical characteristics of the first and the second argument of a transitive construction.

First argument

Second argument

semantic role

agent

patient

grammatical relation

subject

(direct) object

inherent property

human – volitional

(in)animate – non-volitional

referential property

definite – specific

(in)definite – (un)specific

prominence

topic (theme)

comment (rheme)

According to the latest studies mentioned above it is mainly the prominence which triggers the internal development of DOM in a given language. This prominence approach can be specified further using the hierarchy of topicality provided by Givón (1976: 152): a. human > non-human b. definite > indefinite c. more involved participant > less involved participant. agent > dative > accusative) d. 1st person > 2nd person > 3rd person Thus, any extra-linguistic referent sharing the properties, which are found on the left side of the hierarchy, is more likely to be the topic of an utterance than the one exhibiting the characteristics of the right end. This leads us to the individuation scale where we combine the properties of extra-linguistic referents starting from highly individuated entities, such as speaker and addressee, which always have a topical character, over proper nouns to common nouns (see Table 2). This scale combines a number of implicational hierarchies. DOM is supposed to be initiated with personal pronouns, followed by proper nouns and extending to the right. The last category of common nouns, represented as lexical NPs, is the one that shows the greatest variation in object marking. This is due to the fact that the animacy4 and the referential5 scales are applied to the fullest extent. Thus, additional morphological elements have to be added in order to define exactly the degree of prominence of the object referent, i.e. by attributes and determiners such as articles, demonstratives, possessives or oth-

|| 4 [+HUM] > [+ANIM] > [–ANIM]. 5 Personal pronoun > proper noun > definite NP > indefinite specific NP > indefinite nonspecific NP > non argumental.

152 | Hans-Jörg Döhla

er quantifiers. During the diachronic extension process of DOM towards the right of the scale word classes like personal pronouns may become obligatorily marked. On the other hand the varying degree of topicality is still the relevant factor for the overt marking at the right pole of the scale. Tab. 2: The individuation scale (Yamamoto 1999: 28 and Döhla 2014: 268).

[+ANIM]

[‒ANIM]

[+HUM] speaker/ addressee >

bystander/ external referent >

personal pronouns (1, 2, 3)

[‒HUM] proper noun >

common noun >

other inanimated animated noun > noun

proper nouns

common nouns

topical > non topical

definite NP > specific indefinite NP > unspecific indefinite NP > non argumental

3 DOM in Maltese As in all the cases of DOM we have studied there is no written testimony of the incipient stage of DOM where only personal pronouns in object position were marked. The first example so far known probably stems from 1579: (1)

kif fixkil-t l-ohrayn in-fixkil lil-ik as obstruct-1/2SG.PFV OBJ-other.PL 1SG.IPFV-obstruct OBJ-2SG ‘as you obstructed others I obstruct you’ [Vella 2006: 27]

In (1), the object suffix -ik is definitely preceded by the prepositional object marker lil, which is a grammaticalized combination of the preposition li- and the definite article l-6 and which also serves as the indirect object marker. In the case of lohrayn, Vella acknowledged that the l-, besides the possibility of being an object marker, could also represent the definite article l-.

|| 6 In colloquial speech lil may be shortened to l-.

The origin of differential object marking in Maltese | 153

The following examples are either taken from literary works or from our own field work undertaken in Malta in 2012.7 As can be presupposed for an extended DOM system proper nouns and stressed personal pronouns generally always receive a marking: (2)

hem difn-u lil-u u lil mart-u Sara there bury-3PL.PFV OBJ-3SG.M and OBJ wife-POSS.3SG.M Sarah ‘There they buried him and his wife Sarah’ [Gen 49: 31; bible translation by Fortunato Panzavecchia, 19th c.; Kontzi 1999: 90]

(3)

Minflok laqat lil sieħb-u instead hit:3SG.M.PFV OBJ friend-POSS.3SG.M ‘Instead he hit his friend’ (female Maltese informant)

Object referents found further to the right side of the individuation scale need to be high in prominence in order to be marked: (4)

Minflok laqat lill-annimal feroċi, laqat lir-raġel instead hit: 3SG.M.PFV OBJ.DEF=animal wild hit:3SG.M.PFV OBJ.DEF=man ‘Instead of hitting the wild animal, he hit the man’ (female Maltese informant)

(5)

ji-dfen lil dan il-qattus 3SG.IPFV-bury OBJ DEM DEF=cat ‘he buries that cat’ (female Maltese informant)

In (4), annimal feroċi had been mentioned before and was present at the moment of speaking. At the same time, there is a contrastive focus between annimal and raġel. In (5) the ‘cat’ has certainly been mentioned before. In contrast, in the following utterance the aforementioned cat is not marked since the focus is not on the cat but on the killing and on the woman (her) who lost the cat because of the hunter’s mistake: (6)

talli hu kien qatli=lha l-qattus because 3SG.M be:3SG.M.PFV kill:3SG.M.PFV=3SG.F DEF-cat ‘… because he had killed her the cat’, i.e. ‘… because he had killed her cat’ (female Maltese informant)

|| 7 For this, 13 subjects of different age groups, gender and educational background were interviewed. All the interviews contained the following parts: free speech (1st and 3rd person perspectives); verbalizing a 40 pages picture story; a questionnaire with 97 minimal sentence pairs. The total time of recording is 5h39min. The main purpose of the interviews was obtaining data for a qualitative analysis of DOM in Maltese.

154 | Hans-Jörg Döhla

Thus, besides the factor that the object referent must have a high degree of topicality, it seems that focus also plays a certain role. This is also the case in the next example, where an indefinite unspecific NP receives the marking: (7)

Qatt raj-t lil xi xbejba ever see-1/2SG.PFV OBJ INDEF maiden ‘Did you ever see a maiden’ [Dun Karm ‘Il-musbieħ tal-mużew’ 20th c.; Arberry and Grech 1961: 136]

This marking of unspecific but focal referents, as demonstrated in this example taken from poetry, has reached its fullest extend in some cases of everyday spoken language, such as in the general question ‘Did you see anybody?’, a minimal sentence pair implemented in the questionnaire as Rajt lil xi ħadd? vs. Rajt xi ħadd?. All 13 interviewees opted for the first version, whereas the second one was ungrammatical to them and can thus be marked by *. Summarizing the results of our field work and the studies of Borg and Comrie (1984), Fabri (1993), Borg (1994) and Borg and Azzopardi-Alexander (1997), we get to the following extension of DOM in modern Maltese: Tab. 3: The extension of DOM in modern Maltese.

PP >

PN >

CN [+HUM] >

CN [+ANIM] >

[+TOP] >

[+DEF] >

[‒TOP]

[‒DEF]^[+SPEC] >

CN [‒ANIM]

[‒DEF]^[‒SPEC] > no arg.

(almost) complete

50–80%frequent

< 50% less frequent

sporadic or inexistent

Finally and as a special feature of Maltese DOM, we have to mention the optional indexation of the direct object at the verb: (8)

Raj-t=u see-1/2SG.PFV=3SG.M ‘I saw John’

lil OBJ

Ġanni. John [Borg and Comrie 1984: 113]

The phenomenon, known as “clitic doubling” (cf. Čéplö 2013), is certainly a secondary feature emerging as DOM spreads on the individuation scale and can therefore not be situated in the incipient stages of DOM, which are important in

The origin of differential object marking in Maltese | 155

this paper as it is concerned with the origin of DOM in Maltese. Therefore, and because of the lack of space, we are not able to present further details on the subject, limiting us to mentioning the phenomenon when it becomes relevant in the studied Arabic and Romance varieties. In the following we will address the question whether DOM in Maltese is an internal (Arabic) development or if it is a pattern replication induced by language contact. Therefore, there are two different streams to be discussed: one linking its origin to Levantine Arabic influence (cf. Borg 1996); the other one linking its origin to an Old Sicilian superstrate8 (cf. Borg and Mifsud 2002). But before entering the discussion of the first stream we need to look at Classical Arabic where (differential) object marking is claimed to have existed to a minimal extend using the preposition li- ‘for’.9

4 DOM in Classical Arabic As opposed to some scholars who are of the opinion that object marking – even though spurious – can already be found in ancient varieties of Arabic (cf. p.c. of Ignacio Ferrando in Borg and Mifsud 2002: 43), in this study we adopt the position that it does not exist, especially regarding the differential property of the marking.10 In order to explain why we take this position, we have to introduce some basic features of Arabic grammar. Classical Arabic, and thus modern written Arabic, distinguishes three cases, nominative, genitive and accusative, each one represented by a word final vowel already found in Protosemitic, -u, -i and -a, which can be extended by an -n in order to overtly express indefiniteness, -un, -in and -an11, for example:

|| 8 They call it “adstrate”, but we are of the opinion that, according to the linguistic history of Malta and the proper usage of the terms, we are dealing with a superstrate. 9 At this point it does not matter whether we consider Classical Arabic a constructed literary language which has never been the mother tongue of an Arabic tribe or if we attribute its origin to the Arabic vernacular spoken amongst the tribe of the Qurayš in Mecca and surroundings. 10 For an overview on DOM and the use of the agreement pronoun, i.e. clitic doubling, in Semitic languages cf. Khan (1984). 11 The final -n is superscripted since it does not appear as an individual grapheme at the same level as the other consonants of a noun or adjective. Thus, the exact positioning of the nunation in recitation and transcription is – as well as the right (short) vocalization – subject to interpretation of the reader or scholar, if the text, as in most of the cases, is not vocalized.

156 | Hans-Jörg Döhla

(9)

yu-ġliqu 'l=mudarris-ø-u-ø 'l=bāb-a-ø 3SG.M.IPFV-close DEF=teacher-M.SG-NOM-DEF DEF=door-ACC-DEF wa=yu-waqqidu kitāb-a-n ʕalà 'l=maktab-i-ø and=3SG.M.IPFV-drop book-ACC-INDEF on DEF=desk-GEN-DEF ‘The teacher closes the door and drops a book on the desk’

Besides the many special and particular uses of the case endings, the nominative marker is often used for what corresponds to the subject of a main clause12 (‘the teacher’) and the accusative marker to what corresponds to the direct object (‘the door/a book’). The genitive marker is always used after prepositions (‘on the desk’) and as the ending of the determining noun (nomen rectum) within a possessive noun phrase like baytu 'l-raǧuli ‘the man’s house’, where the head noun loses the final -n mentioned above. This genitive construction is very rigid and inflexible, i.e. “[n]o word can be interposed between the noun in the status constructus and the genitive, and consequently an adjective which qualifies the former must be placed after the latter” (Wright 1898: 222), for example: baytu 'l-raǧuli 'l-kabīru ‘the man’s big house’. The head noun of such a possessive construction is always definite. Nevertheless, if the possessor is indefinite, this referential indefiniteness can be triggered to the whole composition and thus to the head noun (Wright 1898: 225): (10)

bayt-u-ø 'l=raǧul-i-ø house-NOM-DEF DEF=man-GEN-DEF ‘the house of the man’, i.e. ‘the man’s house’

(11)

bayt-u-ø raǧul-i-n house-NOM-DEF man-GEN-INDEF ‘the house of a man’, i.e. ‘a man’s house’

This has to do with the rules for definiteness in Classical Arabic which also apply to modern written Arabic and all the modern dialects, including Maltese. The following nouns are always definite:  a noun with the definite article al-;  a noun with a possessive suffix. bayt-u-hū ‘his house’;

|| 12 For the sake of brevity at this point, we have to say often used since in subordinated clauses or in main clauses with an initial conjunction or particle the marking of the subject may correspond to the accusative ending.

The origin of differential object marking in Maltese | 157

 a noun followed by another noun in the genitive case, i.e. the head noun in a genitive or possessive construction;13  double determination is not permitted. Thus, besides the fact that the nouns of the possessive noun phrase may by no means be separated, the head noun can never be considered grammatically indefinite.14 Therefore, in order to obtain a grammatically indefinite head noun15 the possessive construction has to be paraphrased usually by the preposition li‘for; DAT’ put before the nomen rectum (Wright 1898: 199): (12)

bayt-u-n li='l=raǧul-i-ø house-NOM-INDEF for=DEF=man-GEN-DEF ‘a house of the man’

(13)

bayt-u-n li=raǧul-i-n house-NOM-INDEF for=man-GEN-INDEF ‘a house of a man’

Now, since the rigid genitive construction has been loosened, we are even able to interpose other elements, such as attributes to the head noun, in between the possessive construction: (14)

bayt-u-n ṣaġīr-u-n house-NOM-INDEF small-NOM-INDEF ‘a small house of the man’

li='l=raǧul-i-ø for=DEF=man-GEN-DEF

|| 13 In the last two cases, the head noun is in the “construct state” which involves the loss of the final -n which marks indefiniteness. In the first case, the noun is in the “determinated state” which appears after the definite article al-, the vocative particle yā- and the negation with lā. It also entails the loss of the final -n (Fischer 1972: 77). 14 Since in Arabic a noun can never receive a double determination (*al=-bayt-u-hū ‘his house’ or *al=bāb-u 'l=bayt-i ‘the door of the house’), a noun with the definite article or even with a possessive pronoun could never constitute the head noun of a possessive construction, as for example in Italian la porta della casa ‘the door of the house’, or la sua casa ‘his house’ (cf. Lyons 1999: 24) as opposed to Ar. bayt-u-hū. This supports the assumption that the two languages do not treat the feature of definiteness in the same morphosyntactic way. Besides that, the possessive pronouns have completely different syntactic implications. In Italian they are treated more like adjectives which also allow the definite article. In Arabic, on the other hand, they behave like nouns in the genitive which put the head into the status constructus. 15 Even in English the use of the genitive -s, usually restricted to human referents, transfers definiteness to the head noun: a man’s house ‘the house of a man’, cf. Lyons (1999: 23, fn. 12) for further analysis.

158 | Hans-Jörg Döhla

Coming back to the claimed (differential) object marking in Classical Arabic, we do not know of any Arabic manual that mentions or describes a special kind of object marking in a finite verbal clause besides the ending -a(n) mentioned above, not to speak of any differential marking of the object argument. All the examples we know of where the preposition li- is occasionally used to introduce the direct object, refer to infinite constructions with verbal nouns or participles. In general, in these infinite constructions the noun corresponding to the direct object of the verbal noun may be treated as an Arabic accusative object or, and probably more often, as the nomen rectum in the genitive (genitivus objectivus) in a possessive noun construction with the verbal noun as the head. While the first treatment emphasizes the verbal character of the verbal noun, the second one puts emphasis on its nominal character, without alteration in meaning. So we are dealing with purely stylistic variation: (15)

qātil-u-n kill.PTCP.PRS-NOM-INDEF ‘one who kills people’

'l=nās-a-ø DEF=people-ACC-DEF

(16)

qātil-u-ø kill.PTCP.PRS-NOM-DEF ‘one who kills people’

'l=nās-i-ø DEF=people-GEN-DEF [Wright 1898: 64]

In the following, it suffices to look at Wright (1898: §§29–34) since most later European grammarians are based on his grammar. There we find the following cases where li- is used before the direct object belonging to a verbal noun, here called nomen actionis: (a) When the nomen actionis immediately precedes the object and is undefined […]. (b) When the nomen actionis immediately precedes the object, and is defined by the article […]. (c) When the genitive of the subject is interposed between the nomen actionis and the object […]. [Wright 1898: 62] In (a), we are dealing with a genitive construction where the head noun should be indefinite, not only referentially but also grammatically. As we have seen above, this can only be obtained by a paraphrasal construction involving the preposition li-. Wright (1898: 62) provides the following example: (17)

min ġayr-i-ø taʔammul-i-n from NEG-GEN-DEF consider.INF-GEN-INDEF li=ḫaṭāʔ-i-hī wa=ṣawāb-i-hī for=mistake-GEN-POSS.3SG.M and=correctness-GEN-POSS.3SG.M ‘without considering what was false and true in it’

The origin of differential object marking in Maltese | 159

The nomen regens taˀammul-i-n is indefinite according to the construction it is involved in. Therefore, the direct objects depending on the verbal noun taˀammul- cannot be linked to the noun by means of a genitive construction since this would determine the head noun. In (b), the verbal head noun is already determined by the definite article, a fact which also impedes the appearance of the genitivus objectivus since double determination is ungrammatical in Arabic (cf. footnote 14): (18)

tark-u-ø 'l=muṭālabat-i-ø DEF=pursue.INF-GEN-DEF give.up.INF-NOM-DEF li='l=nās-i-ø for=DEF=people-GEN-DEF ‘giving up pursuing the people’

[Wright 1898: 62]

The reason for using the preposition li- in this case is to be found in muṭālabat; since it is already determined by the article, probably in order to emphasize the verbal semantics of ‘to pursue’, the object noun al-nās- cannot be linked formally to its governing verbal noun by a genitive construction. Instead the writer could also have simply extended the chain of dependent genitives, tark-u muṭālabat-i 'l-nās-i, but he didn’t choose to do so. In (c), the actual subject of the verbal noun is interposed between the verbal noun and the object. Now, the subject is in the genitive (genitivus subjectivus) or is substituted by a possessive pronoun, determining the verbal head noun, and prohibiting thus the formation of another genitive construction, i.e. genitivus objectivus, linked to the same verbal noun: (19)

ruʔyat-u-hū see.INF-NOM-POSS.3SG.M ‘his seeing God’

li='llāh-i-ø for=God-GEN-DEF [Wright 1898: 62]

In (19), the anaphoric pronoun refers to a different entity than ‘God’, namely the subject of the verbal noun. A similar construction can be observed in Borg and Mifsud (2002: 37) taken from contemporary Egyptian Arabic (we have changed the transliteration to our system as will be done with all following examples taken from transliterated sources): (20)

šurb il=banāt li='l=ḫamri DEF=girl.PL for=DEF=liquor drink.INF ‘Girls drinking liquor is shameful’16

ʕēb shame

|| 16 Generally, all the case endings have become lost in the Arabic dialects. Thus, we are not sure if the final -i after il-ḫamr is a reflection of the classical ending or if we are dealing with a different, particularly Egyptian, morpheme.

160 | Hans-Jörg Döhla

Again, we are dealing with the appearance of a subject and an object noun together with the verbal noun as the head of the construction. A chain of possessive constructions would be impossible in this case because it would alter the meaning of the whole sentence: (21)

šurb banāt il=ḫamr ʕēb drink.INF girl.PL DEF=liquor shame ‘the drinking of the liquor girls is shameful’ or ‘the fact that the liquor girls drink is shameful’

Now, the former object noun il-ḫamr ‘the liquor’ does not depend on the verbal noun šurb ‘drinking’ any more, but functions as the determiner (nomen rectum) of the noun banāt ‘girls’, thus giving way to the nominal composition banāt il=ḫamr ‘the girls of liquor’, i.e. ‘the liquor girls’. Therefore, in order to obtain the desired meaning of the clause, the possessive chain has to be disrupted, and the only way of syntactic alignment is introducing the direct object with li-. Finally, Fischer (1972: §203, note 2) remarks that the preposition li- is compulsory when the object of the infinite noun phrase comes to stand before the verbal noun it depends on: (22)

li='l=ḍayf-i-ø for=DEF=guest-GEN-DEF ‘honoring the guest’

mu-krim-u-n AP-honor.PTCP.PRS-NOM-INDEF

This does not surprise at all, since the paraphrase with li- is needed to resolve the rigid possessive construction and to left-dislocate the nomen rectum (= direct object) in order to topicalize it. Summarizing, it can be stated that the reason for the overt marking of direct objects in Classical Arabic infinite phrases is to be found in the grammatical construction itself, which formally corresponds to a genitive construction, where the object argument takes the position, formal marking and grammatical function of the otherwise possessor in a possessive construction. Therefore, in order to avoid this rigid genitive construction which, in many cases, is not flexible enough to react to certain communicative needs, the speakers or writers use the paraphrase with li-, irrespectively of the function of the nominal elements involved: ‘possessed entity + li-possessor’ or ‘verbal noun + li-object argument’. If this sporadic use of li- before the direct object mainly to be found in the older varieties of Classical and written Arabic and limited to infinite phrases, has finally paved the way for the introduction of DOM in finite clauses in some Arabic dialects remains questionable.

The origin of differential object marking in Maltese | 161

As opposed to Classical Arabic, there is DOM in several modern and historical spoken Arabic varieties, namely Andalusian Arabic, Levantine and Iraqi Arabic, Cypriot Arabic and Maltese. In the following we will argue that DOM by means of the preposition li- is not a genuine Arabic feature and that its emergence can be connected to the influence of a contact language featuring DOM as an internally developed marking system.

5 Andalusi Arabic Zammit (2009/10: 36) emphasizes the fact that Andalusi and Maltese Arabic share DOM, among several others, as a common feature. He provides the only example teqcí lal ē͑erhuén ‘you shall cover the naked’ taken from Corriente (1977: 126), the source of which is the Doctrina cristiana in lengua arabiga y castellana (1566) by Martín de Alaya. This late document for Andalusi Arabic from the 16th century intended to teach former Muslims the essence of their new Christian religion in vernacular Arabic. In contrast there are far earlier examples, as from the poetry of Ibn Quzmān (1078–1160) (23)

law ray-t li=akwās dār-ī if see- 1/2SG.PFV OBJ=cup.PL house-POSS.1SG ‘if you could see the cups of my house’. [Corriente 1977: 126]

or Zaǧǧālī (1220–1294), (24)

iḏā azwaj al=šayx when marry:3SG.PFV DEF=old.man ‘when an old man marries a girl’

la=ṣabiyya OBJ=girl [Corriente 1977: 126]

where even inanimate object referents (akwās dārī ‘the cups of my house’) or non-topical indefinite generic objects (ṣabiyya ‘a girl’) may be marked. However we do not agree with the Syriac origin of DOM in Andalusi Arabic given the fact that Syriac impact on Andalusi Arabic is virtually non-existent apart from a possible indirect transfer via Judeo-Arabic or Middle Hebrew.17 Even the influence of any Arabic vernacular exhibiting DOM, such as Levantine

|| 17 Cf. Döhla (2004) for the possible Middle Hebrew origin, through Aramaic influence of the Old Spanish pleonastic use of the possessive pronoun in the locution su fijo del rrey ‘lit. his son of the king’, i.e. ‘the king’s son’. This construction can also be found in Levantine Arabic where it is attributed to an Aramaic substrate: ibno la-l-malik < Aram. breh də-malkā (cf. Río Sánchez 2013: 136).

162 | Hans-Jörg Döhla

or Iraqi Arabic, which have been in contact with Aramaic for at least more than fifteen centuries, can be excluded, since most of the traits found in Andalusi Arabic can be traced back to South Arabian, i.e. Yemeni Arabic, origin (cf. also Zammit 2009/10: 24). Moreover, and to our opinion, most convincingly, the fact that DOM cannot be attested in any North African vernacular, pre- or postHilalian, the emergence of DOM in Andalusi Arabic seems to be a geographically isolated phenomenon not linked to any pan-Arabic internal grammaticalization tendencies. More likely, the emergence of DOM in Andalusi Arabic is a result of language contact between the Arabic and the Romance vernaculars spoken in the Iberian Middle Ages. As a matter of fact, the earliest testimonies by Ibn Quzmān in the 12th century are certainly due to the author’s bilingualism and the use of vernacular Arabic for writing his poems (cf. Corriente 2008). Considering the use of spoken varieties in Arabic poetry it is no surprise at all that the earliest testimony of DOM for the emergent Romance languages in al-Andalus can be found in the special Andalusi muwaššaḥ poem, where the last stanza, called ḫarǧa, was either presented in vulgar Arabic or Romance, or in a combination of the both. The following example is the earliest to our knowledge, found in the ḫarǧa of a muwaššaḥ (11th century) with code-switching between colloquial Arabic and Romance: (25)

aṣṣabáḥ BÓNO GÁRRE ME D+ ÓN BENÉŚ YA LO ŚE KE OTRI AMÉŚ A MÍBE ṬU NO(N) QUÉREŚ ‘Good morning, tell me where you come from’ ‘I already know that you love someone else’ ‘(but) me you don’t love’

[Corriente 1997: 319]

As predicted by the individuation scale in Table 2, the marking of an anaphoric topical object pronoun should be expected in any stage of expansion of DOM along that scale. According to the study of Pensado Ruiz (1995: 224) el C[omplemento] D[irecto] P[reposicional] procede de un giro latinovulgar para la expresión del tópico de la frase AD MIHI, AD TIBI. [‘the prepositional direct complement originates from a Vulgar Latin collocation for the expression of the topic of the frase AD MIHI, AD TIBI’].

And according to evidence found in 10th century Latin documents, displaying influence by Romance vernaculars, DOM must have already spread further on

The origin of differential object marking in Maltese | 163

the individuation scale reaching at least the referential category of proper names: (26)

uenerunt Gundessaluo et alio bassalo et prendiderunt ad Sancio et a Nunnu Gomiz ‘There came Gundesalvus and another vassal and apprehended Sancio and Nuño Gomez’ [Latin document from San Millán (Spain, year 930); Bastardas Parera 1953: 36]

Regarding the continuous expansion of DOM in Old Castilian, starting from an already compulsory use before object pronouns and (male) proper names (cf. Döhla 2014), together with the fact that bilingualism along with code-switching was a wide-spread phenomenon, makes a contact scenario between Andalusi Arabic and the expanding Castilian language the plausible source for the observed type of object marking in Andalusi Arabic.

6 Levantine and Iraqi Arabic Among the eastern dialects Levantine and Iraqi Arabic exhibit DOM. The oldest testimonies can be found in South Palestinian Christian Arabic texts (Blau 1983: 141).18 As far as Iraqi Arabic is concerned the earliest source is the vernacular poetry of Ṣafiyy al-Dīn al-Ḥillī (1278–ca. 1350): (27)

(28)

mā

li=šakr

NEG

ʕašaq=hā love: 3SG.M.PFV-3SG.F ‘He did not love Šakar’

OBJ=Šakar

allāh yi-ʕīn-ū li=qalb-ī ʕalà God 3SG.M.IPFV-help-3SG.M OBJ=heart-POSS.1SG against ˈllaḏī ʔa-hwā-hū REL.3SG.M 1SG.IPFV-love-3SG.M ‘May God help my heart against him whom I love’ [Levin 1975: 270]

As can be observed in (27) and (28), the formally marked direct object is also indexed at the verb by means of the enclitic personal pronoun. This clitic doubling has become a compulsory construction in the modern dialects of the Levante and Iraq, where, as far as 3rd person object referents are concerned, the || 18 The examples gathered by Blau (1983) do not display clitic doubling.

164 | Hans-Jörg Döhla

object-marking with l(a) only co-occurs together with bound person forms at the verb19 (cf. Levin 1987: 33, Siewierska 1999 and Haspelmath 2013): (29)

šif-t-u la=mḥammad il=yōm? see- 1/2SG.PFV-3SG.M OBJ=Muhammad DEF=day ‘Did you see Muhammad today?’ [Syrian Arabic; Brustad 2000: 354]

(30)

min ʕAlī šāf-u who Ali see: 3SG.PFV-3SG.M ‘Whom did Ali see?’ [Beirut Arabic; Koutsoudas 1967: 513]

la

OBJ

(31)

ṭird-ō-hum li=l=ʕummāl kull-hum fire-3PL.PFV-3PL.M OBJ=DEF=worker.PL all-3PL.M ‘they fired all the workers’ [Iraqi Arabic; Erwin 1963: 336]

At the same time, the most relevant factor determining the overt marking and indexing of the direct object seems to be the referential prominence of the object within the utterance, since inanimate objects are marked and indexed in the same way as long as they are definite: (32)

inti fataḥ-t-o la=l=ktāb? 2SG open-2SG.PFV-3SG.M OBJ=DEF=book ‘Did you ever open the book?’ [Galilean Arabic, Nazareth; Levin 1987: 35]

(33)

fallš-ō-ha li=l=madrasa l=ʕttīg-a tear.down-3PL.PFV-3SG.F OBJ=DEF=school DEF=old-SG.F ‘they tore down the old school’ [Iraqi Arabic; Erwin 1963: 336]

As opposed to other languages with DOM, the marking of (stressed) pronouns referring to previously mentioned objects is not as evident, since all Arabic varieties, including Maltese and Classical Arabic, have two sets of pronouns. One set consists of free independent personal pronouns which can never be preceded by a preposition. The second one consists of bound pronouns which are suffixed to the verb. These object pronouns are, except for the first person singular, identical with the possessive pronouns which can also be preceded by a preposition. Thus, in order to say ‘he saw him’ we can only use the indexed verb in order to refer to the object: šāf-a (Iraqi Arabic; Erwin 1963: 275). A construction like *šāf-a li=huwwa, using the independent personal pronoun ‘he’ preceded by the prepositional object marker li-, is ungrammatical. However, this type of

|| 19 This construction involving dependent and head marking (Nichols 1986) cannot be found in Classical Arabic.

The origin of differential object marking in Maltese | 165

construction can be made using a demonstrative pronoun as object referent instead of a personal pronoun: (34)

ma

a-ḥibb-a NEG 1SG.IPFV-like -3SG.M ‘I don’t like him’

l=hāḏa OBJ=DEM [Iraqi Arabic; Erwin 1963: 334]

Therefore it seems that DOM, even though contact induced, was not initiated with the personal pronouns (cf. the individuation scale in Table 2) in Eastern Arabic vernaculars. If a speaker really wants to emphasize the anaphoric object pronoun, there is the option of using the suffixed object pronoun twice in an utterance, at the verb and preceded by the preposition stem ʔil- instead of l(i)-. (35)

rād-ō-ha want-3PL.PFV-3SG.F ‘they wanted her’

ʔil-ha OBJ-3SG.F [Erwin 1963: 335]

Generally and repeatedly, the emergence of DOM in Levantine and Iraqi vernacular Arabic has been attributed to the influence of Aramaic which has been in contact with spoken Arabic for more than a millennium (Blau 1983, Borg 1996: 138, Río Sánchez 2013: 136, and Coghill 2014). There, too, definite topical objects receive an overt marking by means of the preposition l-: (36)

tāman qəḇar-ū l=ʕabraham w=l=sārā ʔā(n)ṯt-eh there bury-3PL.PFV OBJ=Abraham and=OBJ=Sarah wife-POSS.3SG.M ‘There they buried Abraham and his wife Sarah’ [Syriac translated from Hebrew, 2nd c.; Gen 49: 31; Peshitta]

Biblical and later Hebrew follows the same criterion for marking the object using the preposition ʔēṯ in most of the cases: (37)

šāmːā qāḇr-ū ʔǣṯ=ʕaḇrāhām wə=ʔēṯ śārā ʔišt-ō there bury-3PL.PFV OBJ=Abraham and=OBJ Sarah wife-POSS.3SG.M ‘There they buried Abraham and his wife Sarah’ [Biblical Hebrew; Gen 49: 31; BHS]

However the preposition lǣ- can also be found, especially in those biblical books where there is an Aramaic background: (38)

wā=ʔǣ-šləḥāh and=1SG.IPFV-send ‘and I sent Eliezer’

lǣ=ʔǣlīʕǣzǣr OBJ=Eliezer [Ezra 8, 16]

Summarizing it can be stated that DOM in Levantine and Iraqi Arabic is an Aramaic substrate due to a long and intensive language contact, and certainly to

166 | Hans-Jörg Döhla

the fact that both genetically related languages share a lot of structural features as well as phonological and morphosyntactic ones. So far we cannot see a link between the emergence of DOM in Maltese and in the Eastern Arabic vernaculars that can be historically proven. On the other hand, Levantine Arabic certainly exerted influence on Cypriot Arabic in developing DOM: (39)

lakay-t l=umm-ak l=imtine OBJ=mother-POSS.2SG DEF=Nicosia meet-1SG.PFV ‘I met your mother in Nicosia’ [Cypriot Arabic; Borg 1994: 41]

7 Siculo Arabic and Old Sicilian So far, we have shown that the development of DOM cannot be traced back to internal evolution in Classical as well as colloquial Arabic. Moreover for those dialects with DOM a language contact scenario can be reconstructed which provides a solid base for the explanation of the emergence of DOM as contactinduced replication. In the same way, as far as DOM in Maltese is concerned, the influence of a contact language is the most likely reason for the development of object marking. In the past, Siculo Arabic and Old Sicilian have been listed as possible contact languages, which is plausible if we look at the linguistic stratification on Malta across time: Tab. 4: The stratigraphy of the Maltese language (Brincat 2011: xxxv).

Strata

Period

Languages

adstratum 2

1800‒

English

adstratum 1

1530‒

Italian

superstratum

(1090)‒1184‒c. 1900

Sicilian

principal stratum

(870?)‒1048‒1241

(Siculo) Arabic

Especially (Old) Sicilian stands out if we consider the fact that it is one of the aforementioned Romance dialects which display DOM by internal development (cf. Bossong 1991 and 2008). It is even possible to assume that Siculo Arabic (cf. Agius 1996), which certainly entered Malta at some point, had developed DOM due to its intensive contact with Old Sicilian. Unfortunately, the only source for our knowledge about Siculo Arabic, the Taṯqīfu ‘l-lisāni wa-talqīḥ ‘l-ǧināni (‘The

The origin of differential object marking in Maltese | 167

correction of language and the pollination of the gardens’) by Abū Ḥafṣ ʕUmar Ibn Ḫalaf Ibn Makkī al-Ṣiqillī (died in 1107), does not mention any morphosyntactic or syntactic features, limiting its scope to phonological and phonetic matters. Moreover and as opposed to Andalusi Arabic, there is no vernacular Siculo Arabic poetry or other sources for the study of the structure of this particular Arabic vernacular. The other possible source for DOM in Maltese, namely Old Sicilian, is well documented and can be linked historically to the development of Maltese (cf. Table 4). According to Iemmolo (2009), based on a corpus study of 14th century Old Sicilian texts, stressed object pronouns are always marked by the preposition a, which also serves to mark the indirect object, in Old Sicilian: (40)

chach-au a mi de lu soy allipergu chase: 3SG.PST OBJ 1SG from DEF POSS.3SG.M room ‘He chased me out of his room’ [Old Sicilian; DIAL 1302‒1321; Iemmolo 2009]

Proper names are also marked in Old Sicilian, however they are also found unmarked depending on the topicality of the object referent: (41)

auchisi a lu dic-tu Pollidoru kill: 3SG.PST OBJ DEF say-PTCP.PST Pollidoru ‘He killed the aforesaid Pollidoru’ [Old Sicilian; ENE 1316‒37; Iemmolo 2009]

(42)

E Quintu Catulu, astuta-tu e auci-su and Quintus Catulus extinguish-PTCP.PST and kill-PTCP.PST ca appi Marcu Lepidu that have: 3SG.PST.IPFV Marcus Lepidus ‘and Quintus Catulus, after extinguishing and killing Marcus Lepidus’ [VAL MAX 87.113; Iemmolo 2009]

Furthermore, lexical NPs referring to human beings can also be marked if they are topically prominent within the sentence: (43)

Lu

Anchises happy mina-tu Anchises have: 3SG.PST.IPFV threaten-PTCP.PST a figl-u. OBJ POSS.3SG.M son-SG.M ‘Father Anchises had threatened his son.’ [Old Sicilian; ENE 1316‒37 ; Iemmolo 2009] DEF

patri father so

168 | Hans-Jörg Döhla

(44)

ki auchisi dui who kill: 3SG.PST two ‘who killed two of his sons’

soi

figl-i POSS.3PL.M son-PL.M [Old Sicilian; ENE 1316‒37; Iemmolo 2009]

In (44), it is quite clear that the object referents are definite but unspecific which leads the writer to omit the overt marking of the direct object. Summarizing Iemmolo’s (2009) analysis we get to the following distribution of DOM in 14th century Sicilian: Tab. 5: Grammaticalization of DOM in Old Sicilian 14th century according to Iemmolo (2009).

PP >

PN>

CN [+HUM] >

[+TOP] >

[+DEF] >

[‒TOP]

[‒DEF]^[+SPEC] >

CN [+ANIM] >

cn [‒ANIM]

[‒DEF]^[‒SPEC] > no arg.

Animate, but not human objects, are only marked sporadically (Iemmolo 2009: 2), but examples are not provided. Finally, Iemmolo, in a personal communication, made the observation that in his Sicilian mother tongue, the appearance of clitic doubling is quite frequent.20 As to our knowledge, there is no exhaustive study of the phenomenon (cf. Posner 1996: 168–169 and Ledgeway 2011: 434–439). But not only Old Sicilian can be taken into account as a trigger for the replication of DOM in Maltese. There is historical evidence that groups of Italian dialect speakers from the Abruzzi region were settled in Malta in the 13th century: The linguistic situation in Malta changed significantly in 1224 when Frederick II expelled large groups of Muslims and deported them to Lucera in Puglia. Moreover, in order to fill the void, Frederick sent over the inhabitants of Celano from central Italy. […] Historical references say that the town was razed to the ground and Tommaso [da Celano], his followers, and the whole population were sent to Malta and Sicily. (Brincat 2011: 69f.)

|| 20 The same phenomenon can be observed in various Spanish varieties in South America (cf. Belloro 2007 and 2015). Here we present an example from our own field work in La Paz, Bolivia: los quiero harto a mis hijos ‘I really love my children a lot’ (female middle-aged worker; La Paz 2012), where the pronominal clitic los co-occurs with the marked lexical NP a mis hijos.

The origin of differential object marking in Maltese | 169

Unfortunately we do not know how many people migrated to Malta and how many actually stayed there. But the Italo-Romance variety spoken by those new settlers also displayed DOM in the Middle Ages as can be seen in the following example from Rome (13th century), close by Celano: (45)

ke nullo Romano offendea ad that no Roman hurt: 3SG.PRS.SBJV OBJ ‘that no Roman should hurt another [Roman]’

l’-altro DEF-other

[Michel 1997: 273]

Furthermore, during the reign of Frederick II, there were about 150 army soldiers stationed on the Island who were Romance-speaking (Brincat 2011: 70). As a consequence of the demographic policy of the first half of the 13th century we register an increase of the Christian population of Malta whereas the Muslim population was declining and even ceasing to exist. This transition from a Muslim to a predominantly Christian society is accompanied at the colloquial level by the survival of the spoken Arabic vernacular “by adapting itself to the changes in religion, culture, and social life through internal simplification and autonomous divergence and by the steady adoption of Romance terms as the need arose” (Brincat 2011: 72). Thus, the superstrate situation left Romance structural elements within the newly adapted Arabic dialect.

8 Concluding remarks Summarizing it can be stated that DOM is not a genuine Arabic feature which emerged through internal development. Wherever DOM has appeared in Arabic vernaculars a clear language contact scenario can be determined where the contact language exhibits DOM. It has been due to the influence of this contact language that the replication of DOM was induced in the Arabic target vernacular: Iberian Romance > Andalusi Arabic; Aramaic > Iraqi Arabic/Levantine Arabic (> Cypriot Arabic); Old Sicilian > Siculo Arabic?; Old Sicilian > Maltese (cf. Map 1). However, we have to acknowledge that we can detect a certain predisposition for object marking in general within pan-Arabic grammar because of the use of the preposition li- for marking the indirect object and as a means of avoiding chains of rigid and inflexible genitive constructions. In the process of resolving these genitive chains, which, due to their grammatical constraints, may impede the successful transmission of the intended information, even direct objects may be preceded by li-. But this preceding preposition may not be understood as an object marker in general and less as a differential marker in

170 | Hans-Jörg Döhla

particular. Rather, its use is conditioned by the speaker’s or writer’s desire to clearly transmit the intended information or by stylistic considerations. Thus, as for the emergence of DOM in Maltese, we hold that Old Sicilian influence leading to the replication of the special type of object marking is the most probable scenario.

Map 1: DOM in Romance languages and spoken Arabic varieties of the Mediterranean.

Acknowledgements: I am indebted to the Swiss National Fond (SNF) for granting me a 3-year post.doc mobility scholarship (2011–2014) in order to pursue the research project “Differential Object Marking in Situations of Language Contact between Romance and non-Romance Languages”. I also thank the CNRS, ParisVillejuif (group SeDyL, FR 3326, especially Isabelle Léglise and Claudine Chamoreau) for hosting me as chercheur invité during that period.

Abbreviations 1, 2, 3 ACC ANIM CN

1st, 2nd, 3rd person accusative animate common noun

OBJ PFV PL PN

object perfective plural proper noun

The origin of differential object marking in Maltese | 171

DEF DEM F GEN HUM IPFV INDEF INF M NEG NOM

definite demonstrative feminine genitive human imperfective indefinite infinitive masculine negation nominative

POSS PP PRS PST PTCP REL SBJV SG SPEC TOP

possessive personal pronoun present past participle relative pronoun subjunctive singular specific topic

References Agius, Dionisius A. 1996. Siculo Arabic. London/New York: Kegan Paul. Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language & Linguistic Theory 21. 435–483. Arberry, Arthur John & Prospero Grech. 1961. Dun Karm. Poet of Malta. Cambridge: Cambridge University Press. Bastardas Parera, Joan. 1953. Particularidades sintácticas del latín medieval. Cartularios españoles de los siglos VIII al XI. Barcelona/Madrid: Escuela de Filología. Belloro, Valeria. 2007. Spanish clitic doubling: A study of the syntax-pragmatics interface. Ph.D. University of New York at Buffalo. Belloro, Valeria. 2015. To the right of the verb: An investigation of clitic doubling and right dislocation in three Spanish dialects. Cambridge: Cambridge Scholars Publishing. Blau, Joshua. 1983. The influence of living Aramaic on ancient South Palestinian Christian Arabic. In Michael Sokoloff (ed.), Arameans, Aramaic and the Aramaic literary tradition, 141‒143. Ramat-Gan: Bar-Ilan University Press. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London/New York: Routledge. Borg, Albert & Bernhard Comrie. 1984. Object diffuseness in Maltese. In Frans Plank (ed.), Objects, 109‒126. London et al.: Academic Press. Borg, Albert & Manwel Mifsud. 2002. Maltese object marking in a Mediterranean context. In Paolo Ramat & Thomas Stolz (eds.), Mediterranean languages. Papers from the MEDTYP workshop, Tirrenia, June 2000, 33‒45. Bochum: Brockmeyer. Borg, Alexander. 1994. Some evolutionary parallels and divergences in Cypriot Arabic and Maltese. Mediterranean Language Review 8. 41–67. Borg, Alexander. 1996. On some Levantine linguistic traits in Maltese. Israel Oriental Studies 16. 133–152. Bossong, Georg. 1982. Historische Sprachwissenschaft und empirische Universalienforschung. Romanistisches Jahrbuch 33. 17–51. Bossong, Georg 1985. Empirische Universalienforschung. Differentielle Objektmarkierung in den Neuiranischen Sprachen. Tübingen: Narr.

172 | Hans-Jörg Döhla

Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Douglas Kibbee & Dieter Wanner (eds.), New analyses in Romance linguistics, 143–170. Amsterdam/ Philadelphia: John Bemjamins. Bossong, Georg. 2008. Die romanischen Sprachen. Eine vergleichende Einführung. Hamburg: Buske. Brincat, Joseph M. 2011. Maltese and other languages. A linguistic history of Malta. La Valetta: Midsea Books. Brustad, Kristen. 2000. The syntax of spoken Arabic. A comparative study of Moroccan, Egyptian, Syrian, and Kuwaiti Arabic. Washington, D.C.: Georgetown University Press. Čéplö, Slavomír. 2013. An overview of object reduplication in Maltese. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 201–224. Berlin: Akademie Verlag. Chafe, Wallace L. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Charles N. Li (ed.), Subject and topic, 25–55. New York: Academic Press. Coghill, Eleanor. 2014. Differential object marking in Neo-Aramaic. Linguistics 52(2). 335–364. Comrie, Bernhard. 21989. Language universals and linguistic typology. Chicago: The University of Chicago Press. Corriente, Federico. 1977. A grammatical sketch of the Spanish Arabic dialect bundle. Madrid: Instituto Hispano-Árabe de Cultura. Corriente, Federico. 1997. Poesía dialectal árabe y romance en Alandalús. Madrid: Gredos. Corriente, Federico. 2008. Code-switching and code-mixing in Ibn Quzmān revisited. In Hans-Jörg Döhla, Raquel Montero Muñoz & Francisco Báez de Aguilar González (eds.), Lenguas en diálogo. El iberorromance y su diversidad lingüística y literaria. Ensayos en homenaje a Georg Bossong, 65–86. Frankfurt/Madrid: Vervuert/Iberoamericana. Dalrymple, Mary & Irina Nikolaeva. 2011. Objects and information structure. Cambridge: Cambridge University Press. DIAL = 1933 [1302‒1321]. Libru de lu dialagu de Sanctu Gregoriu traslatu di frati Giuvanni Campulu di Missina. A cura di Salvatore Santangelo. Palermo: Tipografia Boccone del povero. Dixon, Robert M. W. 1989. Subject and object in universal grammar. In Doug Arnold, Martin Atkinson, Jacques Durand, Claire Grover & Louisa Sadler (eds.), Essays on grammatical theory and universal grammar, 91–118. Oxford: Clarendon Press. Döhla, Hans-Jörg. 2004. Der gehäufte pleonastische Gebrauch des Possessivpronomens in der altspanischen Konstruktion des Typus SU FIJO DEL RREY – ein syntaktischer Arabismus? In Gero Arnscheidt, Elmar Eggert, Claire-Marie Jeske, Rosanna Pardellas Velay, Stefanie Zaun & Christian Zwinkl (eds.), Enthüllen – Verhüllen. Text und Sprache als Strategie. Beiträge zum 19. Forum Junge Romanistik, 217–229. Bonn: Romanistischer Verlag. Döhla, Hans-Jörg. 2014. Diachronic convergence and divergence in differential object marking between Spanish and Portuguese. In Kurt Braunmüller, Steffen Höder & Karoline Kühl (eds.), Stability and divergence in language contact. Factors and mechanisms, 265–289. Amsterdam/Philadelphia: Benjamins. ENE = 1956 [1316‒1337]. La istoria de Eneas vulgarizzata per Angilu di Capu. A cura di Gianfranco Folena. Palermo: Centro de studi filologici e linguistici siciliani. Erwin, Wallace M. 1963. A short reference grammar of Iraqi Arabic. Washington D.C.: Georgetown University Press. Fabri, Ray. 1993. Kongruenz und die Grammatik des Maltesischen. Tübingen: Niemeyer.

The origin of differential object marking in Maltese | 173

Fischer, Wolfdietrich. 1972. Grammatik des klassischen Arabisch. Wiesbaden: Harrassowitz. García García, Marco. 2010. Differentielle Objektmarkierung bei unbelebten Objekten im Spanischen. Dissertation Universität Freiburg i. Br. Givón, Talmy. 1976. Topic, pronoun, and grammatical agreement. In Charles N. Li (ed.), Subject and topic, 149–188. New York: Academic Press. Haspelmath, Martin. 2013. Argument indexing: A conceptual framework for the syntax of bound person forms. In Dik Bakker & Martin Haspelmath (eds.), Languages across boundaries: Studies in memory of Anna Siewierska, 197–226. Berlin: De Gruyter Mouton. Heusinger, Klaus von. 2008. Verbal semantics and the diachronic development of differential Object Marking in Spanish. Probus 20. 1–31. Hoop, Helen de & Peter de Swart (eds.). 2009. Differential subject marking. Dordrecht: Springer. Iemmolo, Giorgio. 2009. La marcatura differenziale dell’oggetto in siciliano antico. Archivio Glottologico Italiano 94(2). 185‒225.21 Iemmolo, Giorgio. 2010/11. Towards a typological study of differential object marking and differential object indexation. Dissertation Università degli Studi di Pavia. Keenan, Edward L. 1976. Towards a universal definition of subject. In Charles N. Li (ed.), Subject and topic, 303–333. New York: Academic Press. Khan, G.A. 1984. Object markers and agreement pronouns in Semitic languages. Bulletin of the School of Oriental and African Studies, University of London, 47(3). 468‒500. Kontzi, Reinhold. 1999. Wort und Schrift. Des kanonikus Fortunato Panzavecchia Bibelübersetzung ins Maltesische, nach den Handschriften des Kathedralarchivs in Mdina. Tübingen: Narr. Koutsoudas, Andreas. 1967. Object particles in Lebanese. Journal of the American Oriental Society 87(4). 512‒517. Ledgeway, Adam. 2011. Syntactic and morphosyntactic typology and change. In Martin Maiden, John Charles Smith & Adam Ledgeway (eds.), The Romance languages. Volume I: Structures, 382‒471. Cambridge: Cambridge University Press. Levin, Aryeh. 1975. The vernacular poetry of Ṣafiyy al-Dīn al-Ḥillī: A source for the reconstruction of Iraqi Arabic in the 14th century. Israel Oriental Studies 5. 259–276. Levin, Aryeh. 1987. The particle ‘la’ as an object marker in some Arabic dialects of the Galilee. Zeitschrift für arabische Linguistik 17. 31‒40. Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press. Michel, Andreas. 1997. Einführung in das Altitalienische. Tübingen: Narr. Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 62. 56‒119. Pensado Ruiz, Carmen. 1995. La creación del complemento directo preposicional y la flexión de los pronombres personales en las lenguas románicas. In Pensado Ruiz, Carmen (ed.), El complemento directo preposicional, 179–230. Madrid: Visor Libros. Posner, Rebecca. 1996. The Romance languages. Cambridge: Cambridge University Press. Río Sánchez, Francisco del. 2013. Influences of Aramaic on dialectal Arabic. In Pedro Juan Monferrer-Sala & Wilfred G. E. Watson (eds.), Archaism and innovation in the Semitic languages. Selected papers, 129–136. Córdoba: Oriens Academic.

|| 21 I have only seen the draft of this paper. This is why I do not indicate the exact page number when citing examples.

174 | Hans-Jörg Döhla

Siewierska, Anna. 1999. From anaphoric pronoun to grammatical agreement marker: Why objects don’t make it. Folia Linguistica 33(1/2). 225–252. VAL MAX 1321/37 = Valeriu Maximu translatatu in vulgar messinisi per Accursu di Cremona, a cura di Francesco A. Ugolini, 2 vols. Palermo : Centro di studi filologici e linguistici siciliani 1967. Vella, Olvin. 2006. L-għejun ta’ l-istorja ta’ l-ilsien Malti. Ġabra ta’ dokumenti bil-Malti mill-bidu sa Vassalli. MA-thesis University of Malta. Weber, Éloide. 2010. La syntaxe de l’objet en espagnol: la question de la préposition. Limoges: Lambert-Lucas. Wright, William. 1898. A grammar of the Arabic language. Vol. II. Cambridge: Cambridge University Press. Yamamoto, Mutsumi. 1999. Animacy and reference. Amsterdam/Philadelphia: Benjamins. Zammit, Martin R. 2009/10. Andalusi Arabic and Maltese: A preliminary survey. Folia Orientalia 45/46. 12‒60.

David Wilmsen

Polar interrogative -š in Maltese: Developments and antecedents Abstract: The earliest Maltese grammars of the 18th and 19th-centuries attest a polar interrogative enclitic -š. An early 20th-century grammar shows it retaining an interrogative function and marking indirect questions. By the late 20th century, grammars either do not address it or they attest it as being largely restricted to an erstwhile 3rd-person copular pronoun hux in polar interrogatives and indirect questions. Nevertheless, the MLRS Corpus shows it functioning in all personal pronouns and in verbs in indirect and rhetorical questions. Its preservation in Maltese provides a clue to the origin of a similar feature in Levantine, North African and Andalusi Arabics. Keywords: Arabic dialects, copula, interrogatives, personal pronouns, polar questions

1 Introduction In their grammar of twentieth-century Maltese, Borg and Azzopardi-Alexander (1997: 3–5) discuss polar interrogatives, questions whose answer is an expected ‘yes’ or ‘no’, noting that these are usually indicated by a rising tone of voice:1 (1a)

(1b)

L-arloġġ qiegħed fuq il-mejda the-watch located on the-table ‘The watch is on the table’ L-arloġġ qiegħed fuq il-mejda? ‘Is the watch on the table?’

An optional, less frequent manner of forming a polar interrogative is to place an interrogative derived from an erstwhile masculine 3rd-person pronoun huwa

|| 1 Borg and Azzopardi-Alexander present diagrams of the tone contours. For our purposes, the tone of voice will be indicted by the question mark. || David Wilmsen: American University of Beirut, Department of Arabic and Near Eastern Languages, P.O. Box 11-0236, Riad El Solh, Beirut 1107 2020, Lebanon. E-mail: [email protected]

176 | David Wilmsen

either at the beginning of the sentence or as a tag at the end (when it becomes hu). In this function, this erstwhile personal pronoun becomes a copular interrogative, meaning ‘is it?’ in cleft questions and tags: (2a)

(2b)

Huwa l-arloġġ qiegħed fuq il-mejda ? is.it( disat iħmir ‘nine donkeys’ fenek ‘rabbit’, PL fniek > għaxart ifniek ‘ten rabbits’

As noted above, connecting /t/, despite being treated by Maltese’s somewhat conservative orthography as a suffix on the numeral, is more accurately viewed as a prefix on the following plural noun (cf. Schabert 1976: 204). This is widely accepted in descriptions of the cognate element in Arabic dialects (e.g. Levin 2003). The basic piece of evidence supporting this view, as pointed out by Hoberman (2007: 278) for Maltese, has to do with stress. The numerals ‘five’, in the form spelt ħamest, and ‘ten’, in the form spelt għaxart, are both pronounced with stress on the initial syllable. If connecting /t/ were genuinely the final segment of these phonological words, then stress should be on the final CVCC syllable of the numeral in each case. If, however, connecting /t/ belongs to the following phonological word, then initial stress in the numeral is as expected. There is another piece of evidence, however, that concerns the epenthetic /i/. If connecting /t/ were genuinely a suffix on the numeral, epenthesis at the start of the following word would be surprising, since CCC and even CCCC clusters across word boundaries do not otherwise trigger epenthesis in Maltese, unless the second word begins with a SONORANT-CONSONANT sequence (e.g. Stolz et al.’s 2011: 265–267 examples rqiq irqiq ‘very narrow’ vs. fost sħabhom ‘among their comrades’). /i/-epenthesis after numerals would therefore appear to have no synchronic phonological motivation. If connecting /t/ is a prefix on the following CCinitial plural noun, however, then /i/ epenthesis in a tCC sequence is clearly motivated, since CCC onsets are not generally licit in Maltese (except for /s/-STOP-LIQUID sequences). Where CCC onsets arise through other morphological processes, for example prefixing of the definite article l- to a CC-initial noun such as bliet ‘towns’, epenthesis is obligatorily triggered, usually breaking up the first two consonants in the cluster, as in l-ibliet ‘the towns’. /i/-epenthesis is therefore fully expected if connecting /t/ is a prefix on the following plural noun. Returning to the question of conditions on /t/-insertion, it is clear that it is at least possible in contexts other than with CC-initial monosyllabic plurals. Possibly relevant properties of plurals that have been identified in the literature

Connecting /t/ in Maltese numerals | 273

include type of onset, number of syllables, position of stress, whether the plural is ‘broken’ (i.e. stem-altering) or ‘sound’ (i.e. suffixing), and whether it is Arabic- or Romance-derived. Concerning most of these factors there is remarkably little agreement. The one point on which all the most detailed accounts agree is that /t/-insertion is optional with at least some disyllabic CC-initial plurals (Aquilina 1965: 118, Borg 1974: 296, Fabri 1994: 230–231, Ambros 1998: 91, Hoberman 2007: 277–278), as in (7). (7a) (7b) (7c)

żiemel ‘horse’, PL żwiemel > żewġ żwiemel ~ żewgt iżwiemel ‘two horses’ kamra ‘room’, PL kmamar > erba’ kmamar ~ erbat ikmamar ‘four rooms’ barmil ‘bucket’, PL bramel > tmien bramel ~ tmint ibramel ‘eight buckets’

Borg (1974: 296) states, however, that sound plurals never trigger /t/-insertion, even in cases of CC-initial disyllabic plurals, as in (8). (8)

stampa ‘picture’, PL stampi > għaxar stampi ‘ten pictures’ (Ungrammatical according to Borg: għaxart istampi)

Ambros (1998: 91) makes a similar point, as does Fabri (1994: 231), though the latter frames matters in terms of stress, not the sound/broken-plural distinction, claiming that /t/-insertion is never licit with CC-initial di- or polysyllabic plurals that have non-initial stress. He gives the following example: (9)

trakk ‘truck’, PL trakki'jiet > żewġ trakkijiet ‘two trucks’ (Ungrammatical according to Fabri: żewġt itrakkijiet)

Fabri acknowledges, however, that there are also CC-initial plurals with initial stress, such as stampi in (8), for which /t/-insertion appears to be ungrammatical, and he further claims that the broken plural ġranet ‘days’ falls into this category: (10)

ġurnata ‘day’, PL ġranet > żewġ ġranet ‘two days’ (Ungrammatical according to Fabri: żewġt iġranet)

He is then careful to deny the possible conclusion from these allegedly ungrammatical examples that /t/-insertion is only licit with Arabic- rather than Romance- (or English-)derived nouns, giving the example of skejjel ‘schools’ (SG skola), for which he notes that /t/-insertion is optional.

274 | Christopher Lucas and Michael Spagnol

None of the aforementioned authors claim that /t/-insertion is illicit with non-Arabic-derived nouns across the board, but Ambros (1998: 91) and Hoberman (2007: 277) suggest that, for vowel-initial plurals, /t/-insertion is only licit with Arabic-derived nouns, such as ilsna ‘tongues, languages’ and aħwa ‘siblings’. Aquilina (1965: 118), on the other hand, suggests that /t/-insertion is licit with any vowel-initial plural, while Borg (1974: 294) claims the exact opposite: that plurals beginning with a genuine (not epenthetic) vowel are, in fact, ungrammatical with connecting /t/. Since the vast majority of monosyllabic plurals in Maltese start with a consonant cluster, it is rarely considered necessary to specifically address the question of which factor is more important in the ‘core’ context for /t/-insertion: monosyllabicity of a plural form or starting with a consonant cluster. An exception is Ambros (1998: 91), who claims that the only relevant criterion in this case is number of syllables, saying simply that “if the following plural is monosyllabic […], then the t-form is used” (CL & MS’s translation). Since it happens that there are no plural nouns in Maltese that are monosyllabic and vowel-initial, this amounts to a claim that monosyllabic plurals beginning with either one or two consonants necessarily trigger /t/-insertion. Ambros is certainly aware of this, since one of the examples he gives of monosyllabic plurals which necessarily trigger /t/-insertion is jiem ‘days’ (SG jum). However, what Ambros was perhaps not conscious of in formulating this rule is that there are a number of other monosyllabic plurals beginning with a single consonant, including xhur ‘months’ (SG xahr) with a silent , which we already noted (fn. 4) was claimed by Cremona (1938) not to permit /t/-insertion, as well as a number of Englishderived sound plurals in -s, such as films ‘films, movies’ (SG film). For his part, Fabri (1994: 231) identifies jiem ‘days’ as the only plural he is aware of that does not start with a consonant cluster and yet seems to force /t/-insertion. It seems likely, therefore, that the generalization that all monosyllabic plurals trigger /t/insertion is too strong as it stands.

3 Experiment design The previous section has shown that a number of factors potentially determining /t/-insertion, all having to do with properties of the following plural noun, have been discussed in the literature: type of onset, number of syllables, position of stress, whether the plural is broken or sound, and whether it is Arabic- or Romance-derived. There is, however, little agreement concerning which of these factors are essential and which are secondary or irrelevant, and in the case of

Connecting /t/ in Maltese numerals | 275

vowel-initial plurals there is outright disagreement as to whether this type of onset favors or disfavors /t/-insertion. Moreover, it seems highly likely that one or more factors not mentioned thus far could also play a role. For example, if type of onset of the plural noun (V-initial, C-initial or CC-initial) turns out to be a relevant factor, it seems at least plausible that type of coda of the preceding numeral (ending in a vowel, sonorant, or obstruent) should also be relevant. Additionally, at least in those contexts where the literature reports optionality (e.g. with CC-initial disyllabic broken plurals), there will likely be an effect of the ‘string frequency’ (in the sense of Krug 1998) of collocations of 2–10 numerals with specific plural nouns. Our position is that the most fruitful way to shed light on these issues is, as a first step, to conduct production experiments with native speakers, such experiments being a reasonable approximation to observing subjects’ actual speech habits. While we would ideally like to have tested each of the aforementioned potentially relevant factors in a single experiment, this would have necessitated an impractically large number of test items. We therefore took the decision to focus on the factors i) choice of numeral, ii) onset of plural, and iii) number of syllables of plural, with a view to arriving at a near-definitive statement of the contribution of each of these. While this means that other potential factors cannot be tested directly, we nevertheless took care to include a range of different plural types among our test items, including representatives of various classes of broken and sound plurals with differing stress patterns, and, where possible, a balance of Arabic- and Romance-derived plurals. We also chose to include certain items, such as snin ‘years’, jiem ‘days’, ġranet ‘days’, skejjel ‘schools’, stampi ‘pictures’, trakkijiet ‘trucks’ and others, about which specific statements concerning the possibility of /t/-insertion have been made in the literature. In total, 56 plurals were tested (see Table 2), with seven distinct plurals in each of the following eight categories: mono- di- and polysyllabic (3 or more syllables) plurals with CC onsets, mono- di and polysyllabic plurals with singleconsonant onsets, and di- and polysyllabic plurals with vocalic onsets. (Recall that Maltese has no nouns whose plurals are monosyllabic and begin with a vowel). The test items consisted of a pairing of the singular form of each noun from each of the eight categories with one of seven numerals, presented to subjects as figures: 2, 4, 5, 7, 8, 9, 10. The order of the test items was randomized and then presented in the same (randomized) order to each subject. Between each test item was a filler, consisting of a pairing of a singular noun with a numeral, written in figures, between 11 and 19. Note in this connection that nouns modified by numerals above ten remain singular in Maltese, and are therefore

276 | Christopher Lucas and Michael Spagnol

never associated with /t/-insertion. The task subjects were asked to carry out was, for each stimulus they were presented with, to pronounce what they saw in words, as they would normally say it in connected speech. Tab. 2: Plurals tested, by onset and number of syllables.

Monosyllabic CC-initial fniek

Polysyllabic

‘rabbits’

bramel

‘buckets’ kmandamenti

‘commandments’

klieb

‘dogs’

platti

‘plates’

trakkijiet

‘trucks’

bniet

‘girls’

stampi

‘pictures’ flokkijiet

‘shirts’

djar

‘houses’

ġranet

‘days’

‘duties’

ħbieb

‘friends’

skejjel

‘schools’ ħsibijiet

‘thoughts’

snin

‘years’

ljieli

‘nights’

żminijiet

‘times’

bwiet

‘pockets’

kmamar

‘rooms’

studenti

‘students’

CV-initial files

V-initial

Disyllabic

dmirijiet

‘files’

kotba

‘books’

neputijiet

‘grandchildren’

films

‘films’

naħat

‘sides’

larinġiet

‘oranges’

xhur

‘months’

bozoz

‘bulbs’

pappagalli

‘parrots’

jiem

‘days’

widnejn

‘ears’

kategoriji

‘categories’

gowls

‘goals’

żgħażagħ

‘youths’

pajjiżi

‘countries’

fonts

‘fonts’

diski

‘disks’

postijiet

‘places’

toasts

‘pieces of toast’

kelmiet

‘words’

verżjonijiet

‘versions’

ilsna

‘tongues’ arloġġi

‘clocks’

erwieħ

‘souls’

ajruplani

‘aeroplanes’

ulied

‘sons’

operazzjonijiet ‘operations’

uċuħ

‘faces’

għasafar

‘birds’

idejn

‘hands’

appartamenti

‘appartments’

aħwa

‘siblings’ artikli

‘articles’

oqsma

‘fields’

‘individuals’

individwi

35 subjects were recruited (18 female, 17 male), all of whom were students at the University of Malta. Subjects were divided into seven groups of five each. Within a single group, all five subjects were presented with identical stimuli. Between groups, while all subjects were exposed to the same test nouns and fillers in the same order, the combination of numeral and noun in the test items varied, such that every test noun was combined once with every test numeral across the seven groups. The purpose of this element of the design was to allow observa-

Connecting /t/ in Maltese numerals | 277

tion of the effect of numeral choice while holding all other factors other than ‘subject group’ constant. While testing every single subject’s response to all seven numerals in combination with all test nouns would have been ideal, this would have resulted in in 392 test items for every subject, making the experiment impractically long. The reason that ‘three’ (bare form tliet, /t/-form tlitt/tlett) and ‘six’ (both bare and /t/-forms sitt) were not included among the test numerals is that, for the former, the phonetic distinction in context between the bare and /t/-forms is too slight to be able to confidently judge aurally, while for the latter there is no distinction between the forms. For illustration, the first eight stimuli (four fillers and four test items) presented to Groups 1 and 2 are given in (11) and (12), respectively, with test items in boldface. (11)

Group 1 sample stimuli 12 qasba 2 tifla 13 bandiera 7 artiklu 15 għalqa 10 larinġa 11 bejt 5 ħsieb

(12)

Group 2 sample stimuli 12 qasba 4 tifla 13 bandiera 8 artiklu 15 għalqa 2 larinġa 11 bejt 7 ħsieb

278 | Christopher Lucas and Michael Spagnol

4 Results and discussion 4.1 Results in outline The overall results for each of the eight plural types tested are given in Table 3 and illustrated in Figure 1 below. These results show that both onset type and number of syllables are indeed highly relevant to the occurrence of /t/-insertion. Analyzing the data using a linear mixed effects model (with random intercept and slope for test subjects but not test items) reveals a main effect of both these conditions (for onset, z = -8.05, p < 0.001; for number of syllables, z = -11.65, p < 0.001). Numeral choice, on the other hand, seems to be not or hardly relevant. There is no main effect of numeral choice, though there does seem to be a weak interaction between all three conditions (z = 2.09, p = 0.04; see Section 4.5 for further discussion). Tab. 3: Overall /t/-insertion responses by onset and number of syllables.5

Monosyllabic

Disyllabic

Polysyllabic

80% (177/221)

53% (123/230)

2% (4/236)

CV onsets

15% (30/198)

0% (1/209)

0% (1/241)

V onsets

/

40% (93/230)

11% (31/244)

CC onsets

4.2 Implications for previous claims We saw above that monosyllabic CC-initial plurals are widely agreed to be a core context for /t/-insertion. Unsurprisingly, this is strongly supported by our findings: test items in this category triggered /t/-insertion 80% of the time. We also saw that all the most detailed existing descriptions of /t/-insertion agree that it may occur with at least some disyllabic CC-initial plurals. This too is borne out by the data: /t/-insertion occurred 53% of the time with these items.

|| 5 The figures in brackets show instances of /t/-insertion as a proportion of the total number of valid responses for test items in that category. Valid responses are those in which the target plural form was supplied either with or without /t/. Responses in which a non-target plural form was supplied, or it was unclear whether or not /t/ and been inserted, were discounted.

Connecting /t/ in Maltese numerals | 279

200 180 160 140 120

Monosyllabic

100

Disyllabic

80 Polysyllabic

60 40 20 0 CC onsets

CV onsets

V onsets

Fig. 1: Overall /t/-insertion responses by onset and number of syllables.

On the other hand we find that the CC-initial polysyllabic plurals in our sample (those with three or more syllables) virtually never triggered /t/-insertion. There is thus a very clear inverse correlation between number of syllables and prevalence of /t/-insertion, at least as far as CC-initial plurals are concerned. We should be cautious, however, about assuming a necessarily causal link between number of syllables and (lack of) /t/-insertion. Recall, for example, Fabri’s (1994: 231) claim that /t/-insertion is never licit with CC-initial plurals where stress is on a noninitial syllable. Our findings are equally consistent with this generalization, since it happens, in fact, that virtually all CC-initial plurals in Maltese with three or more syllables (and all those among our test items) have non-initial stress.6 Similarly, we noted above Borg’s (1974: 296) claim that sound plurals never trigger /t/insertion, irrespective of other features they might have. This claim too is largely borne out by our results (see below for more discussion), since it also turns out that all CC-initial Maltese plurals with three or more syllables are sound, not broken. We should therefore be open to the possibility that the apparent relevance of number of syllables to the prevalence of /t/-insertion is in fact epiphenomenal, and that other factors, such as stress patterns, the sound/broken-plural distinc-

|| 6 The converse is not true: a few CC-initial plurals with non-initial stress (all of them sound rather than broken) have only two syllables, e.g. dħulin ‘friendly (people)’. No such plurals were included among our test items, however.

280 | Christopher Lucas and Michael Spagnol

tion, or even token and string frequency, all of which happen to correlate with number of syllables, may play a more important role.7 We leave a more thorough investigation of this point for future work. Matters with V-initial plurals are less clear cut, though we can say one thing with confidence: the data do not support Borg’s (1974: 294) claim that /t/insertion is straightforwardly ungrammatical with these. That Borg should have made such a claim is perhaps not so surprising, however, given that V-initial plurals seem to be overall a less favorable context for /t/-insertion than CCinitials (especially of the monosyllabic variety). Interestingly, though, while disyllabic V-initials seem to favor /t/-insertion to approximately the same extent as disyllabic CC-initials, we do not see the same (near-)categorical absence of connecting /t/ with polysyllabic V-initials as we do with the polysyllabic CCinitials. Instead, /t/-insertion seems to be a marginal but genuine possibility with polysyllabic V-initials, at least for some speakers. The clearest results of all come from the CV-initial plurals. With these we can say that /t/-insertion essentially never occurs, regardless of number of syllables, except that monosyllabic CV-initials seem to weakly favor /t/-insertion. If we consider the per-item results given in Table 4, however, we will see that this last impression is misleading, and that CV-initials should be viewed as incompatible with /t/-insertion across the board. Tab. 4: Per-item /t/-insertion responses.

CC-initial

CV-initial

Monosyllabic

/t/-insertion Disyllabic

/t/-insertion Polysyllabic /t/-insertion

fniek

94% (33/35) bramel

74% (25/34) kmandamenti 6% (2/34)

klieb

97% (34/35) platti

7% (2/30)

trakkijiet

0% (0/35)

6% (2/33)

stampi

bniet

88% (15/17)

flokkijiet

3% (1/35)

djar

94% (33/35) ġranet

77% (23/30) dmirijiet

0% (0/32)

ħbieb

94% (33/35) skejjel

56% (19/34) ħsibijiet

3% (1/34)

snin

10% (3/30)

80% (28/35) żminijiet

0% (0/32)

bwiet

76% (26/34) kmamar

71% (24/34) studenti

0% (0/35)

files

0% (0/33)

kotba

0% (0/33)

neputijiet

0% (0/35)

films

0% (0/33)

naħat

0% (0/24)

larinġiet

0% (0/33)

xhur

3% (1/30)

bozoz

0% (0/33)

pappagalli

3% (1/35)

jiem

97% (29/30) widnejn

0% (0/26)

kategoriji

0% (0/35)

ljieli

|| 7 See Zipf (1935) on the (inverse) correlation between number of syllables and token frequency.

Connecting /t/ in Maltese numerals | 281

Monosyllabic

/t/-insertion Disyllabic

/t/-insertion Polysyllabic /t/-insertion

gowls

0% (0/30)

żgħażagħ

3% (1/35)

fonts

0% (0/16)

diski

toasts

0% (0/26)

kelmiet

V-initial

pajjiżi

0% (0/35)

0% (0/35)

postijiet

0% (0/34)

0% (0/23)

verżjonijiet

0% (0/34)

ilsna

33% (11/33)

arloġġi

9% (3/35)

erwieħ

84% (27/32) ajruplani

ulied

45% (14/31)

operazzjonijiet 3% (1/34)

uċuħ

24% (8/34)

għasafar

idejn

26% (9/34)

appartamenti 6% (2/35)

aħwa

45% (15/33)

artikli

9% (3/35)

oqsma

27% (9/33)

individwi

9% (3/35)

20% (7/35)

34% (12/35)

4.3 Exceptional items From Table 4 it can be seen that the results for individual items are rather uniform within each of the eight categories, with the notable exception of the five items (boldfaced in Table 4) jiem ‘days’, snin ‘years’, erwieħ ‘souls’, stampi ‘pictures’ and platti ‘plates’. Let us consider CV-initial jiem ‘days’ first of all. We saw above that both Ambros (1998: 81) and Fabri (1994: 231) view /t/-insertion as obligatory with jiem, and Fabri specifically mentions that jiem is the only item he is aware of that does not start with a consonant cluster but for which /t/insertion is obligatory. We decided to include jiem in the study in order to test these claims and with a 97% /t/-insertion rate for this item, we can say that the claims are clearly correct. But when we compare the rate of /t/-insertion for this item with that of all the other CV-initial plurals – which is at or very close to zero across the board – it is clear that jiem is truly exceptional in its category, and its inclusion among the test items distorts what would otherwise be a very clear finding: /t/-insertion does not occur with CV-initial plurals of any kind, whether broken or sound, Arabic- or Romance-derived, and so on. It is therefore worth reassessing our results with this item excluded. First though, let us consider snin ‘years’, for which the picture is very similar, albeit in reverse. We noted above that snin was already identified by Cremona (1938: 204–205) as being exceptional among CC-initial monosyllabic plurals in not permitting /t/-insertion. We also saw that snin with a preceeding /t/-form numeral does not occur in the MLRS corpus (fn. 3). It is not surprising, then, that the figures for /t/-insertion with snin in our study are so low (though given its categorical absence with this item in the MLRS corpus, it is rather surprising

282 | Christopher Lucas and Michael Spagnol

that the figure in our study was as high as 10%). Since snin, like jiem, is thus clearly exceptional in its category (the mean rate of /t/-insertion for the other monosyllabic CC-initials is 91% and the range is 21), our reassessment of the results should also exclude this item. In case there is doubt that excluding these two items is justified, note that there is a straightforward historical explanation for the anomalous behavior of both snin and jiem. In the early Semitic cognate of this construction, /t/-insertion occurred whenever a numeral modified a masculine noun (and only then), while in the spoken Arabic variety from which Maltese more immediately descends, /t/insertion occurred only with nouns (of both genders) whose plurals belonged to a subset of the vowel-initial broken-plural patterns. The Classical Arabic cognate of Maltese snin (SG sena) is feminine and this item was originally CV-initial (sinīn). It thus never triggered /t/-insertion at any pre-Maltese stage. The Classical Arabic cognate of jiem (SG jum), on the other hand, is masculine, and in Maltese this item has undergone phonological reduction from an original form ayyām~iyyām (i.e. a /t/-inserting pattern in dialectal Arabic). In pre-Maltese it would have therefore always have triggered /t/-insertion. Phonological change has led to these items becoming monosyllabic CC-initial in the case of snin and CV-initial in the case of jiem, but as forms which collocate especially frequently with numerals ‘ten’ and below, they have resisted the expected analogical pressure to bring their /t/insertion behavior into conformity with that of other plurals with these onsets (cf. Hooper 1976). In other words, snin is synchronically CC-initial but still behaves as the CV-initial it originally was, whereas jiem is synchronically CV-initial but still behaves as the (Arabic) V-initial it originally was. A similar argument can be made regarding the item erwieħ ‘souls’, despite the fact that its rate of /t/-insertion (84%) is less strikingly at odds with the other disyllabic V-initials tested (mean rate of /t/-insertion 33%, range 21). In choosing the disyllabic V-initial test items, we sought to include items with as wide a range of vocalic onsets as possible. Erwieħ (SG ruħ ‘soul’) was chosen as an example of an /e/-initial plural. Etymologically, however, this seems to have been a monosyllabic CC-initial plural rwieħ, the initial /e/ being epenthetic.8 As noted in Section 2,

|| 8 It might be assumed that the plural form erwieħ simply maintains the initial vowel of the original Arabic form arwāḥ, but if this were the case it would mean that erwieħ was the only example of a Maltese broken plural that retained the Arabic V-initial pattern aCCāC instead of deleting the initial vowel to give CCieC/CCaC (cf. aḥbāb > ħbieb ‘friends’, alwān > lwien ‘colors’ etc.). This seems unlikely. Note that it is the widespread loss in Maltese of original unstressed short vowels that results in the collapse of the /t/-inserting Arabic broken plural pattern aCCāC and the non-/t/inserting pattern CiCāC into a single Maltese pattern CCieC/CCaC. This in turn enables the analog-

Connecting /t/ in Maltese numerals | 283

initial epenthesis in Maltese is usually only triggered before SONORANT-CONSONANT onsets when these follow anything other than a vowel, and the epenthetic vowel is then /i/. A few (originally) CC-initial words whose first consonant is /r/ behave exceptionally with regard to epenthesis, however. As well as (e)rwieħ, another example is (e)rġajt ‘re-, again’. These words display an (etymologically speaking) epenthetic /e/, not /i/. Synchronically, however, it does not seem accurate to describe this initial /e/ as epenthetic, since it often occurs even after vowels. For example, there are eight attestations in the MLRS corpus of ħafna erwieħ ‘many souls’ and only one of ħafna rwieħ. Nevertheless, since the behavior of erwieħ with respect to /t/-insertion seems to be more in line with its origin as a monosyllabic CC-initial than its synchronic status as a disyllabic V-initial, it seems reasonable to exclude it along with jiem and snin in a reassessment of the results. Doing so gives us the figures shown in Table 5 and illustrated in Figure 2. Tab. 5: /t/-insertion responses by onset and number of syllables, excluding jiem ‘days’, snin ‘years’ and erwieħ ‘souls’.

Monosyllabic

Disyllabic

Polysyllabic

CC onsets

91% (174/191)

53% (123/230)

2% (4/236)

CV onsets

1% (1/168)

0% (1/209)

0% (1/241)

V onsets

/

32% (66/198)

11% (31/244)

These exclusions do not change the essential features of the results, they merely accentuate the trends we have already observed. CV onsets are now extremely resistant to /t/-insertion across the board and monosyllabic CC-initials become even more favorable, while disyllabic V-initials become slightly less favorable. Analyzing these data with the same linear mixed effects model as before unsurprisingly gives us the same main effects (onset: z = -9.79, p < 0.001; syllables: z = -9.96, p < 0.001; no main effect of numeral choice). There is still an interaction between onset, number of syllables and numeral choice (z = 2.5, p = 0.01), and there is now also an interaction between onset and number of syllables (z = 2.45, p = 0.01).

|| ical extension of /t/-insertion to all plurals of this pattern (except snin), including those which originally belonged to the non-/t/-inserting pattern CiCāC, e.g. klieb ‘dogs’ < kilāb.

284 | Christopher Lucas and Michael Spagnol

200 180 160 140 120

Monosyllabic

100

Disyllabic

80 Polysyllabic

60 40 20 0 CC onsets

CV onsets

V onsets

Fig. 2: /t/-insertion responses by onset and number of syllables, excluding jiem ‘days’, snin ‘years’ and erwieħ ‘souls’.

4.4 Broken vs. sound plurals As well as jiem, snin and erwieħ, two items whose rate of /t/-insertion is strongly out of line with the rest of their category are the disyllabic CC-initial sound plurals stampi ‘pictures’ and platti ‘plates’, which triggered /t/-insertion just 6% and 7% of the time respectively. By contrast, frequency of /t/-insertion with the other CCinitial disyllabic items in our sample, which are all broken plurals, ranges from 56% for skejjel ‘schools’ to 80% for ljieli ‘nights’, with a mean rate of 72%. This is clearly insufficient evidence to draw strong conclusions about the relevance of the sound vs. broken plural distinction to /t/-insertion generally, but it is certainly suggestive that Borg’s (1974: 296) claim that sound plurals do not trigger /t/insertion is on the right track, and that, as discussed above, this distinction may be more important than number of syllables, at least for the CC-initials. Turning to the V-initials, however, no such clear trends are discernible. With the exception of erwieħ ‘souls’, none of the V-initial plurals tested are especially favorable to /t/-insertion. The highest rate is 45% for both ulied ‘children’ and aħwa ‘siblings’. These are both broken plurals, but neither the fact of being a sound plural, such as idejn ‘hands’ (26%), nor being polysyllabic, as in ajruplani ‘planes’ (20%), results in the across-the-board absence of /t/-insertion that we have observed with polysyllabic and sound CC-initial plurals. On the

Connecting /t/ in Maltese numerals | 285

other hand, the one polysyllabic V-initial broken plural included among our test items, għasafar ‘birds’,9 had the highest rate of /t/-insertion in that category (34%) – the remaining items had a mean rate of 9% and a range of 17. Additionally, the one V-initial plural tested with six syllables (and final stress), operazzjonijiet ‘operations’, had the lowest rate of /t/-insertion in its category (3%), suggesting that perhaps number of syllables and the broken vs. sound plural distinction are both weakly relevant to the likelihood of /t/-insertion with Vinitial plurals. Evidently, this is a topic about which definitive statements are not possible at this stage.

4.5 Numeral choice The final aspect of the results to be considered concerns the effect of numeral choice on /t/-insertion. As noted above, we found no main effect of numeral choice, but there was an interaction between numeral choice and both the other conditions. Inspection of the per-numeral results for each of the eight categories suggests that numeral choice becomes relevant only with disyllabic V-initial plurals. The per-numeral results for this category (excluding erwieħ ‘souls’) are given in Table 6, which also shows the phonological class of each numeral’s final segment, as well as each numeral’s token frequency (with figures for bare- and /t/forms aggregated). It is difficult to say at this point exactly what properties of the numerals have this effect on the rate of /t/-insertion among the disyllabic Vinitials, but frequency and phonology are plausible candidates. Of particular note in Table 6 is the fact that the numerals whose last segment is an obstruent – żewġ ‘two’ and ħames ‘five’ – have noticeably higher rates of /t/-insertion than the others. Also, the much higher frequency of żewġ as compared to ħames corresponds to a higher rate of /t/-insertion for the former than for the latter.

|| 9 The digraph is unpronounced in contemporary Standard Maltese, but was earlier realized as a voiced pharyngeal or uvular fricative, as is still the case in certain Maltese dialects today (see Cassola 2014). Its loss in Standard Maltese is often compensated for with vowel lengthening or, as in this case, preservation of unstressed short vowels in positions where they were otherwise lost. Compare /ɐˈsɐːfɐr/ ‘birds’ with /ˈbrɐːmɛl/ ‘buckets’, which both go back to a single original broken plural pattern CaˈCāCiC.

286 | Christopher Lucas and Michael Spagnol

Tab. 6: Per-numeral responses for disyllabic V-initial plurals.

Numeral

/t/-insertion rate

Final segment

MLRS frequency per million words

żewġ10

50%

Obstruent

893.42

erba’

30%

Vowel

293.27

ħames

43%

Obstruent

276.2

seba’

31%

Vowel

113.49

tmien

33%

Nasal

93.04

disa’

25%

Vowel

73.72

għaxar

21%

Liquid

130.63

5 Inter- or intra-speaker variation? Concerning those categories in our study that produced figures approaching zero or 100% (i.e. the monosyllabic and polysyllabic CC-initials and all the CVinitials), there was obviously strong agreement between our test subjects as to the obligatoriness of (lack of) /t/-insertion in each case. But where /t/-insertion rates were closer to 50% (as with the disyllabic CC-initials and V-initials) there was clearly much less agreement between subjects. The question then arises: is this apparent inter-speaker variation a reflection of different subjects having different usage tendencies? Or is it rather that for many of our subjects /t/insertion is simply optional with the relevant test items, and their choice of whether or not to insert /t/ in these cases was therefore essentially random? In order to shed some light on this issue, we reran our experiment with seven of the original subjects (one from each of the seven groups) approximately 12 months after the first session. We found that none of these seven subjects were consistent in their use of connecting /t/ with all test items across both runs of the experiment. The most consistent subject exhibited the same /t/-insertion behavior across both runs with 47 of the 56 test items (84%), while the least consistent || 10 The figure of 50% /t/-insertion with żewġ includes the few tokens of the alternative form for ‘two’ ġiex/ġixt that were produced by a minority of our subjects, and which we accepted as valid responses to the relevant stimuli. Note that ġiex, like żewġ, is obstruent-final. The MLRS token-frequency figure, however, refers only to instances of żewġ(t). The frequency of ġiex/ġixt is hard to ascertain precisely, since the vast majority of the 3433 tokens of the string in the corpus are not the numeral ‘two’, but the verb ġie ‘come.3MS.PRF’ with the (usually) negative suffix -x, but it is clearly very low: there is only one token of ġixt in the whole corpus.

Connecting /t/ in Maltese numerals | 287

subject was consistent just 50% of the time (28 out of 56 test items). Per-item figures for consistency across all seven retested subjects are given in Table 7. Tab. 7: Per-item consistency of /t/-insertion responses.

CC-

CV-

V-

Mono-

1st test Consistency Di-

1st test Consistency Poly-

1st test Consistency

fniek

94%

86%

bramel

74%

86%

kmandamenti 6%

klieb

97%

100%

platti

7%

100%

trakkijiet

0%

86%

bniet

88%

100%

stampi

6%

71%

flokkijiet

3%

86%

djar

94%

86%

ġranet

77%

33%

dmirijiet

0%

100%

ħbieb

94%

100%

skejjel

56%

57%

ħsibijiet

3%

100%

snin

10%

100%

ljieli

80%

86%

żminijiet

0%

100%

bwiet

76%

86%

kmamar 71%

83%

studenti

0%

100%

files

0%

100%

kotba

0%

100%

neputijiet

0%

100%

films

0%

100%

naħat

0%

100%

larinġiet

0%

100%

100%

xhur

3%

83%

bozoz

0%

100%

pappagalli

3%

100%

jiem

97%

100%

widnejn

0%

100%

kategoriji

0%

100%

gowls

0%

100%

żgħażagħ 3%

100%

pajjiżi

0%

100%

fonts

0%

n/a

diski

0%

100%

postijiet

0%

100%

toasts

0%

100%

kelmiet

0%

100%

verżjonijiet

0%

100%

ilsna

33%

86%

arloġġi

9%

83%

erwieħ

84%

57%

ajruplani

20%

ulied

45%

60%

operazzjonijiet 3%

uċuħ

24%

86%

għasafar

idejn

26%

71%

appartamenti 6%

86%

aħwa

45%

100%

artikli

9%

86%

oqsma

27%

57%

individwi

9%

71%

34%

57% 100% 71%

The general trend is clear: the more consistently an item is treated across both tests by the retest subjects, the more consistently it was treated across all subjects in the first test. Hence, almost all the items which have consistency figures at or near 100% had overall figures close to zero or 100% in the first test. This strongly suggests that a large proportion of the inter-speaker variation observed in the first test is in fact a reflection of the fact that /t/-insertion is simply optional for individual speakers with many of the test items. This is especially the

288 | Christopher Lucas and Michael Spagnol

case with items in the disyllabic CC-initial category and in the di- and polysyllabic V-initial categories.11

6 Summary and remaining questions This article has shown that the key factor determining /t/-insertion in Maltese is the type of onset of the plural noun that follows a numeral between ‘two’ and ‘ten’. Apart from the specific lexical exceptions jiem ‘days’, snin ‘years’ and erwieħ ‘souls’, no plurals with CV onsets ever allow /t/-insertion, plurals starting with vowels often permit /t/-insertion but never favor it particularly strongly, and CC-initial plurals very strongly favor /t/-insertion when monosyllabic. The number of syllables in the plural also plays a role, in that, in our study, Vinitial plurals with three or more syllables triggered /t/-insertion markedly less often than V-initial plurals with two syllables only. Also, CC-initial plurals with three or more syllables essentially never permit /t/-insertion, while disyllabic CC-initials do so less overall than their monosyllabic counterparts. However, we suggested that the distinction between sound and broken plurals, and perhaps other factors, might play a more important causal role in determining rates of /t/-insertion than number of syllables, at least for CC-initial plurals. This is because a) all monosyllabic CC-initial plurals in Maltese are broken, b) all polysyllabic CC-initials are sound, and c) disyllabic CC-initials (platti ‘plates’ and stampi ‘pictures’ in our study) appear to be as resistant to /t/-insertion as their polysyllabic counterparts. This is clearly far from being the final word on this topic. The most pressing task for future work is to collect further data specifically to determine whether it is indeed the case that there is a sharp distinction between sound and broken CC-initial plurals in determining the possibility of /t/-insertion. Another likely

|| 11 This is not to suggest, however, that there is no inter-speaker variation in the use of /t/. There almost certainly is, and it seems likely that it will correlate with social variables of some sort. Note, however, that our sample of 35 subjects was comprised of 18 females and 17 males, and there were no significant differences in /t/-insertion behavior between the genders. Regarding age, all 35 subjects were in their late teens or early twenties. In order to get a preliminary indication of whether age is a relevant social variable for /t/-insertion behavior, we also recruited five subjects aged over 40 to take the test. Although five is clearly too few to state findings with any confidence, the results for the older speakers were so similar to those for the younger speakers that we chose not to investigate any further the question of age-related variation in /t/-insertion.

Connecting /t/ in Maltese numerals | 289

fruitful line of inquiry would be to investigate the role of token frequency of specific plural nouns and the string frequency of specific numeral–noun collocations in determining whether /t/-insertion is felt to be obligatory or merely optional. Finally, it would be interesting to know to what extent /t/-insertion behavior varies across the different Maltese dialects, and to explore the historical evolution of this construction from its origins in early Semitic. These are all questions that we intend to tackle in future work. Acknowledgments: We would like to express our sincere gratitude to Albert Gatt for his invaluable help with the statistical tests used in this work, and to the two reviewers for their helpful comments on an earlier draft. Naturally though, any errors or omissions are entirely our own. This research was supported by a Humboldt postdoctoral fellowship.

References Ambros, Arne. 1998. Bonġornu, kif int: Einführung in die maltesischen Sprache. Wiesbaden: Reichert. Aquilina, Joseph. 1965. Teach yourself Maltese. London: The English Universities Press. Borg, Alexander. 1974. Maltese numerals. Zeitschrift der Deutschen Morgenländischen Gesellschaft 124. 291–305. Cassola, Arnold. 2014. A note on the dating of ħ, għ, and x in Maltese. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 13–22. Berlin: Akademie Verlag. Cremona, Antonino. 1938. Tagħlim fuq il-kitba maltija, vol. 2. Oxford: Oxford University Press. Fabri, Ray. 1994. The syntax of numerals in Maltese. In Joseph M. Brincat (ed.), Languages of the Mediterranean. Proceedings of the conference held in Malta, 26–29 September 1991, 228–239. Msida: University of Malta. Hoberman, Robert D. 2007. Maltese morphology. In Alan S. Kaye (ed.), Morphologies of Asia and Africa, vol. 1, 257–281. Winona Lake, IN: Eisenbrauns. Hooper, Joan. 1976. Word frequency in lexical diffusion and the source of morphophonological change. In William M. Christie Jr. (ed.), Current progress in historical linguistics, 95–105. Amsterdam: North Holland. Krug, Manfred. 1998. String frequency: A cognitive motivating factor in coalescence, language processing and linguistic change. Journal of English Linguistics 26. 286–320. Levin, Aryeh. 2003. From suffix to prefix. Some synchronic and diachronic aspects concerning the t- prefix preceding the counted noun from three to ten. In Ignacio Ferrando & Juan José Sanchez Sandoval (eds.), AIDA 5th conference proceedings (Cádiz, September 2002), 227–238. Cádiz: Universidad de Cádiz, Servicio de Publicaciones. Schabert, Peter. 1976. Laut- und Formenlehre des Maltesischen anhand zweier Mundarten. Erlangen: Palm und Enke.

290 | Christopher Lucas and Michael Spagnol

Stolz, Thomas, Cornelia Stroh & Aina Urdze. 2011. Total reduplication: The areal linguistics of a potential universal. Berlin: Akademie Verlag. Zipf, George. 1935. The psychobiology of language. Cambridge, Mass.: M.I.T. Press.

| Part III: Contact, Bilingualism & Technology

George Farrugia

Languages in contact Is the gender assignment system in Maltese undergoing change? Abstract: The nominal gender system serves two basic functions: (a) the classification of nouns as belonging to a particular category; (b) the triggering of grammatical agreement that is apparent in those words of different grammatical categories associated with the controlling noun. This paper focuses on (a). Being an offshoot of Arabic, Maltese inherited a system comprising two gender categories, masculine and feminine. This means that every noun, whether animate or inanimate, native or borrowed, has to be assigned a gender on the basis of some criterion, even if the language of origin of a loan word does not itself have the grammatical category of gender. Numerous nouns of Romance origin were introduced in Maltese through contact first with Sicilian and subsequently with Italian, two languages that also have a masculine/feminine-based gender system. Some of these nouns introduced new noun endings, but most of them fit in with the established native system. However, the more recent contact, with English has complicated matters. English is a language whose gender system is conceptual. Inanimate nouns are not marked for gender. Although, at a glance, nouns borrowed from English appear to have adopted the model previously established by nouns of Arabic, Sicilian and Italian origin, closer inspection of a corpus of almost 23,000 nouns reveals novel tendencies that are being introduced by English nouns that deviate from the established gender assignment patterns in Maltese. Keywords: grammatical gender, gender assignment, borrowed nouns, language contact, classification of nouns, Maltese

1 Introduction According to Weinreich (1951), Ferguson and Gumperz (1960), and Gumperz (1964), language contact is not an individual enterprise but a historical product of social forces. Contact is mostly the result of socio-historical events such as

|| George Farrugia: University of Malta, Maltese Faculty of Arts, Old Humanities Building (OH), Msida MSD 2080, Malta. E-mail: [email protected]

294 | George Farrugia

social inequality resulting from wars, conquests, colonialism and migrations. Such events have woven a rich history and a unique language for the people of the Maltese islands. Language contact can be moderate at times but it can also be intensive, such as in cases where historical situations have produced long-term stability and acceptance by the bilingual or multilingual population on all linguistic levels. Quite often the morphological level is considered as the most resistant within a language contact situation (Haugen 1954: 385–386; Winter 1973: 144; Givón 1979: 26; Bickerton 1981: 50). It is not easy for bound morphemes to be transferred from one language to another. In contrast, it is common for lexical borrowings to be adapted to the morphological categories of the receiving language.1 Such borrowings may involve the borrowing of nouns into a language with a specific gender or noun-class system. This process may include a reconfiguration of the borrowed material to make it fit already established categories in the receiving language in accordance with a number of gender assignment criteria or else the borrowed material may challenge the already established strategies. According to Corbett (1991: 71) “The simplest hypothesis would be that borrowings (or ‘loanwords’) will be assigned to a gender by the normal assignment rules and that they are therefore like any other nouns.” But is it possible that these borrowings, due to intensive language contact, provide additional mechanisms and strategies that can challenge seriously an already established native system? Maltese, a Semitic language which has been in intensive language contact with Indo-European languages for the last thousand years or so, serves as a good example to study this hypothesis. This paper deals with the integration of hundreds of loan nouns of different origin in Maltese and investigates whether these nouns influenced an already established gender system. Section 3 outlines briefly the general rules which regulate gender assignment in Maltese. The methodology and data of this study are explained in Section 4. Sections 5–8 deal with strategies of gender assignment in Maltese for nouns of various origins, namely Arabic, Sicilian, Italian and English. Finally, Section 9 contains some concluding remarks.

|| 1 See Haspelmath and Tadmor (2009a). Their work is based on the Loanword Typology Project (LWT) (Haspelmath and Tadmor 2009b) a loanword database consisting of vocabularies of 41 languages from around the world contributed by various authors or author teams between 2004 and 2008. The database provides comprehensive information about the loanword status of each word and a systematic basis for crosslinguistic comparison.

Languages in contact | 295

2 Grammatical gender Two factors are relevant in the analysis of grammatical gender: the classification of nouns in a particular category, and the grammatical agreement displayed by the words of different grammatical categories with the controlling noun. These two factors are closely related since agreement plays an important role in defining gender while it is itself determined by gender, as shown in the following example from Maltese: (1)

Ra x-xemx nieżl-a wara see(3M.IPFV) the-sun descend-F.SG behind ‘He saw the sun setting beyond the hills.’

l-għoljiet. the-hills

This example shows how the present participle nieżla which is marked for gender (and number) and which, following Corbett (1991), can be referred to as “target element”, indicates that the noun xemx ‘sun’ is feminine. At the same time, the fact that the noun xemx (the “controller”) is feminine determines the feminine form of the present participle. Thus, grammatical gender is both a syntactic phenomenon, whereby the class or category of the controller is reflected on its target elements, and a morphological one, since grammatical agreement is marked through inflectional mechanisms primarily affixes or clitics. Gender systems typically make a three-way distinction and are found in many Indo-European languages, such as German, Serbo-Croat and Icelandic, as well as in the languages of the northwest Caucasus. In other Indo-European languages, such as Dutch2, French and Italian, the gender system distinguishes two categories. The latter binary system is also found in the Afro-Asiatic languages, including Maltese, which also distinguishes two categories, the masculine and the feminine. Gender systems are not static. Given that languages undergo continuous development caused by a variety of factors, amongst which, contact with other languages, the gender system they incorporate will also undergo change, for example by expanding to incorporate a new gender category. Gender systems can emerge as seems to be the case in Chamorro. According to Stolz (2012: 128) the Spanish influence on Chamorro was strong enough to introduce structural principles into the recipient language, namely, (i) the obligatory overt sex-marking by morphological means and (ii) NP-internal agreement. A language can also lose a gender category, for example, because of a change in its phonetic structure. Mod-

|| 2 See Van Berkum (1996: 22).

296 | George Farrugia

ern Persian has completely lost the Indo-European gender system which written Old Persian originally had, while Modern English only carries traces of the original Germanic system. The distribution of nouns across gender categories can also change, so that a noun which is assigned a particular gender can lose it and acquire a different one. As a result, a particular category can expand or contract, either because new nouns are included in it or because nouns drop out, because they do not remain in use or because they change their gender. In most cases, gender assignment depends on two sources of information, namely, meaning (i.e. semantics) and form (i.e. phonologically or morphologically). Frequently, gender assignment is influenced by a combination of these factors. According to Corbett (1991: 8), all gender systems are semantic, in the sense that they all include a semantic element in the way gender is assigned. In some languages, this factor alone suffices to assign a particular gender to a noun. In systems where this factor is strong, the meaning of the noun determines its gender; conversely, the gender of the noun can give indications as to its meaning. Such systems are sometimes also referred to as “natural gender systems”. Formal systems are normally divided into two types, namely, morphological systems that are sensitive to word structure, including inflection and derivation, and phonological systems that are sensitive to sound (Corbett 1991: 33). These systems can be combined to different degrees in different languages. Morphological systems quite often include a semantic component. This is often the case with derivational morphology. Thus, nouns in Maltese which end in -iż are masculine, for example, Il-Franċiż mhux ħafif ‘French is not an easy language’. Nevertheless, in certain contexts these nouns are assigned masculine gender because they refer to a male, as in, for example, Il-Franċiż ta’ ħdejna miet ilbieraħ ‘Our French neighbor died yesterday’. It is therefore justifiable to claim that even formal rules for gender assignment can have some degree of semantic motivation. There is near universal agreement among researchers that every gender assignment system is semantically motivated to some degree (Corbett 1991: 33f., Aikhenvald 2003: 25). In those cases where a language does not apply semantic principles only, it has to rely on formal ones; when this is the case, the language is referred to as having a mixed gender assignment system. Such languages frequently exhibit different strategies for gender assignment which can interact or be in competition.

Languages in contact | 297

3 Grammatical gender in Maltese Maltese is an example of a language with a mixed gender system in which several sub-systems act in tandem, in a complex interaction that determines gender assignment. These include semantic rules (e.g., nouns having a male referent are masculine) and morphophonological rules (e.g., the majority of nouns ending in -a are assigned feminine gender). In Maltese, each noun needs to be assigned masculine or feminine gender.3 This means that every borrowed noun also needs to be assigned gender on the basis of some criterion, be it semantic or formal, even if the language of origin does not have the grammatical category of gender. According to C. Stolz (2009: 330) Maltese employs yet another assignment strategy which she calls Gender Copy, that is when the borrowing language emulates not only semantic properties of a given noun of the source language (such as lexical meanings or components thereof), but also grammatical ones, such as the affiliation with a given gender category.

Nouns that refer to people or animals normally take gender according to the referent’s sex, while the most important criterion for inanimate nouns is not semantic but the formal, namely, their declension (i.e., the right edge of the word). In general, descriptive grammars of Maltese agree that inanimate nouns that end in a consonant or -u are masculine while those that end in -a are feminine, with a few exceptions in both cases. The vocalic ending -i can indicate either a feminine or a masculine noun. Nouns that end in -o and -e, mostly borrowed from Italian, are not discussed in any detail in traditional descriptive grammars.

4 A quantitative study In a quantitative study of the relationship between noun endings and their associated gender, Farrugia (2010) collected a large corpus of nouns. These nouns were harvested from Aquilina’s Maltese-English dictionary (1987 and 1990), as well as from the concise version of this dictionary published in 2005. A total of

|| 3 Some nouns, such as ministru (‘minister’) and pulizija (‘police officer’) can take either masculine or feminine gender. Normally, in grammars of Maltese, these are referred to as nouns with a “common” or “dual” gender. Fabri (2009) refers to them as nouns which are underspecified for gender.

298 | George Farrugia

22,867 nouns were collected, namely, all nouns from the dictionary lemmas for which a grammatical gender is given. These are distributed as follows according to their origin: 4,701 nouns are of Arabic origin, 13,255 of Sicilian or Italian origin, 1,887 of English origin, 2,311 of unknown or uncertain origin and 713 whose origins varied. A chi-square test confirms that there is a very strong relationship between the noun ending and its gender. In fact, 93% of the sample of nouns under investigation can be assigned gender on a morphophonological basis, according to their ending. This is shown in Table 1. It would be interesting to note whether this applies in a similar way to nouns of different origins in Maltese, starting with nouns of Arabic origin, being the main stratum of Maltese, followed by nouns of Sicilian origin which form the superstratum, and subsequently by nouns borrowed from Italian and English, both adstrata of Maltese.4 Tab. 1: Noun ending and gender.

Noun ending

Masculine

Feminine

Common

Total

a, à, a’5

132 (1.32%)

9,397 (94.58%)

407 (4.10%)

9,936

e, è

84 (32.68%)

172 (66.93%)

1 (0.39%)

257

i, ì

311 (17.50%)

1,439 (80.98%)

27 (1.52%)

1,777

o, ò

97 (93.27%)

6 (5.77%)

1 (0.96%)

104

u,+ ù

2,415 (99.26%)

14 (0.58%)

4 (0.16%)

2,433

-cons.

8,055 (96.35%)

214 (2.56%)

91 (1.09%)

8,360

|| 4 For this stratification of the Maltese language see Brincat (2000: 11) 5 Realized as /ɐ/ or /ɐ:/. While the author of the present work is aware that it is the sound of the final segment of the word which is relevant in the present context, orthographic conventions are being used for convenience.

Languages in contact | 299

5 Nouns of Arabic origin Since Maltese is an offshoot of vernacular Arabic, it inherited a gender system comprising two categories, masculine and feminine. However, as we shall see, in Maltese the system is modified so that Maltese and Arabic do not share exactly the same criteria for the classification of nouns. In the Arabic gender system, meaning dominates form in the case of animate nouns. The feminine of animate nouns is generally formed from the masculine through the affixation of -at, which is the most frequent feminine suffix, or via a different word that may or may not have a feminine suffix, for example, kalb ‘dog.M.SG’ kalb+ at ‘dog.F.SG’; jamal ‘camel.M.SG’ nāqa+ at ‘camel.F.SG’; ’ab ‘father’ ’umm ‘mother’. The feminine is the marked form and the suffix -at, also inherited by Maltese, is the general feminine marker. In the case of inanimate nouns, form has precedence over meaning, although there are a number of exceptions. The gender of these nouns in Arabic is mostly conventional, e.g., qamar ‘moon.M.SG’, šams ‘sun.F.SG’. Some nouns can be either masculine or feminine, e.g., sikkīn ‘knife’, sūq ‘market’. The following are some examples of Maltese nouns of Arabic origin that can be both masculine and feminine in Arabic, but in Maltese take only one gender: abt ‘armpit’, boton ‘litter of rabbits’, ġewnaħ ‘wing’, ħanut ‘shop’, sliem ‘peace’, sellum ‘ladder’, sur ‘bastion’, saba’ ‘finger’, għasel ‘honey’, għonq ‘neck’, għanqbut ‘cobweb’, qaws ‘bow’, ilsien ‘tongue/language’, lejl ‘night’, melħ ‘salt’ (all masculine in Maltese) and darsa ‘molar’, triq ‘street/road’ (both feminine in Maltese). According to Ibrahim (1973), the most “problematic” cases in Arabic are feminine nouns that are unmarked. There has never been unanimous agreement on the gender of these nouns. Procházka’s (2004) study of the Arabic dialects confirms that, in general, there is a lack of agreement on gender assignment in these dialects. However, surprisingly, they are quite homogeneous when they attribute gender to a number of nouns that take the feminine but have no marker.6 Procházka classifies these nouns into specific semantic categories. Normally, discussions of nominal gender in Arabic tend to state that body parts that occur in pairs are usually feminine. This is especially true of Classical Arabic (CA) and Modern Standard Arabic (MSA) although, according to Procházka, it does not apply to contemporary Arabic dialects. In these dialects, such nouns are usually considered masculine, with the exception of ‘ayn ‘eye’,

|| 6 According to Procházka (2004: 254), unmarked feminine nouns are more numerous in Western dialects of Arabic, which include Maltese.

300 | George Farrugia

’udun ‘ear’, yad ‘hand’, rijl ‘leg’, /sāq ‘foot’, which are practically always feminine. It is worth noting that riġel (‘leg’) in Maltese is masculine, which is also the case in some parts of Morocco.7 In some other dialects, there are other unmarked nouns which are feminine, such as lisān ‘tongue’ and sinn ‘tooth’. The case of the noun sinn in CA is interesting, as it is an unmarked feminine noun which became marked in some varieties of Arabic, among them that spoken in Cairo, as well as in Maltese sinna. Other examples of nouns that underwent the same change are ‘arūs, ajūz and sikkīn, which in Maltese became għarusa ‘bride/fiancée’, għaġuża ‘old woman’ and sikkina ‘knife’. Agius (1996) in Brincat (2000: 34) notes a similarity between Maltese and the Arabic spoken in Sicily in the case of the nouns għarusa and għaġuża. He also mentions għanqbuta (< Ar. ‘ankabu:t), which is feminine in Maltese and Siculo Arabic but it can take both genders in CA. Another category of nouns which are not marked in the feminine but tend to take feminine gender in CA is that of nouns that refer to the sky, the earth and the elements. Among these are šams ‘sun’, ’arḍ ‘earth’, nār ‘fire’ and turāb ‘soil’. In Maltese, the words nar ‘fire’ and trab ‘dust’ did not maintain their feminine gender, most probably, due to the fact that they end in a consonant. The word samā’ ‘sky’, though similar in form to feminine nouns, is masculine in many regions, especially when its meaning is that of ‘roof’. In CA and MSA, it can be either feminine or masculine. However, in sedentary dialects, among them those of Syria, Egypt and the Central Maghreb, it is feminine. This noun is considered masculine in Maltese grammars and dictionaries, but is in practice treated as feminine by many speakers.8 Among other nouns that are not marked but are considered feminine in many dialects of Arabic, one finds rīḥ ‘wind’, balad ‘city’, dār ‘house’,9 bāb ‘door’, bayt ‘room’, furn ‘oven’, ṭarīq ‘street’, bi’r ‘well’, fār ‘rat’, fikrūn ‘tortoise’, ‘asal ‘honey’, laḥm ‘meat’, xubz ‘bread’, šaḥm ‘fat’, ṣūf ‘wool’. Maltese seems to exhibit a tendency to regularize and avoid exceptions. Thus, all of these nouns, with the exception of belt, dar and triq, are masculine in Maltese, the likely reason being precisely that they are not marked for the feminine and end in a consonant. Another case that strengthens this observation is the noun mūsā ‘penknife’, which is feminine in Classical Arabic. In dialectal Arabic this noun is

|| 7 Procházka (2004: 239). 8 Aquilina’s Maltese-English dictionary (1987–1990: 1290) marks the noun sema ‘sky’ as masculine, but notes that it is sometimes considered feminine by some speakers. 9 This noun is masculine in other varieties of Arabic, such as those spoken in Central Syria and Cilicia.

Languages in contact | 301

likely to occur without the ending -at but often maintains feminine gender. This is not the case in Maltese, where, probably because it has lost its ending and now ends in a consonant, it is considered masculine. This tendency is in fact mentioned by Procházka (2004: 248), who observes that, barring a few exceptions, the tendency for feminine nouns to become masculine is found in modern dialects of Arabic that simplify the system. Speakers of Arabic dialects detect no rationale as to why such nouns should be feminine. Thus, one might conjecture that many of these unmarked feminine nouns are used frequently and, as a result, resist the transformation from feminine to masculine. All in all, only a few nouns are both feminine and unmarked in the majority of Arabic dialects, including Maltese. In fact, the only inanimate nouns of Arabic origin that belong to this category and that are listed in Aquilina’s dictionary are aħbar ‘news’, art ‘land’, belt ‘city’, dar ‘house’, għajn ‘eye’, id ‘hand’, mewt ‘death’, qalb ‘heart’, qmis ‘shirt’, ras ‘head’, ruħ ‘soul’, sieq ‘leg’, triq ‘street/ road’, xemx ‘sun’ and żaqq ‘belly’. The tendency displayed by Maltese to simplify the system can also be seen in some nouns that end in the sound /t/ in Maghribi dialects which changed their gender from masculine to feminine, most likely on the basis of a phonological analogy with the -t ending, which is one of the markers of the feminine in Arabic. Among others, one finds the following nouns: bayt ‘house’, zayt ‘oil’, mawt ‘death’, ḥūt ‘fish’, waqt ‘time’, ḥanūt ‘shop’, and ‘ankabūt ‘spider’. With the exception of the noun mewt ‘death’, all the nouns in this list are masculine in Maltese, because the -t ending is no longer considered a strong marker of the feminine and therefore, probably has the status of a consonant like any other; hence, it is associated with masculine gender.

6 The relationship between noun endings and gender in nouns of Arabic origin In the corpus compiled by Farrugia (2010), there were 4,701 nouns of Arabic origin. Table 2 shows a strong relationship between (i) the endings -a and -a’, and feminine gender and (ii) consonantal endings and -u, and the masculine. The -i ending in the case of nouns of Arabic origin is mainly associated with the masculine. This relationship between the noun endings and a particular gender could be labeled as native and should be considered the default case in gender assignment to inanimate nouns in Maltese.

302 | George Farrugia

Tab. 2: Gender assignment to nouns of Arabic origin with different endings.

Nouns of Arabic origin

N = 4,701

Noun ending

Gender assigned Masculine

Feminine

Common (M/F)

Total

a + a’

21 (0.81%)

2,558 (98.57%)

16 (0.62%)

2,595

i

29 (90.63%)

3 (9.37%)

− (0.00%)

32

u

47 (100%)

− (0.00%)

− (0.00%)

47

consonant

1,982 (97.78%)

24 (1.19%)

21 (1.03%)

2,027

7 Nouns of Sicilian and Italian origin Although the Normans took possession of Malta in 1091, for a long period the Arabs living in Malta were practically undisturbed. However, after the Muslims were expelled from Malta in the first half of the thirteenth century, the Arabic spoken in Malta came in contact first with Sicilian, and later on with Italian during the rule of the Order of St. John (1530–1798). The contact with Sicilian and Italian served as an appui to the gender assignment system already established in the language. All nouns in Sicilian are either masculine or feminine, depending on the word ending. Generally speaking, in Sicilian, nouns ending in -a are feminine, for example, rota (‘wheel’), porta (‘door’), carta (‘paper’), accetta (‘hatchet’). Whereas Italian uses -o as the ending for masculine nouns, Sicilian generally uses -u, for example, omu (‘man’), libbru (‘book’), nomu (‘name’), santu (‘saint’), ferru (‘iron’), prunu (‘plum’), and catusu (‘large pipe’). In Sicilian, in general, there are no words ending in the unaccented vowels -e or -o. The unstressed vowel -e of Latin becomes unstressed -i in Sicilian and nouns ending with -i can be either masculine or feminine, for example baruni (‘baron’) and ucchiali (‘eye-glasses’), which are masculine, and chiavi (‘key’) and sorti (‘luck’), which are feminine.10 Brincat (2000: 62) notes that Italian words ending in unaccented -o, in Sicilian end in -u while, in Maltese, they either end in || 10 Sicilian Grammar by “U Sicilianu” http://www.docstoc.com/docs/27330806/SICILIAN-GRAMMAR 1/05/2014.

Languages in contact | 303

-u or else lose the final vowel completely and are considered to be masculine nouns, e.g., Italian velo (‘veil’) → Sicilian velu → Maltese velu; Italian filo (‘thread’/‘string’) → Sicilian filu → Maltese fil. The unaccented -e ending in Italian, which becomes -i in Sicilian, in Maltese is either retained or else dropped, e.g., Italian voce (‘voice’) → Sicilian vuci → Maltese vuċi; Italian colore (‘color’) → Sicilian kuluri, Maltese → kulur. If such words end in a consonant in Maltese, they are treated as masculine nouns and, if they end in -i, they are generally feminine. Contact with Italian, later on, did not create any serious problems for the native gender assignment system because Italian also has a gender system consisting of two categories, masculine and feminine. As a general rule, animate nouns in Italian normally take gender according to the sex of the referent, that is, those of male sex are masculine and those of female sex are feminine. This occurs even when the noun ending does not correspond to the gender it is generally associated with. As for inanimate nouns, those which end in -o are masculine, and those that end in -a are feminine. Nouns that end in -e can be either masculine or feminine. As a general rule, masculine nouns of Italian origin ending in -o either drop this vowel and finish with a consonant in Maltese, e.g., monumento → monument (‘monument’), tempo → temp (‘time’, ‘weather’), atto → att (‘action’, ‘deed’, ‘act’), or else they change the -o to -u, as in acquario (‘aquarium’) → akkwarju, anniversario (‘anniversary’) → anniversarju, disastro (‘disaster’) → diżastru. In both cases the nouns are treated as masculine according to the native rules. Nouns ending in -a fit exactly within the pre-established native system and continue to be treated as feminine nouns in Maltese, e.g., carta (‘paper’) → karta, storia (‘story’, ‘history’) → storja; avventura (‘adventure’) → avventura. Nouns ending in -e in Italian, normally either drop this vowel in Maltese and tend to be regarded as masculine nouns since they finish with a consonant e.g. ambiente (‘environment’) → ambjent, vapore (‘steamer’) → vapur, giornale (‘newspaper’, ‘journal’) → ġurnal,11 or they change the final -e to -i and are considered generally to be feminine nouns, e.g., ambizione (‘ambition’) → ambizzjoni, attenzione (‘attention’, ‘care’) → attenzjoni, classe (‘class’, ‘classroom’) → klassi.12 It is interesting to note that nouns that were borrowed from Italian and that are counted as exceptions in this language tend to become regularized in Maltese and take gender according to their ending. Thus, for example, radju (‘radio’), which is feminine in Italian (radio), becomes masculine in Maltese be|| 11 An exception is reliġjon ‘religion’ which is feminine. 12 Some exceptions are eżami, forti, kalċi, torri.

304 | George Farrugia

cause it ends in -u, while sistema (‘system’), telegramma (‘telegram’), klima (‘climate’), dilemma (‘dilemma’), diploma (‘diploma’), fonema (‘phoneme’), morfema (‘morpheme’), tema (‘theme’), problema (‘problem’), gwardarobba (‘wardrobe’), pjaneta (‘planet’), piġama (‘pyjamas’), which are all masculine in Italian, change their gender to feminine in Maltese because they end in -a. Tab. 3: Gender assignment among nouns of Sicilian and Italian origin with different endings.

Nouns of Italian and Sicilian origin

N = 13,255

Noun ending

Gender assigned Masculine

Feminine

Common

Total

a

75 (1.39%)

4,967 (92.14%)

349 (6.47%)

5,391

à

3 (0.58%)

512 (99.03%)

2 (0.39%)

517

e

36 (22.78%)

122 (77.22%)

− (0.00%)

158

è

8 (88.89%)

1 (11.11%)

− (0.00%)

9

i

154 (10.20%)

1,343 (88.94%)

13 (0.86%)

1,510

ì

3 (75%)

1 (25%)

– (0.00%)

4

o

39 (97.50%)

− (0.00%)

1 (2.50%)

40

ò

5 (100%)

− (0.00%)

− (0.00%)

5

u

2,003 (99.55%)

5 (0.25%)

4 (0.20%)

2,012

ù

8 (66.67%)

4 (33.33%)

− (0.00%)

12

consonant

3,554 (98.81%)

22 (0.61%)

21 (0.58%)

3,597

In the corpus under scrutiny, there are 13,255 nouns of Sicilian and Italian origin. This was the largest group of nouns of a particular origin. As in the case of words of Arabic origin, within this group, the endings -a and -à are strongly associated with feminine gender, while nouns that end in -u or a consonant evince a strong tendency to be masculine. It is important to note that nouns

Languages in contact | 305

with these endings together constitute almost 87% of all the nouns of Italian and Sicilian origin. This means that most of the borrowed nouns fitted in the already established native gender system with the exception of nouns ending in -i which are likely to be treated as feminine (88.94%). Contact with Italian also introduced new endings into the Maltese nominal lexicon, such as -e and -o. The -e ending is mainly associated with feminine gender while the -o ending is associated with masculine. C. Stolz (2009) observes that Maltese and Italian, though genetically unrelated, have quite similar gender systems, both functionally and formally. However this resemblance is not due to contact; Maltese inherited its gender system from its Semitic ancestor while Italian from Latin. Studying her Maltese sample of loan nouns, she observes that the gender categorization of the majority of Maltese loan nouns borrowed from Italian coincides with that of their ItaloRomance models, that is, they have the same gender. Her sample, based on Aquilina’s Maltese-English dictionary (1987–1990) consisted of 630 nouns of which 520 were of Italo-Romance origin. Her observation leads her to ask a very important question: Does Maltese really use Gender Copy as a gender assignment strategy or are similar gender assignments of Maltese and Italo-Romance that are so striking in the sample caused by close, but accidental similarities between the gender assignment strategies of both languages? (C. Stolz 2009: 335)

Her conclusion (C. Stolz 2009: 335) is that “Maltese loan nouns of Italo-Romance origin predominantly use Gender Copy as the preferred integration strategy.” This happens regardless of whether there are overt gender markers on the ItaloRomance models or not. Stolz (2009: 335) believes that The presence of overt gender markers correlates with a rocketing of the frequency of Gender copies to almost 100% and therefore seems to trigger Gender Copies. However, in the absence of overt gender markers on the Italo-Romance models Gender Copy is still the preferred gender assignment strategy.

My contention is that the Gender Copy strategy mentioned by Stolz could have influenced the gender assignment only of a sub-category of nouns of ItaloRomance origin, namely those nouns ending with -i (not very common in Maltese), -o and -e (not native). However for the majority of borrowed nouns, which are C-final -u final and -a final, the already established native gender assignment criteria, based on the final segment of the noun, were used. Owens (2010: 110) is also of the same opinion: “In cases where there is no model, as a rule gender-copy is operative, that is, the Maltese takes over the

306 | George Farrugia

gender of the source language noun.” He claims that the association of final loanword -a with the category of feminine noun in Maltese is a typical Arabic feature and virtually all varieties of spoken Arabic today treat a final loanword -a as feminine, as e.g. gang-a ‘drum-F’ in Nigerian Arabic < Kanuri (genderless language) ganga, with the regular feminine plural -aat, gangaay-aat, Iraqi Arabic poost-a (PL poost-aat) ‘mail, post’, Egyptian gonell-a ‘blouse’13 (< It.), and so on.14 According to Owens, it would have been remarkable if Maltese had used a different strategy of classifying -a final loans. It would appear that, through contact with Sicilian and Italian, several nouns were introduced that were assigned common gender, that is, can take either the masculine or feminine gender. This accounts for the notable increase of common gender nouns in Maltese when compared to the number of such nouns of Arabic origin. Such nouns in Italian are likely to end as follows: -e, for example, nipote (‘nephew’/‘niece’), custode (‘custodian’); -ante, for example, amante (‘lover’), cantante (‘singer’); -ista, for example, artista (‘artist’), dentista (‘dentist’); -ente, for example, cliente (‘client’), agente (‘agent’); -cida, for example, omicida (‘homicide’) [used for the perpetrator]; -iatra, for example, pediatra (‘pediatrician’), psichiatra (‘psychiatrist’); -a, for example, atleta (‘athlete’), collega (‘colleague’), which are of Greek origin. Among these nouns that are assigned common gender, those that have been borrowed into Maltese are treated either i) in the same way, that is considered to have common gender, for example, psikjatra (‘psychiatrist’), pedjatra (‘pediatrician’), atleta (‘athlete’), kollega (‘colleague’); or ii) distinguished through the creation in Maltese of two separate forms, one for the masculine and one for the feminine, for example dilettant/a (‘dilettante’/‘beginner’), kantant/a (‘singer’), klijent/ a (‘client’), pensjonant/a (‘pensioner’). Thus, on the whole, nouns borrowed from Sicilian and Italian confirm the general patterns of gender assignment identified previously among nouns of Arabic origin. Borg and Azzopardi-Alexander (1997) also claim that the gender of nouns of Romance origin, whose use in Maltese is well established, is normally stable, whereas in the case of nouns borrowed recently from English, there is more variation in gender.

|| 13 It is interesting to note that Maltese also borrowed this word from Italian għonnella F. 14 Holes (1990: 158) also notes that in Gulf Arabic loan nouns ending in -a are treated as feminine.

Languages in contact | 307

8 Contact with English The more “recent” contact, with English, is the result of the British rule in Malta (1800–1964). English is a language whose gender system is not grammatical but mostly conceptual, where gender marking is by and large reserved for animate, mostly human, referents and where, with some exceptions, there is a reasonably clear relationship between the sex of the referent and the gender of the noun. Inanimate nouns are not marked for gender (Corbett 1991: 102, Greenbaum 1996: 107) except for a few nouns denoting objects like ships, cars, pets and mother countries and which, according to C. Stolz (2009: 325) are treated as “honorary animates” and may be assigned masculine or feminine gender. However, in Maltese, all nouns, borrowed from English, have to be assigned a gender. The corpus mentioned in Section 4 contained 1,887 nouns of English origin. At first glance (see Table 4), nouns borrowed from English appear to have adopted the model previously established by nouns of Arabic, Sicilian and Italian origin, i.e., nouns that end in the sound [a] are associated mainly with feminine gender (84.56%), while those that end in the sound [u] or in a consonantal sound are associated with masculine gender (98.55% and 88.48%, respectively). This view is also shared by C. Stolz (2009: 349): As to the integration of English nouns into Maltese, however, Gender Copy is hardly an option. Maltese integrates English loan nouns into its gender system predominantly with the help of formal criteria, such as the nature of the final segment.

Tab. 4: Gender assignment among nouns of English origin with different endings.

Nouns of English origin

N = 1,887 Gender assigned

Noun ending

Masculine

Feminine

Common

Total

a

5 (3.09%)

137 (84.56%)

20 (12.35%)

162

e

2 (33.33%)

4 (66.67%)

− (0.00%)

6

è

4 (100%)

− (0.00%)

− (0.00%)

4

i

60 (68.97%)

25 (28.74%)

2 (2.29%)

87

ì

6 (85.71%)

1 (14.29%)

− (0.00%)

7

308 | George Farrugia

Nouns of English origin

N = 1,887 Gender assigned

Noun ending

Masculine

o

6 (85.71%)

1 (14.29%)

− (0.00%)

7

ò

2 (66.67%)

1 (33.33%)

− (0.00%)

3

u

68 (98.55%)

1 (1.45%)

− (0.00%)

69

ù

5 (100%)

− (0.00%)

− (0.00%)

5

consonant

1,360 (88.48%)

134 (8.72%)

43 (2.80%)

1,537

Feminine

Common

Total

In contrast with nouns of Italian or Sicilian origin, nouns of English origin that end in the sound [i] are often masculine (68.97%), so that in this case, the trends are similar to those among nouns of Arabic origin. The [e] sound ending is mainly associated with feminine (66.67%) and the [o] sound ending is associated mainly with masculine (85.71%). The data displays a similar trend to that observed among nouns of Italian and Sicilian origin. However, a closer inspection of the data in Table 5 below, which shows the relationship between noun endings and gender, depending on the noun’s origin in the sample under investigation, reveals that nouns borrowed from English which end in a consonant are less integrated within the established system. While 97.78% of nouns of Arabic origin and 98.81% of those of Italian and Sicilian origin which end in a consonant are masculine, in the case of nouns of English origin, the percentage is 88.48% (please refer to Diagram 1). In contrast, while only 1.19% of nouns of Arabic origin and 0.61% of those of Italian and Sicilian origin ending in a consonant are feminine, among nouns of English origin, the percentage rises to 8.72%. These differences are even more pronounced in a study by Farrugia (2003) based on a random sample of 664 nouns borrowed from English and used by informants in their everyday language. Out of a total number of 601 nouns whose gender is stable, 26.12% of those ending in a consonant are feminine (see Diagram 2). This could mean that nouns of English origin are gradually causing a rupture in the association between the consonant ending and masculine gender, which means that more and more nouns ending in a consonant are now assigned feminine gender. At the same time, the native feminine marker par excellence, i.e., /–ɐ/, which was strengthened through contact with nouns of Romance origin, seems now to have become less productive since the majority

Languages in contact | 309

of English loan nouns end in a consonantal sound. Thus, contact with English seems to have weakened the tendency in Maltese to regularize gender assignment, and avoid apparent exceptions to the system, a tendency which was a common practice among nouns borrowed from Sicilian and Italian. Tab. 5: Noun endings and gender, depending on origin in the sample under consideration.15 Gender assigned

Noun ending Arabic

Italian + Sicilian

English

M%

F%

COM %

M%

F%

COM %

M%

F%

COM %

a + (a’)

0.81(1)

98.57

0.62

1.39

92.14

6.47

3.09

84.56

12.35

à

−

−

−

0.58

99.03

0.39

−

−

−

e

−

−

−

22.78

77.22

−

33.33

66.67

−

è

−

−

−

88.89

11.11

−

100

−

−

i

90.63

9.37

−

10.20

88.94

0.86

68.97

28.74

2.29

ì

−

−

−

75.00

25.00

−

85.71

14.29

−

o

−

−

−

97.50

−

2.50

85.71

14.29

−

ò

−

−

−

100

−

−

66.67

33.33

−

u

100

−

−

99.55

0.25

0.20

98.55

1.45

−

ù

−

−

−

66.67

33.33

−

100

−

−

cons.

97.78

1.19

1.03

98.81

0.61

0.58

88.48

8.72

2.80

(1) The percentages in each row are based on the total number of nouns with each ending: in the sample of nouns of Arabic origin: ending -a + a’: N=2,595; ending -i: N=32; ending -u: N= 47; consonant ending: N=2,027. in the sample of nouns of Italian/Sicilian origin: ending -a: N=5,391; ending -à: N=517; ending -e: N=158; ending -è: N=9; ending -i: N=1,510; ending -ì: N=4; ending -o: N=40; ending -ò: N=5; ending -u: N= 2,012; ending -ù: N=12; consonant ending: N=3,597. in the sample of nouns of English origin: ending -a: N=162; ending -e: N=6; ending -è: N=4; ending -i: N=87; ending -ì: N=7; ending -o: N=7; ending -ò: N=3; ending -u: N=69; ending -ù: N=5; consonant ending: N=1,537.

|| 15 The figures do not include nouns of diverse origin or nouns of uncertain or unknown origin.

310 | George Farrugia

These results confirm those reported by Farrugia (2003), namely, that gender among nouns borrowed from English is affected by factors other than the final segment of the word, including: (i)

(ii)

the gender of the hyperonym when nouns are assigned gender possibly by analogy with their native hyperonym for example ‘grapefruit’, ‘pineapple’ and ‘coconut’ are feminine because they are associated with frotta ‘fruit’; ‘jeans’, ‘slacks’, ‘shorts’ and ‘pants’ have a tendency to be regarded as masculine nouns, possibly by analogy with their native hyperonym qalziet ‘trousers’, which is masculine; ‘bedroom’, ‘livingroom’, ‘showroom’, and ‘washroom’ are feminine, possibly by analogy with their native hyperonym kamra ‘room’. gender assignment by analogy, that is, the process whereby a borrowed noun or part of it is assigned the same gender as the native noun that it replaces or with which it is associated. For example, the following nouns are considered feminine although they end in a consonantal sound because the native nouns with which they are associated are feminine: ‘bicycle’ ‘bedspread’, ‘quilt’ ‘suit’ ‘train’ ‘table-cloth’ ‘postcard’ ‘lesson’ ‘washing-machine’

rota kutra/gverta libsa ferrovija dvalja kartolina lezzjoni magna tal-ħasil

‘blade’ ‘band’ ‘raincoat’ ‘speedboat’ ‘face cloth’ ‘sandpaper’ ‘crisps’, ‘chips’ ‘bus’

xafra banda inċirata dgħajsa ċarruta, biċċa karta patata karozza (tal-linja)

Farrugia (2003: 228) claims that analogical gender in the case of inanimate nouns borrowed from English is becoming a very strong factor and often the other factors that influence gender assignment apply only if, for some reason, analogy does not apply. C. Stolz (2009: 350–351) is also of the opinion that in Maltese, the gender of English inanimate nouns, although predominantly assigned on formal (i.e. phonological) grounds, could also be assigned on the basis of an admixture of semantic criteria, namely, by analogy to a next lexical equivalent, cognate, and possibly lexical field. The assignment of gender by semantic analogy to borrowed nouns was also reported in various works such as those by Poplack et al. (1982), Surridge (1982) and Gregor (1983).

Languages in contact | 311

Diagram 1: Masculine nouns ending in a consonant.

Diagram 2: Feminine nouns ending in a consonant.

312 | George Farrugia

9 Conclusion Due to these new gender assignment strategies brought about through contact with English, many nouns of English origin still have an unstable gender in Maltese. This means that they can take either the masculine or the feminine gender as, for example, ‘ashtray’, ‘burner’, ‘case’, ‘cooker hood’, ‘cushion’, ‘gearbox’, ‘pencil box’, ‘shutter’, ‘hamburger’, ‘essay’, ‘skip’, and ‘hotel’. This gender instability is not uncommon in cases of language contact and it is also attested by Corbett (1991: 75) and Poplack and Sankoff (1984: 124) who also consider it a transitory stage. The gender assignment system in Maltese seems to be undergoing change, especially because of the numerous English borrowings. Corbett (1991: 82) believes that the accumulated effect cause by borrowings may lead to considerable changes in the gender system of the language which receives them. It remains to be seen whether this situation, in the case of Maltese, is only an ephemeral one, lasting only until these nouns are assigned a specific gender, or whether it will develop into what is usually termed a “delayed effect” of language contact (Hickey 2001, Thomason 2006). Such an effect does not involve a structural upheaval in the recipient language but a gradual change due to prolonged contact between two languages. These new strategies are challenging an already established system in which the noun-ending criterion for inanimate nouns has always played a major role in gender assignment.

Abbreviations Ar. CA F IPFV

It. M

MSA PL SG

Arabic Classical Arabic feminine imperfective Italian masculine Modern Standard Arabic plural singular

Languages in contact | 313

References Agius, Dionisius 1996. Siculo Arabic. (Library of Arabic Linguistics, Monograph 12). London: Kegan Paul International. Aikhenvald, Alexandra Y. 2003. Classifiers. A typology of noun categorization devices. Oxford: Oxford University Press. Aquilina, Joseph. 1987–1990. Maltese–English dictionary, vol. I & II. Malta: Midsea Books. Aquilina, Joseph. 1999–2000. English–Maltese dictionary, vol. I–IV. Malta: Midsea Books. Aquilina, Joseph. 2006. Concise Maltese–English, English–Maltese dictionary. Malta: Midsea Books. Bickerton, Derek. 1981. Roots of language. Ann Arbor: Karoma. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London: Routledge. Brincat, Joseph M. 2000. Il-Malti. Elf sena ta’ storja. Malta: PIN. Corbett, Greville G. 1991. Gender. Cambridge: Cambridge University Press. Fabri, Ray. 2009. To agree or not to agree: Suspension of formal agreement in Maltese. In Ray Fabri (ed.), Maltese linguistics: A snapshot. In memory of Joseph A. Cremona (1922–2003). (Il-Lingwa Tagħna – Our Language 1), 35–61. Bochum: Brockmeyer. Farrugia, George. 2003. Il-Ġens Grammatikali li Jieħdu fil-Malti n-Nomi ta’ Nisel Ingliż [Grammatical gender of Maltese nouns of English origin]. Malta: Mireva Publications. Farrugia, George. 2010. Il-Ġens Grammatikali fil-Malti. Ph.D.Thesis, University of Malta. Ferguson, Charles & Gumperz, John J. 1960. Introduction. In Ferguson, Charles & Gumperz, John J. (eds.), Linguistic diversity in South Asia, 1–18. Bloomington: Indiana University Research Center in Anthropology, Folklore and Linguistics. Givón, Talmy. 1979. Prolegomena to any sane creology. In Ian F. Hancock (ed.), Readings in creole studies, 30–35. Amsterdam/Philadelphia: John Benjamins. Greenbaum, Sidney. 1996. The Oxford English grammar. New York: Oxford University Press. Gregor, Bernd. 1983. Genuszuordnung. Das Genus englischer Lehnwörter im Deutschen. (Linguistische Arbeiten 129). Tübingen: Niemeyer. Gumperz, John. 1964. Linguistic and social interaction in two communities. American Anthropologist 66(6, Part 2). 137–153. Haspelmath, Martin & Uri Tadmor (eds.). 2009a. Loanwords in the world’s languages: A comparative handbook. Berlin: Mouton de Gruyter. Haspelmath, Martin & Uri Tadmor (eds.). 2009b. World loanword database. Munich: Max Planck Digital Library, 1344 entries. http://world.livingsources.org/vocabulary. Haugen, Einar. 1954. Review of Weinreich (1953). Language 30. 380–388. Hickey, Raymond. 2001. Language contact and typological difference. Transfer between Irish and Irish English. In Dieter Kastovsky & Arthur Mettinger (eds.), Language contact and the history of English, 131–169. Frankfurt: Peter Lang. Holes, Clive. 1990. Gulf Arabic. London: Routledge. Ibrahim, M. H. 1973. Grammatical gender, its origin and development. The Hague: Mouton. Owens, Jonathan. 2010. What is a language? (Review of Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics. Amsterdam: Benjamins). Journal of Language Contact 3. 103–118. Poplack, Shana, Alicia Pousada & David Sankoff. 1982. Competing influences on gender assignment: Variable process, stable outcome. Lingua 57. 1–28.

314 | George Farrugia

Poplack, Shana & David Sankoff. 1984. Borrowing: The synchrony of integration. Linguistics 22. 99–135. Procházka, Stephan. 2004. Unmarked feminine nouns in Modern Arabic dialects. In Martine Haak, Rudolf de Jong & Kees Versteegh (eds.), Approaches to Arabic dialects. A collection of articles presented to Manfred Woidich on the occasion of his sixtieth birthday, 237–262. Leiden: Brill. Stolz, Christel. 2009. A different kind of gender problem: Maltese loanword gender from a typological perspective. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 321–353. Amsterdam/Philadelphia: Benjamins. Stolz, Thomas. 2012. Survival in a niche. On gender-copy in Chamorro (and sundry languages). In Martine Vanhove, Thomas Stolz, Aina Urdze & Hitomi Otsuka (eds.), Morphologies in contact, 93–140. Berlin: Akademie Verlag. Surridge, Marie E. 1982. L’attribution du genre grammatical aux emprunts anglais en français canadien: le rôle des homologues et des monosyllabes. Glossa 16. 28–39. Thomason, Sarah G. 2006. Language change and language contact. In Keith Brown (ed.), Encyclopedia of language and linguistics, vol. 6, 339–346. Oxford: Elsevier. U Sicilianu, Sicilian grammar, Web. 1 May 2013. Van Berkum, Jos J. A. 1996. The psycholinguistics of grammatical gender. Utrecht: Nijmegen University Press. Weinreich, Uriel. 1951. Research problems in bilingualism, with special reference to Switzerland. Columbia University dissertation. Winter, Werner. 1973. Areal linguistics: Some general considerations. In Thomas A. Sebeok (ed.), Current trends in linguistics. Vol. 11, 135–147. The Hague: Mouton.

Bernard Comrie and Michael Spagnol

Maltese loanword typology Abstract: Etymological analysis of the Maltese lexicon within the project Loanwords in the world’s languages shows Maltese to be a “high borrower”, with a proportion of loanwords (35.1%–30.3% from Romance, 4.8% from English) roughly comparable to that found in Japanese (34.9%) and slightly lower than that found in English (41.0%). The breakdown according to semantic field and part of speech yields largely expected results, though Maltese has an unusually high percentage of borrowed kinship terms. A number of individual examples are discussed to illustrate particular methodological points. Keywords: Maltese, language contact, loanwords, typology

1 Introduction The two authors are currently conducting a research product on the etymological composition of the Maltese lexicon within the project Loanwords in the world’s languages (see Section 1.2). The present article is an interim report on certain aspects of this project and its results. We emphasize that the project is still in progress, so that details of the results, especially of the statistics in Section 3, may change, but probably only slightly, given the advanced stage of the project.

1.1 The etymological composition of the Maltese lexicon The basics of the etymological composition of the Maltese lexicon are uncontroversial, with an original Arabic base to which loanwords from Romance (Sicilian/Italian) and more recently English1 have been added. The structural implications of the Arabic versus Romance divide are set out in Aquilina (1959), while

|| 1 We include only English loans that occur in non-code-switched Maltese, irrespective of degree of assimilation to Maltese phonology. || Bernard Comrie: University of California Santa Barbara, Department of Linguistics, Santa Barbara CA 93106-3100, USA. E-mail: [email protected] Michael Spagnol: University of Malta, Old Humanities Building, Room 207, L-Imsida MSD 2080, Malta. E-mail: [email protected]

316 | Bernard Comrie and Michael Spagnol

Brincat (2011) provides a detailed history of the language emphasizing its cultural environment. The brief chronology that follows is based on the latter, which includes a summary table (Brincat 2011: xxxv). There is no clear evidence for the survival of any linguistic material in the Maltese language predating the establishment of Arabic in the islands. While Arabic may have had a continuous presence in Malta from the invasion of 870, Brincat’s work suggests that the starting point may rather have been the immigration of 1048–49 from Sicily, and it also makes clear that the Arabic component of Maltese must be viewed in the context of the general development of Arabic in Sicily (Siculo Arabic; Agius 1996). Close contact between Sicily and Malta on the one hand and the rest of the Arabic-speaking world on the other would have largely come to an end during the first half of the thirteenth century with the expulsion/conversion of Muslims, while strong Romance influence, initially from Sicilian, might have started with the Norman conquest of Malta in 1090, certainly with the consolidation of Norman rule and the Christian religion during the twelfth century. While Italian has had a presence on the islands as an official language from its cession to the Knights of Malta in 1530, the shift from Sicilian to Italian among bilingual Maltese did not really take off until the later eighteenth century, so that Sicilian effectively retained a presence until around 1900. The establishment of British administration around 1800 led to the inclusion of English into the already complex linguistic situation, leading by the time of Maltese independence in 1964 to widespread Maltese-English bilingualism and a diminution of the role of Italian, although the geographical and cultural closeness of Italy and the ready availability of Italian broadcasts have been important factors in guaranteeing a continuing presence of the language. In analyzing the etymological composition of the Maltese lexicon, it is therefore necessary to distinguish the following strata, in chronological order although with at times considerable overlaps: Arabic, Sicilian, Italian, English. For reasons that will be made clearer in Section 2.1, it is often difficult to impossible to distinguish between the Sicilian and Italian components, so for most purposes we combine them into a single component labeled “Romance”.

1.2 Maltese in the project Loanwords in the world’s languages The question “What proportion of the vocabulary of language X comes from each etymological source?” is one that seems reasonable to the lay person, but gives immense practical problems to the linguist. There are two issues. First, how do we measure, for instance, the proportion of vocabulary in language X from source Z? Second, how do we provide comparability when we say that

Maltese loanword typology | 317

languages X and Y have each borrowed a certain proportion of their vocabulary from various sources? One way of measuring the proportion of vocabulary from different sources is by counting the relevant number of entries in a dictionary of the language that gives etymologies. On this basis Brincat (2011: xxxv) calculates that about 32.4% of Maltese vocabulary is of Arabic origin, about 52.5% borrowed from Romance, and about 6.1% borrowed from English, leaving a residue of about 9%. These are certainly meaningful statistics. But nonetheless, their significance needs to be interpreted. Larger dictionaries typically contain lots of words that are specialized, archaic, or otherwise restricted in their use and therefore known to only a small percentage of the speakers of a language. This is clearly the case with such large dictionaries as the Oxford English dictionary, with its large number of rare Latinate vocabulary (among others), but is also true, even if to a lesser extent, of the larger Maltese dictionaries, such as Aquilina (1987– 1990, 1999–2000) and Serracino Inglott (1975–1989). This is of course an inevitable feature of such dictionaries, which aim for comprehensiveness in documenting the language’s vocabulary over recorded history. But it does mean that while the proportions identified in some sense characterize the language, they do not characterize the average, or indeed perhaps any individual speaker’s knowledge of the language. Different dictionaries of the same language will give different breakdowns, for instance if they are of different sizes, even if they follow the same general lexicographic principles. And of course results culled from dictionaries of different languages are likely to be incomparable, especially if they are of different sizes, reflecting, for instance, different amounts of accumulated research on the lexicon of each language. A second way would be to examine the frequency of words of different etymological origins in a corpus of a language. Especially given the rapid development of corpus linguistics over the past couple of decades, this may well be the best single answer that can be given to the questions posed at the beginning of Section 1.2. In making comparisons across languages, however, one needs to ensure that the corpora are comparable. There might well be problems, for instance, in taking a written corpus of one language and comparing it with a spoken corpus of another language, although even different policies regarding the composition of, say, two written corpora could lead to problems of comparability. While the development of Maltese corpora, and digital language resources more generally, is currently making rapid progress, the authors of this article did not have at their disposal, while writing the article, materials for Maltese that could be considered adequate to the task. In a few years this situation will change, and it will then be possible to take a new look, from a different perspec-

318 | Bernard Comrie and Michael Spagnol

tive, at the general question of the etymological composition of the Maltese lexicon. In order in particular to achieve cross-linguistic comparability, we situated our research on the etymological composition of the Maltese vocabulary within the framework of the project Loanwords in the world’s languages, led by Anthony Grant, Kim Schulte, and Uri Tadmor;2 this is itself a successor to the project Loanword typology, which gave rise to the publications Haspelmath and Tadmor (2009a, 2009b). This project operates with a fixed list of about 1500 lexical meanings, belonging to various semantic fields and reflecting various parts of speech. The list is based ultimately on Buck (1949), via the Intercontinental Dictionary Series (Borin et al. 2014). While claims about the general “representativeness” of the list would require extensive investigation and substantiation, its use does satisfy a number of requirements for cross-linguistically comparable etymological investigations: a) the list of meanings is the same irrespective of the language under investigation; b) the list covers a wide range of vocabulary in terms of both semantic field and part of speech; c) the list includes both lexical meanings that are known to be prone to borrowing as well as those that are known to be resistant to borrowing, cross-linguistically, as well as various intermediate possibilities – for justification of this last point, see Tadmor (2009: 65–74). The basis on which all statistical claims are made in the present article is thus this list of 1500 lexical meanings translated into Maltese, with subsequent investigation of the etymology of each Maltese item. We plan to publish the full database, in as close to final form as we can get, and an overview article within the Loanwords in the world’s languages project. To avoid confusion, we emphasize that the aim of the Loanwords in the world’s languages project is to achieve cross-linguistic comparability, whence the use of the same list for all languages. While many of the words in the list belong to unequivocal core/basic vocabulary that reflect universals of the human situation, the need to consider words highly susceptible to borrowing for comparison has necessarily meant the inclusion of words outside such core/ basic vocabulary. Moreover, where realia differ across cultures, it is necessary to ensure that the selection of words does justice to this variation; thus, if ‘horse’ is included because of its salience in the Old World, it is balanced by ‘tapir’ because of its salience in Amazonia. In practice, the investigation followed the following stages. First, and prior to formal collaboration between the two authors, Comrie made a rough version

|| 2 See https://lingweb.eva.mpg.de/loanwords/index.php/Main_Page, last consulted on 2015 November 20.

Maltese loanword typology | 319

of the Maltese word list by translating the English labels of the 1500 lexical meanings into Maltese, using primarily Aquilina (1999–2000), though rechecking with Aquilina (1987–1990) and Serracino Inglott (1975–1989) to ensure, for instance, that the correct meaning of polysemous English words was taken. When Spagnol joined the project in 2011, he reviewed this list thoroughly and made substantial changes, so that the Maltese word list on which this article is based is his list (though with input from Comrie on some policy issues regarding choices); it thus reflects essentially the vocabulary used by younger urban Maltese-speakers. While this meant abandoning a number of the terms given in Aquilina (1999–2000), this is not of course in any way a criticism of this dictionary: first, the Maltese language has changed in the years since Aquilina’s work on his magna opera,3 with old words going out of use and new ones coming into being; second, Aquilina’s dictionaries are intended as comprehensive accounts of the Maltese lexicon across its recorded history, and not just as a description of contemporary usage. In general, where there is a single Maltese lexical item that is clearly the best translation of the given lexical concept, we went with this item. However, we did not attempt to simplify situations where more than one Maltese word serves as a legitimate translation in contemporary usage, so that some of the 1500 entries have more than one Maltese equivalent, each of which counts for the relevant fraction of a point in statistical analyses, i.e. if there are four Maltese equivalents for a given lexical item, each counts as 0.25 of an item. Where there are preferred and less preferred Maltese equivalents, it might have been considered desirable to weight these differently, e.g. one of two equivalents might be 0.75 and the other 0.25. But this would have involved us in subjective judgments that would have been hard to justify, at least in the current absence of extensive corpora that would justify the weighting. It will be useful to give some examples of items for which Spagnol’s contemporary native-speaker knowledge led us either to change our Maltese entry, or at least to weight alternatives differently in our own minds. For ‘bicycle’, Comrie had initially selected the English loan bajsikil, to which Spagnol was able to add the correction that rota is the more usual word; it also means ‘wheel’ and derives from Italian/Sicilian rota ‘wheel’.4 For ‘lion’, Comrie had both Romance-origin ljun and Arabic-origin dorbies, to which Spagnol added the observation that the latter is definitely archaic; it was therefore excluded, given our restriction to contemporary usage. In the case of ‘live, dwell’, the list of equiva-

|| 3 Joseph Aquilina died in 1997; his last works are posthumous. 4 The semantic extension from ‘wheel’ to ‘bicycle’ may be specifically Maltese; Joseph Brincat points out to us that this extension seems unknown in standard Italian or in dialects of Italy.

320 | Bernard Comrie and Michael Spagnol

lents was extended to four – Arabic-origin joqgħod, għex, and għammar, and Romance-origin abita – though noting that abita is more formal, while għammar is both formal and somewhat archaic. For ‘vomit’ we likewise ended up with a list of four items, Italian-origin5 rremetta and rriġetta, and Arabic-origin qala’ (also ‘to get’) and tella’ (also ‘to raise’). Sometimes the correction did not affect the etymology but nonetheless may be important not only for accuracy but also for detailed historical analysis. Thus, on the basis of Arabic ḥājib ‘eyebrow’, Comrie had assumed that the Maltese reflex ħaġeb would have a long first vowel. In fact, the word is rarely used in the singular, as opposed to the plural (pseudo-dual) ħuġbejn, where the vowel is anyway short as unstressed, and many speakers are unsure of the quantity of the stressed vowel in the singular or even prefer a short vowel, as already noted by Serracino Inglott (1975–1989: IV, 21).6 In some cases Spagnol’s insider knowledge enabled us to clarify an etymology. For instance, while Maltese tapir ‘tapir’ is clearly attested before the mid1990s – it is listed by both Serracino-Inglott (1975–1989: VIII, 348)7 and Aquilina (1987–1990: 1400)8 – its entrenchment in the Maltese lexicon is certainly due to the satirical “Tapiro d’Oro” ‘Golden Tapir’ award, introduced into the popular (also in Malta) Italian television program Striscia la notizia ‘The news slithers’ from 1996 (http://en.wikipedia.org/wiki/Striscia_la_notizia#The_Tapiro_d.27 Oro, last consulted on 2014 September 29). This has led us to identify the Maltese word, as it occurs in current usage, as of Italian origin. Incidentally, Serracino-Inglott opts for an Italian origin, while Aquilina opts for English. The next step was to ascertain the etymology of each Maltese item. By the “etymology” we mean, incidentally, the immediate source; thus, Maltese dixx ‘dish’ is a loan from English, even though the English word is in turn a loan from Latin discus (into Proto-West Germanic), which is in turn a loan from Greek dískos. The Maltese word arose in the contact between Maltese and English, and this is what is of concern to us here. Although the project actually only requires identification of the source, e.g. Arabic, Romance (or more specifically: Sicilian, Italian), English, we tried wherever possible to find specific words in the source languages that can be related via plausible phonological and semantic links to the current Maltese items. Various problems that arose in this part of the project are discussed in Section 2, although we note here that it was in general possible

|| 5 The origin seems to be specifically Italian, since this sense is not recorded in Sicilian for either item in Piccitto et al. (1977–2002). 6 This volume was published in 1977. 7 This volume was published in 1984. 8 The word occurs in the second volume, published in 1990.

Maltese loanword typology | 321

to identify Maltese items as being unequivocally of Arabic, of Romance, or of English origin. Cases of words that could, as far as we can tell, equally be of more than one of these origins are exceedingly few. One is forn ‘oven’. The word is clearly ultimately from Latin furnus, but as the word is widely attested in Arabic as furn, it could have entered Maltese along with the language’s original Arabic word stock, but equally it could have been a later loan from Romance. In such cases, we assign 0.5 to each of Arabic-origin and Romance-origin. A similar case is banana ‘banana’, where we are aware of no information favoring English or Italian as the donor language. In some cases, even for some quite common Maltese words, the etymology is uncertain, e.g. ċpar ‘mist’, and such words we assign to a category “Other”. However, this category is much broader than just words of uncertain origin, which accounts for some of the high percentages of “Other” in the tables in Section 3, and means, for better or worse, that this category must also be interpreted carefully, or perhaps better: not interpreted at all, as it is too heterogeneous to be useful. In principle, it would include items that are not from any of the three sources (Arabic, Romance, English), though on our definition of these terms (see below) there are in fact no such items in the list. It also includes lexical meanings in the list for which we found no specific Maltese equivalent, either because the referent is too exotic to Maltese culture to have merited a specific designation, or because the Maltese equivalent would be a paraphrase rather than a distinct lexical item; this, incidentally, accounts for the large percentage in Table 3 under Kinship, given that the list of lexical meanings includes many such terms as ‘maternal uncle’ for which Maltese, like English, has no specific lexical item.9 Finally, it also includes derivatives of borrowed Romance words (and in principle English words, though we found no such examples) that were created within Maltese, since although the root is Romance, the resulting word is not itself a loan from Romance. Good examples would be neputija ‘niece’ and pejjep ‘to smoke (e.g. a pipe)’. Both neputi ‘nephew’ and pipa ‘pipe’ are loans from Romance, but no derivative that could serve as a source for neputija is known to us in Romance, more specifically Sicilian or Italian, while pejjep is even more transparently a Maltese creation, reinterpreting Romance pipa as a triliteral Semitic root √PJP. While the cases discussed in this paragraph are relatively clear, there are borderline cases, although we have tried to be consistent in our judgments, and the number of residual problems is likely to be very small relative to the overall list. Incidentally, where Romance

|| 9 At least, this is true of our reference variety of Maltese, that of younger urban speakers; Joseph Brincat points out to us that rural dialects do have a distinct term barba, from Italian.

322 | Bernard Comrie and Michael Spagnol

items have undergone semantic shifts (such as ‘wheel’ > ‘bicycle’, discussed above), we count these as Romance-origin, even in the new meaning, on the basis that the lexical item itself is as a whole of Romance origin. We take as being “Arabic” the whole of the lexicon that would have characterized Siculo Arabic as spoken in Malta, plus later derivatives. This thus includes loanwords that had already been assimilated into this variety of Arabic, such as gendus ‘bull’, ultimately from Berber (cf. Kabyle agenduz ‘calf’). It should be noted that our identification of items as being specifically Arabic-origin, i.e. indigenous, departs from the general principles of the Loanwords in the world’s languages project by being more specific. In the project, items are identified as being loans or not, with the former then further subdivided according to the specific source, but items that are thus qualified as “not loans” include not only indigenous vocabulary but also most of what we classify as “Other”. In other words, the identification of specifically indigenous vocabulary is not required by the project, but is, in our judgment, useful additional information. Any study of the history of the Maltese language is limited by the paucity of information on the language going back more than a couple of centuries. On the one hand, this means that one must make full use of those sources that are available, as when Aquilina (1987–1990: 161) correctly notes that Maltese ċajta ‘joke’ cannot be a loan from English chat since the word is already attested in Vassalli’s dictionary of 1796, thus predating contact with English.10 Current work being carried out on materials in Maltese archives promises to enrich considerably our understanding of the detailed linguistics of earlier stages of the island’s history, and may well resolve some issues that we have had to leave open.

2 Recent work on the components of the Maltese lexicon In general, we were able to establish unequivocally the etymologies of Maltese items corresponding to the lexical meanings of the project list, in particular identifying items as being Arabic-, Romance-, or English-origin. For English and Italian, this occasionally involved going beyond our own combined knowledge of the languages to consult standard dictionaries and other sources. Nonetheless, there do remain some problem cases, and in this section we note both in-

|| 10 As pointed to us by Joseph Brincat, the word is in fact from Sicilian ciàita, attested in Piccitto et al. (1977–2002).

Maltese loanword typology | 323

stances where access to resources that have only recently become available enabled us to make firm or at least highly plausible decisions, as well as discussing some of the remaining problem cases.

2.1 The Romance component While the identification of items on the list as being of Romance origin is generally unproblematic, major problems arise if one tries to identify items as being specifically Sicilian versus Italian in origin.11 The problem is that the two languages are very close in their vocabularies, in both form and meaning, so that especially after items have undergone phonetic transformation into Maltese it is often difficult or even impossible to decide on linguistic grounds that a particular item is of the one rather than the other origin. Until recently there was the added problem that there was no really reliable dictionary of Sicilian, with purported Sicilian dictionaries often being Italian words translated into Sicilian rather than genuine compilations of the lexical resources of Sicilian. The appearance of Piccitto et al. (1977–2002) provides us with an invaluable resource in investigating the Sicilian origins of Maltese vocabulary. In some cases, a close reading of this dictionary, in comparison with lexical resources for Italian, enables a reasonably clear decision in favor of a Sicilian etymon. Consider, for instance, Maltese brazzuletta ‘bracelet’. Sicilian bbrazzulettu ‘bracelet’ (Piccitto et al. 1977–2002: I, 444) provides the closest match. While Maltese often retains and even extends Sicilian geminate initials in verbs (and extends them to Italian and even English loans), they are frequently lost in nouns, so the initial consonant is not a serious problem (cf. Mifsud 1995: 142–168). The gender is different in Maltese and Sicilian, but since both Sicilian and Italian have masculine gender (versus feminine in Maltese), this does not make Sicilian origin any less plausible than Italian. The medial zz [tts] is decisive in favor of Sicilian rather than Italian, which has braccialetto (where cci is [ttʃ]), reinforced also by the pretonic vowel u rather than a. In some cases where we were initially uncertain the decision fell in favor of Italian. The Maltese adjective sorpriż ‘surprised’ is initially puzzling in comparison with Italian sorpreso, both in that Maltese has intervocalic [z] while at least standard Tuscan has [s] – though the realization of earlier intervocalic single s in several varieties of Italian, including the Tuscan-based standard, is notori-

|| 11 We do not exclude the possibility that some Romance items might be from neither Sicilian nor Italian, as argued for instance by Baglioni (2014).

324 | Bernard Comrie and Michael Spagnol

ously variable (Rohlfs 1966 [1949]: 281–285) – and even more so through the stressed vowel quality [i] in Maltese corresponding to [e] in Italian. But we found nothing comparable in Sicilian in Piccitto et al. (1977–2002). A more detailed search of the Italian materials revealed an archaic alternative Italian form sorpriso, with intervocalic [z] (Dogliotti and Rosiello 1988: 1846), which is an exact match for the Maltese form. And in some cases we are left with forms that are clearly Romance but which do not match any of the Italian or Sicilian forms known to use. Thus, Maltese sufarina or sulfarina ‘match (for lighting)’ is not an exact match to any of Italian solfanello, solfino, zolfanello, zolfino (Dogliotti and Rosiello 1988: 2190), or Sicilian suffareḍḍu, suffarellu, suffarièḍḍu, sulfaloru, sulfarolu, surfaloru, surfareḍḍu, surfarièḍḍu, surfarolu (Piccitto et al. 1977–2002: V, 426–427, 434, 460, 461) This still leaves a huge set of clearly Romance-origin Maltese words where we are unable to decide on any principled and reliable basis whether the word is more specifically of Italian or Sicilian origin. This is the reason why in the statistics we have given “Romance” as a single category. As an experiment, Comrie took all the words of Romance origin and then classified them as either Italian or Sicilian, first using a forced-choice methodology (i.e. every word had to be classified one way or the other), then using a non-forced choice, which meant that words could remain as undecidedly Italian/Sicilian. The first procedure resulted in 47.8% Italian to 52.2% Sicilian; the second to 11.7% Italian, 7.4% Sicilian, and a whopping 80.9% undecided. It should be emphasized that this was only an experiment, although it does suggest that for the words on the list Sicilian and Italian are roughly equally likely contributors to the Loanwords in the world’s languages Maltese vocabulary list, though with a wide margin of error.

2.2 The Arabic component With the Arabic component, it is relatively rare for a previously unclear word to receive a clear etymology, but one such example is dgħajsa ‘boat’, which finds an exact counterpart in Sfax (Tunisia) dɣīsa ‘kind of sailing boat used in Sfax for commercial purposes, and which linked this city with Eastern Mediterranean ports’ (Ziwārī and Šarfī 1998: 259; Zammit 2014: 38). The further etymology of the word remains unclear, but at least we can be sure that it would have been part of the Arabic that took hold in Malta. Another example is Maltese liema ‘which?’, clearly to be related to Tunisian Arabic ǟmǟ ‘which’ (Ben Abdelkader et al. 1977: 443) – again, the deeper etymology remains unclear, but at least this is no longer a specifically Maltese problem.

Maltese loanword typology | 325

In a number of cases recent detailed work on Arabic vernaculars has provided closer fits to Maltese semantics than are to be found in the usual Arabic lexical resources. For instance, Maltese maħmuġ ‘dirty’ (with long u) seems clearly relatable to Arabic xamaja ‘to spoil, rot, decay’, but Moroccan Arabic məxmūž ‘dirty’ (Iraqui Sinaceur 1993–1994: 473) provides an exact semantic parallel. Sometimes the parallel is formal, as with Maltese sħun ‘warm’ (with long u), relatable to Arabic forms like suxn, sāxin ‘hot, warm’, but with an exact formal counterpart in Tunisian Arabic sxūn (Ben Abdelkader et al. 1977: 190). Maltese artab ‘soft’ deviates both semantically and formally from Arabic raṭb ‘moist, fresh, succulent’, but finds an exact match on both counts in Tunisian arṭib ‘soft’ (Ben Abdelkader et al. 1977: 368). For the common Maltese word wisq ‘too’, Aquilina (1987–1990) notes possible links to two Arabic roots, √WSʕ ‘wide’ and √WSQ ‘to load’, though the former is problematic in its third consonant; Moroccan Arabic wāsəq ‘full to bursting, jam-packed’ (Iraqui Sinaceur 1993– 1994: 2056) provides an intermediate link between the senses ‘to load’ and ‘too’. Expanding lexical resources on Arabic vernaculars is a prime desideratum for Arabic studies, and further developments in this area will surely enable us in the long term to locate Maltese better among regional variation in Arabic as a whole.

3 Results After these methodological concerns, we can now turn to the statistical results of the etymological investigation. It should again be emphasized that the project is still in progress, so there may be some changes before the final version is ready, though these are unlikely to affect the general trend of the figures. In terms of the total list of about 1500 items, the vocabulary breaks down as shown by the percentages in Table 1. A clear majority of the vocabulary is of Arabic origin, just under a third of Romance origin, just under 5% of English origin. Tab. 1: The etymological composition of the Maltese lexicon: Overview.

Arabic

56.0

Romance

30.3

English

4.8

Other

8.9

326 | Bernard Comrie and Michael Spagnol

The total percentage of loanwords, 35.1%, puts Maltese clearly among the “high borrowers” in the terms of Tadmor (2009: 56), from whom the other figures in Table 2 are taken (p. 56). It falls just above Japanese, well known for its intensive traditional borrowing from Chinese, as well as more recent borrowing from European languages, in recent decades primarily English. Tab. 2: Maltese among other “high borrowers”.

Tarifiyt Berber

51.7

Gurindji

45.6

Romanian

41.8

English

41.0

Saramaccan

38.3

Coq Wong

37.0

Maltese

35.1

Japanese

34.9

Indonesian

34.0

Bezhta

31.8

Kildin Saami

30.5

Imbabura Quechua

30.2

Archi

29.5

Sakha

29.0

Vietnamese

28.1

Swahili

27.8

Yaqui

26.5

Thai

26.1

Takia

25.9

Table 3 shows the distribution of indigenous and borrowed vocabulary across the 24 semantic fields of the Loanwords in the world’s languages project. Comparison with the “world averages” table in Tadmor (2009: 64), or indeed with received wisdom on the susceptibility of different semantic fields to borrowing, reveals few surprises, perhaps the most striking one being the high percentage of borrowed kinship terms in Maltese: 27.5%, against a world average of only 15.0%; the fact that Maltese has borrowed its word for ‘father’, missier, from Romance really is the tip of an iceberg.

Maltese loanword typology | 327

Tab. 3: The etymological composition of the Maltese lexicon: Semantic fields.

Arabic

Romance

English

Other

Quantity

84.2

13.2

0.0

2.6

Miscellaneous function words

82.8

5.6

0.0

11.6

Sense perception

75.5

18.6

2.0

3.9

The body

67.3

18.9

2.5

11.3

Basic actions and technology

66.9

26.1

2.6

4.4

Emotions and values

65.7

32.3

1.0

1.0

Motion

65.7

19.8

6.1

8.5

Food and drink

64.3

22.3

8.2

5.3

Spatial relations

63.2

26.5

1.3

9.0

Speech and language

62.1

31.0

2.3

4.6

The physical world

59.9

25.3

8.4

6.3

Possession

58.8

32.1

2.2

6.9

Time

58.2

30.7

0.0

11.1

Cognition

54.6

33.1

1.1

11.2

Agriculture and vegetation

52.7

34.1

6.2

7.0

Social and political relations

48.4

48.4

0.0

3.2

Animals

47.8

29.1

13.9

9.1

The house

44.8

43.5

9.6

2.1

Religion and belief

43.3

46.0

2.7

8.0

Kinship

38.8

26.9

0.6

33.8

Clothing and grooming

38.7

47.2

10.4

3.8

Law

36.0

50.0

0.0

14.0

Warfare and hunting

28.8

65.0

2.5

3.8

Modern world

3.0

65.3

22.8

9.0

Finally, Table 4 shows the distribution across parts of speech (excluding adverbs, represented by only five items). They confirm the nearly, but not quite universal cross-linguistic tendency for function words to be less borrowable than content words, and for verbs to be less borrowable than nouns (Tadmor 2009: 59–63).

328 | Bernard Comrie and Michael Spagnol

Tab. 4: The etymological composition of the Maltese lexicon: Parts of speech.

Arabic

Romance

English

Other

Function word

84.7

6.2

0.0

9.1

Verb

75.3

14.1

1.3

9.2

Adjective

65.2

28.5

0.3

6.0

Noun

44.7

39.6

7.2

8.6

4 Conclusions In this article we hope to have shown that it is possible to analyze loanwords in Maltese against a cross-linguistic, typological background, thus providing meaningful comparisons with other languages, both with respect to the overall percentage of loanwords and to individual parts of speech and semantic fields. It is also clear where future work needs to fill in gaps in our knowledge of the history of the Maltese lexicon. Acknowledgments: An earlier version of this material was presented at the Fourth International Conference on Maltese Linguistics held at the University of Lyon II in 2013, in turn building on an even earlier version presented by Comrie at the Third International Conference on Maltese Linguistics held at the University of Malta in 2011. We are grateful to all those who participated in the ensuing discussions, and especially to Joseph Brincat and Martin Zammit for hints on treatment of the Romance and Arabic components respectively. Joseph Brincat also provided detailed comments on an earlier version of the manuscript. We also thank the two anonymous referees for their helpful suggestions. All remaining errors and deficiencies are our own responsibility. This work was supported by the Max Planck Society.

Maltese loanword typology | 329

References Agius, Dionisius A. 1996. Siculo Arabic. London: Kegan Paul International. Aquilina, Joseph. 1959. The structure of Maltese. Malta: University of Malta Press. Aquilina, Joseph. 1987–1990. Maltese-English dictionary. 2 volumes. Malta: Midsea Books. Aquilina, Joseph. 1999–2000. English-Maltese dictionary. 4 volumes. Malta: Midsea Books. Baglioni, Daniele. 2014. Italian loanwords in Maltese and the vocabulary of “Levant Italian”. In Albert Borg et al. (eds.), 45–58. Ben Abdelkader, Rached, Abdeljelil Ayed & Aziza Naouar. 1977. Peace Corps English-Tunisian Arabic dictionary. Washington DC: Peace Corps. Borg, Albert, Sandro Caruana & Alexandra Vella (eds.). 2014. Perspectives on Maltese linguistics. Berlin: Akademie Verlag. Borin, Lars, Bernard Comrie & Anju Saxena. 2014. The intercontinental dictionary series – A rich and principled database for language comparison. In Lars Borin & Anju Saxena (eds.), Approaches to measuring linguistic differences, 285–302. Berlin: De Gruyter Mouton. Brincat, Joseph. 2011. Maltese and other languages: A linguistic history of Malta. (Maltese Social Studies Series 19). Malta: Midsea Books. Buck, Carl Darling. 1949. A dictionary of selected synonyms in the principal Indo-European languages: A contribution to the history of ideas. Chicago: University of Chicago Press. Dogliotti, Miro & Luigi Rosiello (eds.). 1988. Il nuovo Zingarelli: Vocabolario della lingua italiana di Nicola Zingarelli. 11th edition. Bologna: Zanichelli. Haspelmath, Martin & Uri Tadmor (eds.). 2009a. Loanwords in the world’s languages: A comparative handbook. Berlin: De Gruyter Mouton. Haspelmath, Martin & Uri Tadmor (eds.). 2009b. The world loanword database. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wold.clld.org/ (checked 2015 February 13). Iraqui Sinaceur, Zakia. 1993–1994. Le dictionnaire Colin d’arabe dialectal marocain. 8 volumes. Rabat: Al-Manahli. Mifsud, Manwel. 1995. Loan verbs in Maltese: A descriptive and comparative study. Leiden: Brill. Piccitto, Giorgio, Giovanni Tropea & Salvatore Carmelo Trovato. 1977–2002. Vocabolario siciliano. 5 volumes. Catania etc.: Centro di Studi Filologici e Linguistici Siciliani. Rohlfs, Gerhard. 1966. Grammatica storica della lingua italiana e dei suoi dialetti: Fonetica. Torino: Einaudi.12 Serracino Inglott, Erin. 1975–1989. Il-miklem Malti. 10 volumes. Malta: Klabb Kotba Maltin. Tadmor, Uri. 2009. Loanwords in the world’s languages: Findings and results. In Martin Haspelmath & Uri Tadmor (eds.), 55–75. Zammit, Martin R. 2014. The Sfaxi (Tunisian) element in Maltese. In Albert Borg et al. (eds.), 23–44. Ziwārī, ‘Alī az- & Yūsuf aš- Šarfī. 1998. Mu‘jam al-kalimāt wa-t-taqālīd aš-ša‘bīyah bi-Ṣafāqis [Dictionary of folk words and traditions in Sfax]. Sfax: Al-Mu’allifūn.

|| 12 This Italian edition was revised by the author from the German-language original published in 1949.

Marie Azzopardi-Alexander

Language profiling: The weaving of Maltese and English in Maltese children’s conversations Abstract: The conversations of Maltese children aged between 1;10 and 5;6 are analyzed and five groupings are proposed according to the way Maltese and English are woven into their utterances. The children can be seen as occupying one of 5 stages along a Maltese to English continuum. At one end Maltese is used with only a sprinkling of English but most utterances in the conversations use only Maltese. At the other end English is used with more than a sprinkling of Maltese but most utterances are in English only. Two further groups using either Maltese only or English only – could be posited. The work points to practical applications. Keywords: language mixing, code-switching, children’s conversations, bilingualism continuum, morphophonology, syntax, language competence

1 Introduction – Maltese-speaking children This study focuses on the ways young Maltese children use English in their spontaneous conversation with adults. An attempt is made to consider the different ways English1 is woven into the conversation as pointers to places on the continuum of bilingualism of children in Maltese families. The children are grouped according to the way Maltese and English are threaded into their utterances as they converse with adults who address them in Maltese, Maltese which sometimes also includes some English element(s).2 There are many more Maltese-speaking children than those who live in Malta and Gozo. However, these are not included in this study. The kind of codemixing or code-switching (see 2.2 below) these “other” children use may or may || 1 The English referred to here is that spoken in Malta, conventionally labeled ‘Maltese-English’. 2 The data studied includes that collected by students working on language acquisition fieldwork. || Marie Azzopardi-Alexander: University of Malta, Institute of Linguistics, Tal-Qroqq, Msida MSD 2080, Malta. E-mail: [email protected]

332 | Marie Azzopardi-Alexander

not be very different from that seen in the data used here. However, there are added factors – and languages – that may affect the way they converse. The study does not include the growing number of trilingual, originally migrant, families who adopt Maltese and English without casting their original language aside when they settle in Malta. This group of Maltese-speaking children would need a separate study.

2 Languages in Malta The view adopted here is that one needs to consider language mixing “in context and as context” (Lanza 2004: 1). Therefore, I will first list a number of issues that must be raised in any study of the kind. First of all, it must be said that in spite of the fact that English (a Germanic language) is very different from Maltese (a Siculo Arabic language), and possibly partly because it is so different, it has influenced Maltese in a number of ways. The most obvious is the use of English vocabulary particularly for equipment and appliances, science and technology, a variety of subjects within educational programs and young people’s lifestyles. In his study of loan verbs in Maltese, Mifsud (1995: 26) considers the “English adstratum” as being made up “mainly of lexical material adopted by Maltese since 1800” and considers it as “practically the exclusive source of new terminology connected with modern life, especially in the field of science and technology”. Mori (2011) reports that in Maltese Eurolect, whilst Italian remains the main source for legal terms, “English is the major lexical source for verb creation.” Moreover, there is also “the widespread use in EU legal Maltese of Semitic-origin technical words and of a good number of compound noun phrases where Italian loanwords are combined with Arabic words.” (Mori 2011: 121) Over and above the vocabulary that can now be considered as part of Maltese and can be found written in Maltese orthography (e.g. pushchair or puxċer3), English words enter into everyday conversation across social classes although they are not needed to fill any gap since Maltese lexical items exist and are in common use for the same referent (e.g. washing machine for magna tal-ħasil, baby for tarbija, teacher for għalliema). Nevertheless, the bilingual speaker’s access to two lexicons may not be equal in all ways at all times – if ever. The fact that some situations require the use of one language rather than another means that both frequency and recency of use of certain lexical items

|| 3 All Maltese equivalents of English loanwords are written as listed in Aquilina (2006).

Language profiling | 333

will differ. This plays an important part in what I would like to refer to as spontaneous “casual retrievability”.

2.1 Code-mixing and code-switching Different educators, researchers and the general public have different views and attitudes to code-mixing and code-switching. There is a strong belief by some that people who code-mix or code-switch do so because they lack the vocabulary, or because they do not have full competence in the language or do not read enough. Others consider it an attempt by speakers to make an impression related to favoring English. Few regard code-mixing as a positive conversational strategy which bilingual Maltese (or other) speakers rely on for positive communicative reasons (Baker 2001: 101). In academic circles, the term code-switching is occasionally considered as a positive strategy expounded, for example, in the Abstract Level Model (Myers-Scotton and Jake 2014: 511, 516, 523). I hold this view where code-switching is used between speakers who are comfortable using both languages. This is the view reflected here. Switching from Maltese to English by adults will be considered as constituting code-switching and assumes that adult speakers can switch to the appropriate language with no disruption should some conversation participant be considered to be uncomfortable with English or judgmental about code-switching where Maltese items exist for the same referent. The phenomenon of children’s use of both lexicons in the same utterance will be referred to as code-mixing mainly to avoid making a decision as to the child’s selection from the “other” code even where children move from one code to the other with relative ease. In some cases, children alternate Maltese lexical items with the equivalent English items within the same conversation for the same referent, showing that the use of both languages in the same utterance does not occur only where there is a gap in one of the languages as was the case reported by Deuchar and Quay (2000: 113). This alternation sometimes appears to be linked to a trigger by another conversational participant. However, the question of conscious use will not be taken up in any depth. What concerns us is how, from a very early age, Maltese children code-mix in different ways. On the basis of this code mixing I would like to propose profiling that could inform both language assessments and clinical work and the grouping of children for educational purposes. It is important to note that the local adult population uses certain words in English in their everyday usage. Whereas most Maltese adults can use numbers in Maltese it is unlikely that they will do so when they count, refer to dates, do actual

334 | Marie Azzopardi-Alexander

calculations, or refer to numbers in casual situations. English is the language of Mathematics even in the Primary School. Only those needing to declare numbers – such as tellers in banks – will use both languages with some customers. It may be contradictory to state that this is a study of the range of infant bilingualism when we have, at one end, children whose use of English is limited and very predictable, and at the other end, children whose use of Maltese is limited. Nevertheless, the little use made of the “other” language marks the start of a journey that leads to a significant degree of competence in the spoken and, nowadays, the written forms of the languages. Naturally, their receptive competence is far more developed than their expressive competence. However, the gap will narrow considerably in the course of their language development. They will, moreover, go well beyond code-switching and their behavior would be better described as translanguaging: Translanguaging differs from the notion of code-switching in that it refers not simply to a shift or a shuttle between two languages, but to the speakers’ construction and use of original and complex interrelated discursive practices that cannot be easily assigned to one or another traditional definition of a language, but that makes up the speakers’ complete language repertoire. (Garcia and Wei 2014: 22)

2.2 Family groups Camilleri (1995: 84) identifies 4 types of families primarily on the basis of whether they acquire a dialect and whether Maltese and English are first or second languages: A. 1. Dialect 2. Standard Maltese 3. English B. 1. Standard Maltese 2. English C. 1. Standard Maltese and English D. 1. English 2. Standard Maltese This grouping will not be used in this study as the focus is entirely different. It is not clear whether all dialect speakers learn Standard Maltese – it may well be that a neutralized non-dialectal Maltese is involved (see Azzopardi-Alexander 2011 and Borg 2011). One could ask whether children speaking a dialect actually learn non-dialectal Maltese any differently from the way they learn English. Exposure to both increases considerably as the child grows older and interacts with more people outside the home. It should be noted that the children in the groups identified in this study include children from dialect-speaking families as well as from each of the other family types.

Language profiling | 335

It must be said that the number of Maltese parents who speak only English in the home is relatively small. Maltese is still used, possibly with the extended family and visitors. Parents who consider themselves as English speakers may start a conversation in English and continue it in Maltese in certain settings. In fact the children in these homes, though certainly more competent in English than the other groups at the age of 5, actually learn a great deal of Maltese from exposure in the home and in fact understand a significant amount because they come into contact with it in a meaningful way.

2.3 The media as language exposure It is important to note that the influence of cable television and films on DVD or as downloads are now an integral part of most Maltese households. Children therefore have extensive exposure to British, American, Australian and MalteseEnglish throughout the time they spend in the home whether the parents speak dialectal or Standard Maltese or English or all three.

3 The grouping of children on the basis of the language mixing in their conversations This study analyses the language of the conversations of children aged between 1.10 and 5.6 in the home. Five distinct groups are posited that redimension the outlook provided by Camilleri (1995) for different purposes. No distinction is made on the basis of whether dialectal or Standard Maltese or both are learned and used. Dialect speakers are included in every group. The language profiles that are presented here point to distinct groups on the basis of a primarily qualitative nature. Unfortunately, no claim can be made regarding the extent of coverage of the entire population. Before moving into the groupings it should be made clear that there is a number of English words used in ordinary adult conversation even where the adult chooses to use Maltese almost exclusively. These include 1) words for which no native word is used on a regular basis: terms for various items across subjects both where hardware is concerned (e.g. monitor, mouse, clutch, steering wheel, filter) as well as activities (names of sports as well as terms like fieldwork, invigilation) kinship terms (mummy, auntie) and 2) words commonly used as

336 | Marie Azzopardi-Alexander

alternatives to Maltese words in some settings e.g. baby, teacher. What determines the choice of one rather than the other merits a separate study.

3.1 Group 0.1 Maltese There are families for whom English does not feature as a regular part of their lives since they only meet speakers of English incidentally. Other Maltese they meet would not consider it appropriate to use English when addressing them even if English would be their choice at other times. For children in these groups English is just a school matter and they only engage with homeworkrelated English. Although this increases as they move up in school, they still would prefer to do everything in Maltese. At the ages of 0 to 5.11, English probably does not really feature in the home and they do not seem to use any English when addressed in Maltese, unlike the other groups considered here. These children therefore appear outside the five groups. This does not mean they do not join the journey to bilingualism which, for them, starts with kindergarten.

3.2 Group 1 – Maltese with restricted English The first group of children seems to be a very small group in comparison with most of the other groups. These children are addressed in Maltese very consistently at home. The parents do not appear to use the motherese first noted in Ellul (1978) but restrict themselves to Maltese. Some of these parents are very conscious of the code-switching that is widely used by their own peers and have made a conscious decision to use Maltese as the home language. They are keen to avoid language mixing of any kind. Nevertheless they are usually competent users of English even if they differ in the place on the continuum of MalteseEnglishness of their English. They often need to use English extensively at work and elsewhere outside the home though in some cases their work involves a great deal of Maltese too. They do not restrict their children’s exposure to English on the media and are not in any way opposed to the use of English around them. However, those who offered their opinion were very clear about wanting their child to use Maltese before worrying about English and were confident that the child would have no difficulty learning to use English competently. They encourage the child to watch local TV programs for children as well as some appropriate programs in English rather than having their child watching the TV at any time. They also read stories both in Maltese and in English to the chil-

Language profiling | 337

dren. Once the children attend school, exposure to English increases extensively but the exposure to Maltese remains strong. Whilst some of the parents of this group do not seem to be confident users of English and avoid using it, others are highly competent and confident users of English. Nevertheless, they all provide rich language exposure to their children in Maltese – the home language is Maltese. Moreover, they welcome opportunities for their children to listen to English in the media or to interact with English speakers. Whatever code-switching found in their children’s conversation seems to be limited to the usual numerals, colors, animals. Likewise, the children’s code-switching is very contained and predictable. The following examples are representative of these children’s use of English in their conversations with adults: (1.1)

jien għandi sixteen jien għand-i sixteen I have.IPFV-1SG sixteen ‘I have sixteen’ (crayons)

(1.2)

red dan red this-SG.M ‘this is red’

(1.3)

u noqogħdu nagħmlu tal-ġurdien Nicole l-ġurdien u n-oqogħd-u n-agħml-u ta-l-ġurdien Nicole l-ġurdien and 1-stay.IPFV-PL 1-do.IPFV-PL of-the-mice Nicole the-mouse jiena l-pussycat jiena l-pussycat I the-pussycat ‘and we play the mouse game .... Nicole plays the mouse and I cat’

(1.4)

u dak (tal-) u dak (tal-) and that.SG.M (of-the) ‘and that is the baby’s’

u u and

play the

tal-baby tal-baby of-the-baby

Maltese equivalents exist for these and hence the English versions can be considered as realizations in the alternative code. The children in this group not only understand but actually use the “equivalent” numbers, colors and animal names in Maltese in their conversations. It seems that their use of these in English is often associated with specific contexts such as direct questions by older persons that require a one-word answer e.g. being asked their age, or how many

338 | Marie Azzopardi-Alexander

objects they have, or what color something is or to label a picture of some animal. When they actually report the information themselves, as part of the conversation in which they are active participants, they tend to use the Maltese word(s). The small number of examples given to illustrate the use of English within these children’s Maltese utterances reflects their infrequent use of English words, i.e. some conversations are totally in Maltese. The examples in fact are typical of all of the utterances containing English in their conversations. The bulk of the utterances in their conversations is Maltese. The fact that these children understand and use numbers, colors and animal names in both languages indicates that they are already aware of two codes and possibly recognize their perception of the suitability of English for the more direct answers to certain questions which they may associate with formal education. Alternatively, it may be a reflection of the use of English for numerals, colors and animal names in language addressed to children by various adults even if not by their parents. The beginning of exposure to English through contact with school can be posited as a determining factor for the 3-year-olds and older. Younger children may also be influenced by siblings who have started school. The children still learn a great deal of English vocabulary particularly from the media. They become conscious of the “other” language, English, and become interested in it as though they are aware they can access alternative words for the same referents. This is clear from utterances like (1.5) and (1.6):4 (1.5)

x’jgħidulha xemgħa x-j-għid-u-l-ha xemgħa what-3-tell.IPFV-PL-OM-SG.F candle ‘what is candle in English granny?’

(1.6)

x’inhuma chindrin mummy? x-inhuma chindrin mummy? what-they ‘children’ mummy? ‘what are ‘children’ mummy?’

b’Ingliż b(l)-Ingliż with-(the)-English

nanna? nanna? granny?

Children in this group tend to use the Maltese terms mama and papa to call or refer to their parents. Nevertheless, kinship terms are often used in English – thus the alternative use of mummy as in (1.6), as well as daddy. Grandparents are normally addressed in Maltese as nanna and nannu, whereas auntie and uncle are often used alongside zija and ziju.

|| 4 Examples are taken (with permission) from the data collected for Vassallo (2006).

Language profiling | 339

Most utterances in the conversations are fully Maltese with no further codeswitching. As these children grow they also learn to converse in English. However, when addressed in Maltese they do not resort to further use of English elements as found in other groups. So we could say children in group 1 use Maltese as their dominant language with predictable English words possibly triggered by certain contexts and often followed by translations of the English words in Maltese. One 3yr 2mth old child used various animal names in Maltese and some also in English as she labeled the animals in cards: (1.7)

xadin-a monkey monkey-SG.F monkey ‘monkey monkey’

(1.8)

ġiraffa giraffe-SG ‘giraffe

(1.9)

popputamu (ippopotamu) hippopotamus-SG ‘hippopotamus

giraffe giraffe giraffe’ hippopotamus hippopotamus hippopotamus’

As already indicated, most of these children’s utterances contain no English elements in them whatsoever. The children in this first group are not many.

3.3 Group 2 – Maltese with English woven into the utterance The use of English elements in the conversations of the second group of children is quantitatively different from that of the first group. These children use words from many more lexical groups than numerals, colors, animals and kinship terms. In fact they do not seem to have a limit to what they could use except that the English words are usually used nominally, and include the names of objects, people, places or institutions as well as numerals, colors, animals and kinship terms that go beyond those of group 1. The following examples are representative of the language of this second group of children: (2.1)

għalaqt two għalaq-t two close.PFV-1SG two ‘I am two years old’

340 | Marie Azzopardi-Alexander

(2.2)

jien ħa nuża jien ħa n-uża I FUT I-use.1PVF.SG ‘I will use the green one’

l-green l-green the-green

(2.3)

ara lion look-IMP.SG lion ‘look it’s a lion’

(2.4)

dik sejra auntie Mary dik sejr-a auntie Mary auntie Mary that-SG.F go-PP-SG.F ‘she’s going to Auntie Mary’s’

(2.5.) teacher dak teacher that.SG.M ‘he’s a teacher’ (2.6)

dak that.SG.M ‘that’s an

office office office’

(2.7)

qed qed

(n)in-pinġi (n)inpinġ-i PROG 1-paint.IPFV-SG ‘I’m painting a nose’

(2.8)

mur hemm go.IMP.SG there ‘go away car’

(2.9)

monster aja (ara) jobot (robot) jibża j-jobot (r-robot) monster aja (ara) jobot (robot) j-ibża j-jobot (r-robot) 3-fear.IPFV-SG the-robot monster look.IMP.SG robot ‘A monster – look - the robot is afraid (of the monster)’

nose nose nose

car car

(2.10) fejnhom is-sausage fejn-hom is-sausage the-sausage where-3PL ‘where are the sausages?’ (2.11) u filgħodu mmorru l-ischool u filgħodu m-morru l-ischool and morning 1-go.IPFV-PL the-school ‘and in the morning we go to school’

Language profiling | 341

(2.12) Jonathan ħadu l-photo l-iskola Jonathan ħa-d-u l-photo l-iskola Jonathan take-PFV-3PL the-photo the-school ‘Jonathan got his photo taken at school’ The language of this group of children is characterized by the use of English for an extensive – and hence unpredictable – number of lexical groups within a decidedly Maltese frame. This occurs in all conversations and is not necessarily triggered by the adult’s use of English. These children, however, use mainly English nominals embedded into the Maltese frame and are subjected to the phonological rules for Maltese. This is particularly noticeable in that the English words are treated as though they were Maltese and, wherever the Maltese article is used with the English noun, phonological assimilation occurs whenever required (wherever a coronal obstruent occurs, except for /ʤ/). We see the use of the regular il- in (2.2) and (2.11) for the non-coronals, the epenthetic vowel is added to nouns starting with an s-initiated cluster in (2.11) and (2.12) preventing the need for the article to assimilate to the coronal /s/ and the assimilated article is used in utterance (2.10). Example (2.8) is particularly interesting because although /j/ is not coronal, it is treated as such since it replaces the /r/ which the child does not yet produce so that the assimilated article is pronounced as /ij/ to assimilate with /j/ which replaces the /r/ in robot. So what could be regarded as overextension of the assimilation in ij-jobot rather than the unassimilated il-jobot (as it would be in adult Maltese had this actually been the word) shows the observation of the phonological requirement already in place, making room for the time when /r/ is articulated. Thus the child uses /j/ to cover the work of the two different phonemes /r/ and /j/ of adult Maltese speech but has the phonology in place. Yes and no are used alongside the Maltese iva (together with its variants iwa, eħe) and le. This is in contrast with group 1 who do not seem to use the English expressions of affirmation and negation. It is worth noting that the children in group 2 – like those in group 1 – have a good range of vocabulary in Maltese. Whilst the children in group 1 do not opt for English words except for a few lexical groups, those in group 2 do not hesitate in threading English nominals into their Maltese frame. It is interesting that they sometimes actually use both the Maltese and the English word for the same referent within the same conversational turn as they extend their utterance e.g. karozza – car, ktieb – book, skola – school, kbir – big. This seems to serve for emphasis as well as possibly having the function of ensuring that the conversational partner(s) understands what is being said. Utterance (2.4) seems to be an exception in terms of the slight disruption to the Maltese frame resulting from the omission of għand to link the two sentence

342 | Marie Azzopardi-Alexander

parts, the subject (dik – that) plus verb (sejra – is going) to the location (auntie Mary). This may be explained by the probability that this preposition (għand – at) does not develop very early – it was not used by any of the children aged between 1 and 5.5 years reported in Portelli (2005) where developmental stages of Maltese are worked out. Many adults actually use the same kind of language i.e. code-switching extensively with nouns when addressing young children though they do not usually do this with older children. It is difficult to determine where the childdirected code-switching cut-off point occurs. However, used as motherese, this code-switching is unlikely to occur with children beyond primary school age. There is no explicit evidence that the children are conscious of the two codes they are using as separate codes but it is likely that they are.

3.4 Group 3 – Extensive mixing disrupting grammatical flow There are clear differences between the utterances used by the children of group 2 and those of group 3 both quantitatively and qualitatively. For a start, the uses of nominals in English are often inserted into the Maltese frame but do not always trigger the phonological rule requiring the articles to assimilate to the noun. We would have expected il-cat, t-teacher, ix-shoes, il-cookies in examples (3.1) to (3.5). Instead, there is a disruption of the utterance flow making the switch very noticeable. Yet the regular assimilated article is still used in some utterances such as (3.6). The following are some examples worth noting in the data of this group: (3.1)

ara kif deffes rasu ara kif deffes ras-u look-IMP.SG how thrust.PFV.3SG.M head-3POSS.SG.M ‘look how the cat thrust his head in’

(3.2)

din teacher tiegħi din teacher tiegħ-i this-SG.F teacher of-1SG ‘this is my teacher’

(3.3)

fejnha teacher fejn-ha teacher where-3SG.F teacher ‘where is the/my teacher?’

cat cat cat

Language profiling | 343

(3.4)

ħa ħa

ninża shoes n-inża shoes FUT 1-take.off.IPFV.SG shoes ‘I’m going to take my shoes off’

(3.5)

qed qed

(3.6)

teeth it-teeth it-toothpaste jiġi toothbrush teeth it-teeth it-toothpaste j-iġ-i toothbrush teeth the-teeth the-toothpaste 3M-come.IPFV-SG toothbrush ‘teeth the teeth the toothpaste is put on the toothbrush’

(3.7)

cat qattus cat cat ‘cat cat’

(3.8)

jien irrinsjajtu jien ir-rinsja-jt-u I 1-rinse.PFV-SG-DO.SG.M ‘I rinsed it’

(3.9)

jien għandi karozzi u ktiebs jien għand-i karozz-i u ktieb-s I have-1PRS.SG car-PL and book-PL ‘I keep my cars and books over there’

jippjova (jipprova) jwaqqagħlu j-ippjova (jipprova) j-waqqagħ-l-u PROG 3-try.IPFV-SG 3-drop-IPFV-IO.3SG.M lij-jaġel (lir-raġel) lij-jaġel (lir-raġel) to-the-man ‘he is trying to drop the man’s cookies’

(3.10) għandek horsies għand-ek horsie-s have-2PRS.SG horsie-PL ‘you have small horses’

cookies cookies cookies

hemm hemm there

żgħars żgħar-s small.PL

It may also be worth noting that the use of cookies in (3.5) derives from television exposure rather than the use of the commonly used biscuits or the native word gallettini in the Maltese context. This also explains the use of man in (3.11) and (3.12) as an expression (considered ‘cool’) addressed to a peer in annoyance rather than an alternative to the Maltese raġel used for the adult male referent.

344 | Marie Azzopardi-Alexander

(3.11) tihieli ti-hie-li give.IMP.SG-DO.SG.F-IO-1SG ‘give it to me man!’

man man man

(3.12) u ejja man and come.IMP.SG man ‘oh come on man!’ The use made of English in the conversations of the children in this group is much more extensive than that in the conversations of the children in group 2. There are also qualitative differences in that the use of English goes beyond nouns and includes verbs, though not many. Moreover, there is a lack of grammatical cohesion – both morphological and syntactic. For a start, articles are not always used with the English nouns (see 3.1 and 3.3 to 3.5). However, utterance (3.8) shows the use of an English verb with Maltese morphology as though the verb were being treated as a native Maltese verb. So there is a kind of cohesion that reduces the noticeability of this switch where participants are bilingual, i.e. they would understand without necessarily observing the English rinse. Utterance (3.6) may seem to have more English than Maltese items. However, they are self-corrections preceding the target subject toothpaste and are held together by the Maltese verb jiġi. The spatial relation that would be expressed in the adult utterance through the preposition fuq is missing. The first element retrieved is teeth which is self-corrected to it-teeth – showing phonological assimilation – and then corrected to it-toothpaste which is the target subject of the utterance. The location is toothbrush. This utterance could easily be considered similar to utterances in group 2 since toothpaste and toothbrush are used by adults more frequently than xkupilja for brush or krema for paste though there will be the exceptions who might use these. However, the child begins to process the utterance from the use of an English noun which could not have been retrieved easily from the Maltese snien or is-snien (‘teeth’). Thus the utterance could be considered as fully Maltese. We also find Maltese nouns taking the English morphological plural and hence there is a mix within the word itself. This is commonly found in words like mejda (‘table’) taking the English plural mejdas instead of the adult plural imwejjed or kelb (‘dog’) taking the kelbs plural form and several more as reported by both Murray-Curtis (1997) and Zammit (2008). Moreover the English plural form for sodda (‘bed’) is reported to have been beds in Zammit. However nouns like ktieb (‘book’) are frequently occurring nouns and so the plural kotba is normally accessible to children early on precisely because of its frequency of

Language profiling | 345

occurrence. So it is unusual to find them used with the English morph as we do in ktiebs (pronounced with the word-final voiceless obstruents [ps]) in example (3.9). Nevertheless this could be an extension of loanwords allowing this plural e.g. flatts (‘flats’) and many others which are an integral part of the Maltese lexicon. What is even more unusual is the plural form of the commonly occurring plural form of the adjective żgħir (‘small’) which is also pluralized through the internal vowel change to żgħar along with the same English plural morph as żgħars in (3.10). The lack of integration of the borrowed items into the Maltese utterance frame may be seen as indicating that these children do not distinguish between the two separate languages. However, one could easily explain this as the uninhibited practice of choosing items from the full language resources they are building. However, one can detect some awareness of the separateness of the two codes. It may be that the child’s grammar has not yet processed a distinction between grammatical categories particularly of English elements. Perhaps it is a result of the child’s exposure to English loanwords within Maltese utterances both in the home as well as in the kindergarten or school environment providing the child with a choice of vocabulary from the two codes as though they were one and the same code, or the possibility that the child retrieves the English loanwords rather than the Maltese nouns mirroring the exposure to mixed language use by the adults around him/her. We could say that there is a mix at the morpho-phonological level. The children are inserting the English noun without adapting to the Maltese frame through use of the article and in particular the use of the assimilated article that would give the English noun more of a native Maltese flavor. There is a mix at the morphological level – both with nominals (ktiebs in 3.9) and modifiers (żgħars) in (3.10). Furthermore, other words from English are included as expressions rather than for the neutral referent, man in utterances (3.11) and (3.12). However, the children frequently use the Maltese and English forms alongside each other as already seen with group 1 children in (1.7) to (1.9). Utterance (3.13) below shows that the child’s production of the English word is fully developed as it is composed of 3 syllables whereas the Maltese one may be more difficult for the child to retrieve possibly because it is made up of 4 syllables reduced to 3 syllables in the child’s production in (3.13). However, the use of the alternatives within the same utterance indicates the positive skill to ensure understanding in the bilingual context, part of the set of translanguaging skills talked about by Garcia and Li Wei (2014) and several others.

346 | Marie Azzopardi-Alexander

(3.13) tad-dinosaurs ta-d-dinosaur-s of-the-dinosaurs ‘of the dinosaurs (roars)

diżawri (dinosawri) diżawr-i (dinosawri) dinosaur-PL dinosaurs

dinosaurs dinosaur-s dinosaur-PL dinosaurs’

This points to the children being caught in the continuing development towards the full range of English language competencies rather than a restriction which could be criticized as pidgin-like in nature.

3.5 Group 4 – Maltese with a great amount of English The data from group 4 is more extensive precisely because there is a much greater variety of structures to comment on. (4.1)

nagħmillu hawnhekk in his hand n-agħmil-l-u hawnhekk in his hand 1-do.IPFV.SG-IO.SG.M here in his hand ‘I will do it to him here on his hand’

(4.2)

le m’għandhomx il-babies għax they are still big le m-għand-hom-x il-babie-s għax they are still big no NEG-have-3PL-NEG the-babies because they are still big ‘no they don’t have any babies because they are still (not yet) big’

(4.3)

ija qalgħuh imbagħad il-lady tiegħu ija qalgħ-u-h imbagħad il-lady tiegħ-u yes took.off.PFV-3PL-DO-SG.M then the-lady of-3SG.M qalet “no put that door back” qal-et “no put that door back” said.PFV-3SG.F “no put that door back” ‘yes they took it off then his lady (wife/fiancée) said “no put that door back”’

(4.4) kont insejt ... jgħid kon-t insej-t ... j-għid was.PFV-1SG forget.PFV-1SG 3M-say.IPFV.SG ‘I had forgotten ... he says “no it’s stuck”’ (4.5)

qed qed

tippuxjah ti-ppuxj-a-h PROG 3F-push.IPFV-SG-DO.SG.M ‘she is pushing it’

da(n) da(n) this.SG.M

“no is stuck” “no is stuck” “no is stuck”

Language profiling | 347

(4.6) din l-aħħar waħda hux we are playing din l-aħħar waħd-a hu-x we are playing this.SG.F the-last one-F 3SG.M-NEG we are playing ‘this is the last one that we are playing, right?’ (4.7)

ejja we are doing it hurry up eh di l-ewwel ejja we are doing it hurry up eh di-l-ewwel come.IMP.SG we are doing it hurry up eh this.F-DEF-first ‘come on we are doing it hurry up eh this is the first’

(4.8) Jamie puxjani u waqajt Jamie puxja-ni u waqa-jt Jamie push.PFV.3SG.M-DO-1SG and fall.PFV-1SG I didn’t put some cream I put nothing I put some water only I didn’t put some cream I put nothing I put some water only I didn’t put some cream I put nothing I put some water only ‘Jamie pushed me and I fell – I didn’t put any cream I just put some water’ (4.9) ara ħa nagħmel ara ħa n-agħmel look.IMP.SG FUT 1-do.IPFV.SG ‘look I’m going to jump’ (4.10) qagħdna nagħmlu qagħd-na n-agħml-u PROG.PFV-1PL 1-do.IPFV-PL ‘we stayed dancing’

jumping jumping jumping

dance dance dance

(4.11) int għandek jacket one int għand-ek jacket one you have-2PRS.SG jacket one ‘you have one jacket’ (4.12) ija imma umbagħad huwa ija imma umbagħad huwa yes but then he ‘yes but then he was punished’

sar punishment sar punishment became.PFV-3SG.M punishment

(4.13) ija imma illid (irrid) għandna spiders kbar ħafna ta’ veru alive ija imma il-lid (irrid) għand-na spiders kbar ħafna ta’ veru alive yes but 1-want.IPFV-SG have-1-PL spiders big-PL many of real alive ‘yes but I want … we have big spiders em lots real ones alive’

348 | Marie Azzopardi-Alexander

(4.14) u huma pastażi bil-pastażi jagħmlu hekk u huma pastaż-i bi-l-pastaż-i j-agħml-u hekk and they naughty-PL with-the-naughty-PL 3-do.IPFV-PL thus u bil-blavi (bravi) hekk u bi-l-blav-i (bravi) hekk and with-the-good-PL thus ‘and they are naughty with the naughty ones they do like this and with the good ones thus’ (4.15) [ija] (iva) jsawwat [ija] (iva) j-sawwat yes 3-beats.up.IPFV.SG ‘yes he beats up the bad guys’

il-bad guys il-bad guys the-bad guys

(4.16) [ija] (iva) ħallieh [ija] (iva) ħall-ieh yes leave.PFV.3SG.M-DO.SG.M ‘yes he left it in your brain’

fil-[blejn] (brain) fi-l-[blejn] (brain) in-the-brain

tiegħek tiegħ-ek of-your

(4.17) ara imma dan imma din tagħmel ara imma dan imma din t-agħmel look.IMP.SG but this.SG.M but this-SG.F 3F-do.IPFV.SG ‘look but this … but this does magic’

il-magics il-magic-s the-magic-PL

(4.18) u jismu is your kind is magic u jism-u is your kind is magic and name.IPFV-3SG.M is your kind is magic ‘and it’s called ... it’s a kind of magic’ (4.19) għandha wiċċha pink u qiegħed għand-ha wiċċ-ha pink u qiegħed have.PRS.SG.F face.SG-3SG.F pink and located.SG.M hawnhekk hawnhekk here ‘she has a pink face and she has dots here’

bid-dots bi-d-dot-s with-the-dot.PL

(4.20) Le għax jien mermaid ħa nsir meta Le għax jien mermaid ħa n-sir meta no because I mermaid FUT 1-become.IPFV.SG when nikber jien insir mermaid boy n-ikber jien in-sir mermaid boy 1-grow.IPFV.SG I 1-become.IPFV.SG mermaid boy

Language profiling | 349

‘no because I want to become a mermaid when I grow up … I’ll become a boy mermaid’ (4.21) lanqas kont naf li kien hemm lanqas kont n-af li kien hemm not-even Was.PFV.SG 1-know.IPFV.SG that was.PFV.1.SG there il-puzzles u l-ktibijiet imbagħad sibthom il-puzzle-s u l-ktib-ijiet imbagħad sib-t-hom the-puzzles and the-book-PL then find.PFV-1SG-DO.3PL ‘I didn’t know that there were puzzles and books then I found them’ (4.22) Din l-aħħar waħda hux… we are playing Din l-aħħar waħd-a hu-x… we are playing this.F the-last one-F he-NEG we are playing ‘this is the last one isn’t it… we are playing’ Parents of group 4 children use a great deal of both Maltese and English for a variety of reasons. There may be some inconsistencies when they address their children in Maltese and when they address them in English. They switch from one language to the other frequently in informal bilingual company whilst those in group 3 usually use minimal English elements, usually no more than one in an utterance. The weaving of English and Maltese in the utterances of the children in group 4 is very marked at various linguistic levels. The most obvious difference is the fact that many of the utterances have stretches of words in one language followed by stretches in the other language. These are usually whole phrases or clauses. However, these children are still more comfortable with Maltese but very confident about their ability to speak English though the utterance parts do not always flow smoothly. Some of the utterances use reported speech originally heard in English (utterances 4.3, 4.4). So one could say that memory plays an important role in this. However, most utterances cannot be explained as having been triggered in this way and there does not seem to be a reason for the mix except that the children are well on the way to becoming able communicators in different settings with different speakers. We could consider these utterances as decidedly “mixed” – a positive step in my analysis but often considered negative when this mixture persists into adulthood in appropriate settings. Some of these group 4 utterances are not fully formed since the children have not yet fully developed their grammar. Utterance (4.1) may lack the object to the verb nagħmillu. Children often rely on the information available in the context and on gestures for disambiguation. The verb is followed by the adverb of location hawnhekk (‘here’) which is clarified by the gesture that accompanied

350 | Marie Azzopardi-Alexander

it as the child held his own hand out and then by the English prepositional phrase in his hand which renders hawnhekk more explicit. Some utterances have two parts or more at which a switch occurs e.g. each of utterances (4.2) and (4.3) could be considered as perfect though the rest is uttered in the other language. These utterances simply reflect a more extensive use of English. In the switch to English the child in (4.3) shows what can be considered incomplete development of English grammar in the omission of the obligatory pronoun ‘it’. Various utterances – particularly (4.6) and (4.7) – show a smooth flow from one language to the other without structural disruption. Other examples have English verbs that are inflected as though they were Maltese verbs. This causes no difficulty to bilinguals (though some may cringe). We see this in (4.5) tippuxjah, (4.8) with puxjani where the English verb ‘push’ is being used as though it were a Maltese verb although the Maltese verb imbotta would have been part of the child’s vocabulary and is easily retrievable. Hence the interest in this example since it reflects a choice enabled by the growing English vocabulary and probably triggered by some context. Some of the utterances would jar in many bilingual ears if they are heard in adult speech – and they are attested even across the University campus. Within the framework of translanguaging they could be re-considered as being appropriate to particular settings rather than travesties of linguistic principles. The utterances in focus here are exemplified in the morphological mix seen in (4.5) and (4.8) (just discussed) but even more so in (4.9) and (4.10). The English word used in both utterances interferes with the verb phrases since in Maltese the generic verb għamel would not be used to precede the specific verb. Utterance (4.9) could have had naqbeż rather than the combination of nagħmel (‘I make’) followed by jumping. Utterance (4.10), similarly, would have simply been niżfnu rather than nagħmlu (‘we make’) followed by dance. The combination with nagħmlu (‘we make’) followed by the verb in English causes disruption to the structure. Various other interesting issues emerge at different linguistic levels. In (4.2) the clause they are still big means the opposite of what the child intends to say. The difficulty is not in the lack of understanding of big versus small but in the misuse of still modifying big rather than small. The adult continuation of the utterance would be the alternative not yet big or still small. Alternatively, the use of big can be considered as a performance slip as a result of the animation marking her utterances, acting in an adult role to explain what she thought other conversational participants should know. Again on the semantic level illady tiegħu (‘his lady’) in (4.3) reflects possible lack of specific English vocabulary such as wife or fiancée since mara or għarusa are very likely already easily retrievable from the child’s mental lexicon. Yet even here, the influence of Eng-

Language profiling | 351

lish in motherese for words like lady to cover a variety of meanings (including woman, adolescent as well as wife and fiancée) instead of expressing the more specific intention through the Maltese noun or noun phrase. The use of only at the end of (4.8) renders the syntax very marked but typical of many MalteseEnglish speakers. Moreover the use of the negative structure in (4.8) was followed by some rather than any to combine with the negative verb phrase used. The use of nothing necessitated a positive verb phrase. Utterance (4.5) is an example of a verb with an English stem encased in Maltese morphology, tippuxjah, as was irrinsjatu in utterance (3.8). What seems an incomplete structure no is stuck in (4.4), marked by the omission of the pronoun it, is commonly used in the early years and probably results from the inability to recognize the pronoun it as distinct from it’s because the alveolar fricative followed by the alveolar stop in the reduced structure minimizes the auditory salience of the stop and it is probably not processed by the child until s/he recognizes the use of it in such structures i.e. when it’s is recognized as a reduced form for it is. The homorganic context creates the spillover from the phonological to the grammatical. Hence is often covers both the is and it is usage for children in this group and possibly in other groups as well. Utterances (4.9) and (4.10) reflect utterances commonly heard in motherese as well as in children’s speech. This often happens when a sibling starts school or when parents prepare their child for school. A form of the English verb is placed within the verb phrase so that the structure features the imperative form agħmel (‘do’) followed by a word indicative of a verbal action in the adult’s speech. We see this in (4.9) ħa nagħmel jumping and (4.10) nagħmlu dance. Other similar examples in motherese include structures like agħmel sleep, agħmel eat, and is also extended to words that operate in the context as verbs e.g. agħmel nice (meaning to stroke gently). Such usage may well be found in motherese within group 3 and possibly other groups.

3.6 Group 5 – English with Maltese woven in There is a huge qualitative leap from group 4 to group 5 at different linguistic levels. Children in group 5 could be considered to be the mirror image of those in group 2 in terms of language dominance. Clearly, English rather than Maltese is their dominant language and they are spoken to extensively in English, not just Maltese with code-switching from English. Their parents often converse in Maltese when they are not addressing very young child(ren). However, they often move from using utterances in English to utterances in Maltese as well as using

352 | Marie Azzopardi-Alexander

extensive code-switching from Maltese. The change is often triggered by an addition of a conversational partner – as frequently happens with adults – or a change of subject. They still code-switch in ways similar to the code-switching seen in the group 2 children, without any disruption to the English frame. Often Maltese is used in exclamations and tags as well as endearments at the beginning and end of the utterances rather than within the main clause(s). Thus we see miskina in (5.1) and (5.3), qalbi in (5.2), jaħasra in (5.3) and in (5.14), u ejja in (5.4). This happens too in the conversations of group 4 children (see ejja in 4.7 and 4.23). Most adults outside the family address the children in group 5 in English possibly because the English used by the children does not have extensive Maltese phonetic coloring and indicates the appropriateness of using English rather than Maltese. Many of the children’s utterances are English in their entirety. Nevertheless, sometimes a great deal of Maltese is also used. This occasionally includes whole utterances. Utterance (5.7) is one such example which has been included here to illustrate the exclamation that follows ara which is occasionally used to draw the adult’s attention to something (since children expect the adult to look directly at them throughout the conversation), or which is used as a sentence starter similar to utterance-initial well in English. Since these children still have significant Maltese exposure there are some situations where they use a considerable amount of Maltese particularly in situations where they realize that one of the persons with whom they are interacting is addressing them in Maltese. An unexpected example is utterance (5.9) since numerals are usually said in English rather than Maltese – a rare occurrence for this child most of whose utterances are in English or have very minimal use of Maltese. This is a clear example of a child responding to the adult’s use of numerals in her preceding offer. The use of Maltese often occupies central places in the clause e.g. when Maltese is used for verbs as in utterances (5.1), (5.6) and for total utterances as in the exclamative (5.7) which is entirely in Maltese. No English utterances are listed because they can form the largest number in any conversation with a group 5 child. The amount of Maltese increases in time for different reasons such that children brought up in the same family attending the same school often end up speaking very differently from each other in terms of how much of their spontaneous conversation is in English and how much is in Maltese. Clearly, this does not depend on competence but also on other factors including, in particular, self-confidence. There are some interesting features in these conversations. Firstly, there is so much of both languages that it is not always easy to identify which is the matrix language and which is the embedded one (Myers-Scotton 1997). Secondly, there are the prefixed exclamations as well as some relational words such as

Language profiling | 353

the adverbial issa in (5.5) and the conjunction għax in (5.11). The syntax is often marked Maltese-English syntax, e.g. the object preceding the verb in (5.9). One noticeable characteristic when balancing the complexity of the full utterance is the incomplete morphological components. In (5.11) the third person masculine morph is missing so we have ibki for jibki pointing to different levels of acquisition of the two languages, and also the use of the English possessive pronoun ‘his’ preceding the Maltese noun that is already marked with the third person masculine possessive morph in (5.10). We do not find the use of English verbs inflected with Maltese morphs as we find in group 3 (see utterance 3.8) and group 4 (see utterances 4.5 and 4.8). (5.1)

miskina she fell and kissret miskin-a she fell and kissr-et poor.thing-SG.F she fell and broke.PFV-3SG.F ‘poor thing! She fell and broke the present’

(5.2)

happy birthday qalbi happy birthday qalb-i happy birthday heart-my ‘happy birthday darling!’

(5.3)

he’s sick miskin… jaħasra he’s sick miskin… jaħasra he’s sick poor.thing.SG.M what-a-shame ‘the poor thing is sick – what a shame!’

(5.4)

u ejja let’s play għax we have to go soon u ejja let’s play għax we have to go soon and come.IMP.SG let’s play because we have to go soon ‘come on! Let’s play because we have to go soon’

(5.5)

uuu don’t worry issa you find it and don’t worry now you find it ‘come on! don’t worry – you’ll find it’

(5.6)

senna (stenna) mummy, wait wait wait, senna (stenna) wait-IMP.SG mummy, wait wait wait, wait-IMP.SG ‘wait mummy, wait wait wait, wait’

(5.7)

ara x’wadda (waħda) ara x’wadda (waħda) look-IMP.SG what-one.F ‘look my goodness me!’

din din this.F

il-present il-present the-present

354 | Marie Azzopardi-Alexander

(5.8)

ġewwa Tikka inside Tikka ‘go indoors Tikka (the dog)!

(5.9)

mhux wadda (waħda) – tnejn I want mhux wadd-a (waħda) – tnejn I want NEG one-SG.F two I want ‘not one – I want two’

(5.10) that boy will go (grow) d(the)-imnieħru that boy in (h)is imnieħru that boy will go (grow) d(the)-imnieħr-u that boy in (h)is imnieħr-u that boy will grow the-nose-3SG.M that boy in his nose-3SG.M ‘that boy’s nose will grow – that boy, his nose’ (5.11) leave this alone għax ibki leave this alone għax i-bki leave this alone because 3M-cry.IPFV.SG ‘leave this alone otherwise he’ll cry’

da (għax jibki dan) da (għax jibki dan) this-M.SG

(5.12) clean there mummy it’s dirty – maqmuġ (maħmuġ) clean there mummy it’s dirty dirty-M.SG ‘clean there mummy – it’s dirty (repeated in Maltese)’ (5.13) hux eh! You’re lying you didn’t see Batman he-NEG you’re lying you didn’t see Batman ‘of course not! You’re lying ... you didn’t see Batman’ (5.14) he’s not coming to play cos he’s sick he’s not coming to play cos he’s sick ‘he’s not coming to play cos he’s sick

jaħasra! what-a-shame! what a shame!’

In spite of the alternation of English and Maltese, both words and phrases, the grammatical and the conversational flow are not interrupted. Most of the utterances listed would not be understood by a monolingual English speaker. However, it must be emphasized that the number of utterances using both Maltese and English in the conversation of these children was minimal. Most utterances were in English or repeated the same phrase in Maltese that was first said in English – as in utterance (5.6).

Language profiling | 355

4 Why profiling? Talk about a range of bilingualism in Malta has been going on for many years (Borg 1986). We would expect all of the children whose data has been considered to become fairly or very fluent bilinguals by the time they complete their secondary education, provided that we can accept a flexible definition where “bilingual” are those who can communicate in either of the two languages whenever it is necessary. Nevertheless, most if not all will switch across their two languages particularly in conversations where all partners have at least fairly good competence in the languages. In such contexts there is no question of excluding someone from the conversation, or of not being understood. The degree of consciousness of which language to use differs across groups. So is the degree of confidence with which English is used within groups. The degree of confidence changes markedly when the children become adults. That, however, is the remit of another study. One can posit an educational maxim whereby all students are expected to be able to participate in a Maltese-only or in an English-only conversation where necessary. This would involve conversational partners who are not bilingual or settings which are formal. Yet it should be expected that conversation amongst bilinguals will involve some elements of both languages and the amount of such switching or mixing will depend on a number of factors, among them the dominance of a conversational partner. “... one can think of CS (codeswitching) as a type of “trouble in the works” for the language production system ... what makes CS so remarkable is that this ‘trouble’ results in extremely little disorder in the product.” (Myers-Scotton 1997: 239) More research is required to enable children to be assessed differently in primary education when language competence forms the basis of decisions made in the educational field. When the National Minimal Curriculum was launched (Ministry of Education 1999), many recommendations were made in view of the need to have children at 16 become proficient users of spoken and written Maltese and English. One such recommendation on the basis of this study would be that children in some groups require a one-teacher-onelanguage setting to enable them to learn the lesser used language in the home to be used in partial immersion (i.e. for part of the day) at kindergarten level and in early primary. This would ensure that whilst the child is able to use the home language to learn and interact, the other language would become the natural language to use in meaningful contexts with a teacher who does not revert to the familiar language without creating an alien environment. Thus the teacher

356 | Marie Azzopardi-Alexander

would understand the child where necessary but increasingly less so in order to increase the need for the child to respond in the target language. Firman (2007) showed how language dominance varied across school categories (state, church and independent). To my knowledge, the results shown in her study have not been applied to the local teaching or clinical (and other) assessment contexts. Students in primary schools are taught in similar ways whether they come into formal education from a Maltese-dominant or an English-dominant home or an in-between context. The way they are taught literacy skills does not really take their language competences into account and so does not differ across the groups. It may be possible to wait till oral competence in one language reaches a certain level before literacy in that language is required to start. Time could be spent, instead, on the development of oral skills which could then serve as a firm foundation for literacy. Other settings might benefit from drawing on information related to the difference in language competencies. Where clinical settings acknowledge such differences they should be able to manage children’s remediation programs fairly successfully. One can contemplate parent programs that would help to enable some language compensation in the home, enabling parents to increase their ability in their less dominant language (as in the ESU Arm of Support programs from 2009 on (English Speaking Union 2015).

5 Conclusion I consider bilingual competence to be different from that of two monolinguals because the bilingual responds to different settings where one or both languages can be used for communicative purposes. To some extent, the work of Camilleri (1995) started the journey in this direction. I embrace the view that code-switching is a remarkable strategy in that …bilingual speakers talking to other bilingual speakers may exploit the resources of both of their shared languages for a variety of purposes – and in those contexts, code-switching is a sign of bilingual competence, not incompetence. (Thomason 2001: 54)

A closer look at what kind of translanguaging (Garcia and Wei 2014) exists in Malta and Gozo could contribute a great deal to the formulation of enlightened, forward-looking educational policies and to developing healthy local educational practices.

Language profiling | 357

Abbreviations 1, 2, 3 DEF DO F FUT IMP IO IPFV M

1st, 2nd, 3rd person definite direct object feminine future imperative indirect object imperfective masculine

NEG OM PFV PL POSS PP PROG PRS SG

negative object marker perfect plural possession present progressive progressive present singular

References Aquilina, Joseph. 2006. Concise Maltese-English, English-Maltese dictionary. Sta Venera, Malta: Midsea Books. Azzopardi-Alexander, Marie. 2011. The vowel systems of Xlukkajr and Naduri. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 235–254. Berlin: Akademie Verlag. Baker, Colin. 2001. Foundations of bilingual education and bilingualism. Clevedon: Multilingual Matters. Borg, Albert. 1986. The maintenance of Maltese as a language: What chances? Paper presented to the European Workshop on Multicultural Studies in Higher Education (Council of Europe). Nov 1986 (Unpubl.). Borg, Albert. 2011. Lectal variation in Maltese. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 11–31. Berlin: Akademie Verlag. Camilleri, Antoinette. 1995. Bilingualism in education. The Maltese experience. Heidelberg: Julius Groos. Deuchar, Margaret & Suzanne Quay. 2000. Bilingual acquisition. Theoretical implications of a case study. Oxford: Oxford University Press. Ellul, Sonia. 1978. A case study in bilingualism. Cambridge: Huntington. English Speaking Union. 2015. Website: esumalta.org/portfolio-item/arm-of-support-programme/ Firman, Christine. 2007. A study of word-level decoding skills of 7-year-old Maltese children in a bilingual environment. University of Malta: Unpubl. Ph.D. thesis. Garcia, Ofelia & Li Wei. 2014. Translanguaging – language, bilingualism and education. Hampshire: Palgrave Macmillan. Lanza, Elizabeth. 2004. Language mixing in infant bilingualism. A sociolinguistic perspective. Oxford: Oxford University Press. Mifsud, Manwel. 1995. Loan verbs in Maltese. A descriptive and comparative study. Leiden: Brill. Ministry of Education Malta. 1999. Creating the future: National minimum curriculum. (http://www.curriculum.gov.mt/nmc.htm).

358 | Marie Azzopardi-Alexander

Mori, Laura. 2011. Linguistic variation in legal Maltese: EU directives compared to national implementation laws. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and change: The dynamics of Maltese in space, time and society, 109–128. Berlin: Akademie Verlag. Murray-Curtis, Jackie. 1997. The emergence of plural n 3- to 5-year old Maltese-speaking children. Unpublished B.Sc. (Hons) thesis. University of Malta. Myers-Scotton, Carol. 1997. Duelling languages – grammatical structure in codeswitching. Oxford: Oxford University Press. Myers-Scotton, Carol & Janice Jake. 2014. Nonfinite verbs and negotiating bilingualism in codeswitching: Implications for a language production model. Bilingualism: Language and Cognition 17(3). 511–525. Portelli, Josette. 2005. The mean length of utterance in Maltese-speaking children. Unpublished M.A. (Ling.) thesis. University of Malta. Thomason, Sarah. 2001. Language contact. An introduction. Edinburgh: Edinburgh University Press. Vassallo, Marlene. 2006. The language acquisition of a Maltese toddler – a case study. Unpublished B.Ed (Hons) thesis. University of Malta. Zammit, Raymond. 2008. The use of number and gender markers in the Maltese, in 3 to 3;6 year old children. Unpublished B.Sc. (Hons) thesis. University of Malta.

John J. Camilleri

Digitizing the grammar and vocabulary of Maltese Abstract: The Maltese language is highly under-resourced in digital terms. This work presents the current status of two new computational resources for Maltese. We first describe a computational grammar implemented using the Grammatical Framework (GF), where Maltese is the first Semitic language to be added to its Resource Grammar Library (RGL). We then introduce Ġabra, an open online lexical database for Maltese compiled from pre-existing resources. Using the grammar to generate some 4 million inflection forms, we go on to extend the collection and present it as an open, online platform and basis for a full-form computational lexicon for Maltese. Keywords: computational linguistics, natural language processing, language technology, digital grammar, morphological generation, online lexicon

1 Introduction 1.1 Maltese and the digital age A recent META-NET report on the language (Rosner and Joachimsen 2012) indicates that while a number of linguistic resources do exist for the language; Maltese is in general under-represented in digital terms when compared to other European languages. This study finds that Maltese is best resourced with speech synthesis and text corpora, but very poorly represented in other areas of computational linguistics. Specifically, to date there are no computational grammars, morphological analyzers, or text generation systems for Maltese. A Eurobarometer report (European Commission 2011) found that although 89.5% of the respondents in Malta claimed that Maltese was their mother tongue, only 6.5% use exclusively Maltese on the Internet when reading, consuming content or communicating. An overwhelming 90.6% choose to browse websites in English and 20.1% in Italian, respectively. The digital age has also

|| John J. Camilleri: University of Gothenburg, Department of Computer Science and Engineering, Rännvägen 6 B, 41296 Gothenburg. E-mail: [email protected]

360 | John J. Camilleri

seen a sharp rise in words loaned from English, in particular for internet and social network concepts for which no suitable Maltese equivalent exists. The habits of a language community and speed with which new technical words are loaned mean that the official bodies have a hard time deciding on, and introducing, new guidelines successfully (Il-Kunsill Nazzjonali ta’ l-Ilsien Malti 2008b).

1.2 Digital resources for Maltese 1.2.1 Corpora and terminologies A monolingual corpus for Maltese has been compiled and is hosted by the Maltese Language Resource Server (MLRS) (Rosner et al. 2006). The MLRS Corpus of Maltese (Korpus Malti) (Gatt and Borg 2011) can be described as an opportunistic text collection of nearly 120 million tokens, mostly created from publicly available documents, but also including a limited amount of user-contributed material. Texts in version 2.0 (Beta) of the corpus were pre-processed to remove duplicate material and long stretches of non-Maltese text. They were also processed with a simple dictionary-based spelling correction and part-of-speech (POS) tagged. The MLRS also hosts the Corpus of Learner English in Malta (CLEM). There also exist some parallel corpora for Maltese which have come from the translation efforts of the European Commission. The JRC-Acquis multilingual corpus (Steinberger et al. 2006) is a parallel corpus containing the complete text of the European Union Law (Acquis Communautaire) in 22 languages. Other parallel texts for Maltese are accessible through the OPUS project (Tiedemann 2009). The InterActive Terminology for Europe (IATE) is the European Union’s multilingual term base used by various institutions for the collection, dissemination and shared management of EU-specific terminology (IATE 2015). It was launched in 1999 as a web-based infrastructure for all EU terminology resources, incorporating and standardizing all existing terminology databases of the EU’s translation services into a single database. It also includes a number of legacy databases, and now contains a total of approximately 1.3 million multilingual entries or 8 million terms, covering 24 languages of the EU, including Maltese (43,817 terms).

Digitizing the grammar and vocabulary of Maltese | 361

1.2.2 The Maltilex project There has long been an interest in creating a computational lexicon for Maltese. The Maltilex project was first announced in 1998, identifying the need for such a lexicon and outlining the scope of the project for creating one (Rosner et al. 1998). Rosner et al. (1999) go on to highlight the idea of automatic extraction, involving tokenization and headword identification on a corpus. Dalli (2002) describes a concrete implementation of this, using a weakly-supervised learning approach to identify lemmas in an unstructured list of words without requiring any prior rules. Sadly, the only Maltese corpus available at the time was very small (~2000 tokens) and the end results were very noisy and not practically usable. Despite the large corpus for Maltese available today, it seems this experiment in lexicon extraction has not been repeated. As the Maltilex project evolved into the MLRS, a new description of a lexicon structure was presented in Rosner et al. (2006). This paper describes the technical details in designing a lexicon of Maltese, but presents no implementation. The most recent development on this road towards a computational lexicon is the announcement of a project to create a national online dictionary for Maltese (Il-Kunsill Nazzjonali ta’ l-Ilsien Malti 2012). Rather than focusing on extraction from a corpus, the project will instead be digitizing Aquilina’s MalteseEnglish dictionary and making it available online. At the time of writing, the project is still in the process of updating the original dictionary to include modern entries and had not yet entered the digitization phase (p.c., June 2015).

1.2.3 Software tools As part of the Metanet4U initiative, the Maltese Language Software Services (MLSS 2012) portal has made available a number of statistically-trained language-processing tools for Maltese, covering paragraph and sentence splitting, tokenization, and part of speech tagging. A morphological generator for regular Maltese verbs has also been written by Ramon Casha, which was used to populate a word list which can be used in the open-source spell checker ispell. The quality of this word list is considered good, although its coverage is not large (p.c., May 2013).

362 | John J. Camilleri

1.2.4 Other lexical resources Camilleri and Spagnol (2013) constructed an online database of root-and-pattern verbs, based on the exhaustive enumeration of 1,923 roots and over 4,142 verbs by Spagnol (2011). The original data was converted from Excel format into a relational database, with a web-based interface then built around it. This interface provides the ability to search for roots and verbs quickly using regular expression syntax. Mayer et al. (2013) put together a large corpus of broken plurals in Maltese, containing 654 singular-broken plural pairs mainly based on a list of synchronically used or known broken plurals compiled by Schembri (2012). The study is based on a comprehensive inventory of broken plural forms which an average educated speaker of Maltese would be expected to have. The corpus also includes a small number of collective forms, which share the same CV structure of broken plurals but which differ morpho-syntactically. In her M.A. thesis, Ellul (2015) has collected a list of over 2,500 verbal nouns from the Aquilina dictionary and extended it to include some new ones from other sources. An early online English-Maltese dictionary was originally written by Grazio Falzon, consisting of alphabetically ordered English lemmas with their Maltese translation and pronunciation. It contains some 5,458 English entries and includes gender information for nouns. The collection has since been updated and released as a TEI-compliant XML dictionary file within the META-SHARE project (Falzon 2012).

1.3 Motivation and contributions Despite an academic interest in the area for at least 15 years, the linguistic tools and digital resources available for Maltese are still quite limited. The current work summarizes two recent contributions to the field: a computational grammar, and an online collection of lexical resources for Maltese, both of which were originally presented in an M.Sc. thesis by Camilleri (2013). Section 2 introduces the context in which the grammar was written, and covers some of the most interesting aspects of its development. In Section 3 we discuss the foundations of a computational lexicon for Maltese, and introduce Ġabra: an online collection of lexical resources supplemented with full automaticallygenerated inflectional forms. Finally Section 4 looks ahead, discussing future plans for computational resources for Maltese.

Digitizing the grammar and vocabulary of Maltese | 363

2 Computational grammar 2.1 The Grammatical Framework The Grammatical Framework (GF) (Ranta 2011) is a functional programming language specialized for multilingual grammar applications. When implementing computer languages, it is common to make the distinction between abstract syntax – the language’s hierarchical structure – and concrete syntax – what the language looks like as a string. This is based on the idea that semantics are relevant to the abstract level, while syntactic details are a concrete concern. GF applies this distinction to natural languages too. Abstract syntax in GF is used as a language-independent interlingua for semantic representation, shared between multiple concrete syntaxes which describe the linearization into different languages. A grammar defines how an abstract tree should be linearized into a concrete syntax. From this, GF is able to derive both a generator and a parser, becoming a multi-lingual authoring and translation system which avoids the need to write separate rules for each language pair. DetCN : NP

DetQuant : Det DefArt : Quant NumSg : Num

AdjCN : CN PositA : AP

UseN : CN

red_A : A

house_N : N

NP

NP

Det

CN

Quant

AP

CN

A

N

red

house

the

Det

CN

Quant

CN

AP

N

Quant

A

dar

il-

ħamra

id-

Fig. 1: Abstract syntax tree for a noun phrase (above), together with concrete parse trees for the phrase in English (left) and Maltese (right).

364 | John J. Camilleri

2.1.1 Resource Grammar Library As grammars grow in size having a module system becomes vital. Because the morphological and syntactic details of natural languages can be quite complex, it doesn’t make sense for every new grammar written in GF to have to reimplement these features. Instead, the low-level details of a language can be put into a resource grammar whose abstract syntax models phenomena on the syntactic level. A separate application grammar can then focus on high-level modeling tasks, using functions defined in the resource grammar as needed. This allows code for semantic modeling to not be mixed with code that handles syntax and morphology. Apart from the compiler and runtime system, the GF distribution also includes a standard library known as the Resource Grammar Library (RGL) (Ranta 2009). The RGL provides general linguistic descriptions for natural languages which can be reused by using a common language independent API. This library covers the morphology and basic syntax of over 30 languages from around the world. Programmers who wish to use GF do not need to worry about details like word order and agreement in each language; they may simply use the standard functions provided by the API and the linearization into each language is handled automatically by GF.

2.1.2 Lexical resources in GF The RGL contains a small multilingual lexicon of around 300 words, and is common to all languages in the library. This is mainly intended for testing purposes, and is far too limited for real-world applications. Application grammars typically bear the task of defining the lexicons they require, but there has also recently been a focus on large-scale GF lexicons to complement the library. Because of the work involved in lexicon compilation, large-scale GF lexicons are typically created via extraction from existing sources. Monolingual lexicons in GF tend to be idiomatic, in that they are tailored for each language, and are intended as a resource or for a monolingual application. They do not provide links between senses in different languages and thus cannot be used for translation tasks. Multilingual lexicons, on the other hand, share a common abstract syntax to facilitate multilingual applications. Experiments in extracting such resources have been carried out using the English WordNet and its translations into Bulgarian, German, Finnish and Hindi (Angelov et al. 2013; Virk 2013).

Digitizing the grammar and vocabulary of Maltese | 365

2.2 Morphology 2.2.1 Smart paradigms A morphological paradigm describes the complete set of word forms which are associated with a lexeme, such as the conjugations of a verb and the declensions on a noun. In GF, a paradigm is a function which takes one or more representative word forms of a lexeme and constructs an inflection table for it. Natural languages typically have many paradigms for each word class, and manually selecting the correct paradigm for each entry in a lexicon can be laborious. A smart paradigm (Détrez and Ranta 2012) is a meta-paradigm which inspects the given base form of a lexeme and tries to infer which lower-level paradigm applies. In cases where the correct paradigm simply cannot be determined, more word forms are given for discrimination. In general, the average number of forms needed, can be seen as a measure of the predictability of the morphology in a language.

2.2.2 Noun The noun in Maltese has an inherent gender and inflects for number, including collective and dual forms. The collective is syntactically singular but semantically plural, e.g.: (1)

dan it-tuffieħ this.SG.M DEF-apple.COLL ‘these apples are sweet’

ħelu sweet.SG.M

Some Maltese nouns also have a dual, typically parts of the body and measures of time, e.g. ġimgħatejn (‘two weeks’). Nouns can also have two types of plural: determinate and indeterminate, e.g. triqat/toroq (‘roads’). Nouns typically have some subset of these forms, though never all of them together. The linguistic literature does not identify any paradigms for nouns in this sense, thus some further investigation was needed. By surveying the 186 nouns in the RGL mini-lexicon, four different noun paradigms have been identified. They are summarized in Table 1. The analysis also reveals that very few nouns actually have both a determinate and indeterminate plural form, and when they do one often sounds arcane. While this distinction can have some linguistic importance, this is simplified in the GF implementation by storing only one plural form. This change is made internally in the noun representation; the paradigm constructors can still take

366 | John J. Camilleri

both forms in case the representation is changed in the future. Another solution would be to have an indeterminate plural stored as a variant of the determinate plural. For an in-depth study of these phenomena, refer to Mifsud (1995). Tab. 1: Noun paradigms treated in the grammar.

No.

Singular

Collective

Dual

Det. Plural

Ind. Plural

Gloss

1

qattus

–

–

qtates

–

‘cat’

1x

triq

–

–

triqat

toroq

‘road’

2

nagħġa

ngħaġ

–

nagħġiet

–

‘sheep’

2x

frotta

frott

–

frottiet

frottijiet

‘fruit’

2b

–

persuna

–

persuni

–

‘person’

2bx

–

ħalib

–

ħalibijiet

ħlejjeb

‘milk’

2c

–

ċpar

–

–

–

‘fog’

3

pulizija

–

–

–

–

‘police’

4

rkoppa

–

rkopptejn

rkoppiet

–

‘knee’

4x

għajn

–

għajnejn

għajnejn

għejun

‘eye’

Enclitic pronouns: Some nouns in Maltese can take enclitic pronouns to mark possession. For example, the word id (‘hand’) takes the suffixed form of the pronoun tiegħi (‘mine’) to form idi (‘my hand’). The group of nouns that allow this is relatively small and consists mostly of body parts, but the process is certainly common enough to be handled in the grammar (see Stolz et al. 2008: 44–46). The implementational details of enclitic pronouns are discussed in Section 2.3.3. Implementation: The record type of a lexical category defines the internal structure of its inflection table, together with other bits of information. The surface forms of a noun are modeled as a table from NounNumber to strings, where NounNumber is a parameter with possible values for singular, collective, dual and plural (determinate and indeterminate plurals are collapsed together as variants). The Boolean fields hasColl and hasDual indicate whether a noun has collective and dual forms, while takesPron specifies whether a noun takes enclitic pronouns, as described above. This is specified manually in the paradigm as there is no way it can be determined automatically. Nouns in Maltese of Semitic origin are associated with a consonantal root. For example, xorb (‘[alcoholic] drink’) has root √xrb. However the root-andpattern system is rather unpredictable in the case of nouns and no consistent patterns have been found for constructing broken plurals. Thus any broken

Digitizing the grammar and vocabulary of Maltese | 367

plurals must be specified explicitly, and there are no special paradigms for nouns with Semitic roots. The only ‘smart’ behavior in the noun paradigms are simple vowel changes when the singular form ends in an ‘a’, e.g. rota → roti.

2.2.3 Verb The verb is arguably the most interesting aspect of Maltese morphology. Maltese verbs can be divided more or less cleanly in two groups: those which follow a Semitic root-and-pattern morphology, and those which display a concatenative morphology (Spagnol 2011). While all the Semitic verbs follow a root-andpattern morphology, there is a division in the Romance class of verbs. Some have become strongly-integrated into the root-and-pattern system and behave as quadriliteral verbs with the weak j as final radical. Examples include kanta (‘he sang’), vinċa (‘he won’) and serva (‘he served’), from the Italian cantare, vincere and servire respectively. The remainder of Maltese verbs from Romance origin are only looselyintegrated and follow a concatenative morphology of stems and affixes. Examples include żviluppa (‘he developed’), ipperfezzjona (‘he perfected’), antagonizza (‘he antagonized’). The third group of verbs has an English origin, and follows the same morphology as the loosely-integrated Romance verbs. Examples include ibbukja (‘he booked’), ipparkja (‘he parked’), iffittja (‘he fitted’). Despite their different etymology, these two latter groups behave identically from a morphological perspective and thus are not distinguished between in the grammar; they are both treated generically as loan verbs. Classification: The classification of Maltese verbs is covered significantly elsewhere (Borg and Azzopardi-Alexander 1997; Brother Henry F.S.C. 1980; Spagnol 2011; Fabri 2009). While broadly following this linguistic tradition, we are also interested in keeping the representations compact and efficient. Table 2 shows the verb classification as used in the resource grammar. The liquidmedial class corresponds to Fabri’s ITb paradigm (Fabri 2009) (where regular is ITa), which require an epenthetic vowel to be inserted in the imperfective paradigm (Spagnol 2011: 38). As noted by Fabri (2009), għ is not weak, thus defective verbs are technically strong. However, they behave inflectionally as weak verbs and are thus classified as such in our case. Derived verbs: Root-and-pattern verbs in Maltese also exhibit derivational morphology. Derived verbs generally indicate transitivity and/or passiveness with respect to a first-form verb. The different derivational processes are well understood but they occur sporadically; that is to say, no verb root has instanc-

368 | John J. Camilleri

es of all derived forms. The list of such verbs is closed in Maltese and is no longer productive. These have been exhaustively enumerated by Spagnol (2011). Tab. 2: Verb classes treated in the grammar and their characteristics.

Class

Description

Example

Strong Regular

All radicals strong & distinct

qatel (‘he killed’) (√qtl)

Liquid-medial C2 is liquid Geminate C2 & C3 are identical

żelaq (‘he slipped’) (√żlq) ħabb (‘he loved’) (√ħbb)

Weak Assimilative Hollow Lacking Defective

C1 is weak C2 is weak; V1 is long C3 is weak C3 is silent għ

waqaf (‘he stopped’) (√wqf ) dam (‘he delayed’) (√dwm) beka (‘he cried’) (√bkj) qata’ (‘he cut’) (√qtgħ)

Quadriliteral Strong Weak-final

All radicals strong (incl. liquids) C4 is weak

ħarbat (‘he disrupted’) (√ħrbt) pinġa (‘he drew’) (√pnġj)

Irregular

Missing radicals

ħa (‘he took’) (√'ħd)

Loan

Concatenative morphology

ipparkja (‘he parked’)

In the resource grammar, derived verbs are treated simply as separate lexemes in the lexicon. A single parameter in the record type indicates what derived form a verb is in, which determines its morphological paradigm. Concatenative verbs: Loosely-integrated Romance verbs and English loan verbs in Maltese follow an alternative, concatenative morphology. In contrast to the root-and-pattern verb inflection, the stem remains intact throughout the entire table. Mifsud (1995) identifies four classes of loan verbs (types A to D), of which types C and D are not integrated into the Semitic root-and-pattern system. These represent the non-integrated verbs of Romance origin (e.g. ippretenda ‘he expected’) and English loan verbs (e.g. iffitja ‘he fitted’). Despite Mifsud’s distinction, we treat these two classes of loan verbs with the same paradigm in the resource grammar. Participles: Maltese verbs may also have participle forms. Most verbs have a past (passive) participle which is often indistinguishable from the adjective. Examples include miktub (‘written’) from kiteb (‘he wrote’) and misruq (‘stolen’) from seraq (‘he stole’). Very few verbs also have a present (active) participle; they are generally intransitive and describe actions, for example ħiereġ (‘going out’) from ħareġ (‘he went out’) and rieqed (‘sleeping’) from the verb raqad (‘he slept’).

Digitizing the grammar and vocabulary of Maltese | 369

A past participle may be shared by verbs of different derived forms, such as mwaqqaf (‘stopped’) for both waqaf (‘he stopped’) (form I) and waqqaf (‘he stopped [s.t.]’) (form II). An analysis of participles of all the verbs in the resource grammar’s lexicon module can be found in (Camilleri 2013: Appendix B.2). Implementation: Verbs are implemented as tables of the parameter VForm, which distinguishes the different inflection forms that a verb may have, covering the perfective and imperfective aspects and the imperative mood with corresponding subject agreement features. Rather than a table of strings, we instead use a record of strings representing the different stems that may be needed when the base verb form has enclitic pronouns attached to it. In addition to these, we also include in the verb table the present and past participle forms of the verb, which is common practice in the RGL. Since not all verbs have these participles, we also include the two Boolean fields hasPresPart and hasPastPart to indicate their presence. Paradigms: mkV is the smart paradigm for all non-derived verbs. This operator minimally takes the base form of a verb along with a root (in the case of root-and-pattern verbs). The root class is determined and the correct paradigm is applied. If no root is specified as an argument, the verb is treated as a loan verb (with concatenative morphology). Vowel changes are often hard to determine, thus the singular imperative form of a verb is often required as a third argument. For irregular verbs, we can take all 16 forms (7 perfective, 7 imperfective, 2 imperative) together with class, derived form, root and vowel sequence. Verb participles are specified not in the mkV paradigm, but by using the supplementary functions presPartV and pastPartV which add participles to already-constructed verbs. Derived verbs have their own set of paradigms, as they are treated as separate entries in the lexicon. Tab. 3: Verb paradigm examples and how they are constructed in the grammar.

Description

Example call to paradigm

Gloss

Geminate verb

mkV "ħabb" (mkRoot "ħ-b-b")

‘he loved [s.o.]’

Requiring an extra form

mkV "talab" "itlob" (mkRoot "t-l-b")

‘he prayed’

Loan verb

mkV "ipparkja"

‘he parked’

Form II (derived) verb

mkV_II "daħħal" (mkRoot "d-ħ-l")

‘he inserted [s.t.]’

370 | John J. Camilleri

2.2.4 Adjective By comparison to nouns and verbs, the adjective in Maltese is rather straightforward. It inflects for number (singular or plural), and gender in the singular. When an adjective has an inflectional comparative, it does not inflect for further categories. Otherwise, the comparative is formed syntactically with the structural words iktar (‘more’) or inqas (‘less’). In this case the correctly inflected positive form of the adjective needs to be selected. The superlative is always formed syntactically by adding the definite article to the comparative. Paradigms: The smart paradigm for constructing adjectives can take between one and four strings, depending on the predictability of the inflected forms and whether it has a comparative form or not. A second paradigm sameA exists for adjectives which are invariant (and have no comparative form). Tab. 4: Adjective paradigm examples. Description

Example call to paradigm

Gloss

Predictable inflection

mkA "bravu"

‘clever’

Explicit plural

mkA "nadif" "nodfa"

‘clean’

Explicit feminine and plural

mkA "aħmar" "ħamra" "ħomor"

‘red’

Explicit comparative

mkA "dejjaq" "dejqa" "dojoq" "idjaq"

‘narrow’

Invariant inflection

sameA "blu"

‘blue’

2.3 Syntax 2.3.1 Article Absence of an article in Maltese indicates the indefinite: (2)

rajt baqra saw.PFV.1SG cow.SG ‘I saw a cow’

The standard definite article il- is glued to the succeeding word: (3)

rajt il-baqra saw.PFV.1SG DEF-cow.SG ‘I saw the cow’

When that word begins with one of the coronal consonants, it is assimilated: il+ dar becomes id-dar (‘the house’). If the word begins with a vowel, the i from

Digitizing the grammar and vocabulary of Maltese | 371

the article is dropped: l-ajruplan (‘the airplane’). Similarly if a word begins with a consonant cluster with a coronal consonant at its onset, a euphonic i is inserted and the vowel form of the article is used, e.g. l-iskola (‘the school’). The implementation of articles in the resource grammar uses the pre construct, which is specifically designed for modeling variations of a word based on the succeeding letter.

2.3.2 Preposition With indefinite noun phrases, prepositions appear as separate words as in English, e.g. ġo banju (‘in a bath’). When the noun phrase is definite however, the preposition will always assimilate with the article, e.g. ġol-banju (‘in the bath’). This is implemented in the grammar by having both the definite and indefinite forms in the linearization record.

2.3.3 Pronoun Pronouns in Maltese are often enclitic, joining with verbs to indicate direct and indirect objects and with nouns to indicate possession. In the case of verbs, two pronouns can also join together to form a clitic-chain. The suffix forms which the personal pronouns can take are listed in Table 5. While this can be seen as a morphological process, for efficiency reasons the implementation in the resource grammar treats this as a syntactic process via a special bind token. Tab. 5: Whole and suffixed forms for pronouns. Square brackets indicate forms which precede another enclitic pronoun in the dative. Negated forms are not included here.

Agreement NOM

Noun/PREP

Verb ACC

Verb DAT

P1 SG

jien

-i, -ja

-ni

-li

P2 SG

int

-ek, -ok, -k

-ek, -ok, -k

-lek, -lok

P3 SG.M

hu

-u, -h

-u, -h [-hu]

-lu

P3 SG.F

hi

-ha

-ha [-hie]

-lha

P1 PL

aħna

-na

-na

-lna

P2 PL

intom

-kom

-kom

-lkom

P3 PL

huma

-hom

-hom

-lhom

372 | John J. Camilleri

Implementation: A pronoun is implemented with an inherent agreement field, and a table of three strings for personal, possessive, and suffixed form. This suffixed form of the pronoun is used when attaching to nouns and prepositions. The pronoun suffixes for verbs however are not stored as part of the Pronoun record. Rather, the agreement field in the pronoun is consulted directly and the correct suffixed form is chosen within the linearization function of a verb phrase. Pro-drop: Maltese is a pro-drop language, meaning that personal pronouns in the subject position are often dropped since the verb inflection is enough to indicate the subject (Saade 2012). This dropping is optional, so the grammar includes subject pronouns by default and allows them to be omitted explicitly via the ProDrop function defined in the Extra module.

2.3.4 Tense-anteriority system Maltese verbs inflect for mood (indicative and imperative), and in case of the former for aspect (perfective and imperfective). Tenses are formed by combining these with auxiliary verbs kien (‘he was’), ikun (‘he would be’), future marker se/ser/sa (‘he will’) and progressive marker qed (‘he is’). The tense system in the GF Resource Grammar Library is based on features for anteriority (simultaneous, anterior), temporal order (present, past, future, conditional) and polarity (positive, negative). This yields a total of 16 distinct temporal forms. Table 6 lists these tenses and shows how each of them is expressed in Maltese. Tab. 6: The tense-anteriority system used in the Resource Grammar Library and its correspondence to Maltese tenses, using raqad (‘he slept’) as an example.

Anteriority

Tense

Polarity Description

Example

Simultaneous

Present

Pos.

Imperfective

jorqod

Simultaneous

Present

Neg.

Imperfective

ma jorqodx

Simultaneous

Past

Pos.

Perfective

raqad

Simultaneous

Past

Neg.

Perfective

ma raqadx

Simultaneous

Future

Pos.

Prospective

se jorqod

Simultaneous

Future

Neg.

Prospective

m'hux se jorqod

Simultaneous

Conditional

Pos.

Past imperfective

kien jorqod

Simultaneous

Conditional

Neg.

Past imperfective

ma kienx jorqod

Digitizing the grammar and vocabulary of Maltese | 373

Anteriority

Tense

Polarity Description

Example

Anterior

Present

Pos.

(as Simultaneous Past Pos.)

Anterior

Present

Neg.

(as Simultaneous Past Neg.)

Anterior

Past

Pos.

Past perfect

Anterior

Past

Neg.

Past perfect

ma kienx raqad

Anterior

Future

Pos.

Future perfect

se jkun raqad

kien raqad

Anterior

Future

Neg.

Future perfect

m'hux se jkun raqad

Anterior

Conditional

Pos.

Past prospective

kien jorqod

Anterior

Conditional

Neg.

Past prospective

ma kienx jorqod

2.3.5 Phrases Noun phrase: A noun phrase combines all kinds of determiners, pronouns, proper and common nouns, and relative clauses. Most of these elements are linearized into strings in the formation of the noun phrase. The noun phrase also carries information about its agreement features which is needed when combining with verb phrases, as well as whether it is a pronoun and/or in the definite form. In cases where the noun phrase is preceded by a preposition, this preposition may join with the definite article in the NP, or completely assimilate with the phrase if it is a pronoun. This solution builds on that developed in Zammit (2012). Verb phrase: The verb phrase is the most complex aspect of the Maltese resource grammar. A lot of information needs to be propagated up to the VP level and potentially affixed to the main verb. First and foremost, the entire verb record is stored in the verb phrase record without flattening. Secondly, a VP contains a complement, which is typically a non-pronominal object. Two more records fields are needed to indicate whether there are direct and/or indirect objects which need to be joined to the main verb. These fields can encode potentially non-existent information by using the Maybe type, which is modeled on the Haskell equivalent. Adjectival phrase: Adjectival phrases are by comparison quite simple, containing a simple string which inflects for gender/number and a Boolean field to indicate if the AP should go before the noun phrase (superlatives) or after it (comparatives and positives).

374 | John J. Camilleri

2.3.6 Clauses The RGL covers three kinds of clauses: declarative, question and relative clauses. In each case, the clause needs to contain versions of the underlying statement to cover all combinations of tense and polarity. Since question clauses and relative clauses can be made from declarative clauses, the Clause type needs to contain all possible word orders. In other words, it is a function of an Order parameter. Question clauses can also be linearized in different orders depending on whether they are direct or indirect. Variable word order: Intonation in spoken Maltese can be very important in determining whether a statement is a declaration or a question. As a result of this, word order is relatively free and a phrase with a given order can be interpreted in different ways. Similarly, a predicate with a single meaning can be linearized using different word orders. Notwithstanding this, in the resource grammar we take a simpler approach and handle only one word order in each type of clause: SVO for declaratives and VOS for questions. These are arguably the standard orders which are used most commonly. This of course has the effect of reducing the parsing coverage of the grammar.

2.4 Evaluation 2.4.1 Coverage The general methodology for testing GF resource grammars is using treebanks, comparing the linearization against a gold standard (Ranta 2011). A number of treebank files were written during development to test different specific aspects of the grammar, containing in all some 1,307 trees and linearizations. In addition to these treebanks mentioned above, a small treebank of Maltese verb inflections was also compiled during this work. This treebank consists of full inflection tables of 74 Maltese verbs of different paradigms, which have been tabulated in comma-separated value (CSV) format. This amounts to a total of 70,448 unique word forms which have been manually checked. Further details of the evaluation of the resource grammar can be found in (Camilleri 2013, Section 2.4).

Digitizing the grammar and vocabulary of Maltese | 375

2.4.2 Problems encountered Large inflection tables: Our initial approach to verb morphology was to store enclitic pronouns within the inflection table of the verb. Combining inflections for subject, direct objects, indirect objects and polarity yields an inflection table with roughly 1000 unique strings for each verb. This allows very precise morphological control since compile-time strings can be pattern matched and modified. However, while technically possible, this leads to an explosion in the time and space required for grammar compilation. As a result, the grammar was refactored such that the verb inflection table was reduced to just stems, and all enclitic pronouns were suffixed via the bind token at the syntax level. Binding of runtime tokens: Binding works by introducing a special bind token &+ in between tokens which should be linearized as a single word during unlexing. For example, the noun missier (‘father’) may be suffixed with the first person singular enclitic pronoun -i (indicating possession) by producing a list of three tokens: ["missier", "&+", "i"]. When passed through the correct unlexer, this is displayed as a single string "missieri". The lexing process however – which requires the re-insertion of the bind token between stems and suffixes – can be problematic. Lexing must occur as a pre-processing step before input to GF, thereby working without any access to the rules defined in the grammar. The suffix -u for example may be an enclitic pronoun indicating possession in a noun, e.g. qalbu (‘his heart’), as well as a regular conjugational ending of a verb: e.g. talbu (‘they prayed’). The former is composed of the stem qalb and pronominal suffix -u, which should be lexed into tokens ["qalb","&+","u"]. However, a naive external lexer may also try to lex talbu in the same way, which would be incorrect. Without further developments to the GF runtime to allow grammar-based lexing, this resource grammar is of limited use for parsing. Pattern matching on runtime strings: The inability to pattern match on strings at runtime has a big effect on the design of the grammar. This becomes especially evident when stem changes occur because of stress-shifts caused by affixation (also known as Sandhi). The verb fetħet (‘she opened’) for example, undergoes an internal vowel change from e → i when suffixed by the enclitic pronoun -hom to become fetħithom (‘she opened them’). The joining of verb stems and suffixes occurs at the syntactic level, yet by this point it is too late to analyze the verb stem and perform any required vowel changes. Instead, all possible verb stems must be computed within the morphological paradigm and stored in the inflection table. The selection of the correct stem must then be based on parameter values, which requires a lot more work.

376 | John J. Camilleri

Non-existent forms: It is often the case that certain slots in inflection tables simply do not exist. A prime example in Maltese nouns is the dual form, which only a small number of nouns actually have. When present, this form is used to replace both noun and determiner, e.g. xagħarejn (‘two months’). For nouns with no dual form, one uses the numeral for ‘two’ together with the plural form of the noun, e.g. żewġ fliexken (‘two bottles’) (see Gauci and Camilleri 2014). As all nouns have a slot in their linearization table for a dual form, the natural solution would be to set this to the empty string in cases where no dual exists. However, empty strings cause problems during parsing. Furthermore, GF does not allow the checking of strings at runtime. To solve this, an extra nonstring field must be added to the linearization type, such as a Boolean hasDual field, which must then be consulted in the syntax-level functions at runtime.

3 Digital lexicon This section discusses the idea of a full-form computational lexicon for Maltese and introduces Ġabra (‘collection’), an online database of lexical resources with full inflection forms which have been automatically generated by the resource grammar discussed in the previous section. Details about accessing this lexicon are included at the end of this article.

3.1 Available lexical resources Our approach to building a computational lexicon for Maltese is based on the idea of collecting existing lexical resources into a single database. While the number and size of currently available resources is not necessarily enough to cover the whole language, by building an open platform we will be able to add further resources in the future as they become available. The lexical resources used in the building of Ġabra were: – An exhaustive list of all 4,142 root-and-pattern verbs (including hypothetical forms), from the verbal roots database (Camilleri and Spagnol 2013). – A corpus of 654 broken plurals for both nouns and adjectives (Mayer et al. 2013). – A list of over 2,500 verbal nouns found in the Aquilina dictionary and other sources (Ellul 2015). – A Basic English-Maltese dictionary containing some 5,454 English entries (Falzon 2012).

Digitizing the grammar and vocabulary of Maltese | 377

–

A small lexicon (1,776 entries) and basic morphological generator for Maltese from the open-source translation system Apertium (Forcada et al. 2011).

3.2 A full-form lexicon Traditional print dictionaries are organized by head word or lemma, where different word forms for that lemma are specified merely as suffixes. Take the following entry from the Serracino-Inglott (2003: 218) dictionary as an example: ħtieġa n.f.s., pl. -t, -ijiet bżonn; neċessità; siwi; ukoll: eżiġenza; rieda

The entry gives the singular form ħtieġa (‘need’) as a head word, together with the suffixes which give the plural forms. These affixes however cannot be blindly appended to the head word; the correct plural forms in this case are in fact ħtiġiet and ħtiġijiet (not ħtieġat and ħtieġaijiet). So even though the dictionary gives us some information about the other word forms for this lemma, it takes some further knowledge of the language in order to apply the rules correctly. Storage vs. generation: When building a full-form lexicon, one can either store all word forms as individual entries, or store only the headword and produce inflected forms in real-time using morphological generation. The former option is generally more demanding in terms of space requirements, while the latter depends on the morphological predictability of the language in question and the availability of reliable morphological generators. While verb morphology is quite predictable in Maltese, the prevalence of unpredictable broken plurals for nouns and adjectives makes it hard to rely on automatic morphological generation. This is a clear indication that some method of storing full forms is required. We opt to store all inflectional forms in a single digital database, without the use of any real-time morphological generator. This makes the system design simpler, means that the lexicon can more effectively searched, and allows its contents to be more easily converted into other formats. Size calculations: Dalli (2002) estimates that at least 30,000 lemmas must be identified in order to have significant coverage of Maltese. By comparison, Serracino-Inglott’s Maltese dictionary (Serracino-Inglott 2003) contains roughly 26,000 entries, while the Maltese-English volumes of Aquilina’s dictionary (Aquilina 1987, 1990) contain some 80,000. By considering the worst-case inflectional forms of each of the major parts of speech, together with the distribution of the different parts of speech in the Aquilina dictionary, it is estimated that a full-form Maltese lexicon would contain roughly 86 million forms (see Camilleri 2013: Table 3.1).

378 | John J. Camilleri

3.3 Generating full forms with GF 3.3.1 Creating GF modules From the machine-readable word lists mentioned in Section 3.1, a monolingual GF dictionary module can be constructed (following RGL convention, this consists of matching abstract and concrete modules named DictMltAbs and DictMlt respectively). 1. For each lemma in the lexicon, a valid and unique function identifier is generated. This is often an ASCII-ized version of the lemma suffixed with a unique identifier, POS and other relevant information. For example, rikeb (‘he rode’) becomes rikeb_RKB_1_9451_V. 2. A function is defined in the abstract syntax module for this identifier, with a category to match its POS. 3. A linearization for this function is defined in the concrete syntax module, which uses a smart paradigm for the particular POS and supplying (minimally) a lemma form. In the case of Maltese verbs, the root is also supplied since this information is available in the lexicon. It is common to require some post-editing of the generated dictionary grammars since the wrong paradigm will invariably be chosen in some cases. As dictionaries are often quite large however, this may turn into a somewhat time-consuming task. While GF smart paradigms are generally designed to work with minimal information, all relevant data that exists in the source word list can and should be used in order to reduce the guessing required of the smart paradigm.

3.3.2 Producing inflection tables This dictionary module can be used with the GF runtime in order to produce full forms for each of our lexemes. Nouns: The inflection table of nouns can be obtained directly by using the linearize -table command. In cases where nouns take a postclitic pronoun to indicate possession, the following template tree is constructed and linearized with binding in order to produce the correctly inflected forms: DetCN (DetQuant (PossPron ) NumSg)

where is the pronoun indicating the direct object (possessor) and is the noun under analysis.

Digitizing the grammar and vocabulary of Maltese | 379

Adjectives: Simply using the linearize command on the lexical entries is enough to get the inflections for gender and number, as well as the forms for comparative and superlative. Verbs: Since many morphological features of verbs are produced by syntactic constructors in the grammar, we again need to use template trees. At the top level, sentences in the different tenses and polarities are generated using these templates for perfective, imperfective and imperative respectively: UseCl (TTAnt TPast ASimul) (PredVP (UsePron ) ) UseCl (TTAnt TPres ASimul) (PredVP (UsePron ) ) UttImpSg (ImpVP ) UttImpPl (ImpVP )

where stands for polarity, is the subject pronoun, and is a clause, which may be defined in various ways. A clause using both direct and indirect object clitics, for example, is produced using: AdvVP (ComplSlash (SlashVa ) (UsePron )) (PrepNP for_Prep (UsePron ))

where is the verb under analysis, and and represent direct and indirect object pronouns respectively.

3.4 Database design Traditional relational databases work with a strict schema system, where the structure of all data is fixed at design time. As we are dealing with lexical resources from different sources with different structures, this is not ideal. In addition, in order to maximize adaptability to future resources, the database for this lexicon must be flexible enough to support heterogeneous data as it is found. So rather than define a highly specific data schema, we instead outline only a top-level structure for the database. Lexemes are entries as we think of them in a dictionary. Each is represented minimally by a headword (lemma) and POS tag, but can include an English gloss and other features common to the entire entry. Word forms represent the inflected forms of a lexeme, and include the lemma as word form. They only need to consist of a single string indicating the surface form, but ideally include a list of the features that they inflect for. Roots are modeled as separate entities, as they are shared by multiple lexemes. In addition, all entries should include a field indicating the citation references of the sources of that entry. This exists so

380 | John J. Camilleri

that the information stored in the database does not become disconnected from its original authors. The database engine chosen for implementing this lexicon is MongoDB, a NoSQL document-oriented system which stores data as JSON-style documents with dynamic schemas.

3.5 Web application Ġabra 3.5.1 User interface Figure 2 shows a screenshot of the Ġabra web interface in use. The search field in the top menu of the screen allows one to search the entire database, including all lemmas, inflected word forms and English glosses. Clicking on an entry in the results page takes one to the detailed view page, such as that for the entry ħareġ in Figure 2. This page includes gloss and root information, a list of related entries, and all the word forms for the lexeme entry in a table which can be filtered.

Fig. 2: Viewing the details and inflectional forms of the entry for ħareġ.

Digitizing the grammar and vocabulary of Maltese | 381

As mistakes may occur during importing or automatic inflection, Ġabra also includes a system for users to report errata or omissions they may come across. These reports can then be checked manually and corrected as necessary. Apart from reporting errors, users may also suggest new entries to be included in the collection. Submissions are also placed in a queue which are checked manually before inclusion.

3.5.2 Web service API and offline use Apart from a user interface, Ġabra can also be accessed as web service, allowing the resource to be queried from other scripts or applications. The web service has a fully documented API which anyone may use without registration. For users who wish to perform computationally intensive tasks on the lexicon, the entire database can also be downloaded for offline use.

4 Conclusion 4.1 Contributions This paper presents an overview of two significant contributions to the Maltese computational linguistics. The GF resource grammar for Maltese discussed in Section 2 represents the largest known effort, to date, of representing Maltese grammar rules computationally. Despite being considered complete in terms of GF’s resource grammar library, it does not have complete coverage of Maltese itself, and cannot be used for general purpose parsing or translation. It is however very well-suited for morphological generation and natural language verbalization from abstract semantic representations. GF itself is also an ideal tool for defining multilingual controlled natural languages (CNLs). The second contribution is a digital collection of lexical resources called Ġabra, built out of a need for a computational word list for Maltese which is openly available. The collection does not claim to cover all of Maltese, but it already contains some 15,861 lexeme entries and over 4.5 million inflectional forms. The varying detail of these entries is a direct result of the heterogeneous nature of the different sources used in building the collection. It does however provide a flexible platform which is open to the addition of future resources, which could become the basis of a complete, full-form lexicon for Maltese.

382 | John J. Camilleri

4.2 The future There are a number of shortcomings in the resource grammar which should be fixed in the short term. For those, see (Camilleri 2013: Chapter 4). The lexical content of Ġabra can be improved by manually completing the missing annotations of many of the current entries (including POS tags, root information, and English glosses). Furthermore, there are other existing resources available from which new data could be imported into the database, such as the revised spellings from the document Deċiżjonijiet 1 (Il-Kunsill Nazzjonali ta’ l-Ilsien Malti 2008a), online terminologies like IATE, word lists extracted from the MLRS corpus, and user submissions from the Ġabra web application itself. In 2014, a committee for Information Technology was set up by the National Council for the Maltese Language, with the aim of publishing a roadmap for Maltese language technology and supporting its development. Working towards the goal of establishing a national resource for the languages of Malta, substantial work is currently being carried out on correcting and extending the data in the Ġabra lexicon and launching it is a national dictionary of Maltese (p.c., November 2014). Acknowledgements: The research work disclosed in this publication was supported by the MOLTO project, via funding from the European Union’s Seventh Framework Program (FP7/2007–2013) under grant agreement no. FP7-ICT247914, and the Strategic Educational Pathways Scholarship (Malta), which was part-financed by the European Union – European Social Fund (ESF) under Operational Program II – Cohesion Policy 2007–2013, “Empowering People for More Jobs and a Better Quality of Life”.

Abbreviations ACC

CLEM COLL

CV DAT DEF F

GF M

MLRS NOM

accusative Corpus of Learner English in Malta collective comma-separated value dative definite feminine Grammatical Framework masculine Maltese Language Resource Server nominative

Digitizing the grammar and vocabulary of Maltese | 383

PFV PL

POS PREP

RGL SG

perfective plural Part-of-Speech preposition Resource Grammar Library singular

Links and licenses Maltese GF resource grammar The source code for the grammar is released under the GNU Lesser General Public License. This license allows free use for any purpose, including use within non-open source applications. URL: http://www.grammaticalframework.org/lib/src/maltese/ Ġabra The lexical data contained in Ġabra is available via web service and for download. It is covered by a Creative Commons Attribution (CC-BY) license, which allows any use as long as attribution is given. URL: http://mlrs.research.um.edu.mt/resources/gabra/ Verb inflection tables These manually-checked verb inflection tables are made available under the MIT license, allowing any use without restriction. URL: https://github.com/johnjcamilleri/maltese-verb-inflections/

References Angelov, Krasimir, Lauri Carlson, Ramona Enache, Inari Listenmaa, Aarne Ranta, & Shafqat Mumtaz Virk. 2013. Lexicon extraction in MOLTO. Aquilina, Joseph. 1987. Maltese-English dictionary. Vol. 1, A–L. Valletta/Malta: Midsea Books. Aquilina, Joseph. 1990. Maltese-English dictionary. Vol. 2, M–Z. Valletta/Malta: Midsea Books. Borg, Albert, & Marie Azzopardi-Alexander. 1997. Maltese. London: Routledge. Brother Henry F.S.C 1980. Grammatika Maltija. 6th edn. Gżira/Malta: De La Salle Brothers. Camilleri, John J. 2013. A computational grammar and lexicon for Maltese. M.Sc. Thesis, Chalmers University of Technology. Camilleri, John J. & Michael Spagnol. 2013. An online database of root-and-pattern verbs. Talk at the 4th International Conference on Maltese Linguistics (GĦILM 2013). Lyon, France, 2013. Dalli, Angelo. 2002. Computational lexicon for Maltese. M.Sc. Thesis, University of Malta. Détrez, Grégoire & Aarne Ranta. 2012. Smart paradigms and the predictability and complexity of inflectional morphology. In Proceedings of the 13th Conference of the EACL, 645–653. Avignon/France: Association for Computational Linguistics. Ellul, Leanne. 2015. Verbal nouns in Maltese. M.A. Thesis (forthcoming), University of Malta. European Commission. 2011. User language preferences online: Analytical report. Fabri, Ray. 2009. Stem allomorphy in the Maltese verb. Ilsienna 1. 1–20. Falzon, Grazio. 2012. Basic English-Maltese Dictionary. META-SHARE. Last updated 31/01/13. http://metashare.metanet4u.eu/repository/browse/basic-english-maltese-dictionary/ 13fc5802abc511e1a404080027e73ea2a210be7dd5c44a3b9dd47afb4b2a34ef/

384 | John J. Camilleri

Forcada, Mikel L., Mireia Ginestí-Rosell, Jacob Nordfalk, Jim O’Regan, Sergio Ortiz-Rojas, Juan Antonio Pérez-Ortiz, Felipe Sánchez-Martínez, Gema Ramírez-Sánchez & Francis M. Tyers. 2011. Apertium: A free/open-source platform for rule-based machine translation. Machine Translation 25(2). 127–144. Gatt, Albert & Claudia Borg. 2011. Using the MLRS corpus. Tech. rept. Institute of Linguistics, University of Malta. Gauci, Phyllisienne & Maris Camilleri. 2014. Syncretism in Maltese: The case of the dual. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 117–128. Berlin: Akademie Verlag. InterActive Terminology for Europe (IATE). 2015. Last updated 12/05/2015. http://iate.europa.eu/ Il-Kunsill Nazzjonali ta' l-Ilsien Malti. 2008a. Deċiżjonijiet 1 tal-Kunsill Nazzjonali tal-Ilsien Malti dwar il-Varjanti Ortografiċi. Floriana, Malta. Il-Kunsill Nazzjonali ta' l-Ilsien Malti. 2008b. Innaqqsu l-inċertezzi 2 – Seminar fuq il-kliem ta’ nisel Ingliż fil-Malti. Il-Kunsill Nazzjonali ta' l-Ilsien Malti. 2012. The permanent programme for research on the National Maltese dictionary. Mayer, Thomas, Michael Spagnol & Florian Schönhuber. 2013. Fixing the broken plural in Maltese. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 129–158. Berlin: Akademie Verlag. Mifsud, Manwel. 1995. Loan verbs in Maltese: A descriptive and comparative study. Leiden/ The Netherlands: E.J. Brill. MLSS. 2012. Maltese Language Software Services. Accessed 23/11/2015. http://metanet4u.research.um.edu.mt/ Ranta, Aarne. 2009. The GF Resource Grammar Library. Linguistic Issues in Language Technology 2(2). Ranta, Aarne. 2011. Grammatical framework: Programming with multilingual grammars. Stanford, CA: CSLI Publications. Rosner, Michael, Joe Caruana & Ray Fabri. 1998. Maltilex: A computational lexicon for Maltese. In Michael Rosner (ed.), Computational approaches to Semitic languages: Proceedings of the Workshop held at COLING-ACL98, 97–105. Université de Montréal. Rosner, Michael, Ray Fabri, Joe Caruana, M. Loughraïeb, Matthew Montebello, David Galea & Gordon K. Mangion. 1999. Linguistic and computational aspects of Maltilex. In Arabic translation and localisation symposium: Proceedings of the Workshop, 2–10. Rosner, Michael, Ray Fabri, Duncan Paul Attard & Albert Gatt. 2006. MLRS, a Resource Server for the Maltese language. In John Abela, Angelo Dalli & Kristian Guillaumier (eds.), Computer Science Annual Workshop (CSAW), 90–98. University of Malta. Rosner, Mike & Jan Joachimsen. 2012. The Maltese language in the digital age. IL-Lingwa Maltija fl-Era Diġitali. (META-NET White Paper Series). New York: Springer. Saade, Benjamin. 2012. Pro-drop systems in Maltese and Italian: A quantitative approach. Ilsienna – Our Language 2. 67–78. Schembri, Tamara. 2012. The broken plural in Maltese: A description. (Il-Lingwa Tagħna 3). Bochum: Brockmeyer. Serracino-Inglott, Mario. 2003. Id-Dizzjunarju Malti. 2nd edn. Blata l-Bajda/Malta: Merlin. Spagnol, Michael. 2011. A tale of two morphologies: Verb structure and argument alternations in Maltese. Ph.D. Thesis, University of Konstanz.

Digitizing the grammar and vocabulary of Maltese | 385

Steinberger, Ralf, Bruno Pouliquen, Anna Widiger, Camelia Ignat, Tomaž Erjavec, Dan Tufis & Dániel Varga. 2006. The JRC-Acquis: A multilingual aligned parallel corpus with 20+ languages. The 5th International Conference on Language Resources and Evaluation. Proceedings, 2142–2147. Stolz, Thomas, Sonja Kettler, Cornelia Stroh & Aina Urdze. 2008. Split possession. Amsterdam/Philadelphia: Benjamins. Tiedemann, Jörg. 2009. News from OPUS – A collection of multilingual parallel corpora with tools and interfaces. In Galia Angelova, Kalina Bontcheva, Ruslan Mitkov & Nikolai Nicolov (eds.), Recent advances in natural language processing, 237–248. Amsterdam/ Philadelphia: Benjamins. Virk, Shafqat Mumtaz. 2013. Computational linguistics resources for Indo-Iranian languages. Ph.D. Thesis, Chalmers University of Technology/University of Gothenburg. Zammit, Angelo. 2012. Extending the GF framework towards coverage of the Maltese language. B.Sc. Dissertation, University of Malta.

Index of Authors Abela, Giovanni Francesco 49, 51, 53ff., 62f., 65, 69f., 72f., 81f., 86 Ackerman, Farrell 116 Agius, Dionisius A. 55, 66, 85, 166, 194, 300, 316 Aikhenvald, Alexandra 296 Aissen, Judith 150 Alaya, Martín de 161 Ambros, Arne A. 223, 273f., 281 Anderson, Stephen R. 117 Angelov, Krasimir 364 Angoujard, Jean-Pierre 19 Aquilina, Joseph 3, 5, 28, 37, 42, 49, 74, 77, 85, 178, 201, 222f., 273f., 297, 300f., 305, 315, 317, 319f., 322, 325, 332, 361f., 376f. Arberry, A. J. 154 Aronoff, Mark 117, 122f., 139, 143f. Avram, Andrei A. 49, 53, 55, 85f. Azzopardi-Alexander, Marie 7, 24ff., 28, 35, 37, 42, 154, 175ff., 185f., 189, 195, 201, 211f., 215, 222f., 225f., 228, 230, 232, 234, 237, 242, 306, 334, 367 Backley, Phillip 20, 26, 29f. Baerman, Matthew 116f., 119, 129 Baglioni, Daniele 323 Baker, Colin 333 Bakker, Dik 199 Bastardas Parera, Joan 163 Belloro, Valeria 168 Ben Abdelkader, Rached 190f., 324f. Bhat, Darbhe N. 199f. Bickerton, Derek 294 Blau, Joshua 163, 165 Boersma, Paul 95 Bonelli, Luigi 59, 63, 66, 71, 73 Borg, Albert 7, 24ff., 28, 42, 129f., 150, 154f., 159, 175ff., 185f., 189, 195, 201, 211f., 215, 222f., 225f., 228, 230, 232, 234, 237, 242, 306, 334, 355, 367 Borg, Alexander 7f., 10, 23, 25ff., 33, 35, 42, 49f., 52, 55, 57, 63, 65, 72ff., 80, 82ff., 86f., 154f., 165f., 195, 273f., 279f., 284

Borg, Claudia 360 Borin, Lars 318 Bossong, Georg 150, 166 Boucherit, Aziza 13 Brame, Michael K. 5f., 19, 23, 27, 31, 39 Brennan, Susan E. 94, 110 Bresnan, Joan 217f. Brincat, Joseph M. 166, 168f., 298, 300, 302, 316f., 319, 321f. Broadbent, Judith M. 30 Brother Henry F.S.C. 367 Brustad, Kristen 164 Buck, Carl Darling 318 Buhagiar, Victor 224 Buonamico, Giovan Francesco 52f., 58, 62, 65, 71, 73 Busuttil, V. 53f., 74, 77 Butterworth, Brian 94 Cachia, Lawrenz 52, 71 Camenzuli, Anthony 49, 51f., 58f., 63, 66, 71, 73 Camilleri, Antoinette 8, 334f., 356 Camilleri, John 362, 369, 374, 376f., 382 Camilleri, Maris 32f., 116, 118f., 121, 124, 129ff., 136, 200, 212, 376 Campione, Estelle 91, 93 Cantineau, Jean 82, 86 Cardinaletti, Anna 217ff. Cardona, Tony 50, 52, 63, 72 Carletta, Jean 95 Caruana, Josephine 179 Cassar, Mario 49 Cassola, Arnold 52, 55, 285 Caubet, Dominique 194 Caxaro, Peter/Caxaru, Pietru 11, 52, 58, 65, 77, 79ff., 84, 86f. Čéplö, Slavomír 154 Chafe, Wallace L. 150 Clark, Herbert 93 Clements, George Nick 24 Coghill, Eleanor 165 Cohen, David 1ff., 19, 27, 31, 50, 52, 54f., 72, 84ff.

388 | Index of Authors

Comrie, Bernard 131, 150, 154, 318ff., 324 Corbett, Greville G. 116f., 129, 294ff., 307, 312f. Corriente, Federico 161f. Couper-Kuhlen, Elizabeth 93 Cowan, William 4, 12, 52, 79f., 84, 86 Cremona, Antonino 270f., 274, 281 Cruttenden, Alan 91ff., 101, 108 Cutler, Ann 93 Cysouw, Michael 199 Dalli, Angelo 361, 377 Dalrymple, Mary 150 Darden, Bill 126 de Soldanis, Giovanni Pietro F. A. 55 Dessoulavy, C. L. 49, 54 Détrez, Grégoire 365 Deuchar, Margaret 333 Dixon, R. M. W. 150 Dogliotti, Miro 324 Döhla, Hans-Jörg 152, 161, 163 Eklund, Robert 93f., 110 Ellul, Leanne 362, 376 Ellul, Sonia 336 Erwin, Wallace M. 164f. Fabri, Ray 118, 154, 201, 223, 273f., 279, 281, 297, 367 Falzon, Grazio 362, 376 Farrugia, George 297, 301, 308ff. Farrugia, Paulseph-John 92, 95 Favereau, Francis 220f. Ferguson, Charles 293 Ferrando, Ignacio 155 Firman, Christine 356 Firth, John Rupert 3 Fischer, Wolfdietrich 157, 160 Forcada, Mike L. 376 Fox Tree, Jean E. 93 Fsadni, Michael 12, 52, 179 Galea, Joe 224 García García, Marco 150 Garcia, Ofelia 334, 345, 356 Garvice, Charles 224 Gatt, Albert 360

Gauci, Phyllisienne 212, 376 Gibson, Maik 190f. Givón, Talmy 150f., 200, 294 Goldenberg, Gideon 20 Goldman-Eisler, Frieda 94 Grech, Prospero 154 Greenbaum, Sidney 307 Gregor, Bernd 310 Grosz, Barbara 93 Guaïtella, Isabelle 93f. Gumperz, John 293 Halle, Morris 19 Hammett, Sandra 195 Harris, John 20 Haspelmath, Martin 131, 164, 179, 199, 208f., 212, 294, 318 Haugen, Einar 294 Hayes, Bruce 19, 93 Heine, Bernd 2 Heusinger, Klaus von 150 Hickey, Raymond 312 Hirschberg, Julia 93 Hoberman, Robert D. 122f., 139, 143f., 272ff. Holes, Clive 193, 306 Hoop, Helen de 150 Hooper, Joan 282 Hulst, Harry van der 20, 24 Hume, Elizabeth 19, 23f., 38 Ibrahim, M. H. 299 Idsardi, William 20 Iemmolo, Giorgio 149f., 167f. Iraqui Sinaceur, Zakia 325 Isserlin, Benedikt S. J. 37, 42 Jake, Janice 333 Jakobson, Roman 24, 29 Joachimsen, Jan 359 Keenan, Edward L. 150 Khan, G. A. 155 Kontzi, Reinhold 153 Koutsoudas, Andreas 164 Krier, Fernande 85f. Krug, Manfred 275

Index of Authors | 389

Ladd, Robert D. 31 Ladefoged, Peter 26 Lanza, Elizabeth 332 Laufer, Asher 32 Launey, Michel 206f. Lazard, Gilbert 220 Ledgeway, Adam 168 Lentin, Jérôme 1f. Levin, Aryeh 163f., 272 Lindsey, Geoff 20 Lonnet, Antoine 1f. Lowenstamm, Jean 19f. Lucas, Christopher 188 Lyons, Christopher 157 Maclay, Howard 94 Maddieson, Ian 26 Malouf, Robert 116 Marantz, Alec 19 Mascaró, Joan 34 Mayer, Thomas 362, 376 Megiser, Hieronymus 52ff., 58, 62, 65, 70, 72, 77, 87 Michel, Andreas 169 Mifsud, Manwel 38, 117, 130, 135ff., 142ff., 155, 159, 195f., 323, 332, 366, 368 Moniz, Helena 93 Mori, Laura 332 Mous, Marten 217 Murray-Curtis, Jackie 344 Myers-Scotton, Carol 333, 352, 355 Nichols, Johanna 164 Nikolaeva, Irina 150 Osgood, Charles E. 94 Owens, Jonathan 305f. Pace, Joe Felice 86 Paggio, Patrizia 92, 95 Paster, Mary 34 Payne, John R. 222 Pearson, Mark 93 Peltzer, Louise 220 Pensado Ruiz, Carmen 162 Piccitto, Giorgio 320, 322ff. Poplack, Shana 310, 312

Portelli, Josette 342 Posner, Rebecca 168 Procházka, Stephan 299ff. Puech, Gilbert 8, 13, 19, 22, 26f., 36f., 119, 121, 212f., 228 Pustet, Regina 226 Quay, Suzanne 333 Raimy, Eric 20 Ranta, Aarne 363ff., 374 Río Sánchez, Francisco del 161, 165 Rohlfs, Gerhard 324 Rose, Ralph L. 94 Rosiello, Luigi 324 Rosner, Michael 359ff. Saade, Benjamin 200, 372 Sadler, Louisa 129 Samuels, Bridget D. 19f. Sankoff, Gillian 312 Sant, Alfred 224 Šarfī, Yusuf, as- 324 Saydon, Pietru P. 130 Schabert, Peter 5, 142, 272 Schall, Anton 208 Scheer, Tobias 20 Schembri, Ludovik 52, 54, 59, 74, 77, 79f. Schembri, Tamara 362 Sciriha, Lydia 178 Serracino Inglott, Erin 317, 319f. Serracino-Inglott, Mario 377 Shriberg, Elizabeth E. 92ff., 110 Siewierska, Anna 164, 199, 217 Simeone-Senelle, Marie-Claude 13 Sims, Andrea D. 208 Skippon, Philip 52f., 58, 62, 65, 70f., 73 Spagnol, Michael 115, 117, 129ff., 319f., 362, 367f., 376 Spencer, Andrew 118 Starke, Michael 217ff. Steinberger, Ralf 360 Stolz, Christel 297, 305, 307, 310 Stolz, Thomas 200, 272, 295, 366 Stumme, Hans 3ff., 7, 27, 85, 177, 192, 194ff., 265 Stump, Gregory 117

390 | Index of Authors

Surridge, Marie E. 310 Sutcliffe, Edmund F. 3, 26, 32, 178, 223 Swart, Peter de 150 Swerts, Marc 93

Vella, Olvin 53, 55, 152 Véronis, Jean 91, 93 Viitso, Tiit-Rein 217 Virk, Shafqat Mumtaz 364

Tadmor, Uri 294, 318, 326ff. Thomason, Sarah G. 312, 356 Thornton, Anna M. 200 Tiedemann, Jörg 360 Twist, Alina 20

Weber, Éloide 150 Weenink, David 95 Wehr, Hans 53, 77 Wei, Li 334, 345, 356 Weinreich, Uriel 293 Wettinger, Godfrey 12, 49ff., 55ff., 64ff., 72, 74ff., 179 Williams, Maurice 94, 110 Wilmsen, David 176, 179, 187, 190, 195f. Winter, Werner 294 Wolf, Matthew 19, 23, 34 Wright, William 156ff.

Ulfsbjorninn, Shanti J. 23 Ussishkin, Adam 20 Vaissière, Jacqueline 91, 93 Van Berkum, J. J. A. 295 Van den Toorn, M. C. 217 Van Driem, George 217 Vanhove, Martine 8, 11, 13, 19, 27, 52, 54, 84ff., 139, 179 Vassalli, Michael A. 32 Vassalli, Michelantonio 177f. Vassalli, Michele Antonio 177 Vassalli, Mikiel Anton 66, 72, 84, 322 Vassallo, Marlene 338 Vella Gera, Alex 224 Vella, Alexandra 38, 92f., 95, 99, 103, 105 Vella, Francis 77, 177, 179, 190

Yamamoto, Mutsumi 152 Zaliznjak, Andrej A. 119 Zammit, Angelo 373 Zammit, Martin R. 161f., 324 Zammit, Raymond 344 Zipf, George 280 Živanović, Peter 24 Ziwārī, 'Ali az- 324

Index of Languages Arabic 2ff., 6ff., 12, 27, 32, 50, 55, 74, 77, 79f., 82, 85, 149, 155ff., 161f., 164ff., 169f., 176, 179, 190, 192, 194ff., 272f., 281f., 293f., 298ff., 304, 306f., 309f., 315ff., 319, 321, 324f., 327 – Anatolian 195 – Andalusi 161ff., 167, 169, 179f. – Beirut 164 – Classical 9, 149, 155f., 160f., 164, 166, 208f., 213, 282, 299f. – Cypriot 161, 166, 169 – Egyptian 159, 176, 306 – Galilean 164 – Gulf 306 – Iraqi 161ff., 169, 306 – Judeo-Arabic 161 – Levantine 161, 163, 165f., 169, 177, 179, 190, 193, 195f. – Maghribi 7, 9 – Moroccan 176, 179, 194, 325 – Nigerian 306 – Old 63 – Omani 195 – Siculo 166f., 169, 300, 316, 322, 332 – Syrian 164 – Tunisian 177, 190, 192f., 195, 325 – Yemeni 162, 195 Aramaic 161f., 165, 169, 171ff. Archi 326 Berber 322, 326 Bezhta 326 Breton 220f. Bulgarian 364 Castilian 163 Chamorro 295 Coq Wong 326 Dutch 217, 295 English 49, 157, 166, 224, 273, 293f., 296, 298, 305ff., 315ff., 319ff., 325ff., 331ff., 341f., 344ff., 349ff., 359f., 362ff., 367f., 371, 376f., 379f., 382 Estonian 217

Finnish 364 French 3f., 295f. German 295, 364 Gurindji 326 Hebrew 161, 165 Hindi 364 Icelandic 295 Imbabura Quechua 326 Indonesian 326 Iraqw 217 Italian 7, 49, 157, 166, 200, 218, 293ff., 297f., 302ff., 309f., 315f., 319ff., 329, 359, 367 Japanese 315, 326 Kabyle 322 Kanuri 306 Kildin Saami 326 Latin 162, 302, 305, 320 Limbu 217 Nahuatl 206f. Persian 296 Romanian 326 Sakha 326 Saramaccan 326 Serbo-Croat 295 Sfax 324 Sicilian 7, 166, 293f., 298, 302, 304, 306f., 309f., 315f., 319ff. – Old 149, 155, 166ff. Spanish 161, 168, 295 Swahili 326 Tahitian 220f. Takia 326 Thai 326 Tuscan 323 Vietnamese 326 Yaqui 326

Index of Subjects adstrate 149, 155 amodal glides 19, 31f., 41 Arabic dialects 7f., 26, 30, 86, 159f., 176, 179, 187, 190, 195f., 272, 299, 301 aspect 2, 13, 372 auxiliaries 2, 13, 372

grammatical gender 295, 297f. grammaticalization 2, 162, 168, 186

bilingualism 162f., 294, 316, 331, 334, 336, 355 borrowing 7, 86, 293f., 297, 306, 311f., 317f., 321, 326f. – lexical ~ 294, 297f., 303, 305ff., 309ff.

imaala/imāla 7f. 10ff., 30, 50, 73, 76, 79f., 86f. imperfective 4, 13, 115, 118ff., 126ff., 131ff., 137ff., 177, 242, 367, 369, 372, 379 inflection 212, 295f., 362, 365f., 368f., 372, 374ff., 378ff. inflectional morphology 115ff., 130ff., 141, 144f. interrogatives 175ff., 182, 185f., 188ff., 194f. ipsity 252f., 257f., 261

Cantilena 1, 11ff., 52, 54, 58, 62, 65, 70, 72, 77, 79f., 84, 86, 179 classification of nouns 293, 295, 299 clipping see truncation clitic doubling 154f., 163, 167f. code-mixing/language-mixing 332f., 335f., 342, 355 code-switching 162f., 331, 336f., 342, 351f., 356 comparative 370, 373, 379 comparative linguistics 1f., 4ff., 11f., 200, 265 computational linguistics 2, 359, 381 contact language 6, 149, 161, 166, 169 conversation (adult/child) 331ff., 344, 350, 352, 354f. copula 176f., 223, 226ff., 234 corpus studies 49, 65, 92, 95, 167, 179ff., 186ff., 191, 199, 201f., 224f., 228, 230, 232ff., 261f., 264, 317, 361ff.

hesitation lenghthening 91ff., 96f., 101ff., 105ff. historical phonology 1, 49, 84

language contact 2, 7ff., 11, 149, 155, 162, 165f., 169, 293f., 302f., 305ff., 309, 311f., 316, 320, 322, 335 loanwords 38f., 55, 77, 86, 213, 294, 315f., 322, 326, 328, 332, 345 Maltese dialects 4, 35ff., 42, 71, 73, 86, 119, 156, 285, 289, 321 Maltese Language Resource Server (MRLS) 38, 134, 140f., 178, 180, 191, 218f., 262ff., 270, 281, 283, 286, 360f., 382 Maltese phonetics/phonology 1ff., 19, 27, 212, 315 markedness 150, 209f., 212 melodic elements 19f., 28ff., 36f., 40f.

element theory 20, 26, 28, 40 equipollent verbs 131ff., 136, 140ff.

nicknames 49ff., 55f., 58ff., 63, 66ff., 71, 73f., 82, 85 non-Semitic Maltese 3, 118, 120ff., 135 number 21, 176, 201, 205f., 208ff., 212, 221, 295, 333f., 337f., 365f., 370, 373, 379 numerals 222, 269ff., 275, 277, 282, 285, 337ff., 352

gender assignment 293f., 296f., 299, 301ff., 308f., 311f. gender system 293ff., 299, 303, 305, 307, 312

object marking 149, 151, 155, 158, 163f., 166, 169f. onomastic (data) 49, 84, 86f.

derivation 23, 76, 80, 130, 296 devoicing of obstruents 49f., 52f., 65, 84 diachrony 49, 117 digraph 27, 66, 68, 225, 285

Index of Subjects | 393

pausal diphthongs 20, 37f., 42, 44 pauses – filled/unfilled 91ff., 101, 103, 107ff., 236ff., 246, 249f., 253, 255ff. – silent 92ff. perfective 4, 6, 13, 21, 117ff., 131, 133ff., 137ff., 177, 186, 191, 369, 372, 379 personal pronouns 151ff., 164f., 177ff., 186, 200ff., 205ff., 217, 221ff., 226, 228f., 243, 247, 251, 254, 265, 371f. phonological changes 8, 49ff., 282 phonological expressions (PE) 19f., 23, 26f.,. 29f., 34ff., 38, 41 phonological rules 9, 34, 49, 210f., 232, 297, 341f. place-names 49ff., 63, 75ff., 86f. plural – broken ~ 269, 273, 275, 279, 282, 284f., 288, 362, 366, 376f. – sound~ 273ff., 279, 284f. polar interrogatives/questions 175ff., 185, 188ff., 192, 194ff. prosodic components 19ff., 34f., 37ff., 91f., 97, 109, 217, 260 regressive voicing assimilation 43, 49f., 52f., 84 Romance dialects/varieties 162, 166, 169 Romance loans 52, 55, 77, 86, 121f., 135f., 140, 273ff., 281, 293, 305f., 309, 315ff., 319ff., 367f.

sandhi 219ff., 236f., 240, 242, 244, 248, 254f., 259ff., 375 Semitic Maltese 3, 117f., 120, 122ff., 138, 294, 332, 366ff. short-long distinction 206, 210, 213, 217, 221f., 225, 233, 261f., 265 stem-alternation 116, 121, 144 stress 5, 7f., 19f., 23, 35f., 38ff., 99, 105, 108, 153, 164, 167, 215, 217f., 220, 264, 272ff., 279, 285 superstrate 149, 155, 169 surnames 49ff., 55, 58ff., 66f., 71, 74, 77, 82f. syllabic structure 5, 7, 20, 23, 27, 41, 95, 198, 212, 295 syntax 12, 351, 353, 361, 363f., 370, 375f., 378 translanguaging 334f., 350, 356 truncation (aka clipping) 207f., 209ff., 225, 236 turn-taking 94, 108, 110 verbal paradigm 115ff., 127, 144 verbal system 2, 9, 13 vocalic system 1, 3f., 7, 10 vowel harmony 8f., 12, 38, 40, 50, 83f., 87 vowel length(ening) 6, 19f., 32, 35, 38, 40, 206, 285 word stress 20, 35, 38f., 217f.