Maltese: Contemporary Changes and Historical Innovations 9783110783834, 9783110783766

Table of contents :
Preface
Contents
Part I: Lexicon
The semantic structure of adjectives referring to the holiness in Maltese and Arabic languages
The lexical ta’ construction
Nunation from Arabic to Maltese
Part II: Syntax
The verb sequence V1+V2 in Maltese
Differential object indexing in Maltese – a corpus based pilot study
Part III: Morphosyntax
Arabic prepositions and their Maltese equivalents
On WH-PREPS. A contrastive grammatical sketch of Maltese fejn and Spanish donde/donde
Can frequency predict length? A crosslinguistic investigation of Zipf’s law for European adpositions
Part IV: Phonology
Regressive voicing assimilation in Maltese: Diachrony and typology
Index of Authors
Index of Languages
Index of Subjects

Citation preview

Przemysław Turek and Julia Nintemann (Eds.) 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 (Marburg), 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 30

Maltese

Contemporary Changes and Historical Innovations Edited by Przemysław Turek and Julia Nintemann

ISBN 978-3-11-078376-6 e-ISBN (PDF) 978-3-11-078383-4 e-ISBN (EPUB) 978-3-11-078387-2 ISSN 1617-2957 Library of Congress Control Number: 2022932652 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. © 2022 Walter de Gruyter GmbH, Berlin/Boston Cover image: Alpha-C/iStock/Thinkstock Printing and binding: CPI books GmbH, Leck www.degruyter.com

Preface In this volume, we present the collection of articles documenting the contributions to the 7th International Conference on Maltese Linguistics (7 Lingwistika Maltija) held in Kraków, Poland, from July 10 to 11, 2019. The conference was organized by GĦILM (L-Għaqda Internazzjonali tal-Lingwistika Maltija/International Association of Maltese Linguistics) with the local support of the Institute of Middle and Far East of the Jagiellonian University and is part of GĦILM’s conference series on Maltese linguistics which has established itself as an important forum for scholars of Maltese since 2007. The theme of the conference was “Contemporary Maltese: Linguistic Challenges in Communication and Translation”. The program included results of the research projects presented by fifteen scholars from Great Britain, Germany, Malta, Poland, Romania, and Slovakia covering phonology, morphology, lexicon and syntax of the Maltese language. This volume features nine articles representing various areas of study which reflect the growing international interest in Maltese not as a peripheral dialect of Arabic but as an independent Semitic language. The volume is divided into four thematic sections with wide scope: I Lexicon, II Syntax, III Morphology, and IV Phonology. Part I, which is mostly dedicated to the Maltese lexicon, opens with Sebastian Bednarowicz’s paper on The semantic structure of adjectives referring to the holiness in Maltese and Arabic languages. The author makes a comparison of Maltese and Arabic adjectives and demonstrates that the Maltese lexicon lacks a specialized Semitic term concerning God’s holiness. In his paper on The lexical ta’ construction, Ray Fabri categorizes various types of constructions involving the preposition ta’ ‘of’ and explores the expression tal-laħam ‘the butcher’ (literally ‘of the meat’) as a typical example of a lexical ta’ construction. It is thereby shown that the lexical ta’ construction behaves like a compound noun rather than a prepositional phrase. The paper by Christopher Lucas and Michael Spagnol titled Nunation from Arabic to Maltese discusses Maltese words containing an innovative final /n/, suggesting that a non-systematic phonological change – addition of word-final /n/ – can be triggered by the speakers’ tendency to make phonologically anomalous words conform more closely to prototypical Maltese phonological words. Part II deals with the syntax of Maltese: Kirsty Azzopardi’s contribution on The verb sequence V1+V2 in Maltese opens Part II of this volume, which presents two papers dealing with syntactic topics of Maltese linguistics. The author focuses on a construction in Maltese which consists of a sequence of two or more finite verbs in order to explain the acceptability or unacceptability of some

https://doi.org/10.1515/9783110783834-202

VI | Preface

combinations. The paper Differential object indexing in Maltese – a corpus based pilot study by Erika Just and Slavomír Čéplö presents – as the title already suggests – the first corpus based study of differential object indexing in Maltese. The results show that the strongest predictor for object indexing in Maltese is word order, but they also reveal that there is a strong correlation with the identifiability of the referent. Part III is dedicated to morphosyntactic matters. In his contribution on Arabic prepositions and their Maltese equivalents, Przemysław Turek analyzes Arabic prepositions in Classical/Modern Standard Arabic and Arabic dialects and contrasts them with their Maltese equivalents to show how many Classical Arabic and/or dialectal Arabic prepositions have survived in Maltese and how many have disappeared. The paper by Thomas Stolz and Maike Vorholt On WH-PREPS: A contrastive grammatical sketch of Maltese feijn and Spanish donde/ dónde analyzes the structural and functional similarities and differences the spatial interrogatives have in Maltese and Spanish. The authors discuss a property that is shared by these two languages, i.e. the possibility to use spatial interrogatives of Place (and Goal) in prepositional function, with the focus on selected issues of the synchronic morphosyntax of Maltese fejn and Spanish donde/dónde. In another contribution titled Can frequency predict length? A crosslinguistic investigation of Zipf’s law for European adposition, Maike Vorholt investigates the adpositions of sixteen European languages including Maltese and examines the relationship between length and frequency as stated by Zipf’s law of abbreviation. It is shown that length can be predicted by frequency for the majority of the discussed languages. The volume is closed with Part IV on phonology and Andrei A. Avram’s paper on Regressive voicing assimilation in Maltese: Diachrony and typology. The author reconstructs the diachrony of voicing assimilation in consonant clusters on the basis of the spelling used in extant records of earlier Maltese. We hope that this volume, like the previous publications of the International Association of Maltese Linguistics, provides insight into important current works being carried out on various aspects of the Maltese language and that it will serve to inspire and further strengthen research in Maltese linguistics. We are grateful to the Jagiellonian University in Kraków for hosting the 7th International Conference of Maltese Linguistics and for supporting the conference financially. Benjamin Saade kindly lent us a hand when it came to the practical side of organizing the conference. Our thanks also go to the De Gruyter publishing house for accepting this edited volume on its program of scholarly publications. Cornelia Stroh deserves a special word of thanks for the much needed editorial assistance she gave us. Kraków/Bremen, Przemysław Turek and Julia Nintemann, April 2021

Contents Preface | V

Part I:

Lexicon 

Sebastian Bednarowicz  The semantic structure of adjectives referring to the holiness in Maltese and Arabic languages | 3 Ray Fabri  The lexical ta’ construction | 23 Christopher Lucas and Michael Spagnol  Nunation from Arabic to Maltese | 47

Part II: Syntax  Kirsty Azzopardi  The verb sequence V1+V2 in Maltese | 83 Erika Just and Slavomír Čéplö Differential object indexing in Maltese – a corpus based pilot study | 105

Part III: Morphosyntax  Przemysław Turek Arabic prepositions and their Maltese equivalents | 135 Thomas Stolz and Maike Vorholt  On WH-PREPS. A contrastive grammatical sketch of Maltese fejn and Spanish donde/donde | 163 Maike Vorholt  Can frequency predict length? A crosslinguistic investigation of Zipf’s law for European adpositions | 197

VIII | Contents

Part IV: Phonology  Andrei A. Avram  Regressive voicing assimilation in Maltese: Diachrony and typology | 233 Index of Authors | 263 Index of Languages | 267 Index of Subjects | 269

| Part I: Lexicon

Sebastian Bednarowicz

The semantic structure of adjectives referring to the holiness in Maltese and Arabic languages Abstract: The present paper is an analysis of the Maltese and Arabic adjectives related to the notion of ‘holiness’. The research was conducted through the componential analysis of meaning used in terms of the contrastive description. The study showed: 1) the extension of meaning of the Maltese adjective mqaddes (in comparison to their etymological counterpart in Arabic muqaddas) what resulted in changes within the lexical field of holiness, which in Maltese reflects rather the opposition of the Italian adjectives santo and sacro. 2) the general lack of the lexico-semantic parallelism between Maltese and Arabic adjectives bearing the notion of ‘holiness’. Keywords: adjectives; Arabic; holiness; Maltese; semantics

1 Introduction The majority of previous studies on the Maltese lexicon were focused on etymological investigations (including the use of synonyms of Semitic or Romance origin) (Aquilina 1971, 1972, 1976; Trimble 1973; Krier 1976: 107–116), though the semantic (semasiological) method was also used in some works, especially those concerning the polysemy or semantic changes that took place in Maltese words compared with their Arabic or Italian counterparts (Krier 1976: 116–123; Aquilina 1997: 15–41). Building on the latter studies, in our modest contribution of the present paper we would like to analyze a number of adjectives referring to the notion of holiness in the Maltese and Arabic languages.

2 Relation between Maltese and Arabic A comparison of languages is only possible, if there are differences between them. For that reason the influence of phylogenetic relations between linguistic

|| Sebastian Bednarowicz: Kazimierz Wielki University in Bydgoszcz, ul. Grabowa 2, 85-601 Bydgoszcz, Poland, E-Mail: [email protected]

https://doi.org/10.1515/9783110783834-001

4 | Sebastian Bednarowicz

varieties being compared on the process of the contrastive analysis is indisputable. It is clear that non-related languages share less similarities than the cognate ones do. Maltese and Arabic represent undoubtedly the latter situation. The origin of literary Arabic is still a subject of discussion among Semiticists. The theory of classical Arab grammarians, who claimed that the spoken dialects are merely a corrupted form of the literary or Quranic Arabic, was rejected by the majority of Arabists.1 A more plausible hypothesis is that the literary variety of Arabic emerged from vernacular dialects, which were a normal way of communicating among the tribes living in the Arabian Peninsula. This process began after the Islamic empire grew strong, and was, on the one hand, an answer for the need of an official language, which might be used in the state administration. On the other hand, the standardized Arabic was to be a language of the new religion, Islam, with its main sacred text, the Quran (Danecki 2009: 11–23). According to this hypothesis, Maltese should not be treated as a descendant of written Arabic in terms of direct genealogical relationship. Among linguists there is agreement that Maltese, as a Semitic language, emerged from Arabic dialects spoken in Sicily and North Africa after the Arab conquest of these regions. According to Brincat, who quotes the Arab author alḤimyarī, the people of Byzantine Malta were massacred or sold into slavery in 870, and the island remained deserted until 1048–1049, when the Arab colonists from Muslim Sicily and North Africa repopulated the archipelago. The power of Islam, however, had to capitulate soon in the face of Normans, who conquered the island in 1091. Henceforth the history of Malta and the Maltese language were closely attached to Latin Christianity (Brincat 2011: 33–47). The dialectal classification assigns Maltese to the Maghrebian group of contemporary spoken Arabic. In the past this group was represented as well by Siculo-Arabic and Andalusian Arabic, which are known from some written sources (e.g. poetry), though there is a substantial difference between them and the literary variety of Arabic. A relatively short period of the Arab and Islamic domination in Malta and the subsequent influx of Romance speaking settlers caused the Maltese language to differ substantially from other Arabic varieties. Such peculiarities are observable first of all in the vocabulary. The analysis of the etymology of the 41,016 words in Aquilina’s Maltese-English dictionary indicates that 32.41 % are of Semitic origin, 52.46 % are from Sicilian or Italian, and 6.12 % are from English (Brincat 2011: 407). || 1 One of the supporters of the view that Arabic dialects emerged from the standardized language is Versteegh. According to him the dialects came into being as a result of the pidginization and creolization of classical Arabic (Versteegh 1984).

The semantic structure of adjectives referring to the holiness | 5

3 Searching for the plane of comparison One of the most important conditions for the effective comparison of two languages is to find an appropriate point of reference (tertium comparationis) which may indicate both differences and similarities observable in the linguistic items being compared. In other words, tertium comparationis serves as a basis of the comparative matrix and is intended to abstract a linguistic category or class present in these languages, wherein it is assumed that the plane of linguistic comparison should be semantic in its nature (Krzeszowski 1990: 15–21). Considering these premises, in this study we accept the Christian conception of the holiness as a departure point for comparing the semantic content of adjectives expressing such a meaning in Maltese and Arabic languages. Such delimitation seems reasonable, because the Islamic perception of the sacredness does not agree in basic arguments (e.g. God’s inaccessibility postulated by Muslims) with the meaning of holiness, as it is accepted by Christian communities. The method used in our investigations is componential analysis2 consisting in determining a set of distinctive markers, which were defined by Katz and Fodor (1964: 518) as having the function of giving each well-formed string the conceptual content that permits them to be represented in terms of the message they communicate to speakers in normal situations. They are concerned with different kinds of selection and they express different aspects of a language.

The markers will be used in differentiating lexemes according to the degree of their synonymy as well as the semantic collocability. The adjectives compared in that way are categorized with respect to particular features of nouns that they modify.

4 Adjective in Arabic and Maltese According to Baker (2004: 191) adjectives, in their linguistic functions, may be reduced to: 1) direct attributive modifiers; 2) complements of degree heads; 3) resultative secondary predicates. || 2 The componential analysis as a research method emerged from the lexical (semantic) field theory and postulates that: 1) a field must belong to the same sub-system or variety of the language, 2) the members of the field must belong to the same word-class or syntactic context, 3) the words must share at least one semantic component with one another (Hartmann 1973: 4–5).

6 | Sebastian Bednarowicz

This presupposition, however, does not mean that the grammatical category of adjective is universal for all languages of the world. In fact, there are languages which normally use nouns or verbs to render the main semantic function of adjectives, i.e. to qualify nouns. Such a phenomenon is observable for instance in many African languages, which actually have a distinguishable class of adjectives but which is represented by very few of its members in the vocabulary (Baker 2004: 248). Consequently, when approaching the analysis of adjectives, the unique status of this part of speech in a particular language must always be taken into consideration. Arabic and Maltese are Semitic languages, for which Givón (1970: 836–837) stated that “the category ADJECTIVE is indistinguishable from the stative or participial form of the verb and the morphology furnishes clear derivational channels for obtaining all adjectives from (or tracing to) noun stems or verb stems.” This view was shared by Gai (1995: 1–9), who applied the syntactic criteria for delimiting nouns and adjectives in the Semitic languages. On the other hand, Edzard (2001: 43–46 and 48), after perusal of nominalized verb phrases and relative clauses in modern Ethiopic languages, came to the conclusion that they fulfill de facto the role of adjectives and the whole category of “adjective” in Semitic is rather unstable.

4.1 Adjectives in Arabic The classical Arab grammarians divided parts of speech into three sections: ʾism (noun), fiˁl (verb), ḥarf (particle). The adjectives in this division were included in the category of ism (nouns), mainly because of their declensional properties. The most common Arabic term used by Arab grammarians to denote the adjective is ṣifa (lit. ‘description’), which, however, appears as well in another context, namely to call a relative clause with an indefinite antecedent. Thus, the ṣifa refers rather to syntactic and functional properties of linguistic units than to its morphology (Versteegh 1977: 49–50; Owens 1990: 66; Guillaume 1992: 60–65). The priority of syntactic features was also the basis for the definition of Arabic adjectives elaborated recently by Michalski (2009). According to him Arabic adjectives have to meet following conditions (Michalski 2009: 42–43): 1) Capability of linearly following another noun in order to qualify it and agree with it with respect to state (definite/indefinite), number, gender and case; 2) Incapability of linearly preceding another noun, while: (i) retaining agreement (concord) with it with respect to number, gender, case and state and (ii) retaining the same meaning as when following the noun.

The semantic structure of adjectives referring to the holiness | 7

4.2 Adjectives in Maltese In the oldest descriptions of the Maltese language, the category of adjective (aġġetiv) is presented according to the classical division of the parts of speech inherited from the Greek and Roman grammatical tradition. De Soldanis (1750: 80) includes the adjective in the category of nouns and specifies its three degrees. Similarly, Vassali (1791: 108) considers the adjectives to be nouns together with nomina propria and appellativa. On the other hand, Vella (1831: 19), following English grammarians, lists the adjective as one of the nine separate parts of speech. The same division appears in Panzavecchia’s (1845) grammar, where the adjectives are subdivided additionally into primitives and derivatives. The latter group, in turn, embraces the qualificative (e.g. għaref ‘wise’), national (e.g. għarbi ‘Arab’) and other proper (e.g. sajfi ‘summery’) adjectives, which are collectively called aggetivi di pertinenza. As far as the deadjectival derivation is concerned, Panzavecchia (1845: 18–24) mentions diminutives, terms of endearment, augmentatives and derogatives. Sutcliffe (1936: 56–65), Barbera (1939: 70–77) and Aquilina (1965: 43–56, 62–68) focus mainly on the word formation of adjectives and the presentation of various adjectival patterns. All of them have the tendency to consider adjectives and nouns as one nominal group. In more recent works, one can see attempts to include lexical semantics to the description of Maltese adjectives. A rough semantic classification of adjectives may be found in a grammar by Scicluna et al. (2018: 264). More detailed are Fabri’s (2007: 220–229) investigations concerning the adjective + noun compounds, which shed light on the lexical collocations of Maltese adjectives too.3

5 Corpus of texts The adjectives in the present study were selected from Arabic texts written in the 17th and 18th centuries and those composed in Maltese in the 18th and 19th centuries. Owing to the religious character of the textual basis, the occurrence of the adjective ‘holy’ in the works is relatively frequent. Moreover, all these texts express the Christian (Catholic) way of the understanding of the subject of holiness, what will allow us to avoid the interreligious misunderstandings of terms and to focus only on the lexical use of the adjectives in a given context.

|| 3 The lists of adjectival collocations were presented also in Aquilina (1965: 66–67) and Scicluna et al. (2018: 275–277).

8 | Sebastian Bednarowicz

1.

2.

3.

The Arabic adjectives were excerpted from the following texts: Tafsīr wāsiˁ ˁalā al-taˁlīm al-masīḥī (334 pp.). The book is an Arabic translation of the very popular Italian catechism written by Cardinal Roberto Bellarmin (Dichiarazione più copiosa della dottrina cristiana). The translator, John Hesronita, was a member of the Maronite clergy and published his work in 1627. The book was reprinted several times. In 1770 a new Arabic translation, or, more probably, a revision of the existing one, appeared (411 pp.). Its author is unknown and it differs from the former versions. Kitāb al-iqtidāˀ bi-l-Masīḥ (561 pp.) The Arabic edition of De Imitatione Christi by Thomas á Kempis was published in 1663. It was translated by a Carmelite friar Celestino a S. Liduina, who was the brother of Jacobus Golius, a famous 17th-century orientalist (Loop 2013: 114). As a popular manual of the Christian life the book was republished several times. Kitāb al-kātīkīzmū rūmānū. This is the Arabic edition of the fundamental for the Catholic Church Tridentine catechism (Catechismus Romanus). The translation was published in 1786 (first volume, 976 pp.) and 1787 (second volume, 855 pp.). Both parts were translated from Latin by Maronite priests Jacob Arutin and Dionysius Haǧǧar.

The language of these texts may be classified as classical written Arabic with some Middle Arabic features (Khan 2011) concerning phonology and orthography (e.g. the omission of hamza at the end of words) or morphology (very rare and erratic, e.g. the inflectional suffix of 3M.PL -ū for the Imperfect tense in the place of the common -ūna). In turn, the Maltese sources for the present study were: 1. Il-Priedki bil-Malti ta’ Ignazio Saverio Mifsud (750 pp.). The collection of Maltese sermons preached by Ignazio Saverio Mifsud (1722–1773) between 1739 and 1746. In 1752 the homilies were gathered in one volume, which, however, was not published until 2008. The text is considered to be the oldest testimony of Maltese prose. 2. Priedki bil-Malti ta’ Franġisku Saverju Baldacchino (11 pp.). Two sermons on the saints delivered by Franġisku Saverju Baldacchino (1774–1860). 3. Ctieb is-salmi tas-sultan David u il cantici (377 pp. of Latin-Maltese text). The edition of the Latin psalms as well as some Church hymns, which were coupled with their Maltese translation. The translation was made by Richard Taylor and published in 1846. 4. Offizju tal Gimgha il-Cbira (380 pp.). The Holy Week Latin liturgy translated into Maltese by Richard Taylor in 1848.

The semantic structure of adjectives referring to the holiness | 9

5.

Taghlim ghal min li irid isalva ruhu (382 pp.). The Maltese translation of the famous book by St. Alfonso Maria Liguori (Massime eterne). The translator is unknown, however, the book was published with the approval of the ecclesiastical authorities in 1868. 6. Il-Haija tal-Apostlu Missierna San Paul (94 pp.). A booklet containing the extended biography of St. Paul. It was published for the first time in 1858 and then reprinted. The Maltese language used in these works differs stylistically according to the literary genre and the period they were composed in. The differences are observable also with regard to the orthography (e.g. the adjective imqadssa may be spelled: imqazza, imkazza, imkadsa or imcadsa). The common feature of the sermons is a large share of Romance vocabulary or even whole Italian phrases. Besides, all these texts were written before the puristic campaign initiated by the Semitic Society (ix-Xirka Xemija) in the 80’s of the nineteen century, which resulted in the emergence of standardized Maltese. To avoid misunderstandings and longer explanations, in the following analysis we decided to apply the modern Maltese orthography to spell the examples excerpted from the prestandardized sources. In the case of the Arabic language, we use the pausal forms and the ISO system of transcription.

6 Adjectives derived from the root QDS The most common Semitic root that appears in words connected to ‘holiness’ is QDŠ. Originally, it carried the meaning of ‘purity’ as it is attested already in the Akkadian language (Black et al. 2000: 282). Later this meaning was extended to a very special state or quality of persons, places, objects and, particularly, divine beings. Such a sense of holiness may be found in the Hebrew Bible (Lachowski 2003: 2): Holiness is the English word for qōdeš, derived from the Hebrew root qdš, common to all Semitic languages and having essentially the same meaning. The concept of holiness is not established etymologically from the root; it comes from the sense in which its derivatives are used. Hence it signifies ‘‘separateness’’ from the nonholy or profane. What is ‘‘clean’’ or ‘‘pure’’ is also related to ‘‘holy’’ in a ritual sense, i.e., free from defilement by the profane and in a potentially holy state. […] In the cultic sense, it is a quality of an object that is withdrawn from the profane and consecrated to God; in the moral sense, it can be ascribed, to God, to angels, or to men.

10 | Sebastian Bednarowicz

In both Arabic and Maltese, the root QDS is attested in adjectives that are connected semantically with the notion of holiness. These adjectives, however, though paralleled derivationally and etymologically, are not attributed equally to the same semantic classes of nouns. The following analysis will be based on the semantic markers, which were expressed through the connectivity of adjectives with semantic classes of nouns. These classes are characterized by the following distinctive features: personal or impersonal4; divine or non-divine, human or non-human. The adjectives are analyzed also with regard to their intensive or non-intensive meaning. As the reader will notice, in our analysis we have omitted the Arabic adjective qudsī, which, though derived from the root QDS, is used exclusively by Muslims (e.g. ḥādiṯ qudsī ‘holy report’), and therefore is irrelevant for our study.

6.1 Arabic adjectives derived from the root QDS 6.1.1 qiddīs The adjective qiddīs is formed after the adjectival pattern C1iC2C2īC3 which is rather rare in Arabic (cf. fiḫḫīr ‘boasting’). It is usually attributed to a human or supernatural persons, but not to God: (1)

ar-rusul

al-qiddīs-ūn

DEF-apostles

DEF-holy-M.PL.NOM

[Catechismus 1786: 130]

‘the holy apostles’ (2)

baˁḍ ˀunās ṣāliḥ-īn some people good-M.PL.GEN ‘some good and holy people’

(3)

biˀanna al-malāˀika that DEF-angels ‘that the holy angels’

qiddīs-īn [Bellarmino 1671: 78] holy-M.PL.GEN

al-qiddīs-īn DEF-holy-M.PL.ACC

[Catechismus 1786: 70]

Exceptions to this rule are rare, and result from the hesitation, whether a given noun has personal or impersonal meaning, e.g.:

|| 4 In the Arabic language the category of ‘animate’ and ‘inanimate’ nouns is not consistent grammatically. For instance in the plural all animate but impersonal nouns (e.g. animals) are treated as singular feminine nouns. That is why we decided to use the markers ‘personal’ and ‘impersonal’ instead of the more common ‘animate’ and ‘inanimate’.

The semantic structure of adjectives referring to the holiness | 11

(4)

al-ˀanfus

al-qiddīs-a

DEF-souls

DEF-holy-F.SG

[Bellarmino 1671: 177]

‘the holy souls’ As we can see, the adjective attributed to the plural noun is put in the feminine singular form, which suggests that the qualified noun is impersonal grammatically. On the other hand, the ‘soul’, according to the Christian theology, is an immanent part of the human person. Probably, that is why the translator decided to choose the adjective qiddīs in order to emphasize this aspect of religious teaching.

6.1.2 quddūs The lexeme quddūs belongs to an archaic group of adjectives derived from the pattern C1aC2C2ūC3, which emphasizes a degree of a particular feature and yields an intensive meaning. Hence, the aforementioned adjective is often translated as ‘the most holy’ or ‘the holiest’. Nonetheless, quddūs may also be used in the simple meaning of holiness, but only when referring to God himself or His attributes. (5)

(6)

ˀayyuhā al-ṯālūṯ al-quddūs DEF-trinity DEF-very.holy o ‘o the [most] holy Trinity’

[Á Kempis 1738: 133]

al-ˀab

al-quddūs

[Á Kempis 1738: 160]

DEF-father

DEF-very.holy

‘the holiest Father’ [i.e. God] (7)

ˀism-u hu al-quddūs -his DEF-very.holy name-NOM ‘His [i.e. God’s] holiest name’

[Á Kempis 1738: 80]

6.1.3 muqaddas As far as the derivation is concerned, the adjective muqaddas is a passive participle of the verb qaddasa ‘to sanctify’. It is used in contexts other than those characteristic for the adjectives qiddīs and quddūs, i.e. as an attribute of nouns referring to animals, plants, places, items, activities and abstracta. (8)

al-ṣalāt

al-muqaddas-a

DEF-prayer

DEF-holy-F.SG

‘the holy prayer’

[Catechismus 1787: 388]

12 | Sebastian Bednarowicz

(9)

wa-bi-kanīsa [...] muqaddas-a and-in-church.F.SG holy-F.SG.(INDF) ‘and in holy […] Church’

(10)

li-l-ˀanfus al-muqaddas-a for-DEF-souls DEF-holy-F.SG ‘for the holy souls’

[Catechismus 1787: 713]

[Bellarmino 1770: 210]

Example (10) is a translation of the same phrase that we have already discussed above. In (4), there is an unusual collocation of the adjective qiddīs. Here, in turn, the plural noun ‘souls’ was classified as the impersonal one. Although the adjective muqaddas is not applied to humans, one can find some isolated exceptions to this general rule, which happens predominantly when the adjective is used in the predicative function as object complement. This use seems to be connected to the primitive function of this adjective, i.e. the passive participle. That is why, in the following examples the lexeme muqaddas may be translated as ‘consecrated’ or ‘sanctified’: (11)

wa-lā yu-dˁà ˀaḥad min-hum and-not PRS.3M.SG-be.called one from-them muqaddas ġayra al-bābā [Bellarmino 1671: 58] holy except DEF-pope ‘and let no one of them be called sanctified except the Pope’

(12)

wa-kān-ū ya-ta-karras-ūn li-l-lāh and-be-PRF-3M.PL PRS.3M-REFL-devote-PL to-DEF-God taˁālà wa-yu-dˁ-ūn muqaddas-īn Exalted.One and-PRS.3M-be.called-PL holy-M.PL.ACC [Catechismus 1786: 263] ‘and they were devoting themselves to the Most High God and were called consecrated’

The extraordinary use of this lexeme in the collocation with the personal nouns is proved by the fact, that in the first volume of the Catechismus Romanus it is attested in such contexts in two instances only (of 325 appearances in this book). Another indication of the peculiarity of this collocation may also be the fact that in the new Arabic version of Bellarmino’s catechism published in 1770 the first example was already translated using the adjective qiddīs: (13)

lā yu-qāl al-ˀab al-qiddīs not PRS.3M.SG-be.said DEF-father DEF-holy ˀillā li-l-bābā [Bellarmino 1770: 78] except toward-DEF-pope ‘one does not say the Holy Father unless referring to the Pope’

The semantic structure of adjectives referring to the holiness | 13

The use of the adjective muqaddas in the context of God’s attributes (but never God himself) is also rare. It appears occasionally with the noun ‘trinity’: (14)

al-ṯālūṯ al-muqaddas DEF-trinity DEF-holy ‘the holy Trinity’

[Bellarmino 1671:19]

6.1.4 ˀaqdas The elative is one of the Arabic adjectival forms. It is usually applied to express the meaning of degree or, additionally, the intensiveness of a feature. The comparative requires the preposition min ‘from’ placed after the indefinite form of the elative and before the object, that the subject is compared with. On the other hand, the superlative is indicated by the definite article put before the elative form of the adjective. The elative derived from the root QDS has the form ˀaqdas, which may be used without regard to the semantic class of the noun it is attributed to. (15)

(16)

ˁīd al-ṯālūṯ DEF-trinity feast(ST.CONST) ‘feast of the most holy Trinity’

al-ˀaqdas

[Catechismus 1787: 805]

DEF-most.holy

al-qurbān

al-ˀaqdas DEF-most.holy ‘the most holy sacrifice’

[Á Kempis 1738: 98]

DEF-sacrifice

6.2 Maltese adjectives derived from the root QDS 6.2.1 qaddis The adjective qaddis is an etymological counterpart of the Arabic adjective qiddīs. Nevertheless, the Maltese qaddis has much broader use than its Arabic equivalent, namely it is not constrained by the personal character of nouns. (17)

id-dar

tiegħek wisq your very ‘your house is very holy’

DEF-house(F)

(18)

raġel qaddis man holy ‘a holy man’

qaddis-a holy-F

[Taylor 1846: 139]

[Vassallo 1870: 76]

14 | Sebastian Bednarowicz

(19)

għax Mulej-na Alla qaddis because Lord-our God holy ‘because our Lord, God, is holy.’

(20)

ħajja qaddis-a life holy-F ‘a holy life’

(21)

kemm ir-rwieħ qaddis-in how.many DEF-souls holy-PL ‘how many your holy souls’

6.2.2

[Taylor 1846: 227]

[De Liguori 1868: 274]

tiegħek your

[Taylor 1846: 329]

imqaddes

The Maltese adjective imqaddes corresponds derivationally to the Arabic passive participle muqaddas. Nonetheless, similarly to the lexeme qaddis, this adjective may be applied in various semantic contexts: (22)

il-knisja

mqadds-a holy-F ‘the holy church’

[Zammit Ciantar 2008: 289]

DEF-church

(23)

Ejja, o mqaddes come(IMP) o holy ‘Come, o Holy Spirit.’

Spiritu Spirit

(24)

fi-l-lok mqaddes in-DEF-place holy ‘in the holy place.’

(25)

għal l-poplu koll-u mqaddes for DEF-people all-his holy ‘for all the holy people of God’

[Zammit Ciantar 2008: 111]

[Taylor 1848: 187]

t’Alla of.God

[Taylor 1848: 258]

6.2.3 eqdes The lexeme eqdes is a derivational form of the adjective qaddis, which corresponds to the Arabic elative, so it may express both the comparative and superlative degrees. It appears very rarely and is usually substituted by the periphrastic construction aktar/iżjed qaddis (lit. ‘more holy’), which was presented as a

The semantic structure of adjectives referring to the holiness | 15

primary comparative degree of this adjective even in the oldest Maltese printed grammar (de Soldanis 1750: 80). It is worth mentioning here, that the form eqdes, though considered to be archaic, is rather uncommon in older texts. One example is delivered by Panzavecchia in his grammar, where, however, this elative has the form iqdes (Panzavecchia 1845: 22): (26)

l-’iqdes

fost in-nisa among DEF-women ‘the holiest among the women’

DEF-most.holy

6.2.4 qdejdes Another infrequently attested adjectival form derived from the root QDS is a diminutive qdejdes. Nevertheless, this lexeme may be found already in 18thcentury sermons of Saverio Mifsud. It differs semantically from other adjectives founded on the aforementioned root through the aspect of tenderness contained in the diminutive formation, but expresses as well a kind of esteem. (27)

iben [...] qdejdes son [...] holy(DIM) ‘a saintly son’

[Zammit Ciantar 2008: 25]

7 Maltese adjectives of the Romance origin As opposed to Arabic, in Maltese, apart from the Semitic adjectives, there is a group of loanwords referring exactly to the notion of holiness. They take Italian/Sicilian feminine or plural suffixes and are often placed before the noun.

7.1 santu It appears usually in the religious term l-Ispiritu Santu ‘the Holy Ghost’. Other uses are rather rare, and concern mainly the attributes of nouns borrowed from Italian, e.g.: (28)

l-luoghi

sant-i ta Gerusalem holy-PL of Jerusalem ‘the holy places of Jerusalem’ DEF-places

[De Liguori 1868: 194]

16 | Sebastian Bednarowicz

(29)

u mis-sant-i and from DEF-holy-PL ‘and from the holy fathers’

Padri fathers

[Friggieri 2006: 37]

7.2 santissimu This adjective is a superlative form of santu. (30)

ir-riġlejn santissim-i DEF-legs most.holy-PL ‘the most holy legs’

[De Liguori 1868: 149]

(31)

o Sidt-i Мaria Santissim-a o lady-my Mary most.holy-F ‘o my Lady, Mary the most holy’

[De Liguori 1868: 368]

7.3 sagru Another Maltese adjective of Romance provenance is sagru. It appears usually as an attribute of the impersonal nouns, which are often Italian or Sicilian loanwords: (32)

id-diskorsi [...]

kienu be:3PL:PRF ‘the talks [...] were sacred’

sagr-i sacred-PL

DEF-conversations

(33)

ta s-sagr-i kantiċi of DEF-sacred-PL canticles ‘of the sacred canticles’

(34)

naqra fi s-Sagr-a read.1SG.PRS in DEF-sacred-F ‘I read in the holy Scripture’

(35)

j-għid

is-sagr-u

PRS.3M.SG-say

DEF-sacred-M

[Friggieri 2006: 35]

[Friggieri 2006: 35]

Skrittura scripture

test text

[Friggieri 2006:36]

[Zammit Ciantar 2008: 251]

‘the sacred text says’ The use of the adjective sagru with personal nouns is uncommon: (36)

iben ġwejjed, sagr-u u qdejdes son good(DIM) sacred-M and holy(DIM) ‘a peaceful, divine and saintly son’

[Zammit Ciantar 2008: 25]

The semantic structure of adjectives referring to the holiness | 17

7.4 sagrosantu Of purely Romance origin is the superlative compound formation sagrosantu ‘the most holy’. It collocates with impersonal nouns: (37)

(38)

il-knisja sagrosant-a DEF-church most.holy-F ‘the most holy church’

[Taylor 1848: 309–310]

ir-ras

sagrosant-a most.holy-F ‘the most holy head’

[De Liguori 1868: 114]

DEF-head

8 Toward a conclusion The analysis of the adjectives referring to the notion of holiness in the Arabic and Maltese languages allows us to make some suggestions concerning the semantic structure of this notion and its representation in the lexical system. The differences may be indicated by semantic markers: +/− human, +/− divine and +/− personal. Moreover, the adjectives are compared in terms of their intensive meaning. The Arabic adjectives may be distributed as shown in Table 1. Table 1: Distribution of Arabic adjectives.

qiddīs

muqaddas

quddūs

ˀaqdas

human

+(−)5

−(+)

−

+/−6

divine

−

−(+)

+

+/−

personal

+

−(+)

+/−

+/−

intensive

−

−

+

+

As we can see the elative form ˀaqdas is the most universal because it covers all contexts of use of the analyzed adjectives. As far as the other lexemes are con-

|| 5 The brackets () serve to point out the rare use of an adjective in a given context. 6 The symbol +/− means that an adjective may be used with nouns characterized by either the presence or the absence of a given semantic marker.

18 | Sebastian Bednarowicz

cerned, they stand in rather clear semantic opposition to each other (with the exception of some very rare uses of them). The intensiveness of meaning is a common feature of the adjectives quddūs and ˀaqdas. By way of contrast, the semantic structure of the Semitic Maltese adjectives looks as shown in Table 2. Table 2: Distribution of Maltese Semitic adjectives.

qaddis

mqaddes

qdejdes

eqdes

human

+/−

+/−

(+)

(+/−)

divine

+/−

+/−

(+)

(+/−)

personal

+/−

+/−

(+)

(+/−)

intensive

−

−

−

(+)

Apart of the rare in use lexemes qdejdes and eqdes, we can see that the adjectives qaddis and mqaddes are interchangeable lexically. In addition, unlike the Arabic lexicon, in Maltese there is no Semitic counterpart for the adjective referring only to God (quddūs), while in Arabic it lacks the diminutive form, which, in turn, is attested in Maltese (qdejdes). The intensiveness of meaning seems to be characteristic to the adjective eqdes only. As far as the Maltese-Romance adjectives referring to the holiness are concerned, they may be presented as in Table 3. Table 3: Distribution of Maltese Romance adjectives.

santu

sagro

santissimu

sagrosantu

human

+/−

−(+)

+/−

−

divine

+/−

−

+/−

−

personal

+/−

−(+)

+/−

−

intensive

−

−

+

+

One can see here a parallelism between the pair santu – sagro, on the one hand, and the intensive formations santissimu – sagrosantu, on the other. Santu and santissimu may be used as attributes of all semantic classes of nouns. By contrast, the other two adjectives collocate with the impersonal nouns (sagrosantu always and sagru usually). This distinction is observable as well in Italian,

The semantic structure of adjectives referring to the holiness | 19

where the adjective santo may be used universally (Terra santa – ‘the Holy Land; parole sante – ‘holy (i.e. absolutely true) words’; Dio santo! - ‘holy God’ (good God!); un uomo santo – ‘a holy man’), while sacro is restricted to impersonal nouns (luoghi sacri – ‘holy places’; Sacra scrittura – ‘holy writing’ i.e. the Holy Bible). The comparison of the Maltese adjectives with the Arabic ones, demonstrates that the Maltese lexicon lacks a specialized Semitic term concerning God’s holiness. One can also observe the extension of meaning of the Maltese adjectives mqaddes and qaddis in comparison to their derivational/etymological Arabic counterparts (muqaddas and qiddīs), which are characterized by a clear specification of use. It also proves the general lack of semantic parallelism between Maltese and Arabic adjectives bearing the notion of ‘holiness’ and derived from the common Semitic root QDS.

Abbreviations 3 ACC DEF DIM F GEN IMP INDF M NOM PL PRF PRS REFL SG ST.CONST

3rd person accusative definite diminutive feminine genitive imperative indefinite masculine nominative plural perfect tense present tense reflexive singular status constructus

References Á Kempis, Thomas. 1738. Kitāb al-iqtidāˀ bi-l-Masīḥ. Halae: Typographia Instituti Judaici. Aquilina, Joseph. 1965. Teach yourself Maltese. London: The English Universities Press. Aquilina, Joseph. 1971. Due epoche linguistiche nella lingua Maltese. Journal of Maltese Studies 7. 1–36. Aquilina, Joseph. 1972. Maltese etymological glossary. Journal of Maltese Studies 8. 1–62.

20 | Sebastian Bednarowicz

Aquilina, Joseph. 1976. Maltese linguistic surveys. Malta: University of Malta. Aquilina, Joseph, 1987–1990. Maltese-English dictionary. Malta: Midsea Books. Aquilina, Joseph. 1997. Papers in Maltese linguistics. Malta: University of Malta. Baker, Mark S. 2004. Lexical categories. Verbs, nouns, and adjectives. Cambridge: Cambridge University Press. Barbera, Giuseppe. 1939. Dizionario maltese-arabo-italiano: con una grammatica comparata arabo-maltese. Vol. 1. Beyrouth: Imprimerie Catholique. Bellarmino, Roberto. 1671. Tafsīr wāsiˁ ˁalā al-taˁlīm al-masīḥī. Roma: Sac. Congregatione de Prop. Fide. Bellarmino, Roberto. 1770. Tafsīr wāsiˁ ˁalā al-taˁlīm al-masīḥī. Roma: Sac. Congregatione de Prop. Fide. Black, Jeremy, Andrew George & Nicholas Postgate. 2000. A concise dictionary of Akkadian. Wiesbaden: Harrassowitz. Brincat, Joseph M. 2011. Maltese and other languages. A linguistic history of Malta. Sta. Venera: Midsea Books. Catechismus Romanus. 1786. Kitāb al-kātīkīzmū rūmānū. Tomus Primus. Roma: Sac. Congregatione de Prop. Fide. Catechismus Romanus. 1787. Kitāb al-kātīkīzmū rūmānū. Tomus Secundus. Roma: Sac. Congregatione de Prop. Fide. Danecki, Janusz. 2009. Współczesny język arabski i jego dialekty. Warszawa: Dialog. De Liguori, Alfonso Maria. 1868. Taghlim ghal min irid isalva ruhu. Malta: Bonello. Edzard, Lutz. 2001. Adjektive und nominalisierte Relativsätze im Semitischen: Versuch einer Typologie. In Andrzej Zaborski (ed.), New data and new methods in Afroasiatic linguistics: Robert Hetzron in memoriam, 39–51. Wiesbaden: Harrassowitz. Fabri, Ray. 2007. Compounding and adjective-noun compounds in Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 207–231. Amsterdam & Philadelphia: John Benjamins. Friggieri, Olivier (ed.). 2006. Aktar Priedki bil-Malti ta’ Franġisku Saverji Baldacchino (1774– 1860). Symposia Melitensia 3. 21–38. Gai, Amikam. 1995. The category ‘adjective’ in Semitic languages. Journal of Semitic Studies 40(1). 1–9. Givón, Talmy. 1970. Notes on the semantic structure of English adjectives. Language 46. 816– 837. Guillaume, Jean-Patrick. 1992. Le statut de l’adjectif dans la tradition grammaticale arabe. Histoire Épistémologie Langage 14(1). 59–74. Hartmann, Reinhard R. 1973. Contrastive lexicology. Paper presented at the Annual Meeting of BAAL (September 1973). https://files.eric.ed.gov/fulltext/ED102813.pdf Katz, Jerrold &. Jerry A. Fodor. 1964. The structure of language: Readings in the philosophy of language. Englewood Cliffs: Prentice Hall. Khan, Geoffrey. 2011. Middle Arabic. In Stefan Weninger (eds.), The Semitic languages. An international handbook, 817–835. Berlin & Boston: De Gruyter Mouton. Krier, Fernande. 1976. Le maltais au contact de l’italien: étude phonologique, grammaticale et sémantique. Hamburg: Helmut Buske. Krzeszowski, Tomasz P. 1990. Contrasting languages. The scope of contrastive linguistics. Berlin & New York: De Gruyter. Lachowski, Joseph. 2003. Holiness (in the Bible). In Berard L. Marthaler (ed.), New Catholic encyclopedia, 7 Hol-Jub. Detroit: Thompson Gale.

The semantic structure of adjectives referring to the holiness | 21

Loop, Jan. 2013. Johann Heinrich Hottinger: Arabic and Islamic studies in the seventeenth century. Oxford: Oxford University Press. Michalski, Marcin. 2009. Adjectives in hypotaxis: Proposed dimensions for a description of syntagms in Modern Written Arabic. Lingua Posnanensis 51. 39–56. Owens, Jonathan. 1990. Early Arabic grammatical theory. Amsterdam & Philadelphia: John Benjamins. Panzavecchia, Fortunato. 1845. Grammatica della lingua maltese spiegata secondo i principij delle lingue orientali e della lingua italiana. Malta: Tipografia di M. Weiss. Scicluna, Karl, Saviour S. Agius & Chris Giordano. 2018. Inħaddem il-Malti tajjeb. Manwal prattiku tal-grammatika. Ħal-Tarxien: Il-Gutenberg Press. Soldanis, Giovanni Pietro Francesco Agius de. 1750. Della lingua Punica presentamente usata da Maltesi. Roma: Per Generoso Salomoni alla Piazza di S. Ignazio. Sutcliffe, Edmund. 1936. A grammar of the Maltese language with chrestomathy and vocabulary. London: Oxford University Press & Humphrey Milford. Taylor, Richard (trans.). 1846. Ctieb is-salmi tas-sultan David u il cantici. Malta: Tipografia Anlo-Maltese. Taylor, Richard. 1848. Offiziu tal Gimgha il Cbira. Latin u Malti. Malta: Tipografia Anlo-Maltese. Trimble, Louis P. 1973. Some linguistic comments on religious terms in Maltese. Journal of Maltese Studies 9. 59–67. Vassalli, Michael Antonio. 1791. Mylsen Phoenico-Punicum sive Grammatica Melitensis. Roma: Antonio Fulgoni. Vassallo, Ġan Anton. 1870. Il ħaija tal Apostlu Missierna San Paul. Malta: Strada Sant’Ursula No. 187c. Vella, Francis. 1831. Maltese grammar for the use of the English. Leghorn: Glaucus Masi. Versteegh, Kees. 1977. Greek elements in Arabic linguistic thinking. Leiden: Brill. Versteegh, Kees. 1984. Pidginization and Creolization. The case of Arabic. Amsterdam & Philadelphia: John Benjamins. Zammit Ciantar, Joe (ed.). 2008. Il-Priedki bil-Malti ta’ Ignazio Saverio Mifsud: edizzjoni kkummentata bi studju kritiku. Sta. Lucija: l-awtur.

Ray Fabri

The lexical ta’ construction Abstract: Fabri (2009) tentatively identifies a number of potential candidates for compound constructions in Maltese. One such candidate, mentioned but not analyzed in Fabri (2009), is the ta’ compound with examples such as tal-laħam ‘the butcher’ and tal-ħobż ‘the bread baker/seller’ (literally ‘of the meat’, ‘of the bread’), which involve the combination of the preposition ta’ ‘of’ with a definite noun. This paper attempts to characterize this construction in terms of its morphosyntactic and semantic (referential) properties. Various types of ta’ constructions are listed and categorized, and the expression tal-laħam ‘the butcher’ is taken as a typical example of the lexical ta’ construction, and explored in great detail. This analysis should serve as a basis for the analysis of other such ta’ expressions. Keywords: compounds; Maltese; morphosyntax; multi-word expressions

1 Introduction The lexicalized ta’ construction is very common in both spoken and written Maltese. The following is an example. (1)

Illum ma ġie-x ta-l-ħobż come.3SG.M-NEG of-DEF-bread today NEG ‘The breadman didn’t turn up today.’

This paper explores this construction in an attempt at a characterization in terms of its morphosyntactic and semantic (referential) properties. The various types of ta’ constructions are listed and categorized, and, in particular, the lexical ta’ construction is compared to the phrasal (prepositional phrase) ta’ construction, which generally has a possessive interpretation (2). (2)

Id-dar

ta-l-ġara n-biegħ-et of-DEF-neighbor PASS-sell-3SG.F ‘The neighbor’s house has been sold.’

DEF-house

|| Ray Fabri: L-Istitut tal-Lingwistika u t-Teknoloġija tal-Lingwa/Institute of Linguistics and Language Technology, University of Malta, 113, Blokka A, Parkeġġ 6, L-Imsida MSD2080, Malta, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-002

24 | Ray Fabri

The main aim is to establish the status of this construction in Maltese grammar as a lexical item (multi-word expression/complex word/compound) or as a phrase (prepositional phrase). This paper contributes towards a further characterization of multi-word expressions, in particular compounds, and complements other work being carried out on the description of prepositions and prepositional phrases in Maltese.1 The analysis provided here also contributes towards the more general discussion concerning the nature of the syntaxmorphology interface, more specifically the place of word formation, on the one hand, and phrase and sentence formation, on the other within a model of grammatical structure. I will not enter into a terminological discussion at this point of whether the expression under investigation is to be labeled as a compound noun, complex word, multi-word expression or something else, and will refer to it as the lexical ta’ construction.2 This is meant to contrast it with the phrasal ta’ construction shown in (2) above. The analysis provided here is based on data obtained through native speaker introspection, the Maltese Language Resource Server corpus (MLRS: https://mlrs.research.um.edu.mt/) and Aquilina’s (1987/1990) two-volume Maltese-English dictionary. I will first distinguish between various types of ta’ constructions in order to then identify what should count as the lexical ta’ construction as opposed to the phrasal ta’ construction. By ‘lexical’ here I mean a construction that forms a ‘tight’ structural and semantic unit and thus behaves more like a word than a phrase.3

2 The ta’ construction The following is a list of different uses of ta’. Note that this is not meant to be an exhaustive list but to give an indication of the variety of uses. The most salient use of ta’ appears to be the expression of a possessive relation, both abstract possession or ownership (1i) and physical possession or part-of relation (1ii). The rest show other meanings that ta’ might obtain, namely, (2) containment, (3) origin, (4) material, (5) purpose, (6) source, (7) wear /carry, and (8) association.

|| 1 See, for example, Stolz and Vorholt (this volume). 2 For discussions and analyses of multi-word expressions, see, among others, Calzolari et al. (2002); Baldwin and Su Nam (2010); Markantonatou et al. (2018). For compounds, see, e.g., Lieber and Štekauer (2009). 3 For discussions of the concept of word, see, among many others, Dixon and Aikhenvald (2002) and Booij (2005).

The lexical ta’ construction | 25

1. i.

Possession (belonging to) Ownership

(3)

a.

b.

il-ktieb ta’ Ġanni DEF-book of John ‘John’s book’ Dak (huwa) il-ktieb that be.3SG.M.PRES DEF-book ‘That’s John’s book.’

ii.

Part-of

(4)

a.

b.

dipartiment ta-l-gvern department of-DEF-government ‘government department’ il-magna ta-l-karozza DEF-machine of-DEF-car ‘the car’s engine’

2. i.

Containment Physical (encompassing; English ‘for’)

(5)

a.

b.

flixkun ta-l-inbid bottle of-DEF-wine ‘a wine bottle’ kaxxa ta-s-sulfarini box of-DEF-matches ‘matchbox’

ii.

Abstract (including)

(6)

a.

b.

ktieb ta-t-talb book of-DEF-prayer ‘a prayer book’ il-klassi ta-l-ħaddiema DEF-class of-DEF-workers ‘the working class’

ta’ of

Ġanni John

26 | Ray Fabri

3.

Origin (produced by/coming from)

(7)

a.

b.

c.

bajd4 ta-t-tiġieġ egg.SG.M of-DEF-chicken ‘chicken eggs’ ħalib ta-l-baqra milk of-DEF-cow ‘cow’s milk’ ħaxix ta-s-serer vegetables of-DEF-greenhouses ‘greenhouse vegetables’

4.

Material (made of)

(8)

a.

b.

flixkun ta-l-ħġieġ bottle of-DEF-glass ‘glass bottle’ bieb ta-l-injam door of-DEF-wood ‘wooden door’

5.

Purpose (English ‘for’)

(9)

a.

b.

(10) a.

b.

magna ta-l-ħasil machine of-DEF-washing ‘washing machine’ nuċċali ta-x-xemx glasses of-DEF-sun ‘sun glasses’ kuruna ta-r-rużarju crown of-DEF-rosary ‘rosary beads’ bejjiegħ ta-l-laħam seller of-DEF-sun ‘meat seller’

|| 4 Note that bajd functions as a collective/uncountable form in contrast with singular countable bajda ‘egg’ and plural countable bajdiet ‘eggs’ but is formally masculine singular in terms of agreement with the adjective and verb. I will, therefore, gloss nouns with collective meaning as SG.M rather than COLL.

The lexical ta’ construction | 27

6.

Source (producer)

(11) a.

b.

7.

Wear/carry (have on)

(12) a.

b.

8.

siġra ta-l-frott tree of-DEF-fruit ‘a fruit tree’ fabbrika ta-l-karozzi factory of-DEF-cars ‘car factory’

ir-raġel ta-l-ġakketta s-sewda DEF-man of-DEF-jacket DEF-black ‘the man in the black jacket’ ir-raġel ta-l-qargħa DEF-man of-DEF-bald.head ‘the bald-headed man’

Association (related to)

(13) a.

b.

c.

siġra ta-l-Milied tree of-DEF-Christmas Christmas tree’ bajd ta-l-Għid egg.SG.M of-DEF-Easter ‘Easter egg’ uniformi ta-l-iskola uniform of-DEF-school ‘school uniform’

3 Phrasal ta’ construction The prepositional constructions listed above are of two types: (1) noun phrases consisting of a noun head and a prepositional phrase, i.e. NP[N [Pta’ + NPdef]], e.g., siġra ta-l-frott ‘tree of-DEF-fruit’ for ‘fruit tree’, and (2) compounds or multi-word expressions consisting of a head noun and a preposition combined with a noun, i.e. N[N [Pta’ + Ndef]], e.g., siġra ta-l-Milied ‘tree of-DEF-Christmas’ for ‘Christmas tree’. One way of distinguishing between the two constructions is by checking the referential and structural independence of the two nouns involved in each case, i.e. the head noun (is-siġra) and the prepositional noun (il-frott, il-Milied). One way of doing this is by topicalizing the head noun, i.e. shifting its position and referring

28 | Ray Fabri

to it by means of the pronominal clitic -ha attached to the preposition ta’, as shown in (14). This results in an acceptable and meaningful construction in the case of il-frott tas-siġra but an unacceptable and meaningless construction in the case of is-siġra tal-Milied. This indicates that the two nouns are semantically and structurally independent in the former case but not in the latter. (14) a.

Il-frott

tagħ-ha dik is-siġra mimli dubbien of-3SG.F that.SG.F DEF-tree.SG.F full.SG.M fly.SG.M ‘The fruit of that tree is full of flies.’ *Il-Milied tagħ-ha din is-siġra kien DEF-Christmas of-3SG.F this.SG.F DEF-tree.SG.F be.SG.M.PFV tajjeb good.SG.M ‘*The Christmas of the tree was good.’

DEF-fruit.3SG.F

b.

There are other differences, but it is not the aim of this paper to explore the difference between these two types (phrase vs. compound) in detail. What is relevant for the purposes of this study is the fact that, in all of the cases shown above, the ta’+N part of the construction does not generally occur on its own, i.e. without the head noun, unless it can be used elliptically, and is, therefore, context sensitive, which is the case with phrasal constructions such as siġra tal-frott ‘fruit tree’. In the compound type, such as siġra tal-Milied ‘Christmas tree’, it seems that the ta’ part cannot appear on its own. The following are examples of elliptical constructions (15b, 15d, 16b) in which the missing information can be recovered from context (15a, 15c, 16a). (15) a.

b.

c.

d.

(16) a.

Ta’ min hu of who be.SG.M.PRES ‘Whose book is that?’ ta’ Ġanni of John ‘John’s’ Ta’ xiex inhi of what be.SG.M.PRES ‘What is that a tree of?’ ta-l-frott of-DEF-fruit ‘fruit’

dak that

DEF-book

il-ktieb?

dik that

DEF-tree

is-siġra?

Għal liema raġel qed t-għid-li? PROG 2SG.IPFV-tell-me for which man ‘Which man are you telling me about? (‘Which man do you mean?’)

The lexical ta’ construction | 29

b.

c.

(dak) ta-l-ġakketta s-sewda that.SG.M of-DEF-jacket DEF-black ‘the one in the black jacket’ (dak) ta-l-qargħa that.SG.M of-DEF-bald.head ‘the bald-headed one’

As an aside, it is interesting to note that noun phrases like examples (16b,c) and (12) above can also be expressed using the preposition bi ‘with’ instead of ta’. (17) a.

b.

ir-raġel bi-l-ġakketta s-sewda DEF-man with-DEF-jacket DEF-black ‘the man with the black jacket’ ir-raġel bi-l-qargħa DEF-man with-DEF-bald.head ‘the man with the bald-head’

However, these constructions are used in different contexts. Placing them in context helps to illustrate the difference. To take an example, the ta’ version can be used in the context of a question with min ‘who’ (18), while the bi version can be used in the context of a question with kif ‘how’ (19), but not vice-versa. (18) a.

b.

c. d.

e. (19) a.

b.

Min ġie? who come.3SG.M.PFV ‘Who came?’ Ta-l-ġakketta s-sewda of-DEF-jacket DEF-black ‘the black jacket one’ *Bi-l-ġakketta s-sewda with-DEF-jacket DEF-black Ta-l-qargħa of-DEF-bald.head ‘the baldheaded one’ *Bi-l-qargħa with-DEF-bald.head Kif ġie? how come.3SG.M.PFV ‘How did he come?’ *Ta-l-ġakketta s-sewda of-DEF-jacket DEF-black

30 | Ray Fabri

c.

d. e.

Bi-l-ġakketta s-sewda with-DEF-jacket DEF-black ‘with the black jacket’ *Ta-l-qargħa of-DEF-bald.head Bi-l-qargħa with-DEF-bald.head ‘with the bald head’

As can also be clearly seen from the translation, the difference is that the ta’ version helps to identify the person being mentioned, while the bi version describes that person. Moreover, the ta’ version functions as a noun, while the bi version functions as an adverbial. Therefore, the ta’ construction in such cases is used as an entity identifier. This observation will come in handy in the main discussion of the examples below.

4 Stand-alone ta’ constructions: lexical ta’+N On the basis of the examples of phrasal ta’+N constructions such as (18b) above, in which the ta’+NP unit can occur on its own (without the head noun), it seems sensible to assume that we are dealing with an elliptical construction in which the head is (phonologically) empty. In a theory which allows empty pronouns in syntax, the head noun position could be occupied by some abstract pronominal corresponding to explicit ‘one’ in English. Indeed, in every one of the above cases of phrasal elliptical ta’+N, a noun can occupy the head position, e.g. raġel in (18b). However, there are also a number of headless ta’+N constructions in which the head is obligatorily absent, and which I call ‘lexical ta’ construction’ here. The following are a few initial examples (see also (1) above). (20) a.

b.

c.

Mar għand ta-l-laħam. go.3SG.M.PFV at of-DEF-meat ‘She went to the butcher’s.’ Ta-l-posta ma ġie-x illum. of-DEF-post NEG come.3SG.M.PFV-NEG today ‘The postman did not come today.’ Dak il-programm ta-l-ġenn. that DEF-program of-DEF-madness ‘That’s an awesome program.’

The lexical ta’ construction | 31

In the following, I will explore the grammatical properties of these constructions in order to try and determine their morphosyntactic status. Are these constructions also elliptical noun phrases at some level of an analysis? If they are nonelliptical, what is their categorial status? Are they complex nouns, i.e. compounds, or are they prepositional phrases? In order to try and answer these questions, in this section we explore their morphosyntactic and semantic properties. I will start by listing and tentatively classifying some examples, and then move on to analyze their properties. 1. i.

Jobs Seller

(21) a.

ii.

ta-l-gass of-DEF-gas ‘the gasman/woman’ Service provider

(21) b.

2.

Neighboring

(22) a.

b.

3.

ta’ fuq-na of on-1PL.OBJ ‘the neighbors upstairs’ ta’ ħdej-na of near-1PL.OBJ ‘our next-door neighbors’

Modifier/quality

(23) a.

b.

4.

ta-l-posta of-DEF-post ‘the postman

ta-l-flus of-DEF-money ‘well off’ ta-l-għaġeb of-DEF-wonder ‘amazing’

Contextual support

(24) a.

ta-l-ewwel of-DEF-first ‘the former’

32 | Ray Fabri

b.

5.

Idiomatic

(25) a.

b.

c.

6. i.

ta-l-aħħar of-DEF-last ‘the latter’

ta’ barra minn hawn of out from here ‘the devil’ ta-ż-żejjed ħu n-nieqes of-DEF-excessiv take DEF-less ‘be treated as a stopgap’ ta’ min of whom ‘worth (it)’

Proper names Place names

(26) a.

ii.

Ta-s-Sliema of-DEF-Sliema ‘Sliema’ b. Ta-l-Mirakli of-DEF-miracles ‘Of Miracles’ Restaurants

(26) c.

d.

iii.

Ta’ Marija Restaurant of Mary restaurant ‘Mary’s Restaurant’ Ta’ Peter Restaurant of Peter restaurant ‘Peter’s Restaurant’

Nicknames5

(27) a.

Ta-l-Melħ of-DEF-salt ‘Of the Salt’

|| 5 On nicknames in Maltese, see Deguara (2019) and references therein.

The lexical ta’ construction | 33

b.

c.

Ta-s-Sikkina of-DEF-knife ‘Of the Knife’ Ta’ Widna of ear ‘Of an Ear’

In order to understand this construction better, in the following I will start by focusing on and exploring in detail one ‘typical’ example, namely, ta-l-laħam ‘the butcher’, which is another example of the individual job construction like those given in (21) above. This should lay the basis for the eventual exploration of the other examples. In terms of distribution, this construction can stand on its own in the same position as any argument NP, i.e., subject and object (direct, indirect, prepositional). (28) a.

b.

c.

d.

Ta-l-laħam fetaħ of-DEF-meat open.3SG.M.PFV ‘The butcher has opened.’ Ra lil ta-l-laħam see.3SG.M.PFV CS of-DEF-meat ‘He saw the butcher.’ Ta-ha lil ta-l-laħam give.3SG.M.PFV-3SG.M.OBJ CS of-DEF-meat ‘He gave it/her to the butcher.’ Mar għand ta-l-laħam go.3SG.M.PFV at of-DEF-meat ‘He went to the butcher.’

One can also use the ‘full’ expression il-bejjiegħ tal-laħam ‘the meat vendor’ but this form sounds very contrived in an informal context (29a), although it is felicitous in a formal context (29b). (29) a.

b.

Mor-t n-ixtr-i min-għand il-bejjiegħ ta-l-laħam go-1SG 1.IPFV-buy-SG from-at DEF-seller of-DEF-meat imma kien magħluq but be.SG.M.PFV closed ‘I went to buy from the meat vendor but he was closed.’ Il-bejjiegħa ta-l-laħam għand-hom i-żomm-u DEF-seller of-DEF-meat have-3PL.OBJ 3-hold-PL

34 | Ray Fabri

l-ħanut nadif DEF-shop clean.SG.M ‘Meat vendors should keep their shop clean.’ This suggests that tal-laħam in (29a) is not merely an elliptical construction but that it is an independent expression, and that it functions as a noun in its own right. Interestingly, it seems that the expression tal-laħam can also be used predicatively (30a), but, when asked, speakers tend to hesitate and prefer an alternative such as (30b). (30) a.

b.

?Ħi-ja kien ta-l-laħam brother-1SG be.SG.M.PFV of-DEF-meat ‘My brother was a butcher.’ Ħi-ja kien i-biegħ l-laħam brother-1SG be.SG.M.PFV 3SG.M.IPFV-sell DEF-meat /kell-u ħanut ta-l-laħam have-3SG.M.PFV shop of-DEF-meat ‘My brother used to sell meat/had a meat shop.’

Note that one reason for the hesitation in accepting (30a) could be attributed to its ambiguity, since it can be interpreted as ‘My brother was made of flesh’. However, you tend to get the same reaction to constructions which are not ambiguous in the same way, such as (31a) with the more acceptable alternative in (31b). (31) a.

b.

?Ħi-ja kien brother-1SG be.SG.M.PFV ‘My brother was a postman.’ Ħi-ja kien brother-1SG be.SG.M.PFV ‘My brother was a postman.’

ta-l-posta of-DEF-post pustier postman

One interpretation of these facts is that lexical ta’ expressions behave like definite nouns, which cannot be used predicatively, and, when used with ‘be’, are interpreted as equatives (e.g. ‘my friend is the butcher’ as opposed to ‘my friend is a butcher’). The same seems to apply to lexical ta’ expressions. Recall that in Section 3 above I pointed out that ta’ expressions identify, rather than describe, an individual, i.e. they refer to an individual entity, and, therefore, they can be considered definite descriptions.

The lexical ta’ construction | 35

5 Pre-nominal elements The issue I would like to explore in this section is the extent to which the lexical ta’ construction behaves like a NP in terms of its morphosyntactic properties. Let’s start with number and gender.

5.1 Gender and number The expression tal-laħam is formally underspecified for gender. It can have either a male or female referent, and, therefore, trigger either masculine or feminine agreement. (32)

Ta-l-laħam qal-li / qal-t-li… of-DEF-meat say.3SG.M.PFV-1SG.OBJ say-3SG.F.PFV-1SG.OBJ ‘The butcher (he/she) told me…’

The same applies to number, i.e. the construction is underspecified for number. (33)

Ta-l-laħam ħareġ/ of-DEF-meat come.out.3SG.M.PFV ‘The butcher/s came out on strike.’

ħarġ-u come.out-3PL.PFV

fuq on

strajk strike

The default interpretation is probably masculine singular but this can be easily overwritten through context.

5.2 Definiteness and specificity The definite article l- cliticized to the noun in an expression like tal-laħam, which, however, does not make the whole expression definite, just as, for example, ‘of the table’ in ‘the/a leg of the table’ says nothing about the definiteness of the head noun, which can be either definite or indefinite. In the examples above, the lexical ta’ expression refers to a specific individual that is known, or given, in the discourse context. However, the example in (34) shows that it can also be preceded by indefinite wieħed ‘a certain’, in which case it refers to a specific individual who, however, is not known to the hearer. (34)

Iltqaj-t ma’ wieħed/waħda meet-1SG.PFV with one.SG.M/one.SG.F ‘I met a (male/female) butcher.’

ta-l-laħam of-DEF-meat

36 | Ray Fabri

Note that, when used without wieħed, it can also have a generic or non-specific interpretation, both in the singular and in the plural, depending on context. Thus, (35) can be interpreted as referring to a generic butcher (as implied by the inclusion of the indefinite article in the translation (cmp. a statement in English such as ‘The lion is a majestic animal’ in which ‘the lion’ has generic reference). (35) a.

b.

Ta-l-laħam i-kun j-af liema huma of-DEF-meat 1SG.IPFV-be 1SG.IPFV-know which be.3PL l-aħjar biċċ-iet DEF-better piece-PL ‘The/A butcher will know which are the best cuts.’ Ta-l-laħam i-kun-u j-af-u liema huma of-DEF-meat 1.IPFV-be-PL 1SG.IPFV-know-PL which be.3PL l-aħjar biċċ-iet DEF-better piece-PL ‘Butchers will know which are the best cuts.’

Here, too, the expression tal-laħam behaves like any other noun in Maltese in terms of genericity since generic nouns are generally definite (taking the definite article (36a–b)), though they can also be indefinite in the singular (36c). (36) a.

b.

c.

Il-qattus huwa annimal domestiku DEF-cat.SG.M be.3SG.M.PRS animal domestic ‘The cat is a domestic animal.’ Il-qtates i-ħobb-u j-ilagħb-u DEF-cat.PL 3.IPFV-love-PL 3.IPFV-play-PL ‘Cats love to play.’ Qattus-a m’għand-ha qatt t-itħalla cat-SG.F NEG.have-3SG.F.OBJ never 3SG.F.IPFV-be.left ‘A cat should never be left alone.’

waħed-ha alone-3SG.F

5.3 Numerals Nouns can be preceded by numerals, specifically by what Fabri (1994) calls C1 numerals, i.e. ‘transitive’ numerals, such as żewġ (see (37a–b)). C1 numerals are transitive in the sense that they must be followed the noun they quantify, in contrast to C2 numerals (37c–d), which are ‘intransitive’ in the sense that they can occur on their own, and can also serve as pro-forms for numeral + noun. The lexical ta’ construction can only occur with C2 numerals (37e–f).

The lexical ta’ construction | 37

(37) a.

b. c.

d.

e.

f.

Ta-ni żewġ kikkr-i give.3SG.M.PFV-1SG.OBJ two.C1 cup-PL ‘He gave me two cups.’ *Ta-ni żewġ give.3SG.M.PFV-1SG.OBJ two.C1 Tnejn kikkr-i ta-ni two.C2 cup-PL give. 3SG.M.PFV-1SG.OBJ ‘He gave me two cups.’ (lit. Cups, two he gave me) Ta-ni tnejn give. 3SG.M.PFV-1SG.OBJ two.C2 ‘He gave me two.’ *Żewġ ta-l-laħam fetħ-u two.C1 of-DEF-meat open.PFV-3PL ‘two butchers’ Tnejn ta-l-laħam fetħ-u two.C2 of-DEF-meat open.PFV-3PL ‘Two butchers have opened.’

Note that a C2 must always share the definiteness value with the noun (38a–b). (38) a.

b.

tnejn kikkr-i two.C2 cup-PL ‘two cups’ it-tnejn il-kikkr-i DEF-two.C2 DEF-cup.C2-PL ‘both cups’

The lexical ta’ construction can occur with both a definite (i.e. l- marked) and an indefinite C2 (39a–b). (39) a.

b.

Ta-l-laħam it-tnejn fetħ-u of-DEF-meat DEF-two.C2 open.PFV-3PL ‘Both butchers have opened.’ Ta-l-laħam tnejn fetħ-u of-DEF-meat two.C2 open.PFV-3PL ‘Two butchers have opened.’

It appears, therefore, that in these cases tal-laħam behaves like a common noun, and refers to an entity which can be known from context (‘both butchers’) or unknown (‘two butchers’) in context. At this point, note that any robust claims about the nature of these constructions with respect to definiteness, requires an analysis of the pragmatic,

38 | Ray Fabri

semantic and morphological properties of the definite article in Maltese, and, in particular, the role of definiteness, specificity and familiarity (known/unknown), which, however, goes beyond the scope of this descriptive study.

5.4 Demonstratives Demonstratives in Maltese govern a definite noun (40a–b). The expression tallaħam can follow a demonstrative but the construction is then marked, so that it is closer in meaning and use to the demonstrative plus a proper noun (40c), which is also possible but marked relative to a demonstrative plus a definite noun. (40) a.

b. c.

d.

Dak ir-raġel telaq that.SG.M DEF-man.SG.M leave.3SG.M ‘that man left’ *dak raġel telaq that.SG.M man.SG.M leave.3SG.M Dak Ġanni vera kiesaħ that.SG.M John truly cold.SG.M ‘That John is really arrogant.’ Dak ta-l-laħam vera j-dejjaq-ni that.SG.M of-DEF-meat truly SG.M-bother-1SG.OBJ ‘That butcher guy really bothers me.’

The difference between a definite noun and a proper noun here is that, with a definite common noun like ir-raġel ‘the man’, dak ‘that’ implies that there is a choice of individuals in context, of which one is being picked out or individuated (so that pointing, e.g., is possible or implied). However, with a proper noun, and a lexical ta’ expression, generally no choice is implied, and pointing, for example, does not make sense. The following examples in each of which individuals are directly contrasted make this distinction more obvious. Note that the hash symbol is used rather than a star because these constructions are grammatical (structurally) acceptable but infelicitous in context. (41) a.

b.

Dak ir-raġel vera kiesaħ that.SG.M DEF-man truly cold.SG.M imma dan ir-raġel vera minn tagħna but this.SG.M DEF-man truly of ours ‘That man is really arrogant but this man is very nice.’ #Dak Pawlu vera kiesaħ that.SG.M Paul truly cold.SG.M

The lexical ta’ construction | 39

c.

imma dan Pawlu vera minn tagħna but this.SG.M Paul truly of ours ‘That Paul is really arrogant but this Paul is very nice.’ #Dak ta-l-laħam vera kiesaħ that.SG.M of-DEF-meat truly cold.SG.M imma dan ta-l-laħam vera minn tagħna but this.SG.M of-DEF-man truly of ours ‘That butcher guy is really arrogant but this butcher guy is very nice.’

With respect to the demonstrative, therefore, an expression like tal-laħam behaves more like a proper noun than a definite common noun. Just like a proper noun, it serves to uniquely identify a person, in this case, through his/her job.

5.5 Superlative The superlative construction in Maltese can be of two types: a) a synthetic form with the adjective generally taking the form VCCV(C), eqdem from qadim ‘old’ (42a), or b) an analytic form with l-aktar ‘the most’ (literally ‘the more’) followed by the noun and the adjective (42b). (42) a.

b.

L-Imdina hija l-eqdem belt ta’ Malta Mdina be.3SG.M.PRES DEF-old.CMP city of Malta ‘Mdina is the oldest city of Malta.’ L-Imdina hija l-aktar belt interessanti ta’ Mdina be.3SG.M.PRES DEF-more city interesting of ‘Mdina is the most interesting city of Malta.’

Malta Malta

It turns out that the construction under scrutiny here does not fit into the typical noun position in either of the superlative constructions. (43) a.

b.

*Kolinku l-ogħla ta-l-laħam f’Ħaż-Żebbuġ Kolinku DEF-high.CMP of-DEF-meat in Ħaż-Żebbuġ ‘Kolinku is the most expensive butcher in Ħaż-Żebbuġ.’ *Kolinku l-aktar ta-l-laħam għoli f’Ħaż-Żebbuġ Kolinku DEF-more of-DEF-meat high.SG.M in Ħaż-Żebbuġ ‘Kolinku is the most expensive butcher in Ħaż-Żebbuġ.’

In this case, the lexical ta’ expression behaves again as if it were a definite noun since a superlative, whether synthetic or analytic, being itself l- marked, always governs a non-definite noun (compare 43 with 44).

40 | Ray Fabri

(44) a.

b.

*L-Imdina hija l-eqdem il-belt ta’ Malta Mdina be.3SG.M.PRES DEF-old.CMP DEF-city of Malta ‘Mdina is the oldest city of Malta.’ *L-Imdina hija l-aktar il-belt interessanti Mdina be.3SG.M.PRES DEF-more DEF-city interesting ta’ Malta of Malta ‘Mdina is the most interesting city of Malta.’

Note that this construction, again, can be rescued by adding wieħed, which renders the expression tal-laħam non-specific. (45)

L-aħjar

wieħed ta-l-laħam qiegħed one of-DEF-meat located.SG.M ‘The best butcher is in Ħaż-Żebbuġ.’ DEF-better

Ħaż-Żebbuġ Ħaż-Żebbuġ

5.6 The quantifier xi One observes a similar effect with the quantifier xi ‘some’, which also requires a non-definite noun, and therefore cannot be combined with tal-laħam unless it is preceded by wieħed. (46) a. b.

*Janice iltaqgħ-et Janice meet-3SG.F.PFV Janice iltaqgħ-et Janice meet-3SG.F.PFV ‘Janice met some butcher.’

ma’ with ma’ with

xi ta-l-laħam some of-DEF-meat xi wieħed ta-l-laħam some one of-DEF-meat

6 Post-nominal and other elements in NP In this section, I will look at combinations of ta’+N with post-nominal and other elements within the NP. Let us first observe the combination of the lexical ta’ construction with post-nominal modifiers, namely, adjectives (47a), relative clauses (47b) and prepositional phrases (47c).

The lexical ta’ construction | 41

6.1 Nominal modifiers (47) a.

b.

c.

Ta-l-laħam il-ġdid nadif ħafna of-DEF-meat DEF-new clean.SG.M much ‘The new butcher is very clean.’ Ta-l-laħam li xtraj-na min-għand-u lbieraħ of-DEF-meat that buy-1PL.PFV from-at-3SG.M.OBJ yesterday nadif ħafna clean.SG.M much ‘The butcher we bought from yesterday is very clean.’ Ta-l-laħam ta’ għajn-ej-h blu nadif ħafna of-DEF-meat of eye-PL-3SG.M.OBJ blue clean.SG.M much ‘the butcher with the blue eyes is very clean.’

The lexical ta’ construction can be modified by adjectives, relative clauses and prepositional phrases, just like any common noun.

6.2 The quantifier kollha As the following example shows, the lexical ta’ construction under observation combines with the quantifier kollha ‘all’. (48)

Ta-l-laħam koll-ha magħluq-in of-DEF-meat all-PL closed-PL ‘All butchers are closed today.’

illum today

Note that kollha (just like the C2 numerals, e.g. tnejn, see above) does not only occur within the NP but can occur in any position within the sentence, including outside of the NP (e.g. Tal-laħam magħluqin kollha llum, Tal-laħam magħluqin llum kollha, Kollha magħluqin tal-laħam illum). Example (48) shows that ta’+N can be quantified, just like any common noun. Note that prepositional constructions can only be quantified in this way if they are understood as elliptical ta’ constructions (49). This clearly shows, once again, that the ta’ construction is nominal and not prepositional. (49)

Ta’ fuq il-mejda koll-ha antik-i of on DEF-table all-PL old-PL ‘The ones on the table are all old.’

42 | Ray Fabri

6.3 The construct Finally, a typical nominal construction is the so-called construct state (50a), by which a noun and a noun phrase are juxtaposed syntactically producing a possessive construction.6 (50b) shows that the possessor noun (Marija) in the construct can be topicalized and marked as such by the pronominal clitic attached to the possessed noun (xagħar). (50) a.

b.

Xagħar Marija twil ħafna hair Mary long much ‘Mary’s hair is very long.’ Marija xagħar-ha twil ħafna Mary hair-3SG.F.OBJ long much ‘Mary, her hair is very long.’

We can easily replace Marija in (50) by tal-laħam so that the lexical ta’ construction can appear as the argument of a head noun (typically, though not necessarily an inalienable entity, namely, family relation or body-part). (51) a.

b.

Xagħar ta-l-laħam twil ħafna hair of-DEF-meat long much ‘The butcher’s hair is very long.’ Ta-l-laħam xagħar-ha twil ħafna of-DEF-meat hair-3SG.F long much ‘The (female) butcher’s hair is very long.’ (‘The (female) butcher, her hair is very long.’)

To summarize: In terms of distribution, the lexical ta’ construction can be found in typical noun positions and function as head noun in the noun phrase. It is underspecified for gender and number and can therefore combine with adjectives in the noun phrase and verbs in the sentence of any gender and number. It behaves like a definite description, in some cases behaving like a proper noun, in others like a definite noun phrase. By default, the construction is interpreted as specific but can also obtain a generic or unspecific interpretation given the right context.

|| 6 For a study on the construct in Maltese, see Fabri (1996).

The lexical ta’ construction | 43

7 Referential properties One question that remains, and which I would like to discuss, in the case of tallaħam is what kind of entity it refers to, more specifically, whether it refers to a person (the butcher) or a place (the butcher’s), or both. Native speaker intuitions vary on this and a corpus search shows that tal-laħam always appears to refer to the person, i.e. the butcher as opposed to ‘the butcher’s’, so that when the shop is meant it is preceded by il-ħanut ‘the shop’ or il-ħwienet ‘the shops’. However, the ‘butcher’s’ reading is possible, and does not appear to be completely excluded, as the examples in (52) seem to indicate. (52) a.

b.

c.

Ta-l-laħam qiegħed wara din il-kantuniera of-DEF-meat be.located.SG.M behind this DEF-corner ‘The butcher is behind this corner.’ Ta-l-laħam sa j-waqqgħu-h għax of-DEF-meat FUT 3SG.F.IMPF-demolish-3SG.M because mibni fuq ġibjun built on cistern ‘The butcher’s is going to be pulled down because it is built on top of a cistern.’ *Ta-l-laħam sa j-waqqgħu-ha għax of-DEF-meat FUT 3SG.F.IMPF-demolish-3SG.F because mibni fuq ġibjun built on cistern

The clear unacceptability of (52c), which forces a ‘person’ reading through feminine gender (and, therefore, a human reference) as opposed to masculine (and, therefore, also unmarked or neutral reading), indicates that, at least for some speakers, the ‘building’ reading is possible, though marked when compared to the ‘person’ reading. Therefore, it seems reasonable to assume that tal-laħam refers to a person by default but can be taken to refer to the building, given the right context. This is indeed also the case with nouns which refer to a profession and can also be associated with a building, such as spiżjar ‘pharmacist’ and ‘pharmacy’. (53)

L-iżpiżjar qiegħed wara DEF-pharmacist be.located.SG.M behind ‘The pharmacist’s is behind this corner.’

din this

il-kantuniera DEF-corner

44 | Ray Fabri

8 Conclusion It is clear from the above that the lexical ta’ expression tal-laħam behaves like a noun and refers to a (specific or generic) individual entity, rather than like a typical prepositional phrase, which generally refers to (spatial, temporal, possessive, etc.) relations between entities. The lexical ta’ construction is an exocentric compound noun and, as such, falls under the more general notion of multiword expression (see, among others, Calzolari et al. 2002: 1934 for multiword expressions). The analysis presented here shows that phrasal constructions, in this case the ta’+N construction can be lexicalized and, in the process, take on properties typical of individual words as opposed to phrases. Once lexicalized, these forms are presumably stored as single structural entities in the lexicon.7 This contribution is meant as a first step in a more detailed discussion of the lexical ta’ construction and its theoretical implications for the grammar of Maltese, as well as a contribution to a theoretical discussion of multi-word expressions and the morphology-syntax interface.

Abbreviations 1, 2, 3 C1, C2 CS CMP COLL DEF F FUT IPFV M

N Ndef NEG

NP OBJ P PASS

1st, 2nd, 3rd person cardinal 1, cardinal 2 case comparative collective definite feminine future imperfect masculine noun definite noun negation noun phrase object preposition passive

|| 7 For a similar example from Maltese involving the construct state, see Fabri (1996).

The lexical ta’ construction | 45

PFV PL PRES PROG SG

perfect plural present progressive singular

References Aquilina, Joseph. 1987. Maltese-English dictionary. Vol. 1: A–L. Malta: Midsea Books. Aquilina, Joseph. 1990. Maltese-English dictionary. Vol. 2: M–Z. Malta: Midsea Books. Baldwin, Timothy & Kim Su Nam. 2010. Multiword expressions. In Nitin Indurkhya & Fred J. Damerau (eds.), Handbook of natural language processing, 267–292. Boca Raton, USA: CRC Press. Booij, Geert. 2005. The grammar of words. Oxford: Oxford University Press. Calzolari, Nicoletta, Charles J. Fillmore, Ralph Grishman, Nancy Ide, Alessandro Lenci, Catherine MacLeod & Antonio Zampolli. 2002. Towards best practice for multiword expressions in computational lexicons. Proceedings of the Third International Conference on Language Resources and Evaluation (LREC’02), 1934–1940. Las Palmas: ELRA. Deguara, Abigail. 2019. Il-Laqam Personali: Analiżi Lingwistika. University of Malta B.A. thesis. Dixon, R. M. W. & Alexandra Y. Aikhenvald (eds.). 2002. Word: A cross-linguistic typology. Cambridge: Cambridge 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. Malta: University of Malta. Fabri, Ray. 1996. The construct state and the pseudo-construct state in Maltese. Rivista di Linguistica 8(1). 229–244. Fabri, Ray. 2009. Compounding and adjective-noun compounds in Maltese. In Bernard Comrie, Ray Fabri, Elisabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 207–231. Amsterdam & Philadelphia: John Benjamins. Lieber, Rochelle & Pavol Štekauer (eds.). 2009. The Oxford handbook of compounding. Oxford: Oxford University Press. Markantonatou, Stella, Ramisch Carlos, Agata Savary & Veronika Vincze (eds). 2018. Multiword expressions at length and in depth: Extended papers from the MWE 2017 workshop. Berlin: Language Science Press. Stolz, Thomas & Maike Vorholt. this volume. On WH-PREPS: A contrastive grammatical sketch of Maltese fejn and Spanish donde/dónde.

Christopher Lucas and Michael Spagnol

Nunation from Arabic to Maltese Abstract: This article discusses Maltese words containing an innovative final /n/, arguing that /n/ addition is motivated by speakers’ expectation that lexical items resemble prototypical Maltese phonological words. /n/ is added to items that deviate from this prototype in containing a word-final stressed open syllable. Syllable closure through consonant addition eliminates this deviation. /n/ is the consonant chosen because of its pre-existing alternation with zero wordfinally. Discussion of the details of this process and the items it does (not) affect sheds light on the history of Arabic and Maltese, as well as on the nature of irregular phonological change in general. Keywords: etymology; loanword adaptation; paragoge; syllable

1 Introduction In his article Maltese etymological notes, Saydon (1965: 72) observes that: [i]t is a fact, hardly noticed by Maltese scholars, that foreign words ending in an accented vowel add a final consonant when they pass into Maltese, thus Ital. gioventù becomes in Maltese ǧuvintur; Eng. sofa becomes sufan; Eng. blue becomes blun; Eng. jury, pronounced ǧurì, becomes ǧurin; Eng. (grain of) coffee, Ital. caffè, becomes kafen(a), etc.

It is true that this phenomenon has received scant attention. In the period since the publication of Saydon’s article it seems to have gone unmentioned other than in Aquilina’s (1987: s.v. N) note in his dictionary that /n/ may appear as “a post vocalic accretion in the final open syllable, mainly in a few Eng[lish] loan-words. Exx. M[altese] blun ‘blue’; M[altese] skrun ‘screw’.” The purpose of the present article is to subject this phenomenon to detailed scrutiny for the first time: to produce as complete a list as possible of Maltese lexical items in which we find an etymologically unexpected final consonant (typically, but not exclusively, /n/);

|| Christopher Lucas: School of Languages, Cultures and Linguistics, SOAS University of London Thornhaugh Street, Russell Square, London WC1H 0XG, United Kindom, E-mail: [email protected] Michael Spagnol: Department of Maltese, University of Malta, L-Imsida, MSD2080, Malta, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-003

48 | Christopher Lucas and Michael Spagnol

and, more importantly, to offer a reconstruction of how and why these consonants came to be added. We will see in the course of our investigation that this topic intersects with and sheds light on some foundational issues in Arabic and general historical linguistics. These include: the nature of nominal inflection in the earliest stages of spoken Arabic, cognitive motivations for loanword adaptation, and how morphological regularities can feed irregular phonological change. The article is structured as follows. Section 2 sets out the data and provides an initial sketch of our analysis. Section 3 then considers the reverse phenomenon, namely instances of inherited word-final /n/ which alternate with zero, these alternations being central to our account of how /n/ later comes to be appended to the items listed in Section 2. In Section 4 we consider the provenance of final /n/ in xejn specifically, an item whose etymology turns out to have important implications for our understanding of the history of Arabic. Section 5 then presents in detail our analysis of the factors motivating /n/ addition in the majority of items presented in Section 2, while Section 6 discusses the details of the etymology of all of these items. Section 7 concludes.

2 Data In this section we first list the items under consideration, together with an initial indication of their likely etymologies, in Tables 1–5. We then provide an outline of our proposed analysis. The items under consideration are arranged as follows. Table 1 presents every example we have been able to find,1 in which the occurrence of an etymologically unexpected final /n/ is a normal part of standard Maltese (or used to be, in the case of items which have fallen out of regular use). As we will see in Section 3, several of the Arabic-derived items in Table 1 (in addition to a number of other Arabic-derived items whose final /n/ is not an innovation) may optionally be realized without this final /n/. These /n/-less realizations are best understood as informal or allegro variants. In Table 2, by contrast, we list all the items we have found in which forms both with and without etymologically unexpected final /n/ are attested in ordinary standard Maltese, while Table 3 presents all the items we have observed in which a standard form with no final /n/ is commonly replaced in informal or low-register speech by a form with final /n/ appended.

|| 1 The items in Tables 1–2 and 4–5, as well as the historical/obsolete forms in Table 3, are all listed as (optionally) featuring final /n/ in Aquilina’s (1987) dictionary. The remaining items in Table 3 are the result of personal observation of informal and low-register contemporary Maltese speech.

Nunation from Arabic to Maltese | 49

Table 1: Items in which presence of final /n/ is (or was) considered standard.

Item

Gloss

Etymon

dan/din/dawn ~ dana/dina/dawna

demonstrative pronoun/determiner

Ar. *ðā/ðī/ðū

Ġaħan

name of folklore character

Ar. *ǧuḥā Ar. *ɣafā ~ ɣafāʔ ‘chaff’ (?) or *ʕafān ‘mould’ (?)

għefien

‘chaff, wheat stalks; dirt’

ħdejn

‘near’

Ar. *ḥiðāʔ ‘opposite to, in front of’

xejn

‘nothing’

Ar. *šayʔ ‘thing’

xulxin

‘each other’

Ar. *šī l-šī ‘one part to another part’

serkin

‘sulky (cart)’

Eng. sulky

skrun

‘propeller’

Eng. screw

sufan

‘sofa’

Sic. sufà

armajn ~ ormajn

‘too late; nearly’

It. ormai ‘by now’

gabrijolin

‘cabriolet carriage’

It. cabriolè

perun ~ pirun

‘something of great value’

It. Perù ‘Peru’ (historically a major silver exporter)

tabin

type of silk

It. tabì ‘silk tafetta’

biskwin

‘biscuit porcelain’

Fr. biscuit

paspartun

‘cardboard picture frame’

Fr. passe-partout

surtun

‘surtout coat’

Fr. surtout

Table 2: Items in which presence of final /n/ is considered optional.

Item

Gloss

Etymon

blun ~ blu

‘blue’

Eng. blue

ġurin ~ ġurì ~ ġuri

‘jury’

Eng. jury, It. giurì

kakin ~ kakì ~ kaki ~ kakir ~ kajki

‘khaki’

Eng. khaki

baxan ~ baxà ~ baxa ~ paxa

‘Pasha’

It. bascià, Eng. Pasha

50 | Christopher Lucas and Michael Spagnol

Table 3: Items in which presence of final /n/ is considered non-standard but is attested in informal, low-register, or dialectal speech.

Item

Standard form

Gloss

Etymon

barbikjun ~ barbikjù

barbikju

‘barbecue’

Eng. barbecue

ġuġun

ġuġù

‘jujube’

Eng. jujube

kangarun

kangarù

‘kangaroo’

Eng. kangaroo

kjun

kju

‘queue’

Eng. queue

owkejn

owkej

‘okay’

Eng. okay

referin

referì

‘referee’

Eng. referee

sprejn

sprej

‘spray’

Eng. spray

stjun

stju

‘stew’

Eng. stew

xampun

xampù

‘shampoo’

Eng. shampoo

Ġisun

Ġesù

‘Jesus’

It. Gesù

tarì

a pre-modern currency denomination

It. tarì

tarin

2

Table 4 presents all the items we are aware of in which a base form ends in a vowel but an /n/ is (at least sometimes) appended to this base form when number suffixes are added. Table 4: Items in which stem-final /n/ appears before singulative or plural morphemes.

Item

Gloss

Base form

Gloss

Etymon

abbatini

‘altar boys’

abbati

‘altar boy’

Sic./It. abbati (but also dim. abbatinu/abbatino)

kafena

‘coffee bean’

kafè

‘coffee’

It. cafè

tenijiet

‘teas’

te

‘tea’

It. tè

kawxuna ~ gawxuna ~ gamxuna

‘horse’s overreach boot’

kawċù

‘hose pipe’ (< ‘rubber’)

Eng. (< Fr.) caoutchouc

|| 2 Aquilina (1987) gives only the form tarì, but the form tarin is widely used for this denomination in works dating from the period it was in circulation (e.g. Snelling 1766: 23–24; Finlay 1803: 167).

Nunation from Arabic to Maltese | 51

Finally, Table 5 presents all the items we are aware of in which a final consonant other than /n/ has been appended to a vowel-final etymon. Table 5: Items which feature an innovative word-final consonant other than /n/.

Item

Gloss

Etymon

dawl

‘light’

Ar. *ð̣aw < *ḍawʔ

dikutell ~ dukutell ~ digudell ‘a stone stood on its narrow dimension, cargo carried sideways, the last horizontal wooden strip of a cart’ ġuvintur

‘young people’

Fr. de côté ‘sideways, on one side’ It. gioventù ‘youth, young people’

What can immediately be seen from this full list of items is that Saydon’s (1965: 72) generalization quoted in Section 1 is broadly correct: stem-final consonant addition has occurred overwhelmingly in the context of items whose etyma have a stressed final vowel (where this includes vowel-final monosyllables and the diphthongs ⟨ej⟩ and ⟨aw⟩). A majority of these etyma are loans from European languages, but several are inherited from Arabic. The only items with a nonetymological final consonant whose etyma appear not to have this property are: Ġaħan [ˈʤɐhɐn] (the name of a folklore character), għefien ‘chaff, wheat stalks’, serkin ‘sulky cart’, ġurin ‘jury’, kakin ‘khaki’, barbikjun ‘barbecue’, and abbati(ni) ‘altar boy(s)’. As we will see in Section 6, a closer look at the precise etymologies and likely adaptation processes of these items shows that they too conform to the generalization that non-etymological consonants are only added in Maltese to items with a stressed final vowel. The question, then, is why final consonants are added in this context. Our answer, which we discuss and substantiate further in Section 5, is that items with a stressed final vowel, being of very low type frequency in the Semitic portion of the Maltese lexicon, are felt by native speakers to be poor exemplars of the typical Maltese phonological word. Final consonant addition is therefore a means of bringing such items into closer conformity with the phonological prototype. To answer the question of how and why it is (almost always) /n/ in particular that has been recruited for this purpose, we first need to understand the pre-existing alternations of word-final /n/ and zero that Maltese inherited from Arabic.3 || 3 A typological pilot study by VanDam (2004), examining a sample of 18 languages, suggests that “if a language permits a single word-final coda, it will be /n/” and that “[i]f word-final

52 | Christopher Lucas and Michael Spagnol

3 Alternations of /n/ and zero in Arabic and Maltese A basic feature of nominal (and adjectival) morphology in Standard Arabic is the grammatically conditioned alternation of /n/ with zero in case and number suffixes. Typical nouns and adjectives without a dual or plural suffix appear in one of three cases, each indicated by a suffix consisting of one of the three short-vowel phonemes /a, i, u/. In the default situation, the case suffix is then followed by a final /n/, in a process called tanwīn by Arabic grammarians, and usually translated as nunation (i.e. ‘the process of adding the letter nūn to a word’) in English. Absence of nunation is caused by various factors, most notably prefixing of the definite article (a)l-, as well as the noun or adjective in question being non-final in a synthetic genitive construction, as illustrated in (1). In a similar but distinct process, the dual and masculine plural suffixes are allomorphic: they appear with a final /nV/ by default, this /nV/ ending being obligatorily absent when the host element is non-final in a synthetic genitive construction (but not when it is prefixed by the definite article), as illustrated in (2). (1) a.

b.

Standard Arabic samiʕa l-walad-u ṣawt-a-n ɣarīb-a-n hear.PRF.3SG.M DEF-boy-NOM sound-ACC-NUN strange-ACC-NUN ‘The boy heard a strange sound.’ samiʕa r-raǧul-u ṣawt-a walad-i-n hear.PRF.3SG.M DEF-man-NOM sound-ACC boy-OBL-NUN ‘The man heard a boy’s voice.’

|| coda segments are added incrementally (synchronically or diachronically) the first segment allowed into word final coda positions will likely be /n/” (VanDam 2004: 132). It may therefore be tempting to speculate that the human language faculty somehow specifies /n/ as the archetypal coda consonant, and that this is thus (part of) the reason why it is /n/ in particular that Maltese speakers use to close final stressed open syllables. Blevins (2004: 159–164), however, shows on the basis of a much more extensive survey that the evidence for a universal preference for sonorant codas is weak, and argues that where languages do exhibit a tendency of this sort, it is best explained through a combination of chance and independently motivated sound changes. Our focus here is therefore on Maltese-specific factors that will have favored the rise of /n/ addition, though we do not exclude the possibility that both Maltese-specific and more general factors may have worked together to have this result.

Nunation from Arabic to Maltese | 53

(2) a.

b.

Standard Arabic samiʕa t-tābiʕ-ūna l-ʔāḫar-īna hear.PRF.3SG.M DEF-follower-PL.NOM DEF-other-PL.ACC ‘The followers heard the others.’ samiʕa tābiʕ-ū-ka ʔaḫaw-ay-ka hear.PRF.3SG.M follower-PL.NOM-2SG.M brother-DU.ACC-2SG.M l-ʔaṣɣar-ayni DEF-small.ELA-DU.ACC ‘Your followers heard your two younger brothers.’

Case inflection on nominal elements is absent in the contemporary Arabic dialects and Maltese, as is nunation of the kind illustrated in (1), although, as we shall see in Section 4, some residues of the latter do survive. But the dual and plural suffixes are preserved, and, in the case of the dual suffix, the grammatically conditioned alternation between /n/ and zero also survives: where the possessor of a noun bearing the dual suffix is pronominal, the dual morph lacks the final /n/. Thus in Maltese we have an alternation between -ejn [ɛɪn] and -ej [ɛɪ] (with the variant -ajn [ɐɪn] vs. -aj [ɐɪ] after guttural consonants), as illustrated in (3), where we also see that this originally exclusively dual suffix has taken on an additional function as a marker of simple plural: the so-called “pseudodual” (see Blanc 1970; Fenech 1996). This pseudodual plural marking occurs with nouns referring to body parts – such as eyes, ears and hands – that come in pairs in humans. Note that reference to such body parts is obviously relatively frequent in ordinary speech, and that the nouns used for this reference are typically both plural (i.e. two or more, thus carrying the pseudodual suffix) and carry a possessive suffix pronoun. (3) a.

b.

Maltese Irrid nibqa’ b’-seba’ għajn-ejn, il-ħin koll-u. want.IMPF.1SG stay.IMPF.1SG with-seven eye-PL DEF-time all-3SG ‘I want to remain totally vigilant [lit. ‘with seven eyes’] the whole time.’ [Korpus Malti v3.0 literature20] Żommu għajn-ej-kom u widn-ej-kom miftuħ-in… hold.IMP.2PL eye-PL-2PL CONJ ear-PL-2PL open.PTCP.PASS-PL ‘Keep your eyes and ears open…’ [Korpus Malti v3.0 literature7]

The /n/~zero allomorphy that Maltese inherits from its spoken Arabic ancestor is thus robustly maintained, as it is in the majority of contemporary Arabic dialects. What is less typical about Maltese is that, in informal speech, the /n/~zero alternation in (pseudo)duals is optionally extended beyond possessive contexts, as illustrated in (4).

54 | Christopher Lucas and Michael Spagnol

(4) a.

b.

c.

Maltese Wie[ħ]ed minn dawn se jkun fost-na ix-xahar one from DEM.PL FUT be.IMPF.3SG.M among-1PL DEF-month id-die[ħ]el! Min hu u [x’ inhu] il-programm DEF-enter.PTCP.ACT who 3SG.M CONJ what PRED DEF-program ta[’] jum-ej[’] se n[ħ]abbru nhar is-Sibt… GEN day-DU FUT announce.IMPF.1PL day DEF-saturday ‘One of these people will be with us next month! We will announce who it will be and what the two-day program will be on Saturday…’ [https://twitter.com/milanclubmalta/status/669147517188243456, accessed 04/06/2020] kemm imma[ċċ]ja najs il-kulu[r]-i ta-l-iskarf how_much match.IMPF.3SG.M nice DEF-color-PL GEN-DEF-scarf [u] it-tinda. It-tnej[’] blun-i. CONJ DEF-tent DEF-two blue-PL ‘How nicely the colors of the scarf and the tent match. They’re both blue.’ [https://daphnecaruanagalizia.com/2012/04/super-one-its-a-bitchfest/, accessed 04/06/2020] [Ż]ew[ġ] saq-aj[’] zopp-i kell-na di-d-darba. two leg-PL lame-PL POSS.PRF-1PL DEM-DEF-time ‘We had two lame legs at that time.’ [http://archive.maltatoday.com.mt/commentary/0603.html, accessed 04/06/2020]

Note that this is not a straightforward, productive phonological rule (n > 0 / Vɪ_#) that is blind to morphology: non-duals dejn ‘debt’ and għajn ‘eye; spring’, for example, never seem to be realized without the final /n/. On the other hand, while this optional deletion rule is mostly restricted to the morphological context of the (pseudo)dual, this is not absolute: there are a handful of non-dual word stems ending in /ɛɪn/, in which the final /n/ is uncontroversially a retention rather than a recent innovation, where we also find optional /n/ deletion. Most notably there are the various compounds involving *ʔayn ‘where’: fejn ‘where’ (< *fī ʔayn ‘in where’), mnejn ‘from where’ (< *min ʔayn ‘from where’), and għalfejn ‘why’, as illustrated in (5). Note that fejn may also take pronominal suffixes. When it does so the final /n/ is typically retained, but not in all cases, as shown in (6).4

|| 4 A search of the MLRS Korpus Malti v3.0 for fej[ja,k,uh,ha,kom,hom] returns 9 distinct matches, versus more than 400 for fejn[i,ek,u,ha,kom,hom]. The 1PL suffix was omitted from both searches as it fails to distinguish between the two stems, giving rise to fejna in either case.

Nunation from Arabic to Maltese | 55

(5)

(6)

Maltese a. Tinsew-x minn fej’ ġej id-dawl forget.IMP.2PL-NEG from where come.PTCP.ACT DEF-light u d-dell-ijiet. CONJ DEF-shadow-PL ‘Don’t forget where the light and the shadows come from.’ [Korpus Malti v3.0 literature11] b. għax ma taf-x il-mnej’ u l-fej’ because NEG know.IMPF.2SG-NEG DEF-from_where CONJ DEF-where ‘[…] because you do not know the where and the whence’ [Similitudni5] c. mi-t-tgawdija ta-l-ħajja insum u from-DEF-happiness GEN-DEF-life abstain.IMPF.1SG CONJ nistaqsi għalfej’. ask.IMPF.1SG why ‘I abstain from the happiness of life and ask myself why.’ [Korpus Malti v3.0 news62852] Maltese U r-raba’ u l-għelieqi fejn-hom? Fej-hom CONJ DEF-countryside CONJ DEF-field.PL where-3PL where-3PL il-biedja u s-sajd? DEF-agriculture CONJ DEF-fishing ‘And where are the countryside and the fields? Where is the agriculture and fishing?’ [Korpus Malti v3.0 news92922]

Final /n/ is also occasionally omitted with bejn ‘between’ (< *bayn), as illustrated in (7), but hardly ever when pronouns are suffixed.6 (7)

Maltese Għidt bejn-i u say.PRF.1SG between-1SG CONJ ‘I said to myself: “This life…”’

bej’ ruħ-i… “Di-l-ħajja…” between soul-1SG DEM-DEF-life [Korpus Malti v3.0 news64774]

|| 5 Similtudi Din, dwar daqs l-Imħabba, a poem by Leanne Ellul, available online at https://www.um.edu.mt/library/oar/bitstream/123456789/37075/21/Similitudni%20din%2C%2 0dwar%20daqs%20l-imhabba.pdf (accessed 04/06/2020). 6 There are 13 distinct strings in the Korpus Malti containing bej functioning as a free-standing preposition, versus more than 200,000 matches for bejn. A search for bejn[i,ek,u,ha,kom,hom] returns over 2500 matches, whereas bej[ja,k,u,ha,kom,hom] returns just seven meaningful matches.

56 | Christopher Lucas and Michael Spagnol

A more complex case concerns lejn ‘to, towards’, which is occasionally realized as lej’ when used with a nominal complement, as in (8). (8)

Maltese z-żjara tagħ-na lej DEF-visit GEN-1PL to ‘our visit to Sugarshake’

Sugarshake PN

[Korpus Malti v3.0 opinion1221]

The precise etymology of lejn is uncertain. Sutcliffe (1936: 203) and Barbera (1939a: s.v. lejn) are both of the opinion that it derives from a directional preposition involving /l/ – presumably ultimately *ʔilay ‘to, towards, until’ – plus *ʔayn ‘where’. It is hard to accept this proposal, however, since lejn, unlike fejn and mnejn, never functions as an interrogative adverb, always only as a preposition ‘to, towards’. Moreover, forms apparently cognate to lejn, but with the meaning ‘until’ (apparently deriving from *ʔila ʔin ‘until’), are widespread across the Arabic dialects, e.g. Bahrain (i)lēn (Holes 2001: s.v. L-Y-N2). It therefore seems likely that the Maltese form lejn is inherited and has shifted its meaning from temporal ‘until’ to directional ‘towards’. This picture is complicated by the fact that when lejn takes pronominal suffixes, /n/ is omitted 99.9% of the time in the Korpus Malti, giving forms such as lejja ‘to me’ and lejhom ‘to them’.7 These suffixed forms are best understood as being directly cognate with Classical Arabic ʔilayya ‘to me’ and ʔilayhum ‘to them’. That is, although unsuffixed lejn and suffixed lej- behave in synchrony as allomorphs of a single lexeme, ultimately they go back to distinct etyma: the compound form *ʔila ʔin (< *ʔilay ʔan) ‘until’ in the case of lejn; and the single preposition *ʔilay ‘to, towards’ in the case of lej-. The form lej’ with nominal complements therefore represents another instance of final /n/ deletion. Thus the rule of n > 0 / Vɪ_#, while not fully productive, does seem to be able to apply optionally with a handful of non-dual lexemes. Strikingly, however, deletion of final /n/ (in words where that /n/ is a retention, not an innovation) does not appear to occur in any other phonological environment. The only apparent exception to this generalization that we have been able to find concerns hawn ‘here’, which is optionally realized as haw’ /ɐʊ/. We discuss this item in Section 6, where we show that the historical relationship of this form to hawn is unlikely to be one of /n/ deletion at all. The wider phenomenon of optional final /n/ deletion in Maltese should therefore be understood as fundamentally a morphological process, albeit with some phonologically-driven ana-

|| 7 The search lej[ja,k,h,ha,kom,hom] returns over 15,000 matches in the Korpus Malti. Lejn[i,ek,u,ha,kom,hom] returns 16 meaningful matches.

Nunation from Arabic to Maltese | 57

logical extension to lexical items whose rhymes happen to be identical to the (pseudo)dual suffix that is the core context for the process. More important for our purposes than the precise analysis of /n/ deletion, however, is the basic observation that the upshot of this process is a rather frequent optional alternation between zero and /n/ (but no other consonants) word-finally in Maltese. As a consequence of this pre-existing alternation, when Maltese speakers later felt the need to repair words with a stressed final vowel by closing the syllable with an additional consonant, /n/ was the obvious candidate. We cannot automatically assume, however, that every etymologically unexpected word-final /n/ in Maltese is necessarily the result of this process. A particularly interesting case in point concerns xejn ‘nothing’, to which we now turn.

4 xejn and its implications for the history of dialectal Arabic Despite being well aware that /n/ may appear as “a post vocalic accretion in the final open syllable”, Aquilina (1987: s.v. xejn; għefien) makes the interesting claim that the final /n/ in xejn ‘nothing’ (< *šayʔ ‘thing’) is not just another instance of /n/ addition in Maltese, but is rather “a residue of the accusative case of [C]lassical Ar[abic]… a sporadic case of nunation in M[altese]”. On the face of it, there are several reasons to treat this claim with skepticism. Notwithstanding these considerations, the surprising truth is that Aquilina’s claim is almost certainly correct: the final /n/ in xejn is indeed a fossilized retention of nunation. In what follows, we first review the reasons for skepticism before turning to the decisive arguments in favor. The first reason for skepticism is that this would indeed be “sporadic”: there are no other instances of word-final /n/ in Maltese that can plausibly be analyzed as retention of nunation. Aquilina (1987) himself suggests only one, namely the (now largely obsolete) item għefien ‘chaff; dirt’, which seems to correspond to Classical Arabic ɣafāʔ (ɣafāʔan with accusative case and nunation) or the alternative form ɣafā (indeclinable ɣafan with nunation), also meaning ‘chaff’. But this is completely implausible.8 As a concrete, noncount, lowfrequency noun, this item will rarely have been indefinite, and thus rarely have

|| 8 See also Zammit (2014: 35) for a critique of Aquilina’s (1987) almost exclusive use of dictionaries of Classical rather than dialectal Arabic for the relevant etymologies in his dictionary of Maltese.

58 | Christopher Lucas and Michael Spagnol

carried nunation in the first place, certainly not frequently enough for it to plausibly be reanalyzed as part of the stem also in definite contexts. If għefien indeed straightforwardly derives from ɣafāʔ ~ ɣafā, this must therefore rather be an instance of the more general Maltese-internal phonological process of /n/ addition that is the primary topic of the present work (but see Section 6 for evidence that għefien in fact derives from a different item in which /n/ is part of the root). This brings us to the second reason for skepticism regarding Aquilina’s claim about xejn, namely that xej’ (without the final /n/) is both a frequently occurring variant form, and the expected Maltese reflex of *šayʔ without nunation. It also provides the precise phonological context for /n/ addition that we identified above: having a stressed vowel (here diphthong) in final position. Absent positive evidence in favor of the retained nunation etymology, it would therefore seem more parsimonious to analyze this as another regular instance of Maltese /n/ addition. Thirdly, the cognate items to xejn in the most closely related Arabic varieties (urban dialects of Tunisia and western Libya) are not reported to have any final /n/ (e.g. Singer 1984; Yoda 2005; Pereira 2010), despite sharing with Maltese (but apparently no other Arabic varieties) the property of having negative meaning (‘nothing’) when used alone in fragment answers (Borsley and Krer 2012). Finally, there is in fact one item in present-day Maltese whose etymon featured nunation, namely the negative determiner ebda ‘no…, not a single…’. Despite its innovative function (apparently not shared by any Arabic variety),9 this item clearly derives from Arabic *ʔabadan ‘(n)ever’, with nunation and accusative case. This is proven by its survival with the same meaning ‘never’ in the folk saying cited by Aquilina (1987: s.v. ebda) and reproduced in (9). (9)

Maltese folk saying Ħanin-i mar Buġebda, u minn hemm ma CONJ from there NEG darling-1SG go.PRF.3SG.M PN jiġi ebda! come.IMPF.3SG.M never ‘My darling went to Buġebda and is never coming back!’ [Aquilina 1987: s.v. ebda]

This item ebda shows what is in fact the expected reflex of nunation (and accusative case) in an element of this kind – /a/ < *ā < *an – also evident in the widespread dialectal Arabic greetings (yā) hala < *ʔahlan (Piamenta 1996: 130)

|| 9 For the development ‘never’ > negative determiner, compare the archaic/dialectal English negative determiner nary < never.

Nunation from Arabic to Maltese | 59

and marḥaba < *marḥaban (Stokes 2020), as well as Tunis Arabic dīma ‘always’ < *dāʔiman. From this point of view, retention of nunation as /n/ in xejn is thus doubly unexpected. On the other hand, it is true to say that a form of nunation in which the /n/ is retained does survive in a number of Arabic dialects: this is the so-called “dialectal tanwīn” illustrated in (9) (see Stokes 2020 for an overview and historical analysis).10, 11 (10)

Najdi Arabic ǧā-na ḥarbiyy-in Harbi-NUN come.PRF.3SG.M-1PL ‘A tall Harbi man came to us.’

ṭuwīl tall [Ingham 1994: 48]

As Stokes points out, however, in all the dialects that exhibit dialectal tanwīn of this type,12 it appears only on indefinite nouns that have some following attributive element, typically an adjective, as in (9), or occasionally a relative clause or similar element. These noun phrases are then typically interpreted as having specific indefinite reference. These conditions will rarely, if ever, hold for uses of an indefinite pronoun like xejn. If we wish to maintain an analysis of the /n/ in xejn as fossilized nunation, therefore, we will not be able to argue that it is an instance of dialectal tanwīn. Instead, we will have to argue, as Aquilina does, that this is a retention of nunation of a Classical type. This is a difficult argument to make, as it is now very widely acknowledged that Classical Arabic can-

|| 10 All Arabic dialects which are spoken in a diglossic relationship with Standard Arabic (i.e. the great majority) have in their lexica a number of borrowings from Standard Arabic. Some of these items are borrowed with nunation, e.g. šukran ‘thank you’, abadan ‘(n)ever’, ʕafwan ‘sorry’. Since Maltese has not been in a diglossic relationship with Standard Arabic since the medieval period, such items are unsurprisingly absent. 11 This feature is absent in contemporary Maltese, and there is no evidence of it in any historical sources. Corriente and Vicente (2014) claim to have found an instance in the fifteenthcentury poem Il-Kantilena, but their analysis is unsustainable. The phrase in question is betiragin mucsule, which all agree translates something like ‘with wet steps’. Corriente and Vicente appear to think that tiragin should be analyzed as the collective ‘steps’ (contemporary Maltese taraġ) plus dialectal tanwīn (-in). But then the final vowel of the agreeing passive participle mucsule (contemporary Maltese maħsula) is unexplained. The correct analysis is instead that tiragin as a whole represents the broken plural form of this item (contemporary Maltese turġien), with which the passive participle regularly agrees by adding the feminine singular suffix -e (contemporary Maltese -a). 12 There are a few highly conservative dialects spoken in the Tihama region of Yemen and southwestern Saudi Arabia which exhibit a form a nunation with a distribution similar to that of Classical Arabic (Behnstedt 2016: 64–67).

60 | Christopher Lucas and Michael Spagnol

not be the ancestor of the contemporary Arabic dialects (and Maltese), and Classical-style nunation, while safely reconstructible to Proto-Arabic (Huehnergard 2017: 20), is already absent13 in the attested Old Arabic inscriptions in the Safaitic script, which likely date to between the 1st century BCE and the 4th century CE. There are thus a number of reasons to be skeptical about Aquilina’s etymology for xejn. The reasons to think that it is nevertheless almost certainly correct are as follows. The key piece of evidence is that, although it is extremely rare, optional retention of nunation with cognates of xejn is in fact attested in at least some Arabic varieties (or was until the mid-20th century). Thus Prémare (1995: 247) reports a form šāyən as being attested in the meaning ‘anything, whatever, nothing’ in certain traditional registers of Moroccan Arabic such as proverbs and oral literature, while Souag (2020: 8) gives the same form for the Algerian coastal dialect of Dellys. These are both varieties for which an explanation for the final /n/ in terms of Classical-style nunation seems inevitable. At the other end of the Arabic-speaking regions, Johnstone (1967: 117) observes that in eastern Arabian dialects “šayyin occurs rather as an occasional variant of šayy in non-final positions”, though the reference to “non-final positions” may be an indication that in this case we simply have an example of the aforementioned dialectal tanwīn. The analysis of final /n/ in xejn as retention of Classical-style nunation is also supported by several other considerations. First, note that, in marked contrast to għefien ‘chaff’ discussed above, xejn is a generalizing indefinite pronoun restricted to negative polarity contexts. That is, it occupies the extreme indefinite end of a continuum of (in)definite contexts. This does not of course guarantee that a marker of indefiniteness such as nunation should be retained in this context, but it is consistent with it being the final context to lose nunation as a general marker of indefiniteness. Second, xejn is listed as having final /n/ (and no /n/-less form is noted) in the early dictionaries of Vassalli (1796) and de Soldanis (2016 [1766]). Of course, this does not rule out the possibility of this being an early instance of phonological /n/ addition in Maltese, but note that the opposite is true of the item whose basic form today is ħdejn ‘near’: this is listed by Vassalli (1796) only as ħdej (see Section 6.4 for more on this item). This discrepancy is consistent with the /n/ in xejn being a retention rather than an innovation. The final piece of evidence supporting the nunation analysis of xejn concerns the etymologically related item x’inhu ‘what is…’. This item has cognates || 13 Though with some vestigial occurrences: see Al-Jallad (2015: 69).

Nunation from Arabic to Maltese | 61

in a wide range of Arabic varieties, including most Western dialects, to which Maltese is most closely related, as well as Sudanese, Iraqi and Gulf dialects. In all of these it functions simply as ‘what’, without the predicative function carried by the Maltese reflex, which was nevertheless clearly the function of the ancestral form. The basic etymology of x’inhu is uncontroversial: *ʔayy šayʔ hū ‘which thing is it’. Formulated like this, however, the medial /n/ is left unexplained. Thus, in a comprehensive survey of the various instantiations of the meaning ‘what’ across the Arabic dialects, Behnstedt and Woidich (2021) observe that seemingly all who have touched on the matter are in agreement that the medial /n/ in x’inhu and its cognates must therefore be traced to nunation on *šayʔ. As Behnstedt and Woidich point out, however, this is far from the only possible analysis.14 Indeed, taken in isolation, a nunation analysis here should be suspicious for the same reasons cited above for xejn: an analysis in terms of dialectal tanwīn is untenable, the Arabic dialects are not descended from Classical Arabic, and nunation is already absent or vestigial in the earliest pre-Islamic Arabic inscriptions. But it is no longer necessary to consider the etymologies of xejn and x’inhu in isolation. When we consider them together we see that they are mutually reinforcing, making inevitable the conclusion, however surprising it may be, that the /n/ in both items is a reflex of nunation of the kind found in Classical Arabic. Note that this conclusion in no way forces us to accept the traditional idea that the Arabic dialects (and Maltese) are in fact directly descended from Classical Arabic. The evidence against this is by now overwhelming (see, e.g., Retsö 2011, 2013; Huehnergard 2017 for some of the key arguments). The significance of our finding here is rather that it constitutes fresh evidence of just how similar the actual ancestor of the Arabic dialects (and Maltese) was to Classical Arabic, especially as regards nominal morphology. In particular, it shows that Owens (1998) is wrong to suggest that evidence of nunation in the Arabic dialects is restricted to dialectal tanwīn (or “linker -Vn” as he calls it), and that dialectal tanwīn is a separate development to the nunation of Classical Arabic. Instead, it must be the case that the system of nunation found in Classical Arabic in fact approximates the system of Proto-Arabic, and both dialectal tanwīn and the /n/ in xejn and x’inhu must be relics of this system.

|| 14 In the context of Maltese x’inhu specifically (though with reference to the ancestral Arabic situation), Ambros (2006: 70) suggests analogical extension of the medial /n/ from min hu ‘who is…’ as an alternative to the nunation retention analysis. Whether or not we accept this proposal, it is clear that the use of inhu as a predicator with kif ‘how’, jekk ‘if’ and other items, which is unique to Maltese, is indeed an instance of analogical extension from x’inhu.

62 | Christopher Lucas and Michael Spagnol

A final issue to be dealt with here concerns the variant form xej’ (without final /n/), as illustrated in (11). (11)

Maltese … [ji]en m’-għand-i 1SG NEG-POSS-1SG ‘… I have nothing left.’

xej’ nothing

aktar. more [Korpus Malti v3.0 news81770]

What is the historical link between xej’ and the full form xejn? There are two possibilities, which are not in fact mutually exclusive. The first is that the form xej’ derives from xejn and represents one instance of the wider process of optional /n/ deletion discussed in Section 3 in relation to (pseudo)duals as well as fejn ‘where’ and bejn ‘between’. The second is that Maltese has always had an /n/-less form of this item, as we see in Libyan and Tunisian Arabic, alongside the form with unexpected retention of nunation, the latter merely happening to conventionalize at an early stage as the default form. Either or both of these scenarios are compatible with the fact that Vassalli (1796) and de Soldanis (2016 [1766]) list only forms with /n/ in their dictionaries.

5 Motivations for /n/ addition We saw in Section 3 that /n/ is unique among the consonants of Maltese in its word-final behavior, in virtue of its frequent alternation with zero in the preRomance, Arabic-derived lexicon. This is why it is the natural choice for Maltese speakers who feel the need to close word-final stressed open syllables (including in monosyllabic items). But why should Maltese speakers have felt this need? The first issue to consider in addressing this question is whether the phonological conditions under which we observe word-final /n/ addition can be sufficiently narrowly defined as to show that this is an exceptionless change of the Neogrammarian type. If this were the case, then we would have a straightforward answer to the question above: word-final stressed open syllables became ungrammatical and had to be repaired. There seems to be no prospect of viewing /n/ addition as a Neogrammarian-type change, however. While nonetymological /n/ is only15 added to word-final stressed open syllables, there are a number of instances where this phonological environment conspicuously fails to trigger /n/ addition.

|| 15 See Section 6 for a discussion of the apparent exceptions to this claim listed in Section 2.

Nunation from Arabic to Maltese | 63

Most notably, there are the Arabic-derived monosyllabic items (mostly function words) listed in Table 6 (together with their token frequencies in the Korpus Malti v3.0).16 Table 6: Frequencies of Arabic-derived vowel-final monosyllabic function words in the Korpus Malti v3.0.

Item

Gloss

Instances per million words in Korpus Malti v3.0

li

complementizer

37,850.2

u

additive conjunction

23,000.32

ta’

possessive marker

20,446.14

ma

negator

6,492.03

se

future marker

3,763.73

jew

disjunctive conjunction

3,358.88

xi

indefinite determiner

2,724.03

hu

3SG.M pronoun

1,556.04

ma’

‘with’

1,427.11

fi

‘in’

1,283.17

hi

3F.SG

741.34

bla

‘without’

321.2

le

‘no’

260.82

sew

‘truth; honest; correctly’

223.48

ġo

‘inside’

71.25

ħu

‘brother’

11.69

In addition to these, there are a very few Arabic-derived items with more than one syllable in which the final syllable is both open and stressed. One such is the predicator inhu [ɪ'nʊ] (discussed in Section 4), as well as its feminine form

|| 16 As well as the items listed in Table 6, there are a few verbs with this phonological profile in their 3SG.M or 3PL perfect forms, e.g. ta(w) ‘he (/they) gave’, ġie/ġew ‘he (/they) came’, as well as the SG.M active participle ġej ‘coming’. These are not included in Table 6 as they are plausibly explained away as counterexamples to a putatively exceptionless rule of /n/ addition: these forms participate in regular inflectional paradigms, in which, for example, 3SG.M perfect is signaled by the bare (unsuffixed) root. That /n/ cannot be added in this context is reinforced by analogy with hundreds of other verbs where /n/ is also not part of the inflection for the 3SG.M perfect.

64 | Christopher Lucas and Michael Spagnol

inhi [ɪ'nɪ]. /n/ is never added to these.17 In addition, there is the whole class of agent nouns or active participles with a weak final root consonant. These items end in a stressed diphthong -ej [ɛɪ], cognate with Classical Arabic -āʔ. Examples include għaddej ‘passing’, mexxej ‘leader; one who walks a lot’, and xerrej ‘buyer’. There is also Mulej ‘Lord (God)’, in which the final diphthong is an innovation of Maltese (compare Tunis Arabic mūla ‘lord, owner’; Singer 1984: 225) – perhaps a backformation from Mulejja ‘my Lord’ – but which nevertheless behaves similarly to the class of participles in never adding /n/. Moreover, it is certainly not the case that all loans into Maltese with the relevant phonological profile (having word-final stressed open syllables) receive an additional final /n/. Unsurprisingly, higher-register or learned items, and items which are known by their users to have (near-)identical counterparts in other European languages (e.g. ineffabilità ‘ineffability’), never take an additional /n/. Similar considerations presumably underlie the fact that only some speakers (sometimes) add /n/ with the items listed in Table 3, and that /n/ addition with these items is considered low-register or sub-standard. But note that there are also very frequent, basic loans to which /n/ is never added by any speakers as far as we are aware, for example diġà ‘already’ and its truncated form ġa, as well as nursery forms such as pipì ‘urine’ and kokò ‘faeces’. It cannot therefore be argued that /n/ addition only fails to apply with the relevant loanwords when these are high-register or learned items. Maltese /n/ addition is therefore not an exceptionless change of the Neogrammarian type. But neither is it totally sporadic. How then can we explain the fact that speakers feel the need to add /n/ in some instances, despite the basic acceptability of the relevant phonological sequences? And what factors intervene in the case of those items where /n/ is never added in the same phonological context? We propose that the first question is best answered in terms of the prototypicality of phonological words. As we have seen, while word-final stressed open syllables do occur in the pre-Romance portion of the Maltese lexicon, words containing such syllables make up a tiny proportion of the overall whole; that is, they have very low type frequency. As such, they are poor exemplars of the prototypical Maltese phonological word. While such items

|| 17 One can, however, make a similar argument here as in fn. 16 for verbs: inhu is an inflecting form (SG.M: inhu, SG.F: inhi, PL inhuma) in which the inflecting element is identical to the 3rd person pronouns. Since the singular pronouns hu and hi do not, for whatever reason, take an additional final /n/, it is not surprising that the homophonous (and etymologically related) inflections of inhu do not take an additional final /n/ either.

Nunation from Arabic to Maltese | 65

have very low type frequency, their token frequency is conversely very high in general, as shown in Table 6. This inverse relationship between type and token frequency is unlikely to be coincidental. As has been clear since at least Hooper (1976), high token frequency has the effect in diachrony of inhibiting analogical change. It also promotes irregular, reductive phonological change, which is why, crosslinguistically, the lexical items with the highest token frequency typically display a range of irregularities (i.e. they have features which are rarely found in the lexicon as a whole). In a discussion of the productivity of competing morphological patterns, Bybee (2001: 118–130) shows that new additions to the lexicon are assimilated to inflectional classes with the highest type frequency, and that high token frequency of an existing pattern actually serves to inhibit its availability as a model for inflectional class assignment (see also Albright 2002 for a similar finding). We suggest that the same principles apply to the kind of non-Neogrammarian sound change under investigation here. When native speakers of Maltese started to be confronted with increasing numbers of words (principally loanwords) which exhibit the very low type-frequency property of having a word-final stressed open syllable, they felt pressure to adapt the phonology of these items to something more prototypical, despite the fact that such sequences do occur in the pre-Romance lexicon. They chose /n/ addition as the means of adaptation for the reasons discussed above. An analogous process of widespread but non-systematic adaptation of loanwords from non-prototypical (but still grammatical) phonological patterns can be found in the history of English.18 As Svensson (2004) shows, English words of more than one syllable loaned from French within the last few centuries exhibit a very gradual tendency towards adaptation of the stress pattern – from the final stress typical of French to the initial stress typical of Germanic languages – a process which appears to still be ongoing, despite the fact that final stress can in no way be described as ungrammatical in present-day English (precisely because of the significant number of French loanwords stressed in this way). For example, British English has ámateur and déficit but cigarétte and colonnáde, while British English tends in general towards greater Germanicization of such loans than does North American English, as can be seen in the differing pronunciations of words such as garage, massage and debris, and names such as Bernard. Although the strategies for improving the phonology of loans differ in the English and Maltese cases, we nevertheless witness directly parallel processes: large numbers of loanwords with non-prototypical phonolo-

|| 18 We are grateful to Steve Rapaport for pointing out to us this parallel.

66 | Christopher Lucas and Michael Spagnol

gy are absorbed; because their phonology is not actually ungrammatical in the recipient language they are non-systematically adapted to a more prototypical phonology; and since many remain unadapted, these have the effect of making the relevant pattern less marginal than it was prior to the influx of loanwords. We now turn to the question of what factors intervene in the case of those items with the relevant phonology where /n/ is never added. Let us note first of all that in the preceding paragraphs we have proceeded as if the phonological process of /n/ addition is essentially restricted to loans from Romance (and English). Table 1 shows that this is not accurate in absolute terms. However, once we set aside xejn (which we saw in Section 4 is a retention), as well as dan/din/dawn and ħdejn (which we will see in Section 6 likely added /n/ for independent reasons), the restriction to loanwords comes closer to being total. Nevertheless, it is not total, and it would be surprising from an acquisitional point of view if it were, given that not all speakers can be expected to identify a word on first exposure as loaned rather than inherited. Instead, we suggest that the apparent bias of /n/ addition towards loans is in fact epiphenomenal. It is not entirely clear what the precise mix of factors is that prevents /n/ from being inserted across the board in word-final stressed open syllables, but we can make the following suggestions. First, as we saw above, it is to be expected that high-register loanwords and loans that are known by speakers to have close equivalents in locally prestigious languages will resist phonological adaptation. This phenomenon can be observed in high-register British English where one can observe (attempts at) non-native nasal vowels in French loans such as genre and restaurant. At the other end of the register scale, the failure of nursery forms (such as pipì ‘urine’ and kokò ‘faeces’ referred to above) to add a final /n/ can be understood in terms of a universal preference in such forms for CV syllables (and reduplication; Jakobson 1971: 25). More generally, since we have established that /n/ addition is not an exceptionless Neogrammarian sound change, it should probably instead be seen as a series of analogical extensions, whose continued semi-productivity is made possible by the fact that /n/ addition is optional for a number of items, for at least some speakers (see Tables 2 and 3). Once cases of the sort discussed in the previous paragraph have been excluded, as a first approximation we can propose that a given item will resist this analogical spread of /n/ addition in proportion to its token (not type) frequency. As noted above, Bybee (2001: 124– 126) shows that lexical items with high token frequency “achieve a certain autonomy from related forms”, plausibly explaining the /n/-resisting behavior of the items in question. As can be seen from Table 7, most, but not all, of the /n/-

Nunation from Arabic to Maltese | 67

resisting items (italicized) are more frequent than the forms which do take /n/ (bold), and vice versa. Table 7: Items with at least one instance per million in the Korpus Malti v3.0 that have nonetymological /n/ or a word-final stressed open syllable (excluding verbs, demonstratives and ħdejn).

Item

Gloss

Instances per million words in Korpus Malti v3.0

li

complementizer

37,850.2

u

additive conjunction

23,000.32

ta’

possessive marker

20,446.14

ma

negator

6,492.03

se

future marker

3,763.73

jew

disjunctive conjunction

3,358.88

xi

indefinite determiner

2,724.03

hu

3SG.M pronoun

1,556.04

ma’

‘with’

1,427.11

fi

‘in’

1,283.17

hi

3SG.F

741.34

bla

‘without’

321.2

le

‘no’

260.82

diġà

‘already’

255.68

sew

‘truth; honest; correctly’

223.48

inhi

3SG.F predicator

181.73

inhu

3SG.M predicator

154.85

ġo

‘inside’

71.25

xulxin

‘each other’

95.12

Ġesù (/Ġisun)

‘Jesus’

34.03 (/ 0)

ġuri (/ġurin)

‘jury’

28.13 (/ < 1)

ħu

‘brother’

11.69

blu (/blun)

‘blue’

11.5 (/ < 1)

kafè (/kafena)

‘coffee (bean)’

11.27 (/ < 1)

ġa

‘already’

8.93

kju (/kjun)

‘queue’

5.69 (/ < 1)

68 | Christopher Lucas and Michael Spagnol

Item

Gloss

Instances per million words in Korpus Malti v3.0

owkej (/owkejn)

‘OK’

~5 (/ < 0)19

sufan

‘sofa’

2.15

Ġaħan

name of folklore character

1.73

Since not all the /n/-resisting forms are more frequent than all the forms which do add /n/ (viz. ħu ‘brother’ and ġa ‘already’), token frequency is unlikely to be the whole story of what promotes or inhibits /n/ addition. An alternative proposal, which seems to offer a closer fit with the data, is that it is especially those items which rarely attract nuclear or contrastive stress – that is, items whose status as independent phonological words is relatively doubtful – that resist /n/ addition. This looks to be a promising line of inquiry, but we leave a more thorough investigation of this proposal to future work. For now, we turn in the following section to a discussion of the etymologies of the individual items in Tables 1–5 (other than xejn, which was dealt with in Section 4).

6 Etymologies of individual items 6.1 Straightforward cases The etymologies of the following items are uncontroversially as listed in Tables 1– 3 and their final /n/ is straightforwardly the result of /n/ addition to close a final stressed open syllable: skrun ‘propeller’, sufan ‘sofa’, ormajn/armajn ‘too late’, gabrijolin ‘cabriolet carriage’, perun/pirun ‘something of great value’, biskwin ‘biscuit porcelain’, paspartun ‘cardboard picture frame’, surtun ‘surtout coat’, blu(n) ‘blue’, ġuġù(n) ‘jujube’, kangarù(n) ‘kangaroo’, kju(n) ‘queue’, owkej(n) ‘okay’, referì(n) ‘referee’, sprej(n) ‘spray’, stju(n) ‘stew’, xampù(n) ‘shampoo’, Ġisun (Ġesù) ‘Jesus’, tarì(n) ‘a pre-modern currency denomination’.

|| 19 This item is mostly spelled in the corpus, but a significant proportion of the sentences containing this string are English, not Maltese.

Nunation from Arabic to Maltese | 69

6.2 Cases with an intermediate stress shift Several items in Tables 1–3 must have undergone a two-stage adaptation process. This can be seen most clearly with kaki(n), borrowed from English khaki as kaki, with a stressed, long initial vowel. The first stage in the adaptation process is that, surprisingly, stress is shifted from the first syllable – a pattern with high type frequency in the Arabic-derived lexicon – to the final, open syllable – resulting in a pattern whose type frequency, as we have seen, is so low in the Arabic-derived lexicon that speakers often feel the need to repair it by adding /n/. So why did speakers not simply retain initial stress here? The answer must be that by the time khaki was borrowed (not earlier than the late nineteenth century), sufficient Romance items with a final stressed open syllable had already been borrowed, and not (yet) repaired by /n/ addition, that speakers came to feel that a prototypical loanword (as opposed to a fully integrated Maltese word) should have a final stressed open syllable, and they then adjusted the stress of kaki – to kakì – accordingly. We can draw a parallel here with the way in which English verbs such as download are adapted once borrowed into Maltese, namely with gemination of the initial consonant, suffixation of a palatal glide to the stem, and conjugation according to the weak-final verbal inflectional class, as in iddawnlowdjajna ‘we downloaded’ (see Mifsud 1995; Lucas and Čéplö 2020: 277–279). These adaptations bring such loaned verbs into line not with inherited, Arabic-derived verbs, but with the highest type-frequency verbs loaned from Italo-Romance. In both this case and that of kakin, we see that, while loaned material is adapted to bring it into closer conformity with the Arabic-derived lexicon (viz. assignment to the weak-final conjugation class and /n/ addition, respectively), the mixed nature of the Maltese lexicon is such that loanwords come to have their own prototypical features, distinct from what is prototypical of Arabic-derived items, and more recent loans are adapted to conform more closely specifically to the loanword prototypes (viz. initial gemination plus glide insertion, and stress shift, respectively). In addition to kaki(n), the items from Tables 1–3 that must have (optionally) undergone this stress shift as a prerequisite for (optional) /n/ addition are serkin ‘sulky cart’ and barbikju(n) ‘barbecue’. The items ġurin ~ ġurì ~ ġuri ‘jury’ and baxan ~ baxà ~ baxa ~ paxa ‘Pasha’ represent a similar case, except here the doublets with final/non-final stress (ġurì vs. ġuri and baxà vs. baxa) result not from an intra-Maltese adaptation process, but from having been borrowed separately from two different sources: English jury and Pasha with initial stress, and Italian giurì and bascià with final stress. It is the latter, with the expected final stressed open syllable that give rise to the forms ġurin and baxan with /n/ addition.

70 | Christopher Lucas and Michael Spagnol

6.3 Table 4 items As far as we are aware, there are no attestations of the base forms of the items in Table 4 – abbati ‘altar boy’, kafè ‘coffee’, te ‘tea’ and kawċù ‘hose pipe’ – that feature an added /n/. But with each of these items a stem-final /n/ is (at least sometimes) inserted when number suffixes are added: abbatini ‘altar boys’, kafena ‘coffee bean’, tenijiet ‘teas’ (nonstandard; standard form tejiet), kawxuna/gawxuna/gamxuna ‘horse’s over-reach boot’. This is best understood as the suffixed and unsuffixed forms exhibiting different layers of borrowing and adaptation. When originally borrowed, kafè, te and kawċù must have undergone /n/ addition in the speech of at least some Maltese speakers (who also made other adaptations to the phonology of kawċù, including substituting /ʃ/ for /tʃ/). Kawxuna, kafena and tenijiet are the outcome of the singulative morpheme -a and the plural morpheme -ijiet being suffixed to these forms that have undergone /n/ addition. The present-day base forms without final /n/ must then represent reborrowings from the respective source languages. Abbati(ni) likely represents a slightly different case, in that both Sicilian and Italian attest diminutive forms of the base a(b)bati ‘altar boy’ with the suffix -ino/-inu, giving abatino/abbatinu ‘little altar boy’, with plural a(b)batini. It seems that both the base and the suffixed forms were borrowed into Maltese, with the base form conventionalizing as the singular and the diminutive form conventionalizing as the plural. A parallel case can be found in uffiċċju ‘office’, whose plural for some (perhaps most) speakers is not the expected uffiċċji, but uffiċini, which is clearly originally the plural of the diminutive form uffiċina.

6.4 Arabic-derived items 6.4.1 xulxin Stumme (1904: 122) links this item, with a dialectal variant xilxin (Aquilina 1987: s.v. xilxin), to an ultimate etymon *šayʔ li-šayʔ ‘(one) thing to (another) thing’. This seems to be correct, but the vowel qualities suggest specifically *šī l-šī as the immediate phrase from which this item was lexicalized. As far as we are aware, no Arabic dialect forms reciprocal pronouns in this way (the usual main etymon being *baʕḍ ‘some, a part’). Additionally, while *šī as an indefinite determiner is very widespread in Arabic dialects and Maltese, it cannot typically stand alone with a pronominal function (i.e. its syntax is closer to that of English every than to all). This points to *šī l-šī being a partial calque on an ItaloRomance form similar to Italian l’un l’altro ‘each other’ (lit. ‘the one the other’).

Nunation from Arabic to Maltese | 71

Whatever the precise details of the etymology, there is no doubt that the final /n/ in this item is a straightforward case of /n/ addition, since once this phrase underwent univerbation it would have been left with a final stressed open syllable, and the word is cited – without /n/ – as xilxi in the Mezzo Vocabolario of c.1765 (Cassola 1996: 170; Kontzi 1999: 421).

6.4.2 Ħdejn Ħdejn ‘near’ could be seen simply as a straightforward case of /n/ addition. But there is reason to think in this case that other factors were responsible for the addition of /n/ than just the presence of a final stressed open syllable. What is clear about this item is the following: it is cognate with Classical Arabic ḥiðāʔ ‘opposite to’ and Tunis Arabic ḥða and baḥða ‘near’; the original form of this item in Maltese is ħada, also meaning ‘near’; the form ħada only appears with nominal complements, not pronominal suffixes; and with pronominal suffixes the form is (almost always20) ħdej. This form is somewhat unexpected. The expected form, on the basis of comparative Arabic and Maltese-internal evidence would be ħdie-. Compare Tunis Arabic ḥdā- (Singer 1984: 627) and how Maltese -na and -ha become -nie- and -hie- before a suffix pronoun, as in jibgħat-ha ‘he sends it’ (send.IMPF.3SG.M-3SG.F) versus jibgħat-hie-lu ‘he sends it to him’ (send.IMPF.3SG.M-3SG.F-DAT.3SG.M). Note, however, that Maltese wara ‘behind’ behaves in the same anomalous way, with the suffixed form being waraj- instead of the expected wara- [wɐˈrɐː] (compare Tunis Arabic uṛā-; Singer 1984: 627). These diphthongal forms presumably arose by analogy, though with which item(s) precisely is not clear. Possible candidates are lej- and the presuffixal form of għal ‘for’, which is għali- [ɐˈliː], but which must earlier have been għalej- (compare Classical Arabic ʕalay- and Cairo Arabic ʕalē-, not ʕalī-). In any case, the form ħdej used independently in place of ħada must have been a backformation from the suffixed forms. Regarding /n/ addition with the independent form, note that once we have a suffixing form lej- and an independent form lejn conceived of as allomorphs of a single item (see Section 3), along with a suffixing form ħdej- and an independent form ħdej, there is clearly strong analogical pressure to create an independent form ħdejn. We take this analogical pressure to have been the primary factor promoting /n/ addition in this case.

|| 20 The search ħdej[ja,k,h,ha,kom,hom] returns over 3000 matches in the Korpus Malti. Ħdejn[i,ek,u,ha,kom,hom] returns 3 distinct matches.

72 | Christopher Lucas and Michael Spagnol

6.4.3 Near demonstrative dan(a) (M), din(a) (F), dawn(a) (PL) All who have addressed the matter agree that these items derive from the Arabic demonstrative elements ðā (M), ðī (F), ðū (PL). Aquilina (1987: s.v. dan) notes of dan that “[f]inal n is often omitted in the spoken language”, though we should probably view this as retention of unsuffixed da rather than secondary loss of /n/. The disagreement among etymologists concerns the provenance of the final /n/. Thus Barbera (1939b: 108) points out that a suffix -ni is added to both personal and demonstrative vowel-final pronouns in Sicilian21 (e.g. chistu ‘this’ > stuni) and suggests that something like this process was replicated for the Maltese demonstratives, resulting in suffixation of /n/ to these. Saydon (1965: 72) asserts that “[n]othing is farther from the truth” than Barbera’s proposal, and instead highlights the widespread addition of /n/ in Maltese, citing the instances quoted in Section 1, and arguing that “da becomes dan, independently of any Sicilian influence.” Aquilina (1987: s.v. dan) also disagrees with Barbera’s suggestion of influence from Sicilian, suggesting instead that the presence of /n/ in these items is in fact a pre-Maltese dialectal Arabic innovation, citing, from O’Leary’s (1923: 160) comparative grammar of Semitic languages, such forms as “Tripoli” hädūn and “Morocco” hādūn. Such forms are indeed attested in North African Arabic varieties (e.g. Jewish Tripoli; Yoda 2005: 127), but only as plurals, where -ūn is a plural suffix. A final /n/ in the singular demonstratives is unknown in any Arabic variety beyond Maltese (see Magidow 2013). Aquilina’s suggestion that this feature is inherited is therefore untenable.22 A crucial point to note here is the fact that, unlike all the other instances of word-final /n/ considered in this article, with these demonstratives it is not merely /n/ that alternates with zero word-finally but also -na. The one discussion of these demonstratives to date that takes account of this fact is Fischer’s (1959: 67–71), which draws a crucial parallel with the various Maltese forms

|| 21 Rohlfs (1968: 468–469) shows that optional suffixation of -ni (or -ne) to a range of functional items with a final stressed open syllable is rather widespread across Italo-Romance varieties going back at least to the thirteenth century. The fact that this suffix is itself always vowel-final (and that it targets especially high token-frequency items) means it cannot be seen as the source of Maltese /n/ addition in general. But the obvious similarity between Italo-Romance -ni/-ne suffixation and Maltese /n/ addition presumably played at least some role in establishing the latter as a stable feature of Maltese. 22 O’Leary (1923: 160) suggests that /n/ in Maltese demonstratives is cognate with demonstratives also ending in a final nasal consonant in Aramaic, Ethiopic, and Modern South Arabian, but this also cannot be correct, given its absence in Arabic varieties beyond Maltese. See also Hasselbach (2007).

Nunation from Arabic to Maltese | 73

expressing ‘here’: hawna, hawn, and haw. Fischer’s only mistake is to assume that all three of these forms derive from presentative *hā prefixed to *hunā ‘here’, and that haw must arise from hawn through deletion of the final /n/. As we saw in Section 3, there is no general rule of final /n/ deletion in Maltese: it is entirely restricted to the pseudo(dual) suffix and to the coincidentally similar fejn ‘where’ (and derivatives) and bejn ‘between’. Instead, while hawn(a) clearly does derive from *hā hunā,23 haw should rather be seen as a retention of a form – derived from *hā hu(wa) ‘here he/it is’ – that existed in Maltese from the beginning and which is still to be found in Tunis Arabic (Singer 1984: 259). Aside from on this point, Fischer’s reconstruction seems correct: once we have the semantically near-identical forms haw and hawna alternating, it is natural to view -na as a suffix and to optionally extend this to the demonstrative pronouns, presumably starting with the plural daw, giving the alternation daw : dawna (and then da : dana and di : dina). The optional loss of the final /a/ in both hawna and the newly extended demonstratives (dana, dina, dawna) is unsurprising in view of the crosslinguistic tendency for phonetic reduction of high token-frequency items, and in any case forms part of a wider tendency in Maltese to (optionally) omit a final /a/ in such items, seen also in hemm ‘there’ (< hemma < *θamma) and jien(a) ‘I’.

6.4.4 Ġaħan and għefien Not enough is known at present about the development of these items to be certain about their etymologies, but the following notes may serve as a starting point for future investigations. In eastern Libyan Arabic, and in all Arabic dialects which preserve stressed short vowels in non-final open syllables, the name of the folklore character under discussion here is ǧuḥa (with initial stress).24 The majority of Western Arabic dialects, including most contemporary Tunisian varieties, which have lost stressed short vowels in non-final open syllables, have žḥa as the regular reflex of ǧuḥa. It is unclear when exactly this loss happened, but it had not yet spread to the Arabic variety that was first brought to Malta, because contempo-

|| 23 This is a pre-Maltese development. Loss of the medial *h gives rise to the diphthong which is preserved in Maltese hawn(a) and regularly monophthongized in Tunisian hūni and Levantine hōn. 24 The local pronunciation of this name in a range of different Arabic dialects can be heard at the following link: https://www.reddit.com/r/arabs/wiki/dialects (accessed 24/01/2022).

74 | Christopher Lucas and Michael Spagnol

rary Maltese retains these vowels; compare Maltese sema ‘sky’ with Tunis Arabic sma. The expected Maltese form of this name would therefore be **Ġoħa, with initial stress. Note that the form of this name in Sicilian and Italian folklore is Giufà or Giucà with final stress, presumably reflecting the adoption of this name from the Western Arabic form žḥa. In addition to the unexpected /a/ in the first syllable, the difficulty with the Maltese form of the name – Ġaħan [ˈdʒɐhɐn] – is that it has initial stress and a short vowel in the second syllable, which matches the expected form **Ġoħa, but suggests that /n/ was unexpectedly added to an unstressed final open syllable. Since there is no evidence for /n/ addition in this context elsewhere in the Maltese lexicon, we must seek an alternative, necessarily speculative, explanation. Our suggestion is that the form Ġaħan is the compromise outcome of competition between the inherited, initially-stressed form **Ġoħa and the Sicilian/Italian form Giufà, which, offering the correct phonological environment of a final stressed open syllable, was supplied with an additional final /n/ when borrowed into Maltese. Għefien presents a somewhat similar problem. We saw in Section 4 that Aquilina’s (1987: s.v. għefien) suggestion that the final /n/ in this item is a retention of Classical Arabic nunation is entirely implausible. What about the rest of his proposed etymology, namely that its etymon is *ɣafā(ʔ)? This looks much more plausible, since the semantic correspondence is perfect: both għefien and Classical Arabic ɣafā ~ ɣafāʔ mean ‘chaff’ (Lane 1863: s.v. ɣfw/ɣfy). There are two problems, however. First, the expected reflex of *ɣafā(ʔ) in Maltese is **għefa, with initial stress. That is, it should not have provided the appropriate phonological context for /n/ addition. The second problem is linked to the fact that orthographic in contemporary standard Maltese represents the outcome of two separate Arabic phonemes: /ʕ/ and /ɣ/. Their merger and subsequent loss (except in orthography) is a relatively recent development of standard Maltese, which has not yet spread to all Maltese dialects (Lucas and Čéplö 2020: 271). At the time that Vassalli wrote his dictionary of Maltese (published 1796), the distinction remained robust and, for the meaning “fieno e paglia che rimane avanti le bestie” [‘hay and straw that remains in front of livestock’] he gives only “∩fŷn” (i.e. /ʕfɪːn/) not “ᴟfŷn” (i.e. /ɣfɪːn/), as would be expected if this item really does derive from *ɣafā(ʔ). This is surprising, since the root √ʕfn is apparently everywhere else in Arabic associated only with the semantic field of mould and decay. In Maltese there has clearly been a shift, however. Aquilina (1987: s.v. għafen) gives the following meanings for the verb għafen: “l. To become desiccated, dry (leaves of wheat, etc.) […] 2. […] To grow more than is normal for a good crop (wheat, corn). 3. […] To grow too densely together leaving no air or space in between causing damage to leaves.” It appears that there

Nunation from Arabic to Maltese | 75

has been a semantic broadening process here, from the specific problem of mould (in crops), to a range of problems that can affect crops, including dryness. Against this background, one would expect Arabic ʕafān ‘mouldiness’ to become something like għefien ‘poor/dry harvest’, and a further shift from there to ‘chaff’ is not implausible. Clearly there is too much uncertainty here to make confident pronouncements, but the evidence suggests that we cannot simply assume either that għefien derives from *ɣafā(ʔ), or that it has undergone the same process of /n/ addition as the other items discussed in the present article.

6.5 Table 5 items The three items in Table 5 – dawl ‘light’ (< Arabic *ð̣aw < *dawʔ), dikutell ‘stone stood on its narrow dimension’ (< French de côté ‘sideways’), and ġuvintur ‘youth’ (< Italian gioventù ‘youth, young people’) – represent all the Maltese items that we are aware of in which there is an etymologically unexpected additional final consonant other than /n/. All three present the basic phonological context for /n/ addition – a final stressed open syllable – but in each case we can point to idiosyncratic details of each item which likely motivated the addition of the consonants we see, rather than /n/. Beginning with dawl, Aquilina (1987: s.v. dawl) proposes that the origin of the final /l/ is the definite article of the following noun in a synthetic genitive construction. It is hardly plausible, however, that speakers would reanalyze the /l/ in a string such as daw l-qamar ‘light of the moon’ as belonging to the first word, as then they would be forced to analyze the possessor noun as indefinite – an analysis for which there would be no obvious motivation. Instead it seems that here the chief motivation for the addition of some final consonant is the fact that without it this word would have been bi- instead of tri-radical, following the regular loss of final glottal stop from the ultimate etymon *ḍawʔ. This pressure towards triradicalization is realized in the cognate item in most Arabic dialects by gemination of the glide: ḍaww/ð̣aww. In Maltese we can instead triradicalize this item with /n/ addition. Why we have a final /l/ rather than /n/ in dawl can be understood as an ease-of-articulation effect: the pharyngealization spreading across the whole word as a result of the originally emphatic *ð̣, combined with the lip-rounding of the /w/, mean that speakers aiming to produce [n] could plausibly have been perceived as having intended /l/. Aquilina (1987: s.v. dikutell) proposes that dikutell ‘a stone stood on its narrow dimension, cargo carried sideways, the last horizontal wooden strip of a cart’ (with variant forms dukutell and digudell) derives from French de côté ‘sideways, on one side’. This is at least phonetically plausible, and since we are

76 | Christopher Lucas and Michael Spagnol

aware of no better candidate etymon in Sicilian or Italian, we accept this etymology as probably correct. Clearly relevant here is that the additional final /l/ is geminate. The motivation for adding a consonant is of course the final stressed open syllable. The addition of the geminate /l/ rather than any other consonant appears to be conditioned by the vowel quality (presumably perceived by Maltese speakers as /ɛ/) and to constitute a sort of folk etymology, assimilating this loan to the many Maltese words – such as kastell ‘castle’ or martell ‘hammer’ – that have the suffix -ell, these being borrowings of Sicilian or Italian words ending in -ello or -ellu, respectively. Finally, a similar process of folk etymology seems to be behind the addition of /r/ in ġuvintur ‘young people’, whose Italian etymon gioventù again has a final stressed open syllable. At issue here is the fact that a great many Maltese words of Romance origin contain the suffix -ur (often -tur; e.g. pittur ‘painter’, ambaxxatur ‘ambassador’), whose agentive meaning will have made it especially plausible to speakers as the correct termination of a word referring to people.

7 Conclusion In this article we have seen that a non-systematic phonological change – addition of word-final /n/ in the case at hand – can be driven by pressure felt by speakers to make phonologically anomalous words conform more closely to a prototype. We have also seen that a range of factors can enable an item to resist this pressure; most fundamentally high token frequency, or the strongly correlated property of phonological dependence on other lexical items. Since it is generally native lexical items that have these properties rather than loanwords, we have seen that Maltese final /n/ addition is largely, but not exclusively, a process affecting loans. An adequate understanding of this phenomenon has only been possible by means of a detailed examination of the etymologies of the items (apparently) affected. In the course of this examination we have seen that some of the items considered contain a final /n/ for reasons quite different to prototype conformity. The most notable of these is xejn ‘nothing’, whose final /n/ we argued represents a fossilized retention of the kind of grammatical nunation familiar from Classical Arabic but generally completely absent from the Arabic dialects. Thus we have further evidence from these investigations of just how rich the Maltese language can be as a source of data that helps us understand not only the history of Arabic dialects, but the nature of language change, and linguistic cognition, in general.

Nunation from Arabic to Maltese | 77

Acknowledgments: Earlier phases of the research for this article were presented at the 7th International Conference on Maltese Linguistics (Lingwistika Maltija) in Kraków, and the 4th Edinburgh Symposium on Historical Phonology, in July and December 2019. Our thanks are due to the audiences at both conferences for their valuable feedback. Note that in one important respect – concerning the etymology of the final /n/ in Maltese xejn ‘nothing’ – the analysis proposed here is quite different from that presented in 2019. This rethink is in large part the result of crucial input from Lameen Souag, to whom we would like to express our sincere gratitude. We would also like to thank Manwel Mifsud, Olvin Vella and Joseph M. Brincat for alerting us to several examples of Maltese /n/ addition that we had previously overlooked, as well as a number of colleagues who gave us very helpful feedback on an earlier draft via academia.edu. Any errors which remain are entirely our own.

Abbreviations 1, 2, 3 * ** ACC ACT

Ar. CONJ DAT DEF DEM

dim DU ELA

Eng. F

Fr. FUT GEN IMP IMPF

It. M NEG NOM NUN OBL

1st, 2nd, 3rd person reconstructed form unattested form accusative active Arabic conjunction dative definite article demonstrative diminutive dual elative English feminine French future genitive imperative imperfect Italian masculine negative nominative nunation oblique

78 | Christopher Lucas and Michael Spagnol

PASS PL PN POSS PRED PRF PTCP SG

Sic.

passive plural proper name possessive predicator perfect participle singular Sicilian

Electronic resources Korpus Malti v3.0: https://mlrs.research.um.edu.mt/

References Albright, Adam. 2002. The lexical bases of morphological well-formedness. In Sabrina Bendjaballah, Wolfgang U. Dressler, Oskar E. Pfeiffer & Maria D. Voeikova (eds.), Morphology 2000: Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000, 5–15. Amsterdam & Philadelphia: John Benjamins. Al-Jallad, Ahmad. 2015. An outline of the grammar of the Safaitic Inscriptions. Leiden: Brill. Ambros, Arne Amadeus. 2006. Bongornu, kif int? Einführung in die maltesische Sprache. Wiesbaden: Reichert. Aquilina, Joseph. 1987. Maltese-English dictionary. 2 vols. Malta: Midsea Books. Barbera, Giuseppe Maria. 1939a. Dizionario maltese-arabo-italiano, con una grammatica comparata arabo-maltese. Vol. 2: F–L. Beirut: Imprimerie Catholique. Barbera, Giuseppe Maria. 1939b. Dizionario maltese-arabo-italiano, con una grammatica comparata arabo-maltese. Vol. 1: A–E. Beirut: Imprimerie Catholique. Behnstedt, Peter. 2016. Dialect atlas of North Yemen and adjacent areas. (Trans.) Gwendolin Goldbloom. Leiden: Brill. Behnstedt, Peter & Manfred Woidich. 2021. Wortatlas der arabischen Dialekte. Vol. 4: Funktionswörter und Phraseologisches. Leiden: Brill. Blanc, Haim. 1970. Dual and pseudo-dual in the Arabic dialects. Language 36. 42–57. Blevins, Juliette. 2004. Evolutionary phonology: The emergence of sound patterns. Cambridge: Cambridge University Press. Borsley, Robert D. & Mohamed Krer. 2012. An HPSG approach to negation in Libyan Arabic. Essex Research Reports in Linguistics 61(2). 1–24. Bybee, Joan L. 2001. Phonology and language use. Cambridge: Cambridge University Press. Cassola, Arnold (ed.). 1996. Il “Mezzo Vocabolario” maltese-italiano del ’700. Said international. Corriente, Federico & Ángeles Vicente. 2014. Again on Peter Caxaro’s Cantilena: Some remarks from the viewpoint of wider Western Arabic dialectology. Journal of Maltese Studies 28. 75–89.

Nunation from Arabic to Maltese | 79

Fenech, Edward. 1996. Functions of the dual suffix in Maltese. Rivista di Linguistica 8. 89–99. Finlay, W. M. 1803. Arithmetical magazine. New York: G. F. Hopkins. Fischer, Wolfdietrich. 1959. Die demonstrativen Bildungen der neuarabischen Dialekte: Ein Beitrag zur historischen Grammatik des Arabischen. The Hague: Mouton & Co. Hasselbach, Rebecca. 2007. Demonstratives in Semitic. Journal of the American Oriental Society. JSTOR 127(1). 1–27. Holes, Clive. 2001. Dialect, culture, and society in eastern Arabia, vol.1: Glossary. Leiden: Brill. Hooper, Joan. 1976. Word frequency in lexical diffusion and the source of morphophonological change. In Jr William M. Christie (ed.), Current progress in historical linguistics, 95–105. Amsterdam: North Holland. Huehnergard, John. 2017. Arabic in its Semitic context. In Ahmad Al-Jallad (ed.), Arabic in context: Celebrating 400 years of Arabic at Leiden University, 3–33. Leiden: Brill. Ingham, Bruce. 1994. Najdi Arabic: Central Arabian. Amsterdam & Philadelphia: John Benjamins. Jakobson, Roman. 1971. Studies on child language and aphasia. Vol. 114. The Hague: Mouton. Johnstone, Thomas M. 1967. Eastern Arabian dialect studies. Oxford: Oxford University Press. Kontzi, Reinhold (ed.). 1999. Wort und Schrift: Des Kanonikus Fortunato Panzavecchia Bibelübersetzung ins Maltesische, nach den Handschriften des Kathedralarchivs in Mdina. Tübingen: Narr. Lane, Edward. 1863. An Arabic–English lexicon. London: Williams and Norgate. Lucas, Christopher & Slavomír Čéplö. 2020. Maltese. In Christopher Lucas & Stefano Manfredi (eds.), Arabic and contact-induced change, 265–302. Berlin: Language Science Press. Magidow, Alexander. 2013. Towards a sociohistorical reconstruction of pre-Islamic Arabic dialect diversity. The University of Texas at Austin PhD thesis. Mifsud, Manwel. 1995. Loan verbs in Maltese: a descriptive and comparative study. Leiden: Brill. O’Leary, De Lacy Evans. 1923. Comparative grammar of the Semitic languages. London: Routledge, Trench, Trübner & Co. Owens, Jonathan. 1998. Case and proto-Arabic, part II. Bulletin of the School of Oriental and African Studies. JSTOR 215–227. Pereira, Christophe. 2010. Le parler arabe de Tripoli (Libye). Zaragoza: Instituto de Estudios Islámicos y del Oriente. Piamenta, Moshe. 1996. More on the Arabic dialect of the Negev bedouins. Quaderni di Studi Arabi 14. 123–136. Prémare, André-Louis de. 1995. Dictionnaire arabe-français. Vol. 7. Paris: L’Harmattan. Retsö, Jan. 2011. Classical Arabic. In Stefan Weninger, Geoffrey Khan, Michael P. Streck & Janet C. E. Watson (eds.), The Semitic languages: An international handbook, 782–810. Berlin & Boston: De Gruyter Mouton. Retsö, Jan. 2013. What is Arabic? In Jonathan Owens (ed.), The Oxford handbook of Arabic linguistics, 433–450. Oxford: Oxford University Press. Rohlfs, Friedrich Gerhard. 1968. Grammatica storica della lingua italiana e dei suoi dialetti: Fonetica. (Trans.) Salvatore Persichino. Torino: Einaudi. Saydon, Peter P. 1965. Maltese etymological notes. Journal of Semitic Studies 10(1). 67–82. Singe, Hans-Rudolf. 1984. Grammatik des arabischen Mundart der Medina von Tunis. Berlin: Walter de Gruyter. Snelling, Thomas. 1766. A view of the coins at this time current throughout Europe. London: T. Snelling.

80 | Christopher Lucas and Michael Spagnol

Soldanis, Giovanni Pietro Francesco Agius de. 2016. Damma tal kliem Kartaginis mscerred fel fom tal Maltin u Ghaucin. (Ed.) Rosabelle Carabott. Malta: Rosabelle Carabott. Souag, Lameen. 2020. Maltese and North Africa linguistics: Common roots and areal divergence. In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese linguistics on the Danube, 3– 25. Berlin & Boston: De Gruyter Mouton. Stokes, Phillip. 2020. A fresh analysis of the origin and diachronic development of “dialectal tanwīn” in Arabic. Journal of the American Oriental Society 140(3). 637–664. Stumme, Hans. 1904. Maltesische Studien: eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache. Leipzig: Hinrichs. Sutcliffe, Edmund. 1936. A grammar of the Maltese language, with chrestomathy and vocabulary. Oxford: Oxford University Press. Svensson, Ann-Marie. 2004. On the stressing of French loanwords in English. In Christian Kay, Carole Hough & Irené Wotherspoon (eds.), New perspectives on English historical linguistics, vol. II: Lexis and transmission, 225–234. Amsterdam & Philadelphia: John Benjamins. VanDam, Mark. 2004. Word final coda typology. Journal of Universal Language 5. 119–148. Vassalli, Mikiel Anton. 1796. Ktieb il-kliem Malti. Rome: Antonio Fulgonio. Yoda, Sumikazu. 2005. The Arabic dialect of the Jews in Tripoli (Libya): grammar, text and glossary (Semitica Viva). Vol. 35. Wiesbaden: Harrassowitz. Zammit, Martin R. 2014. The Sfaxi (Tunisian) element in Maltese. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 23–44. Berlin: Akademie Verlag.

| Part II: Syntax

Kirsty Azzopardi

The verb sequence V1+V2 in Maltese Abstract: This study is concerned with a construction in Maltese which consists of a sequence of two or more finite verbs. This construction is frequent in both spoken and written Maltese. The present study focuses on sequences consisting of two verbs. A sample of 500 chains of type V1+V2 was collected using the Maltese Language Resource Server (MLRS) corpus of written Maltese. Some of these chains are analyzed in order to shed light on the nature of the verb sequence and to explain the acceptability or unacceptability of some combinations. Keywords: Maltese; morphosyntax; semantics; verb sequence

1 Introduction In Maltese, a sequence of finite verbs, here referred to as ‘verb sequence’ or ‘verb chain’, can often be observed. Verb chains vary in length and while the minimal realization consists of merely two verbs, as in (1), longer verb chains, such as that in (2)1, are also common. (1)

Bd-ew j-istudja-w begin-3PL.PFV 3IPFV-study-PL ‘They started studying.’

(2)

Kon-t be-1SG.PFV n-ipprova 1SG.IPFV-try n-agħmel 1SG.IPFV-do

rid-t want-1SG.PFV n-ara 1SG.IPFV-see (biex in.order.to

n-erġa’ n-ibda 1SG.IPFV-repeat 1SG.IPFV-begin x’ n-aqbad what 1SG.IPFV-start n-aqbad il-ħalliel) 1SG.IPFV-catch DEF-thief

|| 1 The x’ in x’naqbad appears to break the chain. However, this is not the case if we consider xi to be the direct object of naqbad and, moreover, consider the verb chain to be made up of verb phrases rather than verbs, as is suggested in Fabri and Borg (2017). This interpretation of the verb chain will be discussed in Section 2.2. || Kirsty Azzopardi: Institute of Linguistics and Language Technology, University of Malta, Car Park 6, Porta Cabins, L-Imsida, MSD2080, Malta, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-004

84 | Kirsty Azzopardi

‘Once more, I had wanted to start trying to see what I could do (to catch the thief).’ This characteristic feature of Maltese was recognized as early as the 19th century by Vassalli in his Grammatica della lingua Maltese (1827). However, it has since been left in the dark and studies devoted, either in part or totally, to the verb chain in Maltese are very few, namely Maas (2009), Stolz (2009), Peterson (2009), Camilleri (2016), Fabri and Borg (2017) and Azzopardi (2019)2. The aim of this study is twofold. Firstly, it looks at verbs occurring in the sample gathered using MLRS (Maltese Language Resource Server) – 500 chains of the type V1+V2 – and categorizes them according to their distribution relative to the two positions, i.e. verbs that only fill the slot V1 belong to one group, those that occupy the V2 slot belong to another group, while verbs that may occur either as V1 or as V2 make up another group. This study also aims to investigate the role of semantics in influencing or determining the allowed or disallowed combinations. This paper is structured as follows: The terminology used across the literature on the verb chain in Maltese is discussed in Section 2, followed by a brief description of the characteristics of the verb chain in Section 3. The morphosyntactic behavior and restrictions will be discussed in Section 4, while the subsequent section will deal with the adopted methodology. Section 6 is concerned with the analysis and discussion of a class of verbs that seem to be somewhat homogeneous in their behavior. A brief summary of the findings discussed in this paper will then be given in Section 7.

2 Terminological issues3 In this section we will have a look at the terminology used in studies about the verb sequence in Maltese. Talk about terminology is divided in two parts. In the first part we will discuss what the Maltese verb chain has been called and the reasons for this. The second part will deal with the identification of the elements making up the chain. Is it simply a stacking of verbs or are we dealing with something other than verbs here?

|| 2 This paper is in great part based on Azzopardi (2019) and its findings. 3 The terms ‘verb chain’ and ‘verb sequence’ in this study are meant as theory-neutral terms used to refer to chains of the type V1+V2.

The verb sequence V1+V2 in Maltese | 85

2.1 Baptizing the verb sequence A conspicuous feature observed across the literature on the Maltese verb chain is the lack of agreement when it comes to terminology. As Stolz (2009: 138) points out, terminological discord does not suggest “disagreement as to the unit-like nature of the Maltese verb chain”, but the choice of one term over another could hold implications for the relationship between the members of the chain. In Fabri (1993: 203), verb chains are called ‘Multi-Verb-Konstruktionen’ (multi-verb constructions), which Fabri argues contain a verb V1 that ccommands another verb V2, where the subject affix of V2 agrees with a complement, that is a subject or a direct object, of V1. Having a neutral term in view, Fabri and Borg (2017) term the chain ‘verb sequence’, while Stolz (2009) uses the term ‘verb chain’ as a neutral term. Stolz (2009: 175) goes on to say that the verb chain in Maltese is not the same as a complex predicate or as a predicative complex. Despite this, the label Maas (2009) utilizes to make reference to the verb chain is ‘complex predicates’, which are defined as “predicates of a single clause that consist of more than one syntactic word” and which “are from the point of view of the syntactic structure of the sentence equivalent to a monoverbal predicate” (Maas 2009: 113). The elements constituting ‘complex predicates’ are described as either the ‘modificans’, the modifying element, or the ‘modificatum’, the modified element (Maas 2009: 113). In light of this, in the following example qabad is the modificans, whereas jiġri is the modificatum. (3)

Qabad j-iġri 3SG.M.IPFV-run catch:3SG.M.PFV ‘He started running.’

The V1 position, referred to by Maas as the modificans, is termed ‘auxiliary’ in Camilleri (2016). However, Maas (2009: 114) argues against classifying the modificans as an auxiliary, as is the tendency in Indo-European languages, and distances himself from this practice by employing the term ‘coverb’ in relation to those verbal elements that function as modifiers of finite verbs in a complex predicate.

2.2 To be or not to be (merely) a verb Stolz (2009: 138) describes the chain as “any sequence of minimally two verb forms in one and the same utterance”. Furthermore, Stolz (2009: 139) asserts that the verb chain is not “an equivalent of the verbal phrase in terms of constituency”. However, both the V1 and the V2 can, optionally, occur along with their complements (Fabri and Borg 2017: 81), as is apparent in (5).

86 | Kirsty Azzopardi

(4)

Mark mar j-ixtri Mark go:3SG.M.PFV 3SG.M.IPFV-buy ‘Mark went shopping.’

(5)

Mark mar sa-l-Belt Mark go:3SG.M.PFV until-DEF-Valletta j-ixtri erba’ flokk-ijiet 3SG.M.IPFV-buy four shirt-PL ‘Mark went to Valletta to buy some tops.’

Fabri and Borg (2017: 82) observe that “[t]he occurrence of complements is a clear indication that, syntactically, we are not simply dealing with a verb sequence, but with a sequence involving a VP with its verbal predicate head and its (optional) complement”; therefore, the complement cannot be interpreted as an element that breaks the chain, since it does not bring about any differences in syntactic relations.

3 Characteristics of the verb chain In this section we will discuss the verb chain characteristics which have been identified in previous studies. These characteristics are somewhat prominent, albeit sometimes problematic. Notwithstanding, the features discussed here help us determine whether what we are looking at is in fact a verb chain.

3.1 Grammatical agreement A sine qua non of the verb chain is that its members must agree in person, number and gender, and therefore they must have a common argument – the subject. This characteristic is identified by Vassalli (1827) when talking about a pair of verbs. He states that the second verb is governed by the first, and therefore it must agree with the first verb in gender, number and person. The following example, written in modern Maltese orthography, is the one Vassalli (1827: 131) provides to sustain his argument. (6)

J-rid j-iġi 3SG.M.IPFV-want 3SG.M.IPFV-come ‘He wants to come.’

The verb sequence V1+V2 in Maltese | 87

This means that in case of a change in subject, the verb corresponding to the new subject would fall out of the chain (Stolz 2009: 147), as in the following example. (7)

Xtaq-t im-mur n-ara=h wish-1SG.PFV 1SG.IPFV-go 1SG.IPFV-see=3SG.M.OBJ ‘I would have liked to go see him perform.’

j-irreċta 3SG.M.IPFV-act

In the example given above, xtaqt, immur and narah are the verbs comprising the chain, whereas jirreċta cannot be regarded in the same way as the verbs that precede it. This change in subject can be attributed to the encliticized bound pronoun in narah. It is also worth noting in passing that a verb chain in the imperative is also possible, in which case both the V1 and the V2 assume the imperative inflectional form. In such a case, “[a]greement in terms of number is obligatory” (Fabri and Borg 2017: 72). (8)

Morr-u go-2PL.IMP ‘Go study!’

studja-w! study-2PL.IMP

The significance of grammatical agreement can be understood through Stolz’s statement when he asserts that “agreement is one of the formal criteria which guarantee a high enough degree of cohesion among the members of a verb chain to justify their being considered a unit of sorts” (Stolz 2009: 149). Fabri and Borg also consider this criterion to be one of the main defining features of the sequence. It should be underlined that some striking cases of agreement can be identified, since agreement is not always as evident as in the aforementioned examples. For instance, this is the case for impersonal verbs and pseudo-verbs, which do not inflect in the same way Maltese verbs traditionally do; rather, they express subject through the addition of a suffix4. This is exemplified in (9) and (10), where irnexxielha is an impersonal verb and beħsiebha is a pseudo-verb. (9)

Irnexxie-lha t-iġi manage:PFV-3SG.F 3SG.F.IPFV-come ‘She managed to come.’

(10)

Beħsieb-ha t-iġi intend:IPFV-3SG.F 3SG.F.IPFV-come ‘She intends to come.’

|| 4 What we are here referring to as traditional verbs take on the bound pronoun as a means for marking the object (Amaira 2014: 33).

88 | Kirsty Azzopardi

3.2 Adjacency Trask (1993: 7) defines ‘adjacency’ as “[t]he linear relation holding between two elements in a sentence which are not separated by any other element”. As a matter of fact, Stolz (2009: 152) observes that “[s]trictly speaking, adjacency requires the members of a verb chain to be immediate neighbours syntagmatically”. Similarly, Maas (2009: 117) claims that “[coverbal modification] is asyndetic” and, therefore, no conjunctions should occur in the verb chain. In spite of this, it seems that the Maltese verb chain is not impenetrable and, as opposed to this, nonverbal elements do sometimes occur (Stolz 2009: 152). In his study, Stolz (2009: 136) investigates “the internal cohesion of the Maltese verb chain with special focus on how the criterion of strict adjacency can be violated and whether or not these violations are rule-governed”. In about 730 sentences, Stolz (2009) came across 750 cases of broken verb chains. It is rather crucial to add that the chain-breakers may not necessarily be made up of just one word, as can be seen in example (11), where the chain is broken by the prepositional phrase ta’ spiss, which takes on an adverbial function. Such occurrences of a non-verbal element in the verb chain suggest that the criterion of adjacency is not as salient as the aforementioned criterion, that is grammatical agreement. (11)

Bd-iet ta’ spiss t-mur begin-3SG.F.PFV of often 3SG.F.IPFV-go ‘She started visiting him very often.’

iż-żur=u 3SG.F.IPFV-visit=3SG.M.OBJ

4 Morphosyntactic behavior and restrictions The verb chain in Maltese is not formed arbitrarily but there are certain restrictions that are imposed on the aspect-sensitive members of the chain. These members may occur either in the perfective or the imperfective form, but “not in free combination” (Stolz 2009: 149). Taken into consideration as two variables, perfective and imperfective would yield four different possibilities: IPFV+IPFV, IPFV+PFV, PFV+IPFV, PFV+PFV. This, however, does not mean that all four combinations are always possible, that is, for every combination of verbs. According to Stolz (2009), a V1 in the perfective can precede either a V2 in the imperfective or one in the perfective, whereas a V1 in the imperfective can only be followed by an imperfective V2. This means that Stolz excludes the

The verb sequence V1+V2 in Maltese | 89

combination IPFV+PFV. However, the following examples, covering the range of combinatorial possibilities, are all acceptable for the native speaker of Maltese. (12)

N-ixtieq im-mur 1SG.IPFV-wish 1SG.IPFV-go ‘I would like to go/attend.’

(13)

N-ixtieq mor-t 1SG.IPFV-wish go-1SG.PFV ‘I wish I had gone/attended.’

(14)

Xtaq-t im-mur wish-1SG.PFV 1SG.IPFV-go ‘I wished to go/attend.’

(15)

Xtaq-t mor-t wish-1SG.PFV go-1SG.PFV ‘I wished I had gone/attended.’

On the other hand, Fabri and Borg (2017: 70) argue that “while V1 can be morphologically either in the perfect … or the imperfect … form, any following verb must be in the imperfect”. However, they also note that the tense marker, kien, and modal verbs like seta’ ‘can’ can be followed by a V2 in the perfective form, yet the combinations that feature in their examples regarding this behavior are of the type PFV+IPFV and PFV+PFV. Therefore, it is quite unclear where Fabri and Borg (2017) stand with regard to the combination IPFV+PFV. In this study, I will show that all four morphosyntactic combinations are used in contemporary Maltese, yet the deeper one goes into the subject, the more evident it becomes that not all four morphosyntactic possibilities are allowed for each V1.

5 Methodology In this section I will discuss the method used in order to gather and then analyze the collected sample of verb chains of the type V1+V2. Some problems encountered during the course of this study will be discussed briefly, along with the way in which they were overcome.

90 | Kirsty Azzopardi

5.1 The collected sample of V1+V2 chains Due to the limited number of studies regarding the verb chain in Maltese, it would not have been practical to start analyzing lengthy verb chains. As Fabri and Borg (2017: 68) state, “insight into the structure and relations within the sequence can best be gained by focusing our analysis on the relationship between specific pairs of predicates rather than examining the whole possible sequence at one go”. Moreover, they argue that “we can … build up any sequence by transitively adding a new predicate at a time: letting ‘a’, ‘b’, ‘c’ stand for a predicate or predicate type, we examine the relationship a+b, then b+c, then a+b+c, etc.” (Fabri and Borg 2017: 68). It is for this reason that I decided to analyze verb sequences made up of only two verbs. In order for this study to be carried out, a sample of 500 verb sequences of the type V1+V2 were gathered using MLRS. The sequences taken as part of the research sample were not chosen haphazardly, rather it was ascertained that the same lexeme combination never occurred more than once. Therefore, whenever two lexemes ‘X+Y’ making up a verb chain were taken as part of the sample, another verb chain with lexemes ‘X+Y’ but with different person, number, gender, and tense and aspect distinctions was eliminated. For this reason, if a sequence like (16) were already part of the sample, (17) would not be accepted, whereas (18) would be. The reason for this is that the lexeme combinations in (16) and (17) are identical; the V1s are different instantiations of the lexeme seta’ ‘can’, while the lexeme wasal ‘arrive’ is being realized in the V2 in both cases. However, the lexeme combination in (18) is different because, although the V1 corresponds to the same lexeme as before, seta’, the lexeme in the V2 corresponds to the second binyan form of √WSL, which would yield the lexeme wassal ‘lead to’ as opposed to wasal ‘arrive’. (16)

N-istgħ-u 1IPFV-can-PL ‘We can arrive.’

n-asl-u 1IPFV-arrive-PL

(17)

Setgħ-et t-asal can-3SG.F.PFV 3SG.F.IPFV-arrive ‘She can arrive.’

(18)

J-ista’ j-wassal 3SG.M.IPFV-can 3SG.M.IPFV-arrive.CAUS ‘He/It can lead to …’

The verb sequence V1+V2 in Maltese | 91

5.2 The syntactic position After gathering the data sample from MLRS, the data was analyzed for different lexemes occupying the V1 position, those occupying the V2 position, and lexemes that show occurrences in both positions. While in the V1 there were occurrences from 55 different verbs, the number of different lexical occurrences in the V2 was almost four times as great, namely, 297. Contrastively, the verbs that appear in both the V1 and V2 amount to 31. These figures suggest that there are restrictions on the choice of the V1, since the V2 slot seems to tolerate more verbs than the V1 slot. This is probably related to the fact that “the most grammaticalized element … occupies the leftmost position whereas the purely lexical element is entitled to the position on the right edge” (Stolz 2009: 150).

5.3 Analyzing the verbs The first step I took in order to analyze the verb chains from the collected data sample was to come up with a paradigm of morphosyntactic possibilities for each V1, that is to judge the acceptability of different morphosyntactic combinations with regard to a particular V1. In order to do so, the main tool was my intuition as a native speaker of Maltese. However, I also consulted other native speakers to see which combinations sound natural to them and which they deem unacceptable. Upon getting a paradigm of possibilities for each verb V1, all the V1s were classified into groups, depending on which combinations are allowed and disallowed. This made it easier to understand which verbs behave similarly, in spite of the idiosyncrasy of each verb. The difficulty arose from the fact that, at times, all four morphosyntactic combinations were possible, but only because specific verbs gave rise to meaning differences. To shed light on this issue, Aquilina’s dictionary was consulted and all possible meanings for the V1 were drawn out. The morphosyntactic possibilities were then tested for each meaning of the same verb and it became clear that different meanings of the same verb may allow different distributions. Context is almost always needed in these cases, as a verb chain could be ambiguous with regard to the meaning of V1. Meanwhile, it is worth noting in passing that verbs other than those in the sample were also considered because they help to provide more robust explanations.

92 | Kirsty Azzopardi

6 Perception verbs in the V1 When looking at the morphosyntactic distribution for each V1, perception verbs were classed in the group that showed most freedom in the acceptability of temporal-aspectual combinations. The preliminary morphosyntactic distribution for verbs such as ħass ‘feel’, mess/messu ‘touch’, and ra ‘see’ is given below, where all combinations seem to be possible at first glance. IPFV+IPFV IPFV+PFV PFV+IPFV PFV+PFV

Figure 1: Preliminary morphosyntactic distribution for perception verbs.

6.1 ħass Rather than immediately looking at the morphosyntactic possibilities for this verb, it is more fruitful to look into verb meaning. One way of doing this is to consult dictionaries. In Aquilina (1987: 515), two of the meanings given for ħass are “To feel by touch” and “To feel, perceive (physical or moral pain or satisfaction); to be sensitive; to be affected by external impressions or presentiments”. Therefore, ħass can carry different meanings, which in turn could have an effect on the possible morphosyntactic realizations. The following is the paradigm showing the morphosyntactic combinations for a chain of the type V1(ħass)+V2, with ħass used in its physical sense.

6.1.1 Physical meaning (V1 + bound pronoun) (19)

T-ħoss-ok t-aħraq 2SG.IPFV-feel-2SG.OBJ 2SG.IPFV-burn ‘You feel like you’re burning up.’

(20)

*T-ħoss-ok 2SG.IPFV-feel-2SG.OBJ

(21)

Ħass-ejt-ek t-aħraq feel-2SG.PFV-2SG.OBJ 2SG.IPFV-burn ‘You felt like you were burning up.’

ħraq-t burn-2SG.PFV

The verb sequence V1+V2 in Maltese | 93

(22)

*Ħass-ejt-ek feel-2SG.PFV-2SG.OBJ

ħraq-t burn-2SG.PFV

On the other hand, the paradigm for V1(ħass) employed in its metaphorical sense is different from the above, as can be seen below.

6.1.2 Metaphorical meaning (V1 + bound pronoun) (23)

T-ħoss-ok 2SG.IPFV-feel-2SG.OBJ ‘You feel scared.’

t-ibża’ 2SG.IPFV-scared

(24)

T-ħoss-ok bża-jt 2SG.IPFV-feel-2SG.OBJ scared-2SG.PFV ‘You feel that you were scared.’

(25)

Ħass-ejt-ek feel-2SG.PFV-2SG.OBJ ‘You felt scared.’

(26)

Ħass-ejt-ek bża-jt feel-2SG.PFV-2SG.OBJ scared-2SG.PFV ‘You felt that you were scared.’

t-ibża’ 2SG.IPFV-scared

The paradigms presented above show that whenever the V1(ħass) carries a physical meaning, some restrictions seem to be imposed on the morphosyntactic combinations. In contrast, when the meaning in the V1 transcends the limits of physicality, the verb chain seems to be totally free, in that all morphosyntactic possibilities, that is all four combinations of imperfective and perfective, can be exploited. It is worth noting that, in all the examples given above, ħass has got to do with the subject’s perception of the self and, in fact, it is not used transitively, but reflexively. This results from the use of the suffixed bound pronoun in the V1, which agrees with the subject of the verb and is coreferential with it. Indeed, ħass can only occupy the V1 slot if it is used reflexively, as is evident in the following examples. (27)

Wara x-xena makabra li ra, after DEF-scene macabre that see:3SG.M.PFV Mark ħass-u j-ibża’ Mark feel:3SG.M.PFV-3SG.M.OBJ 3SG.M.IPFV-scared ‘After the gruesome scene he saw, Mark was feeling scared.’

94 | Kirsty Azzopardi

(28)

*Wara x-xena makabra li ra, after DEF-scene macabre that see:3SG.M.PFV Mark ħass j-ibża’ Mark feel:3SG.M.PFV 3SG.M.IPFV-scared

6.2 mess/messu Among other meanings of the verb mess, Aquilina (1990: 812) gives “To touch, to handle” and “To be the turn of, to fall to [someone]’s lot to do [something]”, as well as “Idiomatic usage equivalent to English ‘should’”. The first definition given here corresponds to the physical sense of the verb, and it seems that mess can never occur in the V1 whilst carrying this particular meaning. This example suggests, as already mentioned in the previous subsection, that the physical meaning of the verb is the sense in which most morphosyntactic restrictions lie. As in the case of ħass, the V1(mess/messu) benefits from full combinatorial freedom whenever its meaning is fully detached from physicality, namely, when it expresses deontic modality. In this case, the verb undergoes change not only on a semantic level but also on a morphological one. Whenever mess/messu is meant to express deontic modality, its form is that of an impersonal verb, as is shown in the examples below.

6.2.1 “Idiomatic usage equivalent to English ‘should’” (29)

Minn-hom kollha, i-miss-u j-itkellem from-3PL all IPFV-should-3SG.M 3SG.M.IPFV-speak ‘He should be the one to speak among them all.’

hu he

(30)

Jekk ma ried-x għaġin, i-miss-u if NEG want:3SG.M.PFV-NEG pasta IPFV-should-3SG.M tkellem speak:3SG.M.PFV ‘He should have spoken up if he didn’t want to have pasta.’

(31)

Mess-u j-itkellem dwar should:PFV-3SG.M 3SG.M.IPFV-speak about x’ qed i-ħassb-u what PROG 3SG.M.IPFV-worry-3SG.M.OBJ ‘He should open up about what is worrying him.’

The verb sequence V1+V2 in Maltese | 95

(32)

Mess-u tkellem aħna u n-iftiehm-u should:PFV-3SG.M speak:3SG.M.PFV we and 1IPFV-agree-PL ‘He should have spoken up while we were planning.’

So far, we have seen morphosyntactic paradigms for V1(mess/messu) that stand on either extreme of the spectrum of combinatorial possibilities, yet this V1 also bears a meaning with a more balanced paradigm. This is the case when mess/messu is used in the sense of “To be the turn of, to fall to [someone]’s lot to do [something]” (Aquilina 1990: 812).

6.2.2 “To be the turn of” (33)

Min i-miss-u j-itkellem? who IPFV-be.one’s.turn-3SG.M 3SG.M.IPFV-speak ‘Whose turn is it to speak?’

(34)

*Min who

(35)

?Min mess-u who be.one’s.turn:PFV-3SG.M ‘Whose turn was it to speak?’

j-itkellem? 3SG.M.IPFV-speak

(36)

*Min who

tkellem? speak:3SG.M.PFV

i-miss-u tkellem? IPFV-be.one’s.turn-3SG.M speak:3SG.M.PFV

mess-u be.one’s.turn:PFV-3SG.M

In this case, it seems like the only possible combination is IPFV+IPFV. While IPFV+PFV and PFV+PFV are not acceptable, the combination of perfective and imperfective, respectively, seems to be somewhat doubtful. Although one might accept this combination, some speakers would probably prefer something like (37), which would lead to a verb chain of the type V1+V2+V3. (37)

Min kien i-miss-u j-itkellem? who be:3SG.M.PFV IPFV-be.one’s.turn-3SG.M 3SG.M.IPFV-speak ‘Whose turn was it to speak?’

6.3 ra Among other definitions, Aquilina defines the verb ra as “To see” (1990: 1165); “To ascertain, to make sure, to find out” (1990: 1166). It is rather interesting to note that Aquilina does not distinguish between seeing with physical eyes and

96 | Kirsty Azzopardi

seeing with the mind’s eye, i.e. perceiving. The closest Aquilina (1990: 1165) gets to the latter is through “To see, in the sense of imagining a situation”, although this may not necessarily be equivalent to perception. For this reason, the sense of sight will be referred to as the physical meaning of ra, whereas perception will be considered metaphorical, in that it is an extension of sight.

6.3.1 Physical meaning As opposed to the V1(mess/messu), ra can take on a physical meaning when in the V1, albeit displaying some combinatorial restrictions. This can be observed in the examples displayed below. (38)

Jekk n-ara n-ikteb f’ dan id-dlam, if 1SG.IPFV-see 1SG.IPFV-write in this DEF-darkness n-iktib=lu ittra 1SG.IPFV-write=3SG.M.OBJ letter ‘If I’ll be able to see well enough in this darkness, I will write him a letter.’

(39)

*Jekk n-ara ktib-t if 1SG.IPFV-see write-1SG.PFV n-iktib=lu ittra 1SG.IPFV-write=3SG.M.OBJ letter

(40)

Lanqas ra-jt n-ikteb f’ dak not.even see-1SG.PFV 1SG.IPFV-write in that ‘I could not see well enough to write in that darkness.’

DEF-darkness

*Lanqas not.even

id-dlam DEF-darkness

(41)

ra-jt see-1SG.PFV

f’ in

dan this

ktib-t f’ write-1SG.PFV in

id-dlam, DEF-darkness

dak that

id-dlam

These examples suggest that a verb chain with ra in the V1 used in its physical sense only permits a V2 in the imperfective and any chain with a perfective V2 is excluded. However, there seems to be an exception to this with regard to the V2(dagħa).

6.3.2 Physical meaning – the exception The rule established as of now supposes that when the V1(ra) carries a physical meaning, all the combinations with an imperfective V2 are acceptable. However, an exception to this rule surfaces when the verbal lexeme in the V2 is dagħa

The verb sequence V1+V2 in Maltese | 97

‘swear’. The following is a combination of lexemes ra and dagħa as given in Flask (2007). (42)

lanqas lampa biex n-ara n-idgħi ma hemm not.even lamp in.order.to 1SG.IPFV-see 1SG.IPFV-swear NEG there ‘There isn’t even a lamp for me to be able to see.’

It is worth noting that although in this case the V1(ra) assumes a physical meaning, we are not actually talking about something one needs to employ their sense of sight for. To a certain extent, the combination nara nidgħi can be considered idiomatic since it is not true to the combined meanings of its components, in that it does not signify not being able to swear due to limited visibility, rather that it is pitch dark. This stands in opposition to a verb sequence such as nara nikteb, which denotes being able to see well enough to write. It is rather clear that this difference in meaning is brought about by the V2. Moreover, it seems that the paradigm of morphosyntactic combinations is quite limited when dagħa is in the V2 and only the combination IPFV+IPFV is totally acceptable. (43)

N-ara 1SG.IPFV-see

n-idgħi 1SG.IPFV-swear

(44)

*N-ara 1SG.IPFV-see

dgħa-jt swear-1SG.PFV

(45)

?Ra-jt see-1SG.PFV

n-idgħi 1SG.IPFV-swear

(46)

*Ra-jt see-1SG.PFV

dgħa-jt swear-1SG.PFV

The main difference between the paradigm given above and that corresponding to V1(ra) used in its physical sense lies in the combination of perfective and imperfective, respectively. While this is allowed in the regular paradigm, it is somewhat doubtful in the paradigm corresponding to the exception. Since none of the morphosyntactic combinations with a verb in the perfective seem to be definitely accepted in the latter, the perfective can be expressed through the addition of an inflected form of beda ‘start’ or kien+qed at the beginning of the verb chain, such that a chain of type V1+V2+V3 is obtained. (47)

Lanqas bde-jt n-ara n-idgħi not.even start-1SG.PFV 1SG.IPFV-see 1SG.IPFV-swear bi-d-dlam li kien hemm with-DEF-darkness that be:3SG.M.PFV there ‘I was not seeing anything because of how dark it was.’

98 | Kirsty Azzopardi

(48)

Lanqas kon-t qed n-ara n-idgħi not.even be-1SG.PFV PROG 1SG.IPFV-see 1SG.IPFV-swear bi-d-dlam li kien hemm with-DEF-darkness that be:3SG.M.PFV there ‘I was not seeing anything because of how dark it was.’

6.3.3 (V1 + pronominal suffix)+V2 The V1(ra) can also occur in the V1 in its physical sense when used reflexively, namely, through the addition of a pronominal suffix congruent with the subject of the verb.

6.3.3.1

Physical meaning (V1 + bound pronoun)

(49)

Fi-l-ħolm ġieli n-ara-ni n-werżaq in-DEF-dreams sometimes 1SG.IPFV-see-1SG.OBJ 1SG.IPFV-scream ‘In my dreams, I sometimes see myself screaming.’

(50)

*Fi-l-ħolm ġieli in-DEF-dreams sometimes

(51)

Fi-l-ħolm ġieli ra-jt-ni n-werżaq in-DEF-dreams sometimes see-1SG.PFV-1SG.OBJ 1SG.IPFV-scream ‘I have sometimes seen myself screaming in my dreams.’

(52)

*Fi-l-ħolm ġieli ra-jt-ni in-DEF-dreams sometimes see-1SG.PFV-1SG.OBJ

n-ara-ni werżaq-t 1SG.IPFV-see-1SG.OBJ scream-1SG.PFV

werżaq-t scream-1SG.PFV

In this case, it appears that only verb chains with a V2 in the imperfective are acceptable. For these restrictions to hold there seem to be two conditions: i) The V1(ra) must function as a reflexive verb ii) A physical meaning is assumed by the V1(ra) If the first condition is fulfilled but the meaning of the V1 no longer pertains to the sense of sight, a different paradigm is obtained. In fact, as opposed to what has been portrayed in the paradigm above, all morphosyntactic combinations are possible when the V1(ra) strays away from its physical sense and conveys a metaphorical meaning, that is perceiving. In the following examples, the V1(ra) does not allude to the sense of sight but to an experienced feeling, or to an inner state. In reference to what we have discussed before, it seems like the combina-

The verb sequence V1+V2 in Maltese | 99

torial possibilities (or restrictions) are highly dependent on the physicality, or lack thereof, in the meaning assumed by the V1.

6.3.3.2 Metaphorical meaning (V1 + bound pronoun) (53)

Meta n-qum kmieni n-ara-ni when 1SG.IPFV-wake.up early 1SG.IPFV-see-1SG.OBJ n-egħja malajr 1SG.IPFV-get.tired quickly ‘Whenever I wake up early, I find myself getting tired quickly.’

(54)

Meta n-ara-ni għeje-jt aktar when 1SG.IPFV-see-1SG.OBJ get.tired-1SG.PFV more mi-s-soltu n-ieqaf from-DEF-usual 1SG.IPFV-stop ‘Whenever I feel that I’ve gotten more tired than usual, I stop.’

(55)

Meta naqqas-t mil-l-ikel ra-jt-ni when reduce-1SG.PFV from-DEF-food see-1SG.PFV-1SG.OBJ n-egħja mi-x-xejn 1SG.IPFV-get.tired from-DEF-nothing ‘When I cut down on food, I found myself getting tired without having done much.’

(56)

Ilbieraħ ra-jt-ni għeje-jt malajr wisq yesterday see-1SG.PFV-1SG.OBJ get.tired-1SG.PFV quickly a.lot ‘Yesterday I found myself getting tired in a really short time.’

6.3.4 Meaning of checking (V1+V2_x) Up to now, we have only looked at the V1(ra) when it means “To see” (Aquilina 1990: 1165), both in the sense of sight and of perception. However, another meaning that ra in the V1 can assume is “To ascertain, to make sure, to find out” (Aquilina 1990: 1166), in other words a meaning of checking, although an undertone of trying seems to be present. It is interesting to note that when the V1(ra) expresses this meaning, -x must be suffixed to the V2. (57)

Aħjar better

t-ara s-sib=hom-x 3SG.F.IPFV-see 3SG.F.IPFV-find=3PL.OBJ-NEG

j-iġr-u 3IPFV-run-PL

100 | Kirsty Azzopardi

ċ-ċwievet qabel t-waħħal fi-n-nies! DEF-keys before 3SG.F.IPFV-blame in-DEF-people ‘It would be better if she were to see whether she herself can find the keys lying around before blaming others.’ (58)

Aħjar t-ara sab-it-x dak li kell-ha better 3SG.F.IPFV-see find-3SG.F.PFV-NEG that that have:PFV-3SG.F t-fittex qabel t-itkellem! 3SG.F.IPFV-search before 3SG.F.IPFV-speak ‘She had better see whether she has found what she was supposed to look for before saying anything.’

(59)

Kieku qatt n-af li ra-t is-sib-x if never 1SG.IPFV-know that see-3SG.F.PFV 3SG.F.IPFV-find-NEG l-għajnuna meta kell-ha bżonn DEF-help when have:PFV-3SG.F need ‘I can’t recall just one time when she tried to seek help when she needed it.’

(60)

Ra-t sab-it-x il-bieb it-tajjeb? find-3SG.F.PFV-NEG DEF-door DEF-good see-3SG.F.PFV ‘Did she check whether she had found the right door?’

Once again here, all combinations are possible.

6.4 Other perception verbs So far, we have only considered the cases of three verbs of perception in the V1, namely, ħass, mess/messu and ra. Despite the behavioral patterns noted in the aforementioned verbs, other verbs that pertain to the same class, such as sema’ ‘hear’ and xamm ‘smell’, display different patterns in their paradigm of morphosyntactic possibilities.

6.4.1 sema’ (V1 + bound pronoun) (61)

N-ismagħ-ni n-istona 1SG.IPFV-hear-1SG.OBJ 1SG.IPFV-sing.out.of.tune ‘I hear myself singing out of tune.’

(62)

?N-ismagħ-ni stuna-jt 1SG.IPFV-hear-1SG.OBJ sing.out.of.tune-1SG.PFV ‘It sounds to me as if I were singing out of tune.’

The verb sequence V1+V2 in Maltese | 101

(63)

Sma-jt-ni n-istona hear-1SG.PFV-1SG.OBJ 1SG.IPFV-sing.out.of.tune ‘I heard myself singing out of tune.’

(64)

Sma-jt-ni stuna-jt sing.out.of.tune-1SG.PFV hear-1SG.PFV-1SG.OBJ ‘It sounded to me as if I were singing out of tune.’

It seems like the V1(sema’) can only be used in its physical sense and this, as argued before, affects the paradigm of morphosyntactic possibilities. Moreover, sema’ seems to occupy a V1 slot only when used reflexively. With regard to xamm, it is worth noting that it could denote not only the meaning “To smell” but also “To smell [something] out by intuition” (Aquilina 1990: 1543). The first meaning drawn out from Aquilina corresponds to that of the V1 in (65); yet xammejtu+jinħaraq does not constitute a verb chain since the two verbs do not share the subject. An example including xamm in its metaphorical sense is provided in (66), although once again no verb chain is formed. Therefore, it looks like none of the meanings of xamm drawn out from Aquilina can correspond to the meaning denoted by this verb in the V1. (65)

L-ikel xamm-ejt=u DEF-food smell-1SG.PFV=3SG.M.OBJ ‘I smelt the food getting burnt.’

j-inħaraq 3SG.M.IPFV-get.burnt

(66)

Xamm-ejt=u biex ħiereġ go.out.ACTP:3SG.M smell-1SG.PFV=3SG.M.OBJ COMP ‘I could tell what he was going to come up with.’

For xamm to occur in the V1, a suffixed bound pronoun coreferential with the subject of the V1 itself is required, which would render the verb reflexive. This is shown in the examples provided hereunder.

6.4.2 xamm (V1 + bound pronoun) (67)

In-xomm-ni n-inten 1SG.IPFV-smell-1SG.OBJ 1SG.IPFV-smell.bad ‘I sense my unpleasant odor.’

(68)

*In-xomm-ni ntin-t 1SG.IPFV-smell-1SG.OBJ smell.bad-1SG.PFV

(69)

Xamm-ejt-ni smell-1SG.PFV-1SG.OBJ

n-inten 1SG.IPFV-smell.bad

102 | Kirsty Azzopardi

‘I sensed my unpleasant odor.’ (70)

*Xamm-ejt-ni smell-1SG.PFV-1SG.OBJ

ntin-t smell.bad-1SG.PFV

Although V1(xamm) seems to have a more restricted paradigm of morphosyntactic possibilities than V1(sema’), neither shows as high a degree of freedom as other perception verbs. This restriction might result from the fact that both xamm and sema’ only occur in the V1 when they have a physiological sense. The sense we have not talked about is taste. The verb tiegħem ‘taste’ does not occur in the V1 and can only be used mono-transitively or intransitively, as shown in (71) and (72). (71)

Il-kafè n-tiegħm=u DEF-coffee 1SG.IPFV-taste=3SG.M.OBJ ‘I don’t like the taste of coffee.’

ħażin bad

(72)

inie-x in-tiegħem minħabba be:1SG.IPFV-NEG 1SG.IPFV-taste because.of ‘I have lost my sense of taste because of this cold.’

M’

r-riħ

NEG

DEF-cold

It is also worth noting that tiegħem does not function as a reflexive verb, in contrast to the other verbs of perception. (73)

*In-tegħm-u-na 1IPFV-taste-PL-1PL.OBJ

(74)

*Tegħim-nie-na taste-1PL.PFV-1PL.OBJ

7 Conclusion It should be evident that it is rather easy to be deluded when coming up with a paradigm of preliminary morphosyntactic distribution with regard to any verb. The morphosyntactic distribution of V1 and V2 verbs can only be ascertained when the polysemous facet of verbs is taken into consideration. All possible meanings of a verb V1 should be considered and tested in order to determine the possible combinations with a V2. As has been shown in the previous section, various meanings of the same verb V1 might give rise to different paradigms of morphosyntactic distribution. A conspicuous feature of perception verbs in the V1 is the divide between the physical and the metaphorical use, which results in different possible com-

The verb sequence V1+V2 in Maltese | 103

binations. If the physical and the metaphorical are regarded as the two ends of one spectrum, we could say that the more the meaning of a verb leans towards the physical, the more restrictions will manifest themselves in its paradigm. In contrast, if a verb meaning seems inclined towards the metaphorical, its paradigm will most likely benefit from a high degree of combinatorial freedom. This divide can be expressed in more general terms, in that in the case of perception verbs, the physical meaning(s) of a verb correspond(s) to its primary meaning(s), whereas any other meanings that stray away from physicality can be regarded as its secondary meanings. Through this approach, generalizations can be made, if and where possible, which would suit other verbs that do not pertain to the class of verbs discussed in this paper, namely, non-perception verbs. Primary meaning/s (physical)

Secondary meaning/s (metaphorical)

Figure 2: Semantic spectrum.

Acknowledgments: I would like to express my sincere gratitude to Albert Borg and Ray Fabri for their invaluable feedback on an earlier version of this paper. Naturally, any errors are solely my responsibility.

Abbreviations 1, 2, 3 ACTP CAUS COMP DEF F IMP IPFV M NEG OBJ PFV PL PROG SG

V VP

1st, 2nd, 3rd person active participle causative complementizer definite feminine imperative imperfective masculine negation object perfective plural progressive Singular verb verb phrase

104 | Kirsty Azzopardi

References Amaira, Mark. 2014. Il-kategoriji grammatikali tal-Malti: Il-verbi, in-nomi, u l-aġġettivi. Malta: University of Malta M.A. thesis. Aquilina, Joseph. 1987. Maltese-English dictionary – vol. 1. Valletta: Midsea Books. Aquilina, Joseph. 1990. Maltese-English dictionary – vol. 2. Valletta: Midsea Books. Azzopardi, Kirsty. 2019. Il-verbi f’sekwenza fil-Malti. Malta: University of Malta B.A (Hons) thesis. Camilleri, Maris. 2016. Temporal and aspectual auxiliaries in Maltese. Essex: University of Essex Ph.D. thesis. Fabri, Ray. 1993. Kongruenz und die Grammatik des Maltesischen. Tübingen: Niemeyer. Fabri, Ray & Albert Borg. 2017. Modifiers and complements within the Maltese verb sequence. In Benjamin Saade & Mauro Tosco (eds.), Advances in Maltese linguistics, 67–86. Berlin & Boston: De Gruyter Mouton. Flask, Wayne. 2007, November 28. Back in the saddle. Wayne Flask – Author Translator Music Writer Public Annoyance. http://wayneflask.com/2007/11/back-in-the-saddle/ Maas, Utz. 2009. Complex predicates in Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 113–132. Amsterdam & Philadelphia: John Benjamins. Maltese Language Resource Server (MLRS): http://mlrs.research.um.edu.mt/CQPweb/ malti03/ Peterson, John. 2009. Pseudo-verbs: An analysis of non-verbal (co-)predication in Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 181–204. Amsterdam & Philadelphia: John Benjamins. Stolz, Thomas. 2009. Splitting the verb chain in modern literary Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 133–179. Amsterdam & Philadelphia: John Benjamins. Trask, Robert L. 1993. A dictionary of grammatical terms in linguistics. London: Routledge. Vassalli, Michelantonio. 1827. Grammatica della lingua Maltese. 2nd edn. Malta.

Erika Just and Slavomír Čéplö

Differential object indexing in Maltese – a corpus based pilot study Abstract: This paper presents the first corpus-based study of DOI in Maltese. In this pilot study, the potential triggering factors were tested as predictors in a descriptive model. The results show that the strongest predictor for object indexing in Maltese is word order, but when taking only semantic referential features into account, the analyses reveal that DOI seems to be strongly predictable by definiteness, as well as by the part of speech of the head of the NP. Our study therefore supports observations from previous investigations, both on Maltese and typological; furthermore, the analysis gives insight into the combined effects of the relevant factors. Keywords: corpus study; indexing; information structure; object marking

1 Introduction 1.1 Background and terminology In the literature on Maltese, the phenomenon under discussion appears under various different names: it is known as “optional direct object agreement” (Fabri 1993), “suffixed object pronoun” (Borg and Azzopardi-Alexander 1997), “pronominal clitic” (Fabri and Borg 2002; Vella 2009), or “object reduplication” (Čéplö 2014). In the literature on Arabic, Romance or Balkan languages, the term “clitic doubling” frequently occurs (see e.g. Aoun 1999; Kallulli and Tasmowski 2008 or De Cat and Demuth 2008); this term also appears to be preferred by linguists of the generative bend and, whether with good reason or not, it is often used as the default. For our analysis, we use the more neutral term differential indexing, with indexes being defined as bound markers on the verbal predicate expressing argument features, most commonly person and number. Indexing (Haspelmath

|| Erika Just: Department of Comparative Language Science, University of Zurich, Thurgauerstrasse 30, 8050 Zurich, Switzerland, E-mail: [email protected] Slavomír Čéplö: Austrian Academy of Sciences, IMAFO-BYZ, Hollandstraße 11–13, 1020 Wien, Austria, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-005

106 | Erika Just and Slavomír Čéplö

2013) is a more neutral term than agreement, as it does not presuppose any syntactic relationship between the marker and the referential noun phrase (Haig and Forker 2018). Later in this chapter, we will make comparisons between differential indexing in Maltese and other languages, Semitic as well as from other families. This would be rather difficult if we chose to discuss the phenomenon under the umbrella of agreement, due to the various presuppositions associated with that term, the expected relationships between controller and target, to say nothing of the theoretically loaded terminology. Also, the morphological status of the index as a clitic or an affix is considered irrelevant. The latter is often equated with obligatoriness of marking, which is unjustified (Haig and Forker 2018: 720), as clitics can be obligatory, just as affixes can be grammatically optional. The term differential was coined by Bossong’s (1982) work on Sardinian and New Iranian languages, originally referring to variation in object case marking. However, differential marking patterns are not restricted to case or adpositional marking, but include indexing as well, as it is likewise a means of encoding arguments. Differential argument marking (DAM) can broadly be defined as any situation where an argument of the predicate with the same semantic argument role is coded differently (Seržant and Witzlack-Makarevich 2018). The argument role in question in the current study is the P argument, i.e. the less agent-like argument of a two-place predicate.1

1.2 Differential indexing in Maltese Whereas subject indexing is typically obligatory in Maltese, be it with or without a co-occurring referential noun phrase, object indexing alongside an overt noun phrase has been considered optional and triggers indexing only if the referential noun phrase is the topic of the clause (Fabri 1993: 92). In Maltese, both the direct and the indirect object can be indexed, but the present study limits itself to the investigation of direct objects in monotransitive clauses. The following sentences exemplify how the presence of the object index for the third person singular, masculine in (1b) does not change the propositional content of the clause compared to (1a):

|| 1 This definition follows the generalized argument roles framework as proposed by Bickel and Nichols (2008), Bickel (2010) and Witzlack-Makarevich (2010).

Differential object indexing in Maltese | 107

(1)

a.

b.

c.

Jien nara l-programm. I see:1SG.IPFV DET-program(M) ‘I am watching the program.’ Jien nara-h il-programm. I see:1SG.IPFV-3SG.M DET-program(M) ‘The program, I am watching it.’ Jien nara-h I see:1SG.IPFV-3SG.M ‘I am watching it.’

[Fabri 1993: 92]

The direct object index in Maltese can also have a purely pronominal function, as shown in (1c), but this does not constitute a differential pattern, so it is the difference between the structures in (1a) and (1b) which interests us in the present study. DOI in Maltese has been given some attention in the literature, but whereas Sutcliffe (1936) and Aquilina (1959) only give descriptions of the phenomenon, it is Fabri (1993) who is the first to associate it with information structure, describing the pragmatic inacceptability of indexed object noun phrases which are in focus, exemplified in (2) below: Helgard, who is the information asked for in (2a) is new information in (2b), therefore in focus and cannot be indexed, which renders sentence (2c) unacceptable in this context. The preposition lil is obligatory with human or human-like direct objects, as well as indirect objects, and has by itself nothing to do with indexing the object. (2)

a.

Lil

min rajt? who see.PFV.2SG ‘Whom did you see?’ Rajt lil Helgard. see.1SG PREP Helgard ‘I saw Helgard.’ *Rajt-ha lil Helgard. see.1SG-3SG.F PREP Helgard. PREP

b.

c.

[Fabri 1993: 145]

Later studies have extended Fabri’s (1993) work and put differential object indexing into the broader context of sentence information structure. Fabri and Borg (2002) investigate the relationship between topicality, focality and word order. The basic word order in Maltese is SVO, but it is by no means the only option. The authors state that also with indexed object noun phrases, other configurations are possible and thus both postverbal and preverbal object noun phrases with and without indexes are possible. Borg and Azzopardi-Alexander (2009) add phonological aspects to these analyses, stating that indexed preverbal object noun phrases form an intonationally separated unit from the re-

108 | Erika Just and Slavomír Čéplö

mainder of the clause. Unfortunately, OV(S) is the only order they consider in their investigation. All the studies on differential object indexing in Maltese come to the conclusion that a full account of the phenomenon is a difficult task to accomplish, and that no referential feature such as humanness, animacy or definiteness alone triggers the construction, but that the role of information structure lies at the core of it. Indexing preverbal objects has been claimed to be the main topicalization strategy in Maltese (Borg and Azzopardi-Alexander 1997: 126), whereas non-indexed preverbal object noun phrases are usually in focus (Fabri and Borg 2002: 359–360). The studies concerned with differential indexing mentioned so far were all based on the authors’ intuitions as native speakers. The first one to go about the matter based on examples from real texts or spoken data is Čéplö (2014) who investigates DOI in Maltese against the background of the different clitic doubling phenomena described in the languages of the Balkan and Romance languages. The study explores various examples of DOI in Maltese, with their various contexts and different word orders. Amongst other things, the findings affirm that preverbal indexed objects are not necessarily marked for definiteness. There are examples of quantified as well as bare nouns which are preverbal and indexed (Čéplö 2014: 206–207), which suggests that it is rather specificity than definiteness which makes it likely for a referent to be indexed in this position. As for postverbal noun phrases, Čéplö (2014: 219) states that indexing them seems to be “optional”, as his investigation shows that, in terms of intonation, the indexed object can form an intonational unit with the verb, like the nonindexed one and that a pause between the verb and the noun phrase is possible but not obligatory, which shows that an indexed postverbal object noun phrase is not necessarily dislocated. He further observes that these indexed in situ objects, although rarer than non-indexed ones, occur frequently in exclamations, exhortations and especially in questions (Čéplö 2014: 219), which can be seen as an indicator for their information-structural prominence. As noted earlier, there is no one-to-one correspondence between word order and the information-structural status of the referent (Fabri and Borg 2002: 359–360). So it seems that in Maltese, neither the placement of the noun phrase, nor intonational cues, nor indexing can on their own be regarded as a means to express the information-structural status of an object referent. Conversely, it is not very fruitful to try and account for DOI in Maltese on the basis of the rather fuzzy labels topic and focus only (cf. Čéplö 2014: 213), even if looking at corpus data reveals that undeniably, indexed objects are highly prominent and to analyze them as topics seems justified and obvious. Also, an important finding from

Differential object indexing in Maltese | 109

Borg and Azzopardi-Alexander’s (2009: 73–74) study is that with preverbal indexed objects, the preposition lil, which has to be used for any referent which can be referred to with a personal name, is no longer obligatory. This is an indicator that indexing is a way of marking highly accessible referents. But then again, not every object referring to a topical referent is indexed, neither in Maltese nor in other languages with DOI. Dalrymple and Nikolaeva (2011: 51–57) in their study on differential indexing phenomena in Romance and Bantu languages propose to speak of topicworthiness rather than topichood when accounting for the referential features which trigger differential indexing. In order to account for a referent’s status with regard to topicality, different versions of hierarchies have been proposed2, at the core of which lie argument properties such as animacy, givenness, identifiability and the full noun vs. pronoun distinction. Clearly, for some languages topicality scales are more important than for others, and the thresholds therein, i.e. from which position in the scale does a referent require special (differential) marking, have to be considered language specifically for each property. In other words, variables such as givenness, animacy or identifiability contribute to topicworthiness to different degrees in each language. One of the goals of this chapter is to set the stage for a fine-grained analysis of the triggering factors, as it is seemingly a highly complex interplay of different variables which renders object referents in Maltese worthy of indexing.

2 DOI crosslinguistically 2.1 Beyond Semitic Differential indexing is not restricted to objects, but can affect subjects as well. But with differential subject indexing, it is very often the absence of an otherwise present index that codes the deviating scenario. Differential subject indexing has up until now been somewhat neglected in the study of differential marking phenomena (but see Just, in preparation). Consider spoken French, where subject indexing (with a verbal proclitic) has become quite common (cf. Culbertson 2010), but it is not possible for all types of subjects. For instance, it is not possible with quantified and indefinite subjects or with noun phrases which

|| 2 See Seržant and Witzlack-Makarevich (2018) for a detailed overview.

110 | Erika Just and Slavomír Čéplö

provide answers to wh-questions (features typically associated with focus). The latter case is as exemplified in (3a). (3)

a.

French (Indo-European, France) [De Cat 2004: 6] Ceux du groupe A (*ils) ont fini leur travail. ‘[Q: Who finished their work?] Those in group A have finished their work.’

However, indexing of such a subject is possible if the co-referential noun phrase is a strong pronoun. In fact, it is obligatory to index strong pronouns in colloquial French: b.

Moi *(je) l’ai appelé. ‘[Q: Who called Jean?] I called him.’

[Culbertson 2010: 115]

In contrast to differential subject indexing, much attention has been devoted to differential object indexing, in language-specific studies (e.g. Muxí 1996 on Catalan; Béjar 1999 on Selayarese or Downing 2018 on Chichewa), family or area specific studies (e.g. Friedmann 2008 on languages of the Balkan; Riedel 2009 on Bantu or Klamer and Kratochvíl 2018 on Alor-Pantar) as well as typological studies (e.g. Arkadiev 2010 or Iemmolo 2011). Differential object indexing is very often associated with topic-related argument properties, an indexed P being the more marked construction as opposed to a non-indexed P. This is not surprising given the fact that, whereas the A role is usually occupied by referents bearing topic features and is located high on the animacy and definiteness scale, the P argument is generally associated with the opposite, often serving to introduce new information (Du Bois 1987; Comrie 1989 inter alia). Therefore, a deviating scenario seems to call for a distinct marking pattern. Before going into some examples, a few words will be said on the notion of topic and focus and how they are used here. The terms involve a great deal of vagueness due to the extent of linguistic diversity, both in form and function of information-structural features involved. Topic is generally connected to the notions of givenness, a high degree of identifiability and also to a high ranking in the person hierarchy, and is assumed to relate to the hearer’s knowledge. Focus, on the other hand, brings about an information update and is associated with such notions as newness or contrastiveness. But as a matter of fact, the meanings conveyed by similar constructions in different languages labeled as “focus construction” or “topic construction” are so manifold that topic and focus should often be considered as interpretive effects of such constructions, and not as being at their core (cf. Matić and Wedgwood 2013). Also, comparable and recurrent structures in different languages do

Differential object indexing in Maltese | 111

not necessarily convey the same information-structural status of a referent (cf. Skopeteas and Fanselow 2010 for a crosslinguistic investigation of noncanonical word order, clefts and focus). This also holds for structurally similar DOI constructions in different languages. With DOI, it is very often the case that an otherwise per-default non-present index shows up if the referent is high in saliency or topicality, i.e. bearing features which are usually not associated with objects, or the P role, respectively (Du Bois 1987). Sometimes, the situation is quite straightforward and object indexing can even become mandatory for subclasses of nouns, such as for all nouns referring to animates, as in the Alor-Pantar language Teiwa spoken in Indonesia (Klamer and Kratochvíl 2018), or all nouns referring to humans in the Madang language Kesawai spoken in Papua New Guinea (Priestley 2008). In other languages, the line between indexing and non-indexing is not that easily drawn. One such example is the Bantu language Sambaa, spoken in Tanzania, where object indexing also has to do with animacy: it is obligatory for proper names, titles – as in (4a) and (4b) – and first and second person pronouns; the index can therefore not be omitted with these referents. It is described as common (but not obligatory) with other types of human referents, less common with animals – as in (4c) – and rare but acceptable with inanimates (Riedel 2009: 45–46). (4) a.

b.

c.

Sambaa (Benue-Congo, Tanzania) N-za-mw-ona askofu 1SG.SBJ-PFV-1.OBJ-see 5.bishop3 ‘I saw the bishop.’ *N-za-ona askofu 1SG.SBJ-PFV-see 5.bishop Int.: ‘I saw the bishop.’4 N-za-(ji)-ona kui. 1SG.SBJ-PFV-(5.OBJ)-see 5.dog ‘I saw the/a dog.’

[Riedel 2009: 45–46]

In another Bantu language, Nkore-Kiga (Uganda), the indexing morphology is quite similar to that of Sambaa: a verbal prefix agreeing with the noun class of the object noun phrase. As a rule, objects are not indexed if the coreferential noun phrase is overt. However, they always have to be indexed if the object is “topicalized” and the noun phrase shifts to the preverbal “topic-position” (Tay-

|| 3 Numbers followed by a period indicate Bantu noun classes. 4 Although askofu ‘bishop’ belongs to noun class 5, there is a noun class 1 index on the verb, as this is the noun class usually used for humans.

112 | Erika Just and Slavomír Čéplö

lor 1985: 78, 91)5. In (5a), the object (enkoni) is postverbal and there is no index on the verb. In (5b), the pronominal object is indexed (gi-). Finally, in (3c), there is again an overt NP referring to the object, now in the preverbal position, and it is also obligatorily indexed: (5) a.

b.

c.

Nkore-Kiga (Benue-Congo, Uganda) [Taylor 1985: 91] omuntu a-kwata enkoni 1.person 1.SBJ-hold 9.stick ‘Someone is holding a stick.’ omuntu a-gi-kwata 1.person 1.SBJ-9.OBJ-hold ‘Someone is holding it.’ enkoni omuntu a-gi-kwata 9.stick 1.person 1.SBJ-9.OBJ-hold ‘As for the stick, someone holds it.’ (translation modified)

A third, again different, Bantu example comes from Ruuli, also spoken in Uganda. Looking at (6a) and (6b) might suggest that DOI in Ruuli works quite similar to that in Nkore-Kiga, however, it is not restricted to preverbal object noun phrases, although it is in fact more commonly found with such than with in situ objects (Just and Witzlack-Makarevich 2018), exemplified by (6c). (6d) shows a non-indexed preverbal object which indicates that Ruuli is different from NkoreKiga in this regard: placing an object noun phrase in a preverbal position does in itself not trigger indexing. (6) a.

b.

c.

d.

Ruuli (ruc, Benue-Congo, Uganda) [Witzlack-Makarevich et al. 2019] Iswe tu-li-ire bunyonyi na obusolo 1PL.SBJ-eat-PFV 14.bird COM 14.animal 1PL ‘As for us, we have eaten birds and animals.’ Obuterega o-bu-maite 14.obuterega 2SG.SBJ-14.OBJ-know.PFV ‘Do you know the obuterega traps?’ Naye we o-bi-maite ebyo? 2SG.SBJ-8.OBJ-know.PFV 8.DEM but 2SG ‘But you, do you know that?’ Amaani mu-ta-ire-mu 6.strength 2PL.SBJ-put-PFV-LOC ‘You have put in a lot of strength.’

|| 5 Bantu languages commonly have SVO as the basic word order.

Differential object indexing in Maltese | 113

These examples from Ruuli suggest that DOI is related to topicality of the referent. However, this is hard to tell without context, and on the basis of selected examples only. A quantitative study of the triggering factors for differential indexing in Ruuli (Just and Witzlack-Makarevich 2018) examined the interplay of these factors. Based on a corpus annotation for variables such as noun class, PoS, animacy, identifiability and textual givenness, the statistically relevant factors were identified using conditional inference. The results of the analyses show that the strongest predictor for DOI in Ruuli is in fact word order, with preverbal noun phrases being more likely to be indexed than postverbal ones; but taking only semantic properties of the referent into account, the analyses reveal that DOI seems to be strongly predictable by textual givenness and humanness. This is not really surprising with regard to the assumption that topicality is involved in DOI in Ruuli, as both givenness and humanness relate to high accessibility and thus topicality. However, as not every human or every given object is indexed, it shows that we are dealing with probabilistic rules, which can be adequately described using descriptive models. We have just seen three examples of languages belonging to the same family, where object indexing is quite similar with regard to its morphological realization. Also, object indexing can be labeled differential in Sambaa, Nkore-Kiga and Ruuli, as in all three languages, it is the same macrorole, the P argument, which becomes indexed only under certain conditions. Although these conditions can be traced back to referential features which are usually not associated with this macrorole, they are not the same from language to language. Also, DOI has grammaticalized to some extent in one of the languages, namely Sambaa, where it has become obligatory for certain nouns. A similar situation can be found in the languages of the Balkans, where DOI has been integrated into the different language systems to various degrees. Starting off as a pragmatic phenomenon expressing the topicality of a referent, it has grammaticalized to various extents (or not at all) in the different languages (Friedmann 2008). But unlike the situation found with the Bantu languages, the languages of the Balkan belong to different language families, which indicate the areality of this phenomenon. Consider the following examples from Macedonian (Slavic), Albanian, and Romanian (Romance); in Macedonian (7), object indexing is obligatory for specific direct objects (Franks and King 2000: 115). In Greek, exemplified in (8), DOI is never obligatory but preferred if the referent is topical: the sentence in (8) can be the answer to the questions ‘Who read the book?’ or ‘What did Ana do to/with the book?’, but not to a question like ‘What did Ana read?’, where ‘the book’ would be in focus (Kallulli 2000: 219). In Romanian, shown in (9), DOI with post-verbal objects is depend-

114 | Erika Just and Slavomír Čéplö

ent on the presence of the special preposition pe, which in turn is conditioned by the semantics as well as by definiteness of the referent (Cojocaru 2003: 34). (7)

Macedonian (Slavic, Northern Macedonia) [Franks and King 2000: 115] Marij *(go)=poznava učenikot/Vlado/toj učenik/nego. Marija 3SG.M.ACC=know.3SG pupil.DEF/Vlado/that pupil/3SG.M ‘Mary knows the pupil/Vlado/that pupil/him.’

(8)

Greek (Greece) I Ana to=diavase to vivlio DEF Ana 3SG.N.ACC=read DEF book ‘Ann did read the book/ read the book’

[Kallulli 2000: 219]

(9)

Romanian (Romance, Romania) O=aştept pe ama. PREP mom 3SG.F.ACC=wait.1SG ‘I’m waiting for mom.’

[Cojocaru 2003: 34]

These examples illustrate what the hitherto largest typological study of DOI, Iemmolo (2011), already suggested, namely that DOI across languages is systematically associated with signaling high salience or prominence of the object referent. Furthermore, the findings from Ruuli (Just and Witzlack-Makarevich 2018) as well as other usage based studies of differential marking phenomena (such as Schikowski 2013 on differential object case marking in Nepali or Schnell 2018 on subject indexing in Vera’a) suggest that every time a structure in any language is described as optional, this should be read with a grain of salt. Before we have another look into DOI in Maltese and our analysis based on bulbulistan corpus data (Čéplö 2018b), we will briefly look into differential indexing in other Semitic languages.

2.2 DOI in Semitic languages Before turning back to DOI in Maltese and our investigation of the factors responsible, we will briefly discuss DOI constructions in other Semitic languages. Amberber (2008) provides an overview of Amharic differential case marking, including an account of differential indexing. In Amharic (Ethiopia), the basic word order is SOV, and direct objects are differentially case marked, with definite objects only receiving the case marker -n (Amberber 2008: 745–746). If a direct object is case marked in this manner, it can also be indexed. By implication, only definite referents can be indexed, but they don’t have to be; the indexing is thus described as “optional” (Amberber 2008: 745). See (10a–b) for an overview of the situation in Amharic:

Differential object indexing in Maltese | 115

(10) a.

b.

Amharic (Semitic, Ethiopia) [Amberber 2008: 745] ləmmɑ t'ərmus-u-n səbbər-ə-(w) Lemma bottle-DEF-ACC break.PFV-3M.SBJ-3M.OBJ ‘Lemma broke the bottle.’ ləmmɑ and t'ərmus səbbər-ə-(*w) Lemma one bottle break.PFV-3M.SBJ-3M.OBJ ‘Lemma broke one bottle.’

In example (10a), the direct object t'ərmus is definite, and is thus obligatorily case marked. Additional to the subject, which is always indexed in Amharic, the direct object too can be indexed, as indicated by the brackets. In the second example, (10b), the direct object is indefinite, not case marked, and thus cannot be indexed. As for varieties of Arabic, DOI is not only attested for Maltese but also for varieties of the Levant, including northern Iraq, and Central Asia (see Souag 2017 for an overview). Like with Bantu languages, the indexes within the Arabic languages are similar in form. Examples (11a–c) show that in Lebanese Arabic, the direct object index – which can also be used pronominally without cooccurring noun phrase – can be absent (as in 11b) and present (example 11c) alongside the object noun phrase, a circumstance which is called “optional” by Aoun (1999:17): (11) a.

b.

c.

d.

Lebanese Arabic (Semitic, Lebanon) [Aoun 1999: 14–17] kariim ʔakal suuʃi Karim eat.3SG.M.PFV sushi ‘Karim ate sushi.’ kariim ʃeef-o Karim see.3SG.M.PFV-3SG.M.ACC ‘Karim saw him.’ kariim ʃeef-o la-saami Karim see.3SG.M.PFV-3SG.M.ACC PREP-Sami ‘Karim saw Sami.’ kariim ɦəkee-lo la-saami ɦkeeye Karim tell.3SG.M.PFV-3SG.M.DAT PREP-Sami story ‘Karim told Sami a story.’

However, in contrast to Maltese, there is not only differential direct object and differential indirect object indexing (as in 11d) in Lebanese Arabic, but also differential marking involving bound forms of possessors, shown in (11e) and of prepositional complements, as in (11f). Both these examples would also be possible without the bound person form.

116 | Erika Just and Slavomír Čéplö

(11) e.

f.

kariim ʃeef Karim see.3SG.M.PFV ‘Karim saw Sami’s book.’ kariim raaɦ Karim go.3SG.M.PFV ‘Karim went with Sami.’

kteeb-o book-3SG.M

la-saami PREP-Sami

maʕ-o with-3SG.M

la-saami PREP-Sami

Another contrast to Maltese is that in Lebanese Arabic, a co-referential lexical noun phrase – be it the object, possessor or prepositional complement – has to be preceded by the preposition la-. This “dummy case assigner” (Souag 2017: 49) can never occur with these noun phrases if there is no additional bound person form, with the exception of indirect objects (Aoun 1999: 17). Differential indexing in the varieties of the Levant has also been associated with information-structural properties of the referent, both topicality (e.g. Cowell 1964 and Brustad 2000 on Syrian Arabic) and focality or emphasis (Levin 1987 for northern Palestinian varieties). Souag (2017) clarifies that although one might be tempted to trace DOI in Arabic varieties back to their shared heritage, the DOI constructions differ greatly from one language to another, and show more similarities with contact languages than with their genetically more closely affiliated languages. This claim has been made before for different regions of the Arabic speaking world, but Souag (2017) provides the first microtypological investigation of the clitic doubling phenomena, which include differential object indexing. He looks at each region where these constructions have been reported, making comparisons and pointing at the links with adstrates languages. For Maltese, he demonstrates the similarities in differential indexing to Sicilian and the dissimilarities to Levant Arabic, concluding that “Maltese clitic doubling is thus better explained as the result of Sicilian superstratum influence than as a retention from some early stage of Arabic” (Souag 2017: 61). This suggests that a closer study of this contact phenomenon is a desideratum, contingent on our full understanding of DOI in Maltese and all the relevant variables.

3 Factors licensing differential object indexing in Maltese: a corpus based pilot study 3.1 Research questions As noted in Section 1.2, DOI in Maltese appears in contexts where the referent can be described as being topical. This is not surprising, as crosslinguistically,

Differential object indexing in Maltese | 117

DOI is generally and systematically associated with signaling high salience or prominence of the object referent (Iemmolo 2011). However, in Section 2 we have shown that saliency and topicality are quite nebulous concepts, and although we find recurring factors being involved (such as identifiability, givenness or animacy), the different variables have a different weight and interact to different degrees across different languages. From these observations, the following research questions arise for Maltese: 1) What are the factors which license differential object indexing in the presence of a co-referential overt NP in Maltese? 2) Which of these factors are the strongest predictor(s), i.e., which make it more probable for an object to be indexed? 3) Are the factors hierarchically ordered? Research question 1) aims at identifying the variables which are basically relevant for topicworthiness in Maltese, as the number one candidate out of the variety of all potential factors discussed in the literature on Maltese, other Semitic languages, and differential argument marking in general. Research question 2) seeks to compare the effect of each of the relevant factors, and with 3), we set out to find out more about how the factors interact; for instance whether one variable, such as animacy, is only statistically relevant if another feature (e.g. givenness or specificity) is involved, too. These three research questions logically build on one another. In the next section, we describe how we went about the corpus annotation which leads to answering question number 1) which in turn forms the starting point for questions 2) and 3).

3.2 Variables In order to find out which variables connected to topicworthiness of a direct object are relevant for Maltese, a special layer of annotation was added to a sample of sentences taken from the bulbulistan corpus (BC; Čéplö 2018b). Based on previous findings on the phenomenon of DOI, transitive and ditransitive clauses were annotated for the following formal and semantic variables, which are displayed in Table 1, along with their values. Below, some of the variables will be briefly discussed in more detail. Clauses with overt object noun phrases, as well as without (i.e. with pronominal reference), were annotated in order to see whether – or for which variables – there are differences with regard to the topicworthiness of overtly expressed and non-overtly expressed referents.

118 | Erika Just and Slavomír Čéplö

Table 1: Features for the quantitative analysis of DOI in Maltese.

Variable

Values

1

index

present absent

2

presence of lil

present absent

3

word order

SVO VOS OVS SOV OSV VSO

4

person, number and gender of the 1SG 2SG referent 3SG.F 3SG.M 1PL 2PL 3PL

5

referent semantics

human kinship animal anthropomorphic physical object event abstract entity

6

PoS of the object

noun pronoun NA (i.e. non-overt)

7

subcategory of the head noun

proper noun common noun personal pronoun impersonal use of personal pronoun possessive pronoun demonstrative pronoun interrogative pronoun

8

modification of the noun phrase

modified not modified

9

subcategory of modification

adjective numeral determiner relative clause possessive multiple

Differential object indexing in Maltese | 119

Variable

Values

10

identifiability

definite specific non-specific

11

textual givenness

given new

12

clause type

main relative clause adverbial clause complement

13

clause polarity

positive negative

14

sentence type

declarative imperative interrogative exhortative

It has been mentioned before that the preposition lil has to be used for referents which can be referred to by a proper name, i.e. for human referents or humanlike referents. This feature was included as being potentially relevant for indexing due to the finding by Borg and Azzopardi-Alexander (2009: 73–74) that with preverbal indexed objects, lil is no longer obligatory for referents which would usually require its presence. Part of speech of the coreferential noun phrase was included due to the findings from other languages, such as Sambaa or French (cf. examples 3 and 4 in Section 2.1), where word classes are directly relevant for indexing. Whether the object noun phrase is modified or not was included for two reasons: firstly, to account for the length of the phrase, although this is admittedly not a very precise way of measurement. And secondly, referents which require modification can be considered less identifiable by the addressee than referents which do not require modification. This brings us to the next variable which requires a brief explanation: identifiability. Together with givenness it was used as a proxy for the information-structural status of the referent. With identifiability we aim at capturing the extent to which a referent can be explicitly identified by the speaker and the hearer. The concept of definite as it is used in our annotation is based on the notions of uniqueness and familiarity and describes that a referent can be identified by both the speaker and the hearer. A specific referent, in turn, is unambiguously identifiable by the speaker only, and referents we labeled non-specific are not identifiable, neither by the speaker nor the hearer (Lyons 1999). Due to the various notions associated with the term givenness and the apparent fuzzi-

120 | Erika Just and Slavomír Čéplö

ness of subdividing categories (cf. Baumann 2012), we decided to code only for the two values new and given within the preceding discourse. So in total, it was decided to add a layer of annotation of fourteen variables, thirteen of which were considered possible independent variables, and the presence vs. the absence of the index as the response variable.

3.3 Corpus, challenges and preliminary solutions Immediately upon starting the annotation, we stumbled across a significant problem: it turns out that in the existing corpora of Maltese, DOI is extremely rare. We initially set out to annotate the Maltese Universal Dependencies Treebank (MUDT, Čéplö 2018a) which, like the larger bulbulistan corpus, is composed of four text types – newspaper, fiction, non-fiction and parliament, i.e. parliamentary debates (Čéplö 2018b: 58–2, 172–176) – which fall into two categories based on their origin: written (newspaper, fiction and non-fiction) and spoken (parliament). When selecting the text samples to conduct analysis, we quickly found that in the fictional texts in MUDT (which we selected for annotation so as to avoid boredom), DOI hardly ever occurs. One explanation for this would be that this phenomenon is much more prevalent in spoken Maltese. To confirm this would require a sufficiently large corpus of spoken Maltese, still a desideratum in Maltese linguistics. This would be quite a finding, since such a split had not been mentioned in previous literature, but would not be surprising given that this is similarly the case in other DOI systems which have not yet fully grammaticalized (cf. Section 2.1 above). The other explanation, of course, is that DOI in Maltese is actually quite rare across the board (cf. Čéplö 2018b: 235), contrary to assessments such as that provided by Borg and AzzopardiAlexander (1997: 126) who describe DOI with preverbal objects as “a wide spread characteristic of Maltese”. To go about this challenge of rare occurrences of the phenomenon under investigation, we abandoned any attempts to annotate MUDT and instead focused on the larger and more general bulbulistan corpus, while limiting ourselves to the parts of it that come closest to naturalistic speech, namely the parliamentary debates transcripts.6 To control for variations in semantics, valency etc., we decided to limit our investigation to a single verb; to ensure there would be enough mate-

|| 6 We refer to these transcripts as “coming close to naturalistic speech” and not “naturalistic speech”, as a comparison of randomly selected transcripts with their audio recordings has made it clear that some editing was performed (Čéplö 2018b: 58).

Differential object indexing in Maltese | 121

rial for the analysis, we selected the verb għamel ‘to do/make’, since one of its forms, nagħmluha, is the most frequent verb with a direct object index in the parliament text type, with 4896 occurrences in total. We therefore extracted all (orthographic) sentences (cf. Čéplö 2018b: 63–64) containing the keyword nagħmlu ‘we do/make’ (without index) and nagħmluha ‘we do/make it(F)’ (with an index); the preceding and the following 1000 characters were extracted as well to be able to account for the context. This, incidentally, had the positive side effect to control for verb semantics at the same time. We randomized their order and then proceeded to annotate all the relevant clauses, meaning all the clauses which contained nagħmlu with an overt object NP, and all the clauses containing nagħmluha with or without an overt object NP. We excluded clauses which contained nagħmluha where the direct object has a pronominal function, more specifically that of an expletive pronoun (something rarely discussed in the literature), such as the very frequent (ħalli) nagħmluha ċara “(let’s/in order to) make it clear”. In this manner, we ended up with a sample subcorpus of the bulbulistan corpus of 73555 words of parliamentary debates which contain 286 relevant clauses for annotation, with 133 instances of nagħmlu plus noun phrase, and 153 occurrences of nagħmluha plus noun phrase. Due to the nature of a text type such as parliamentary debates, we had to face some cutbacks with regard to our variables. Because of the topics discussed in parliament, as well as the limitation to one verb nagħmlu ‘we do/make’, the referent semantics were restricted to non-animate referents. As a consequence, there were also no objects accompanied by lil. Therefore, a more in-depth investigation of referent semantics as described in Section 3.2 as well as the interplay of object indexing and the presence of the preposition lil are two of the objectives reserved for a follow-up study of DOI in Maltese, which should be based on naturalistic data of spoken Maltese.

3.4 Findings As laid out above, DOI in general as well as in Maltese cannot be explained by hard and fast rules, but has to be explained on the basis of a set of variables. This chapter presents the starting point for a quantitative analysis of these variables. In the following, we provide some descriptive statistics of object indexing in Maltese on the basis of our corpus annotation of clauses containing nagħmlu and nagħmluha of those variables we found to be relevant, before presenting the evaluation of the variable interplay based on conditional inference. The first variable we found to have an impact on indexing is givenness, with given referents, i.e. referents which were mentioned within the preceding 1000

122 | Erika Just and Slavomír Čéplö

characters, being more likely to be indexed. Figure 1 shows that the proportion of given referents in clauses with indexed object noun phrases (nagħmluha) is much larger than the proportion of new referents with this verb form. Consequently, with nagħmlu (no index) we find more new than given referents.

Figure 1: Indexing and givenness of referent, clauses with overt objects only.

Similarly, turning to our second proxy for information structure, identifiability, Figure 2 reveals the distribution of indexed and non-indexed objects over definite, specific and non-specific referents. We see that non-specific referents, i.e. referents which are neither identifiable by the hearer nor the speaker, are never indexed. The proportion of definite and specific referents is fairly equal among the non-indexed referents, whereas for indexed referents, the definite referents outweigh specific ones by far. The third variable we found to have a fair impact on indexing is the part of speech of the object noun phrase. In Figure 3 we see that if the referent is realized pronominally, it is very likely to be indexed: the number of non-indexed pronouns is very low. Lastly and unsurprisingly, word order also had a strong impact on indexing. In our data we found no occurrences of preverbal objects which were not indexed, so all preverbal objects are indexed. Additionaly we can see in Figure 4 that postverbal objects are less likely to be indexed.

Differential object indexing in Maltese | 123

Figure 2: Indexing and identifiability, overt objects only.

Figure 3: Indexing and PoS of the object noun phrase.

124 | Erika Just and Slavomír Čéplö

Figure 4: Indexing and order of verb and object.

All other variables, modification of the noun phrase, clause and sentence types and polarity, have not (yet) shown to be of relevance. However, the data base for the present pilot study is still small, and, as has been mentioned above, a corpus of naturalistic spoken Maltese would be ideal. A follow-up study will hopefully not only provide us with more discourse contexts to work with, but also provide the opportunity to look into more variables such as a fine grained analysis of referent semantics, predicate semantics, the presence of lil and register. Although our database is not yet ideal for a thorough investigation on DOI in Maltese, we nevertheless want to get an idea of how the variables described above are weighted against each other. A nice way to go about a probabilistic distribution of a response variable (indexing in our case) is using conditional inference trees (Tagliamonte and Baayen 2012; Levshina 2015). Like a logistic model, a decision tree makes a prediction of an outcome based on given variables. In our case, the outcome is binary, that means we have two alternative responses: indexed P and not indexed P. Tree-based methods have some advantages over other statistical models. Their visualization makes them interpretable in a straightforward way, as the prediction process can be followed quite easily. The order of interactions is mirrored in the trees’ nodes, where the splits occur. Also, tree-based methods can handle missing data quite well and are especially robust in cases with a

Differential object indexing in Maltese | 125

relatively high number of variables compared to the sample size of the data, as in our case. The recursive partitioning of conditional inference trees, as used in the present study, is based on repeated significance tests, providing better predictive performance than simple decision trees (cf. Hothorn et al. 2006). The latter can show high variance and can be prone to overfitting. Once the variable with the strongest association with the response variable is identified, the algorithm makes a binary split and subdivides the dataset into two subsets; this is then repeated with the next variable. The first tree we present in Figure 5 shows the effects of all possible predictors mentioned above: givenness, identifiability, PoS of the noun phrase and order of verb and object.

Figure 5: Conditional inference tree with all potential predictors for indexing

All splits are significant at the level of 0.05. The first split at the first node at the top divides the dataset into two, based on word order. The first subset, with OV (object-verb) word order, branches to the left, and the second subset, which entails clauses with VO (verb-object) order, branches to the right. This means that the variable word order has the strongest association with indexing, and we here see again that in our data, all preverbal objects are indexed (node 2). The

126 | Erika Just and Slavomír Čéplö

strongest predictor for the subset of VO is discourse accessibility (node 3); the probability for given referents (node 4) to be indexed lies at 100% if they are also pronominal (node 5), and at around 20% if the object is a proper noun (node 6). For discourse-new referents, a split occurs (at node 7) on the basis of identifiability. As the splits are always binary, the two values definite and specific are grouped together and opposed to the value non-specific, which is a 100 percent indicator of non-indexing in the data. In a nutshell, the tree shows that the strongest predictor for object indexing in Maltese is word order, followed by discourse accessibility, with given referents being more likely to be indexed. Within the given referents, PoS is the strongest predictor for indexing, and for the new referents, it is identifiability, with no significant split between definite and specific referents. Figure 5 shows that the strongest predictor for DOI in Maltese seems to be word order; but just as object indexing itself, we can assume that word order variation is a differential pattern depicting the argument’s semantic properties. Therefore, we built another tree model, without word order as a potential predictor. This second model in Figure 6 shows that without word order, identifiability is the strongest predictor for indexing (split at node 1).

Figure 6: Conditional inference trees without word order as predictor.

Differential object indexing in Maltese | 127

This time, definiteness outweighs the other two values. The second split within the {specific/ non-specific} subset then again occurs on the basis of this feature, which means that also the difference between specific and non-specific referents is significant, with specific referents having a likelihood of just over 20% of being indexed (node 3). As for the right branch, PoS subdivides the definite referents (nodes 6 and 7). In sum, disregarding word order, identifiability (definite vs. specific or non-specific) is the strongest predictor for indexing in our data, followed by PoS of the object NP, with pronouns (which are inhently definite) always being indexed.

3.5 Summary Looking into systems where object indexing is not grammaticalized, and has often been labeled optional (as is the case in Maltese), it is impossible to account for it on the basis of rules. All one can do is describe it bottom-up and investigate the variables which might correlate with it to various degrees. In this paper, we set out to do just that for Maltese using corpus data. Due to the realities of the available Maltese corpora, we had to deal with several cutbacks with regard to our database of factors: indexing overt objects is nearly non-existent within the written parts of the corpus, and rare in the quasi spoken transcripts of parliamentary debates. In order to find a comparable number of clauses with indexed and non-indexed referents, we extracted a random sample of clauses containing either nagħmlu or nagħmluha from the parliament text type, and added an additional layer of annotation of variables thought potentially relevant for topicworthiness and thus indexing. From the variables which were left, accessibility, identifiability, PoS, and word order turned out to be significant within our subcorpus. Using conditional inference trees, we were able to identify how the variables interrelate. Considering all four variables, word order is the strongest predictor for direct object indexing: in our data, all preverbal objects were indexed. The second strongest predictor among the postverbal objects is accessibility, with discourse-given-referents being much more likely to be indexed than discourse-new ones. However, leaving out word order and only looking at the referential properties, we are facing identifiability as being the strongest predictor for DOI in Maltese, followed by PoS of the noun phrase. In summary, DOI shows a strong correlation with identifiability of the referent. However, the results also show that indexing is not restricted to specific or definite referents, neither is there an absolute obligatoriness for these referents to be indexed. DOI in Maltese is therefore an instance of differential argument

128 | Erika Just and Slavomír Čéplö

marking as defined by Seržant and Witzlack-Makarevich (2018), i.e. that this marking strategy is not caused by the referents’ argument role, but other factors connected to it. These findings are neither surprising nor new. But they confirm what has been said about the inadequacy to try to find hard and fast rules to account for DOI where the phenomenon has not grammaticalized. Our approach shows that the findings of previous studies are in accordance with the outcome of a quantitative – though provisional – corpus study, and that such studies can help at getting a deeper understanding of the interactions of the different variables involved.

4 Outlook A complete investigation of a phenomenon such as DOI has to be done on the basis of spontaneous spoken data, as the triggering factors are often related to topicality or salience of the referent. To this day, there is no corpus of such data for Maltese, so for a preliminary investigation, we settled for a subcorpus of parliamentary transcripts to see what this pilot study might reveal with regard to the different factors triggering DOI. Our next step will be a more in-depth investigation of the matter once a suitable corpus is available. This will make it possible to account for more variables which might be of importance, too, such as animacy of the referent or modification of the noun phrase. Also, different verb semantics have to be investigated along with referential features in order to be able to make more profound statements about the nature of DOI in Maltese.

Note: This chapter appeared as part of EJ’s Dissertation at Leiden University, in Just, Erika. 2022. A functional approach to differential indexing – Combining perspectives from typology and corpus linguistics. Amsterdam: LOT. DOI: https://dx.medra.org/10.48273/LOT0620

Abbreviations 1, 2, 3 ACC COM DAT

1st, 2nd, 3rd person accusative comitative dative

Differential object indexing in Maltese | 129

DEF DEM DET F IPFV LOC M N OBJ PFV PREP SBJ

definite demonstrative determiner feminine imperfective locative masculine neuter object perfective preposition subject

References Amberber, Mengitsu. 2008. Differential case marking of arguments in Amharic. In Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case, 743–755. Oxford: Oxford University Press. Aoun, Joseph. 1999. Clitic-doubled arguments. In Kyle Johnson & Ian Roberts (eds.), Beyond principles and parameters: Essays in memory of Osvaldo Jaeggli, 15–42. Dordrecht: Kluwer. Aquilina, Joseph. 1959. The structure of Maltese: A study in mixed grammar and vocabulary. Malta: The Royal University of Malta. Arkadiev, Peter. 2010. Clitic doubling: Towards a typology. Paper presented at the Workshop on Clitics and Syntactic Typology. https://www.academia.edu/1695634/Clitic_doubling_Towards_a_typology (checked 01/24/2020) Baumann, Stefan. 2012. The intonation of givenness: Evidence from German. Berlin & Boston: De Gruyter Mouton. Béjar, Susana. 1999. Agreement alternations and functional licensing in Selayarese. Toronto Working Papers in Linguistics 16(2). 51–62. Bickel, Balthasar. 2010. Grammatical relations typology. In Jae Jung Song (ed.), The Oxford handbook of language typology, 399–444. Oxford: Oxford University Press. Bickel, Balthasar & Johanna Nichols. 2008. Case-marking and alignment. In Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case, 304–321. Oxford: Oxford University Press. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London: Routledge. Borg, Albert & Marie Azzopardi-Alexander. 2009. Topicalisation in Maltese. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese Linguistics, 71–81. Amsterdam & Philadelphia: John Benjamins. Bossong, Georg. 1982. Der präpositionale Akkusativ im Sardischen. In O. Winkelmann & Maria Braisch (eds.), Festschrift für Johannes Hubschmid zum 65. Geburtstag: Beiträge zur allgemeinen, indogermanischen, und romanischen Sprachwissenschaft, 579–599. Bern: Francke. Brustad, Kristen. 2000. The syntax of spoken Arabic: A comparative study of Moroccan, Egyptian, Syrian, and Kuwaiti dialects. Georgetown University Press.

130 | Erika Just and Slavomír Čéplö

Čéplö, Slavomír. 2014. An overview of object reduplication in Maltese. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 201–222. Berlin & Boston: De Gruyter Mouton. Čéplö, Slavomír. 2018a. Maltese universal dependencies treebank. In Joakim Nivre, Mitchell Abrams, Željko Agić et al., Universal Dependencies 2.3, LINDAT/CLARIN digital library at the Institute of Formal and Applied Linguistics (ÚFAL), Faculty of Mathematics and Physics, Charles University, http://hdl.handle.net/11234/1-2895 (checked 01/20/2020). Čéplö, Slavomír. 2018b. Constituent order in {Maltese}: A quantitative analysis. Prague: Charles University in Prague PhD thesis. Cojocaru, Dana. 2003. Romanian grammar. Durham: Slavic and East European Language Research Center. Comrie, Bernard. 1989. Language universals and linguistic typology: Syntax and morphology. Chicago: University of Chicago press. Cowell, Mark W. 1964. A reference grammar of Syrian Arabic: Based on the dialect of Damascus. Washington: Georgetown University Press. Culbertson, Jennifer. 2010. Convergent evidence for categorial change in French: From subject clitic to agreement marker. Language 86(1). 85–132. Dalrymple, Mary & Irina Nikolaeva. 2011. Objects and information structure. Cambridge: Cambridge University Press. De Cat, Cécile. 2004. On the impact of French subject clitics on the information structure of the sentence. Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe & Petra Sleeman (eds.), Romance languages and linguistic theory, 33–46. Amsterdam & Philadelphia: John Benjamins. De Cat, Cécile & Katherine Demuth (eds.). 2008. The Bantu Romance connection: A comparative investigation of verbal agreement, DPs, and information structure. Amsterdam & Philadelphia: John Benjamins. Downing, Laura J. 2018. Differential object marking in Chichewa. In Ilja A. Seržant & Alena Witzlack-Makarevich (eds.), The diachrony of differential argument marking, 41–67. Berlin: Language Science Press. Du Bois, John W. 1987. The discourse basis of ergativity. Language 63(4). 805–855. Fabri, Ray. 1993. Kongruenz und die Grammatik des Maltesischen. Tübingen: Max Niemeyer. Fabri, Ray & Albert Borg. 2002. Topic, focus and word order in Maltese. In Abderrahim Youssi, Fouzia Benjelloun, Mohamed Dahbi & Zakia Iraqui Sinaceur (eds.), Aspects of the dialects of Arabic today, 354–363. Rabat: Amapatril. Franks, Steven & Tracy H. King. 2000. A handbook of Slavic clitics. Oxford: Oxford University Press. Friedman, Victor A. 2008. Balkan object reduplication in areal and dialectological perspective. In Dalina Kallulli & Liliane Tasmowski (eds.), Clitic doubling in the Balkan languages, 35– 63. Amsterdam & Philadelphia: John Benjamins. Haig, Geoffrey & Diana Forker. Agreement in grammar and discourse: A research overview. Linguistics 56(4). 715–734. Haspelmath, Martin. 2013. Argument indexing: A conceptual framework for the syntactic status of bound person forms. In Dik Bakker & Martin Haspelmath (eds.), Languages across boundaries – Studies in memory of Anna Siewierska, 209–238. Berlin & Boston: De Gruyter Mouton. Hothorn, Torsten, Kurt Hornik & Achim Zeileis. 2006. Unbiased recursive partitioning: A conditional inference framework. Journal of Computational and Graphical Statistics 15(3). 651– 674.

Differential object indexing in Maltese | 131

Iemmolo, Giorgio. 2011. Towards a typological study of Differential Object Marking and Differential Object Indexing. Pavia: Università degli studi di Pavia PhD thesis. Just, Erika. in preparation. Towards a structural and functional comparison of differential A and P indexing. Just, Erika & Alena Witzlack-Makarevich. 2018. A corpus based analysis of P indexing in Ruuli. Paper presented at the Syntax of the world’s languages conference (Paris, 2018). Kallulli, Dalina. 2000. Direct object clitic doubling in Albanian and Greek. In Frits Beukema & Marcel Den Dikken (eds.), Clitic phenomena in European languages, 209–248. Amsterdam & Philadelphia: John Benjamins. Kallulli, Dalina & Liliane Tasmowski. 2008. Clitic doubling, core syntax and the interfaces. In Dalina Kallulli & Liliane Tasmowski (eds.), Clitic doubling in the Balkan languages, 1–32. Amsterdam & Philadelphia: John Benjamins. Klamer, Marian & František Kratochvíl. 2018. The evolution of differential object marking in Alor-Pantar languages. In Marian Klamer (ed.), The Alor-Pantar languages: History and typology, 69–95. Berlin: Language Science Press. Levin, Aryeh. 1987. The particle la as an object marker in some Arabic dialects of the Galilee. Zeitschrift für arabische Linguistik 17. 31–40. Levshina, Natalia. 2015. How to do linguistics with R data exploration and statistical analysis. Amsterdam & Philadelphia: John Benjamins. Lyons, Christopher. 1999. Definiteness. Cambridge University Press. Matić, Dejan & Daniel Wedgwood. 2013. The meanings of focus: The significance of an interpretation-based category in cross-linguistic analysis. Journal of Linguistics 49(1). 127–163. Muxí, Isabel. 1996. Optional participial agreement with direct object clitics in Catalan. Catalan Working Papers in Linguistics 5(1). 127–145. Priestley, Carol. 2008. A grammar of Koromu (Kesawai), a Trans New Guinea language of Papua New Guinea. Canberra: Australian National University Doctoral dissertation. Riedel, Kristina. 2009. The syntax of object marking in Sambaa: A comparative Bantu perspective. Leiden: Universiteit Leiden PhD dissertation. Schikowski, Robert. 2013. Object-conditioned differential marking in Chintang and Nepali. Zurich: University of Zurich PhD thesis. Schnell, Stefan. 2018. Whence subject-verb agreement? Investigating the role of topicality, accessibility, and frequency in Vera’a texts. Linguistics 56(4). 735–780. Seržant, Ilja A. & Alena Witzlack-Makarevich. 2018. Differential argument marking: Patterns of variation. In Ilja A. Seržant & Alena Witzlack-Makarevich (eds.), The diachronie of differential argument marking, 1–40. Berlin: Language Science Press. Skopeteas, Stavros & Gisbert Fanselow. 2010. Focus types and argument asymmetries: A cross-linguistic study in language production. In Carsten Breul & Edward Göbbel (eds.), Comparative and contrastive studies of information structure, 169–197. Amsterdam & Philadelphia: John Benjamins. Souag, Lameen. 2017. Clitic doubling and language contact in Arabic. Zeitschrift für Arabische Linguistik 66. 45–70. Sutcliffe, Edmund. 1936. A grammar of the Maltese language with chrestomathy and vocabulary. Oxford: Oxford University Press. Tagliamonte, Sali & R. Harald Baayen. 2012. Models, forests and trees of York English: Was/were variation as a case study for statistical practice. Language Variation and Change 24(2), 135–178. Taylor, Charles V. 1985. Nkore-kiga. London: Croom Helm.

132 | Erika Just and Slavomír Čéplö

Vella, Alexandra. 2009. On Maltese prosody. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz & Martine Vanhove (eds.), Introducing Maltese linguistics, 47–68. Amsterdam & Philadelphia: John Benjamins. Witzlack-Makarevich, Alena. 2010. Typological variation in grammatical relations. Leipzig: Universität Leipzig Doctoral dissertation. Witzlack-Makarevich, Alena, Saudah Namyalo, Anatole Kiriggwajjo, Zarina Molochieva & Amos Atuhairwe. 2019. A corpus of spoken Ruuli. Makerere University and Hebrew University of Jerusalem.

| Part III: Morphosyntax

Przemysław Turek

Arabic prepositions and their Maltese equivalents Abstract: The aim of this paper is showing how many Classical Arabic and dialectal Arabic prepositions have survived in Maltese and how many have disappeared. Moreover, the etymology of the Maltese prepositions of Arabic origin is determined. The data shows that the number of prepositions in Maltese has been reduced in comparison to e.g. the Tunisian or Syrian dialects of Arabic. Some prepositions have become multifunctional to a much higher degree than their equivalents in other dialects. The analysis shows that Maltese dialectal prepositions can be derived from simple or complex forms of Arabic prepositions or nouns used as prepositions. Keywords: classical and dialectal Arabic prepositions; etymology; Maltese prepositions

1 Introduction Initially a Siculo-Arabic dialect, Maltese gradually transformed into a separate language since approximately the middle of the 13th century. However, it still retains the basic elements of the Arabic layer regardless of phonological and lexical changes. Notwithstanding many competing ideas about the dialect(s) from which the Maltese language has developed, most of its features, especially historical ones, point to a West Arabic Maghrebian origin. It suffices to mention some of the features that Maltese shares with, for example, Andalusian and Tunisian dialects, such as – phonological merger of /ḍ/ and /ḏ¢/ (ẓ), cf. Arabic ẓalām ‘darkness’ and Maltese dlam ‘darkness’ (Corriente 2006: 102–103), – ’imāla, i.e. the fronting and raising of Old Arabic /ā/ toward /ī/ (Levin 2007: 313–314),

|| Przemysław Turek: Institute of the Middle and Far East, Faculty of International and Political Studies, Jagiellonian University, Oleandry Str. 2A, 30-063 Kraków, Poland, E-Mail: [email protected]

https://doi.org/10.1515/9783110783834-006

136 | Przemysław Turek

–

–

retaining of most diphthongs, e.g. sajf ‘summer’, sejf ‘sword’, mewt ‘death’, sawm ‘fasting’, but cf. fuq ‘on, over, above’ (Aquilina 1990: 826, 1256, 1262, 1286; Saari 2003) the conservation of interdental phonemes in the early stages of development in the 16th–17th centuries evidenced by entries in Megiser’s word-list (1610: 11) such as Veheb ‘gold’, cf. Arabic ḏahab and Maltese deheb, 14 – Fne ‘two’, cf. Arabic iṯnāni, genitive-accusative iṯnayni, and Maltese tnejn; (Mifsud 2008: 146).

Some features of the Maltese language, such as the change of the erstwhile uvular stop /q/ to the glottal stop /ʔ/, were treated by some researchers as evidence of the Levantine origin of the dialect from which Maltese supposedly developed. Stumme (1904: 83) claims for instance, that “das Maltesische ist im Grunde genommen kein Maghreb-, sondern ein Syro-Arabisch” [‘Basically, Maltese is not Maghreb-, but Syro-Arabic’]. However, the said change is rather late in origin, as evidenced by various Maltese sources (Stumme 1904: 81–82; Vassalli 1796: 389). The development of the Maltese language was also influenced by the relatively long presence of the Muslim community, which constituted a significant part of the local society of the islands until 1240 (Brincat 2011: 71–72; Vella 1974: 100). It can therefore be assumed that, at least within this community, the diglossia survived: Siculo-Arabic dialect versus Classical Arabic, the latter being used in writing and in obligatory religious rites. The local dialect has been used most probably by all the inhabitants of Malta and Gozo (Muslims, Jews and Christians) in all low register situations. The Maltese language, an isolate Arabic dialect, which eventually developed into an independent language, retains in its lexicon numerous Arabic prepositions common to Classical/Modern Standard Arabic and other Arabic varieties as well.

2 The issue to be raised In this article, I focus on the analysis of Arabic prepositions in Classical/Modern Standard Arabic and Arabic dialects and their Maltese equivalents to show how many Classical Arabic and/or dialectal Arabic prepositions have survived in Maltese and how many have disappeared. Procházka (1993) made a similar analysis for all Neo-Arabic dialects although his work does not cover all Maltese prepositions. Full coverage of Maltese

Arabic prepositions and their Maltese equivalents | 137

prepositions is achieved in Saari’s (2003) PhD-thesis, whose impact on the general discussion is limited because the published version of the thesis is in Hebrew. Schmidt et al. (2020) analyze 57 Maltese prepositions, based on the etymological remarks of Aquilina’s (1987, 1990), dividing them into forms of Arabic and Romance origin respectively. Of the former, there are 44 whereas 12 are classified as Romance – one preposition being of uncertain origin (Schmidt et al. 2020: 249). I divide my analysis into four parts: – Section 3 is devoted to Classical Arabic/Modern Standard Arabic inseparable prepositions; – Section 4 covers separate prepositions; – Section 5 addresses nouns used as prepositions; – Section 6 looks at the Maltese prepositions as continuations of their dialectal Arabic equivalents. Contrary to Procházka’s approach (1993: 10–25, 26–45) who included 36 prepositions of Classical Arabic and 43 new prepositions in Modern Standard Arabic according to their appearance in Wehr’s dictionary (1994), I investigate all Classical Arabic prepositions remaining in use in Modern Standard Arabic. Besides, almost all prepositions marked as characteristic for Modern Standard Arabic, are discussed in Classical Arabic lexicographic sources (cf. e.g. Ibn Manẓūr 1981). Therefore, I treat all the discussed prepositions listed below as a diachronic system in continuity and change. My analysis does not cover (with few exceptions) the compound prepositions or phrasal prepositions (Stolz 2020). In their majority, these compound prepositions follow the classical constructions attested in Arabic lexicographical sources and in the descriptive grammars of Arabic dialects (e.g. Ryding 2005: 366–400; Wright 2005: 129–212). Exceptions are those constructions which involve a Romance element as second component. The three parts include 45 prepositions found in Classical and Modern Standard Arabic, listed in the oldest lexicographic sources, including the Koran (Haywood and Nahmad 1990: 412–421; Ibn Manẓūr 1981: passim; Penrice 1873: passim; Ryding 2005: 366–400; Wright 2005: passim). I omitted such rare nominal forms of prepositions as ‫ ﺇﺑﺎﻥ‬ʔibbāna ‘during’ (Ibn Manẓūr 1981: 12), ‫ ﺇﺛﺮ‬ʔiṯra ‘right after; immediately after’ (Ibn Manẓūr 1981: 27), ‫ ﺇﺯﺍء‬ʔizāʔa ‘facing; in the face of’ (Ibn Manẓūr 1981: 75–76), ‫ ﺃﺛﻨﺎء‬ʔaṯnāʔa ‘during’ (Haywood and Nahmad 1990: 528), as well as the form ‫ ﻋﻘﺐ‬ʕaqiba ‘right after, immediately after’ (Ibn Manẓūr 1981: 3023), and not used e.g. in Arabic translations of the Bible (see below) – all of them absent from Maltese. The Classical/Standard Modern Ara-

138 | Przemysław Turek

bic prepositions as well as the Arabic quotations are given in their full (nonpausal) forms, preserving the case endings. To get the best possible equivalents, I use examples taken from Arabic and Maltese translations of the Bible provided with an English translation of the King James’ version (Il-Bibbja 2013; Al-Kitāb al-muqaddas 1982; Al-Kitāb almuqaddas 1988; Al-Kitāb al-muqaddas 2009; The Holy Bible 1953). The diachronic continuity of Maltese prepositions has been verified on the basis of Maltese lexicographic sources (e.g. de Soldanis 1750; Aquilina 1987, 1990; Vassalli 1796). To emphasize the relationship of Maltese with the Maghreb dialects, I give examples taken from the Tunisian dialect. In addition, I present some morphological phenomena, typical for the Syrian dialect and attested also in Maltese. In Section 6, I analyze 12 Maltese prepositions of Arabic dialectal origin. Therefore, I took into consideration the Maltese translation of the Bible as a primary source; the Arabic translations of the Bible were used to show the dialectal use/development of the Arabic prepositions in Maltese in comparison to the strictly literary use of the prepositions in the above-mentioned Arabic translations.

3 Inseparable prepositions in Classical/Modern Standard Arabic and their Maltese equivalents The following Classical Arabic inseparable prepositions exist in almost all Semitic languages and are of primary origin (Bergsträsser 1995: 19–20; 222–223; Procházka 1993: 7). For the presentation of the data, I have chosen the format of tables with the data from Classical/Modern Standard Arabic in the left column and those of Maltese in the right column (see Table 1). Sentential examples are assigned numbers and are marked as AR for Arabic and MT for Maltese. No abbreviation is used for English. Since my main interest is philological I refrain from providing morpheme glosses for the examples. In this group only one preposition, Arabic ka- ‘as, like’ is not found in Maltese, although it could appear in some Arabic dialects, retaining its literary character. Its Maltese counterpart is usually bħal ‘like, similar to, as’ (MT3, MT47, MT54, MT62, MT63). In Arabic as well as in Maltese the preposition b(i)- is not to be separated from the following noun, (AR1, AR2, MT1, MT2, MT9, MT51). It is attested in the oldest Maltese written sources (15th century).

Arabic prepositions and their Maltese equivalents | 139

Table 1: Inseparable prepositions.

Classical/Modern Standard Arabic

Maltese

‫ ﺑـ‬bi- ‘in, by, with’ etc.1 (AR1) ʕaynun bi-ʕaynin wa-sinnun bi-sinnin (AlKitāb al-muqaddas 1982, II: 9) An eye for an eye, and a tooth for a tooth; (The Holy Bible 1953, II: 7, Matthew, 5:38) (AR2) ʔin kuntu ʔatakallamu bi-ʔalsinati n-nāsi wa-l-malāʔikati wa-lākin laysa lī maḥabbatun, (Al-Kitāb al-muqaddas 1982, II: 283) Though I speak with the tongues of men and of angels, and have not charity, (The Holy Bible 1953, II: 207, I Corinthians 13:1)

bi ‘with; in, by; of’2, cf. Tunisian Arabic b(i)‘with’3 (MT1) Għajn b’għajn u sinna b’sinna. (IlBibbja 2013: 1360) (MT2) Li kont nitkellem bl-ilsna tal-bnedmin u tal-anġli bla ma kelli l-imħabba, (Il-Bibbja 2013: 1586)

‫ ﻙ‬ka- ‘as, like’4 (AR3) li-ʔanna-hu kāna yuʕallimu-hum ka-man la-hu sulṭānun wa-laysa ka-l-katabati (Al-Kitāb al-muqaddas 1982, II: 57) for he taught them as one that had authority, and not as the scribes. (The Holy Bible 1953, II: 120, Mark, 1:22)

not used in Maltese, cf. Syrian Arabic ka- ‘as’5 (MT3) għax beda jgħallimhom bħal wieħed li għandu s-setgħa, u mux bħall-kittieba. (IlBibbja 2013: 1400)

‫ ﻟـ‬li- ‘for, to, because of’, used to express the Dative and denotes possession6 (AR4) ʔaʕṭū mā li-Qayṣara li-Qayṣara wa-mā liLlāhi li-Llāhi (Al-Kitāb al-muqaddas 1982, II: 79) Render to Caesar the things that are Caesar’s, and to God the things that are God’s. (The Holy Bible 1953, II: 120, Mark, 12:17)

lil ‘to (Dative case)’7, cf. Tunisian Arabic l- ‘to, for etc.’8 (MT4) Agħtu lil Ċesari dak li hu ta’ Ċesari, u lil Alla dak li hu ta’ Alla. (Il-Bibbja 2013: 1418)

|| 1 Ibn Manẓūr (1981: 197); Penrice (1873: 14); Wright (2005, II: 156–164). 2 de Soldanis (1750: 113); Aquilina (1987: 113–114); Panzavecchia (1845: 247, 248 b(i)); Vassalli (1796: 51b); Wettinger (2006: 74, 1500); Wettinger and Fsadni (1983: 47–48, Caxaru be); Wettinger and Fsadni (1983: 42, c. 1672). 3 Ben Abdelkader et al. (1977: 33); Zavadovskij (1979: 74). 4 Ibn Manẓūr (1981: 3957); Penrice (1873: 122–123); Wright (2005, II: 177–178). 5 Cowell (2005: 488). 6 Ibn Manẓūr (1981: 4102); Penrice (1873: 129); Wright (2005, II: 147–153). 7 de Soldanis (1750: 82); Aquilina (1987: 750); Panzavecchia (1845: 247 l, li e lil); Vassalli (1796: 428, 440e). 8 Ben Abdelkader et al. (1977: 59, 155, 413); Zavadovskij (1979: 74).

140 | Przemysław Turek

One has to distinguish the Maltese variant of the preposition li- ‘to’, i.e. lil(-), from the shortened form of the relative pronoun illi ‘who, that, which’, i.e. li (Aquilina 1987: 566, 744–745). The Maltese continuation of Arabic li- can appear as inseparable or separate preposition (MT28, MT4, MT44).

4 Separate Classical/Modern Standard Arabic prepositions and their Maltese equivalents On account of the number of prepositional types, the data is distributed over two consecutive tables. Three of ten Arabic prepositions listed in Tables 2–3, i.e. ‫ ﺇﻟﻰ‬ʔilā ‘to, unto, until’, ‫ ﻋﻠﻰ‬ʕalā ‘over, on, against’ and ‫ ﻣﻦ‬min ‘from’, are used in many Semitic languages and represent the oldest Semitic layer (Bergsträsser 1995: 19–20; 222–223; Procházka 1993: 7). The remaining forms are of later origin and can be derived partially from six basic prepositions (Procházka 1993: 11–21). Table 2: Separate prepositions I.

Classical/Modern Literary Arabic

Maltese

‫ ﺇﻟﻰ‬ʔilā ‘to, unto, until’9

’il particle indicating direction, precedes few

(AR5) yā ṣadīqu rtafiʕ ʔilā fawqin (Al-Kitāb al-

adverbs of place10

muqaddas 1982, II: 122)

(MT5) Ħabib, itla f’post aktar ’il fuq. (Il-Bibbja

Friend, go up higher: (The Holy Bible 1953, II:

2013: 1454)

120, Luke, 14:10) ‫ ﺑﻼ‬bi-lā ‘without’11

bla ‘without’12, cf. Tunisian Arabic blǟ or blǟš

(AR6) Man kāna min-kum bi-lā ḫaṭiyyati fa-l-

‘without’13

yarmi-hā ʔawwalan bi-ḥaǧarin (Al-Kitāb al-

(MT6) Min fostkom hu bla dnub jitfgħalha hu

muqaddas 1982, II: 161)

l-ewwel ġebla. (Il-Bibbja 2013: 1485–1486)

He that is without sin among you, let him first cast a stone at her. (The Holy Bible 1953, II: 120, John, 8:7)

|| 9 Ibn Manẓūr (1981: 120); Penrice (1873: 9); Wright (2005, II: 144–146). 10 Aquilina (1987: 564); Panzavecchia (1845: 247 el, il o l); Vassalli (1796: 364). 11 Wright (2005, II: 163). 12 Aquilina (1987: 128-129); Vassalli (1796: 428a). 13 Ben Abdelkader et al. (1977: 372).

Arabic prepositions and their Maltese equivalents | 141

Classical/Modern Literary Arabic

Maltese

‫ ﺣﺘﻰ‬ḥattā ‘up to, as far as’14 (AR7) nafs-ī ḥazīnatun ǧiddan ḥattā l-mawti (AlKitāb al-muqaddas 1982, II: 50) My soul is exceeding sorrowful, even unto death: (The Holy Bible 1953, II: 120, Matthew, 26:38)

not used in Maltese, cf. Tunisian Arabic ḥattǟ ‘until’15 (MT7) Għandi ruħi mnikkta għall-mewt; (IlBibbja 2013: 1392)

‫ ﻋﻠﻰ‬ʕalā ‘over, on, against’16 (AR8) ṯumma naqala min hunāka wa-ḥafara biʔran ʔuḫrā wa-lam yataḫāṣamū ʕalay-hā (AlKitāb al-muqaddas 1982, I: 41) And he removed from thence, and digged another well; and for that they strove not: (The Holy Bible 1953, I: 33, Genesis, 26:22) (AR9) fa-ġtamma ʕalā l-qawli wa-maḍā ḥazīnan (Al-Kitāb al-muqaddas 1982, II: 74) And he was sad at that saying, and went away grieved: (The Holy Bible 1953, II: 56, Mark, 10:22)

għal ‘for, at, on’;17 cf. Syrian Arabic ʕala ‘on, about, to, against’, var. ʕal with the article,18 and Tunisian Arabic ʕlǟ / ʕala ‘on’19 (MT8) Għalhekk telaq minn hemm, u ħaffer bir ieħor u ma ġġildux għalih, (Il-Bibbja 2013: 28) (MT9) Imma għal dan il-kliem ir-raġel qarras wiċċu u telaq b’qalbu sewda, (Il-Bibbja 2013: 1414)

As shown in Table 2, the Arabic compound preposition bi-lā ‘without’ has been preserved in Maltese as bla with elision of the short vowel (MT6). The prepositions ḥattā ‘up to, as far as’ and ʕan ‘from, about, concerning’ which occur in other dialects, are not attested in Maltese. They are replaced by other prepositions, e.g. a functional equivalent of ḥattā in Maltese is sa ‘till, until’ (Aquilina 1990: 1244–1245) = (MT65), and ʕan is rendered by dwar ‘on, about, concerning’ (MT55). In general, the preposition ʔilā ‘to’ has changed to l-, but in Maltese it keeps the form ’il (Bergsträsser 1995: 197; Panzavecchia 1845: 247).

|| 14 Ibn Manẓūr (1981: 768); Penrice (1873: 31); Wright (2005, II: 146–147). 15 Ben Abdelkader et al. (1977: 427); Zavadovskij (1979: 74). 16 Ibn Manẓūr (1981: 3091); Penrice (1873: 100); Wright (2005, II: 166–173). 17 de Soldanis (1750: 104 ghal); Aquilina (1990: 951–953); Panzavecchia (1845: 247 gɧal); Vassalli (1796: 342a Ոal ev ՈALA); Wettinger and Fsadni (1983: 47–48, Caxaru al,’al). 18 Cowell (2005: 476). 19 Ben Abdelkader et al. (1977: 266); Zavadovskij (1979: 74).

142 | Przemysław Turek

Table 3: Separate prepositions II.

Classical/Modern Literary Arabic

Maltese

‫ ﻋﻦ‬ʕan ‘from, about, concerning’20 (AR10) wa-ḫaraǧa ḫabarun ʕan-hu fī ǧamīʕi l-kūrati l-muḥīṭati (Al-Kitāb al-muqaddas 1982, II: 97) and there went out a fame of him through all the region round about. (The Holy Bible 1953, II: 73, Luke, 4:14)

not used in Maltese, cf. Syrian Arabic ʕan ‘about, from’21 (MT10) u l-fama tiegħu xterdet ma’ dawk linħawi kollha, (Il-Bibbja 2013: 1434)

‫ ﻓﻲ‬fī ‘in’22 (AR11) fī-l-badʔi kāna l-kalimatu (Al-Kitāb almuqaddas 1982, II: 145) In the beginning was the Word, (The Holy Bible 1953, II: 109, John, 1:1)

fi ‘in, at, inside, within, into’,23 cf. Tunisian Arabic fī ‘in, at’24 (MT11) Fil-bidu kien il-Verb, (Il-Bibbja 2013: 1473)

‫ ﻟﺪﻯ‬ladā, ‫ ﻟﺪﻥ‬ladun ‘at, by, upon, having; with’ = not used in Maltese (MT12) Jien ninsab quddiem il-qorti ta’ Ċesari; Latin ‘apud’25 (Il-Bibbja 2013: 1546) (AR12) ʔanā wāqifun ladā kursiyyi wilāyati Qayṣara (Al-Kitāb al-muqaddas 1982, II: 238) I stand at Caesar’s judgment seat, (The Holy Bible 1953, II: 175, Acts, 25:10) ‫ ﻣﻊ‬maʕa ‘with’26 (AR13) wa-ḏhab ʔawwalan iṣṭaliḥ maʕa ʔaḫī-ka (Al-Kitāb al-muqaddas 1982, II: 9) first be reconciled to thy brother, (The Holy Bible 1953, II: 7, Matthew, 5:24)

ma’ ‘with, at’27, cf. Tunisian Arabic mʕa/mʕǟ ‘with’28 (MT13) u mur l-ewwel irranġa ma’ ħuk, (IlBibbja 2013: 1359)

‫ ﻣﻦ‬min ‘from’ 29 (AR14) ʔiḏhab min ʕinda-nā li-ʔanna-ka ṣirta

minn ‘from’ indicating place, person, source)30, cf. Tunisian Arabic min ‘from’31

|| 20 Ibn Manẓūr (1981: 3143); Penrice (1873: 101); Wright (2005, II: 139–144). 21 Cowell (2005: 476). 22 Ibn Manẓūr (1981: 3505); Penrice (1873: 113); Wright (2005, II: 153–156). 23 Aquilina (1987: 333–334); Panzavecchia (1845: 247); Vassalli (1796: 195l); Wettinger and Fsadni (1983: 47–48, Caxaru fo, fi); Wettinger and Fsadni (1983: 42, c. 1672). 24 Ben Abdelkader et al. (1977: 33); Zavadovskij (1979: 74). 25 Ibn Manẓūr (1981: 4023); Penrice (1873: 131); Wright (2005, II: 165–166); Ryding (2005: 392). 26 Ibn Manẓūr (1981: 4234); Penrice (1873: 139); Wright (2005, II: 164–165). 27 Aquilina (1990: 780); Panzavecchia (1845: 247 ma e magꜧ); Vassalli (1796: 451l). 28 Ben Abdelkader et al. (1977: 449); Zavadovskij (1979: 74). 29 Ibn Manẓūr (1981: 4281–4282); Penrice (1873: 140–141); Wright (2005, II: 129–139). 30 de Soldanis (1750: 79 men); Aquilina (1990: 833–834); Panzavecchia (1845: 247 min); Vassalli (1796: 493k MYN vel mn, […] mynni); Wettinger and Fsadni (1983: 47–48, Caxaru min(n)); Wettinger and Fsadni (1983: 42, c. 1672 menn). 31 Ben Abdelkader et al. (1977: 160); Zavadovskij (1979: 74).

Arabic prepositions and their Maltese equivalents | 143

Classical/Modern Literary Arabic

Maltese

ʔaqwā min-nā ǧiddan (Al-Kitāb al-muqaddas 1982, I: 41) Go from us; for thou art much mightier than we. (The Holy Bible 1953, I: 33, Genesis, 26:16) (AR15) fa-min šaǧarati t-tīni taʕallamū l-maṯala (Al-Kitāb al-muqaddas 1982, II: 81) Now learn a parable of the fig tree; (The Holy Bible 1953, II: 61, Mark, 13:28)

(MT14) Itlaq minn magħna, għax int tqawwejt wisq iżjed minna. (Il-Bibbja 2013: 28) (MT15) Mis-siġra tat-tin tgħallmu din ilparabbola. (Il-Bibbja 2013: 1420)

‫ ﻣﻨﺬ‬munḏu ‘since’32 (AR16) munḏu wuḍiʕa l-ʔinsānu ʕalā l-ʔarḍi (AlKitāb al-muqaddas 1982, I: 811) since man was placed upon earth, (The Holy Bible 1953, I: 599, Job, 20:4) (AR17) munḏu ṣibā-hu (Al-Kitāb al-muqaddas 1982, II: 72) Of a child (The Holy Bible 1953, II: 64, Mark, 9:21) i.e. since he was a child [my explanation]

mindu ‘since’33 (MT16) mindu deher fid-dinja l-bniedem (IlBibbja 2013: 578) (MT17) Minn ċkunitu (Il-Bibbja 2013: 1412)

According to Table 3, the Maltese temporal adverb mindu ‘since’ is a continuation of Classical Arabic munḏu (originally a composed preposition min + ḏū ‘from’ + ‘possessor, owner’), not used in the dialects, which often replaced it with min ‘from’ (Wright 2005, II: 173). The Maltese form is only used adverbially preceding the verb, similarly to the use in Classical Arabic (Wright 2005, II: 174–175), cf. the Arabic and Maltese sentences above. The Maltese prepositional equivalent of Arabic munḏu ‘since’ is usually minn ‘from’ (MT17), indicating time (Aquilina 1990: 833). The lexicographers mention also the rare form minḏu (Wright 2005, II: 174), hence Maltese mindu (with dental d instead of interdental ḏ). Out of the ten Arabic prepositions of this group, six are used in Maltese in a form that is more or less similar to the original, i.e. ’il indicating direction, bla ‘without’ (MT2, MT6, MT21, MT31), għal ‘for, at, on’ (MT7, MT8, MT9, MT22, MT36, MT38, MT56), fi ‘in, at, inside, within, into’ (MT5, MT11, MT16, MT31, MT32, MT37, MT38, MT39, MT46, MT54), ma’ ‘with, at’ (MT13, MT56), and minn ‘from’. The Arabic prepositions ḥattā ‘up to, as far as’ and ʕan ‘from, about, concerning’, although attested in some Arabic dialects, have no continuation in Maltese. Some Arabic forms, e.g. ʕalā (Maltese għal) have replaced other prepositions and thus expanded their use and meaning (AR7, MT7, AR22, MT22, AR36, MT36).

|| 32 Ibn Manẓūr (1981: 4275); Ryding (2005: 384–386); Wright (2005, II: 173–175). 33 Aquilina (1990: 832); Vassalli (1796: 494d myndu melius mondu, 515s).

144 | Przemysław Turek

5 Classical/Modern Standard Arabic nouns used as prepositions and their Maltese equivalents A large group of prepositions in Classical /Modern Standard Arabic form construct-state nouns in the accusative, used as prepositions denoting time and place (Wright 2005, II: 178–188). The short vowel endings denoting accusative disappear in the pausal form and as such are also used in Arabic dialects. As mentioned earlier, I omitted 5 obsolete Classical Arabic forms; still this group, consisting of 32 prepositions is the largest one. The size of this group justifies the division of its membership over a succession of five tables of almost equal size (= Tables 4–8). The prepositions under review can also appear as elements of compound prepositions both in Arabic and Maltese (Stolz 2020: 446–449; Wright 2005, II: 188–190). Table 4: Denominal prepositions I.

Classic/Modern Literary Arabic

Maltese

‫ ﺃﻣﺎﻡ‬ʔamāma ‘before, opposite’ (of place)34 not used in Maltese (MT18) itellgħukom quddiem gvernaturi u (AR18) wa-tusāquna ʔamāma wulātin waslaten (Il-Bibbja 2013: 1367) mulūkin (Al-Kitāb al-muqaddas 1982, II: 18) And ye shall be brought before governors and kings (The Holy Bible 1953, II: 13, Matthew, 10:18) ‫ ﺑﻌﺪ‬baʕda ‘after’ of time or rank35 (AR19) baʕda hāḏā l-kalāmi bi-naḥwi ṯamāniyati ʔayyami (Al-Kitāb al-muqaddas 1982, II: 110) And it came to pass about eight days after these sayings, (The Holy Bible 1953, II: 95, Luke, 9:28) (AR20) yaʔtī baʕd-ī man huwa ʔaqwā minn-ī (Al-Kitāb al-muqaddas 1982, II: 56) There cometh one mightier than I after me, (The Holy Bible 1953, II: 42, Mark, 1:7)

not used in Maltese, cf. Tunisian Arabic baʕd ‘after’36 (MT19) Daqs tmint ijiem wara dan id-diskors, (Il-Bibbja 2013: 1445) (MT20) Ġej warajja min hu aqwa minni, (IlBibbja 2013: 1399)

‫ ﺑﻴﻦ‬bayna ‘between’37 (AR21) bayna-nā wa-bayna-kum huwwatun

bejn ‘between, amongst’38, cf. Syrian Arabic bēn ‘between, among’39 and Tunisian Arabic

|| 34 Penrice (1873: 9); Wright (2005, II: 187). 35 Ibn Manẓūr (1981: 310); Wright (2005, I: 281, II: 186–187). 36 Ben Abdelkader et al. (1977: 15, 46); Zavadovskij (1979: 74). 37 Ibn Manẓūr (1981: 403); Wright (2005, II: 180–181).

Arabic prepositions and their Maltese equivalents | 145

Classic/Modern Literary Arabic

Maltese

ʕaẓīmatun qad ʔuṯbitat (Al-Kitāb al-muqaddas 1982, II: 126) between us and you there is a great gulf fixed: (The Holy Bible 1953, II: 95, Luke, 16:26)

bīn ‘between’40 (MT21) hemm vojt bla qjies bejnkom u bejnna, (Il-Bibbja 2013: 1457)

‫ ﺕﺟﺎﻩ‬tiǧāha, tuǧāha ‘opposite’41 (AR22) huwa ruṭbun tuǧāha š-šamsi (Al-Kitāb al-muqaddas 1982, I: 800) He is green before the sun, (The Holy Bible 1953, I: 591, Job, 8:16)

not used in Maltese (MT22) Iħaddar ġmielu għal għajn ix-xemx, (IlBibbja 2013: 568)

‫ ﺗﺤﺖ‬taḥta ‘under, below’ of place or rank42 (AR23) wa-lā yūqidūna sirāǧan wa-yaḍaʕūnahu taḥta l-mikyāli (Al-Kitāb al-muqaddas 1982, II: 8) Neither do men light a candle, and put it under a bushel, (The Holy Bible 1953, II: 6, Matthew, 5:15)

taħt ‘under; beneath’43, cf. Tunisian Arabic taḥt ‘below, under’44 (MT23) Anqas ma jixegħlu l-musbieħ u jqegħduh taħt is-siegħ, (Il-Bibbja 2013: 1358)

‫ ﺗﻠﻘﺎء‬tilqāʔa ‘opposite; in front of’45

not used in Maltese (MT24) xħin timmira l-qaws kontrihom. (Il(AR24) tufawwiqu s-sihāma ʕalā ʔawtāri-ka Bibbja 2013: 614) tilqāʔa wuǧūhi-him (Al-Kitāb al-muqaddas 1982, I: 846) when thou shalt make ready thine arrows upon thy strings against the face of them. (The Holy Bible 1953, I: 625, Psalms, 21:12)

Table 5: Denominal prepositions II.

Classic/Modern Literary Arabic

Maltese

‫ ﺣﺬﺍء‬ḥiḏāʔa ‘opposite’ (AR25) huwa ḥiḏāʔa-ka ‘he is near you’ [my translation]46

ħada, ħdej(n) ‘beside, near, close to’47, cf. Tunisian Arabic ḥḏǟ ‘near’48 (MT25) Raġa’ beda jgħallem ħdejn il-baħar, (Il-

|| 38 Aquilina (1987: 95); Panzavecchia (1845: 247 bein); Vassalli (1796: 34a). 39 Cowell (2005: 488). 40 Zavadovskij (1979: 74). 41 Ibn Manẓūr (1981: 4776); Wright (2005, I: 281). 42 Penrice (1873: 22); Wright (2005, II: 182). 43 de Soldanis (1750: 190 tahht giù); Aquilina (1987: 1388); Panzavecchia (1845: 247 taԧt); Vassalli (1796: 65i). 44 Ben Abdelkader et al. (1977: 46, 426); Zavadovskij (1979: 74). 45 Ibn Manẓūr (1981: 4065); Wright (2005, I: 281).

146 | Przemysław Turek

Classic/Modern Literary Arabic

Maltese Bibbja 2013: 1403) And he began again to teach by the sea side: (The Holy Bible 1953, II: 46, Mark, 4:1)

‫ ﺣﺴﺐ‬ḥasaba ‘according to; in accordance with’49 (AR26) li-mā-ḏā lā yasluku talāmīḏu-ka ḥasaba taqlīdi š-šuyūḫi (Al-Kitāb al-muqaddas 1982, II: 68) Why walk not thy disciples according to the tradition of the elders, (The Holy Bible 1953, II: 51, Mark, 7:5)

not used in Maltese (MT26) Dan l-għala d-dixxipli tiegħek ma jġibux ruħhom skont it-tradizzjoni ta’ missirijiethom, (Il-Bibbja 2013: 1409)

‫ ﺣﻮﻝ‬ḥawla ‘round about’50 (AR27) ṯumma naẓara ḥawla-hu ʔilā ǧamīʕihum wa-qāla li-r-raǧuli mudda yada-ka (AlKitāb al-muqaddas 1982, II: 101) And looking round about upon them all, he said unto the man, Stretch forth thy hand. (The Holy Bible 1953, II: 76, Luke, 6:10)

not used in Maltese (MT27) U xeħet ħarsa madwaru fuqhom ilkoll, u qal lir-raġel: “Midd idek.” (Il-Bibbja 2013: 1437)

not used in Maltese ‫ ﺣﻮﺍﻟﻲ‬ḥawālay ‘approximately’51 (AR28) wa-ḥāraba ǧamīʕa ʔaʕdāʔi-hi ḥawālay- (MT28) u għamel gwerra lill-għedewwa tiegħu hi (Al-Kitāb al-muqaddas 1982, I: 449) kollha ta’ madwaru: (Il-Bibbja 2013: 318) and fought against all his enemies on every side, (The Holy Bible 1953, I: 336, I Samuel, 14:47) ‫ ﺧﺎﺭﺝ‬ḫāriǧa ‘outside; outside of’52 (AR29) wa-ḫaraǧa ḫāriǧa l-madīna (Al-Kitāb al-muqaddas 1982, II: 38) and went out of the city (The Holy Bible 1953, II: 29, Matthew, 21:17)

not used in Maltese (MT29) ħareġ barra mill-belt, (Il-Bibbja 2013: 1383)

‫ ﺧﻠﻒ‬ḫalfa ‘behind’53 (AR30) wa-ʔanta qad ʔabġaḍta t-taʔdība waʔalqayta kalām-ī ḫalfa-ka (Al-Kitāb al-

not used in Maltese (MT30) int, li tobgħod it-twiddib, u tixħet kliemi wara dahrek? (Il-Bibbja 2013: 636)

|| 46 Ibn Manẓūr (1981: 814–815); Wright (2005, I: 281). 47 Aquilina (1987: 471); Borg and Azzopardi-Alexander (2012: 265); Panzavecchia (1845: 247 ԧada, ԧdein); Wettinger and Fsadni (1983: 42, c. 1672 chdeina). 48 Zavadovskij (1979: 74). 49 Ibn Manẓūr (1981: 864); Ryding (2005: 391). 50 Ibn Manẓūr (1981: 1055); Wright (2005, II: 188). 51 Ibn Manẓūr (1981: 1055); Ryding (2005: 392). 52 Ibn Manẓūr (1981: 1125); Ryding (2005: 398). 53 Ibn Manẓūr (1981: 1234); Penrice (1873: 44); Wright (2005, II: 187–188).

Arabic prepositions and their Maltese equivalents | 147

Classic/Modern Literary Arabic

Maltese

muqaddas 1982, I: 867) Seeing thou hatest instruction, and castest my words behind thee. (The Holy Bible 1953, I: 640, Psalms, 50:17)

Table 6: Denominal prepositions III.

Classic/Modern Literary Arabic

Maltese

‫ ﺧﻼﻝ‬ḫilāla ‘during’54 (AR31) wa-ḫilāla ḏālika l-waqti, sa-yaṭlubu nnāsu l-mawta fa-lā yaǧidūna-hu. (Al-Kitāb almuqaddas 2009: 1299) And in those days shall men seek death, and shall not find it; (The Holy Bible 1953, II: 291, Revelation, 9:6)

not used in Maltese (MT31) F’dawk il-jiem il-bnedmin jibdew ifittxu l-mewt bla ma jsibuha; (Il-Bibbja 2013: 1711)

‫ ﺩﺍﺧﻞ‬dāḫila ‘inside, within’55 (AR32) bāraka ʔabnāʔa-ki dāḫila-ki (Al-Kitāb al-muqaddas 1982, I: 935) he hath blessed thy children within thee. (The Holy Bible 1953, I: 691, Psalms, 147:13)

not used in Maltese, cf. Tunisian Arabic dǟḫil ‘within’56 (MT32) u jbierek ġewwa fik ’l uliedek. (Il-Bibbja 2013: 708)

‫ ﺩﻭﻥ‬dūna ‘on this side of, under, without’57 (AR33) ʔinna-mā ʔamāma r-rabbi l-laḏī ḫtāranī dūna ʔabī-ki wa-dūna kulli bayti-hi (Al-Kitāb al-muqaddas 1982, I: 935) It was before the LORD, which chose me before thy father, and before all his house, (The Holy Bible 1953, I: 366, II Samuel, 6:21)

not used in Maltese, cf. Tunisian Arabic dūn ‘without’58 (MT33) Il-Mulej kien li għażilni flok missierek u daru kollha, (Il-Bibbja 2013: 346)

not used in Maltese ‫ ﺳﻮﻯ‬siwā ‘other than; except’59 (AR34) ʔanta lam tuḫbir-ni, wa-lā ʔanā samiʕtu (MT34) U anqas int ma għarraftni, u jien ma siwā l-yawmi (Al-Kitāb al-muqaddas 1982, I: 32) smajtx ħlief illum. (Il-Bibbja 2013: 22) neither didst thou tell me, neither yet heard I of it, but to day. (The Holy Bible 1953, I: 26, Genesis, 21:26)

|| 54 Ibn Manẓūr (1981: 1249–1250); Ryding (2005: 398). 55 Ibn Manẓūr (1981: 1341); Ryding (2005: 390). 56 Ben Abdelkader et al. (1977: 449). 57 Ibn Manẓūr (1981: 1460); Ryding (2005: 391); Wright (2005, II: 182–186). 58 Zavadovskij (1979: 74). 59 Ibn Manẓūr (1981: 2163); Wright (2005, II: 209–210).

148 | Przemysław Turek

Classic/Modern Literary Arabic

Maltese

‫ ﺻﻮﺏ‬ṣawba ‘in the direction of, toward, to’60

not used in Maltese (AR35) wa-lākinna-hum rafaḍū stiqbāla-hu li- (MT35) Iżda hemmhekk ma laqgħuhx, billi hu ʔanna-hu kāna muttaǧihan ṣawba ʔŪrušalīma. kien sejjer Ġerusalemm. (Il-Bibbja 2013: 1446) (Al-Kitāb al-muqaddas 1988, II: 102) And they did not receive him, because his face was as though he would go to Jerusalem. (The Holy Bible 1953, II: 84, Luke, 9:53) not used in Maltese, cf. Tunisian Arabic ḏi¢ d ‘against’62 (MT36) Jonsob il-ħażin għall-bniedem ġust, (IlBibbja 2013: 626)

‫ ﺿﺪ‬ḍidda ‘against’61 (AR36) aš-širrīru yatafakkaru ḍidda ṣ-ṣiddīqi (Al-Kitāb al-muqaddas 1982, I: 857) The wicked plotteth against the just, (The Holy Bible 1953, I: 633, Psalms, 37:12)

Table 7: Denominal prepositions IV.

Classic/Modern Literary Arabic

Maltese

‫ ﺿﻤﻦ‬ḍimna ‘within, inside, among’63 (AR37) tawaǧǧaha ʔilā Kafranāḥūma lwāqiʕati ʕalā šāṭiʔi l-buḥayrati ḍimna ḥudūdi Zabūlūna wa-Naftālīma, (Al-Kitāb almuqaddas 1988, II: 4) he came and dwelt in Capernaum, which is upon the sea coast, in the borders of Zabulon and Nephtalim: (The Holy Bible 1953, II: 6, Matthew, 4:13)

not used in Maltese (MT37) u mar joqgħod Kafarnahum, qrib ilbaħar, fl-inħawi ta’ Żebulun u Naftali, (Il-Bibbja 2013: 1357)

‫ ﻁﻮﺍﻝ‬ṭawāla, ṭiwāla ‘during; for’64 (AR38) ʔinna-nā min ʔaǧli-ka nuwāǧihu ḫaṭara l-mawti ṭiwāla n-nahāri. (Al-Kitāb almuqaddas 2009: 1188) For thy sake we are killed all the daylong; (The Holy Bible 1953, II: 633, Romans, 9:36)

not used in Maltese (MT38) Minħabba fik il-jum kollu jikkundannawna għall-mewt; (Il-Bibbja 2013: 1563)

‫ ﻋﺒﺮ‬ʕabra ‘across, over’65 not used in Maltese (AR39) yukrimu l-ʔaḫīru ṭarīqa l-baḥri, ʕabra l- (MT39) hekk issa fl-aħħar isebbaħ it-triq talʔUrdunni, Ǧalīla l-ʔumami (Al-Kitāb albaħar, ’l hemm mill-Ġordan, il-Galilija tal-

|| 60 Ibn Manẓūr (1981: 2520); Wehr (1994: 617). 61 Ibn Manẓūr (1981: 2564); Ryding (2005: 390). 62 Ben Abdelkader et al. (1977: 15). 63 Ibn Manẓūr (1981: 2613); Ryding (2005: 390). 64 Ibn Manẓūr (1981: 2726); Ryding (2005: 397). 65 Ibn Manẓūr (1981: 2782); Ryding (2005: 398).

Arabic prepositions and their Maltese equivalents | 149

Classic/Modern Literary Arabic

Maltese

muqaddas 1982, I: 1002) and afterward did more grievously afflict her by the way of the sea, beyond Jordan, in Galilee of the nations. (The Holy Bible 1953, I: 738, Isaiah, 9:1)

ġnus. (Il-Bibbja 2013: 777, Isaija 8:23b)

‫ ﻋﻨﺪ‬ʕinda ‘with, at’ used of time and place, possessive66 (AR40) kam ʕinda-kum min al-ḫubzi? (Al-Kitāb al-muqaddas 1982, II: 29) How many loaves have ye? (The Holy Bible 1953, II: 22, Matthew, 15:34)

għand ‘with or at’ possessive67, cf. Tunisian Arabic ʕand ‘at, among, etc.’68 (MT40) Xi kemm għandkom ħobż? (Il-Bibbja 2013: 1376)

‫ ﻓﻮﻕ‬fawqa ‘on, over, above’ of place and rank69 (AR41) fa-waqafa fawqa-hā wa-ntahara lḥummā fa-tarakat-hā (Al-Kitāb al-muqaddas 1982, II: 98) And he stood over her, and rebuked the fever; and it left her: (The Holy Bible 1953, II: 74, Luke, 4:39)

fuq ‘on, upon, over, above’70; cf. Syrian Arabic fōʔ ‘above, over, upstairs’71 and Tunisian Arabic fūq ‘over, etc.’72 (MT41) Resaq fuqha, ordna lid-deni, u d-deni ħallieha. (Il-Bibbja 2013: 1435)

‫ ﻗﺒﻞ‬qabla ‘before’ of time73 (AR42) fa-ṣaʕida kaṯīrūna min al-kuwari ʔilā ʔŪrušalīma qabla l-fiṣḥi (Al-Kitāb almuqaddas 1982, II: 170) and many went out of the country up to Jerusalem before the passover, (The Holy Bible 1953, II: 74, John 11:55)

qabel ‘before’74, cf. Syrian Arabic ʔabᵊl ‘before’75 and Tunisian Arabic qbäl ‘before’76 (MT42) u ħafna nies mill-kampanja telgħu Ġerusalemm qabel l-Għid (Il-Bibbja 2013: 1492)

|| 66 Ibn Manẓūr (1981: 1234); Penrice (1873: 102); Wright (2005, II: 178–180). 67 de Soldanis (1750: 104 ghant); Aquilina (1990: 967–968); Panzavecchia (1845: 247 gꜧand); Vassalli (1796: 346d ՈAND); Wettinger and Fsadni (1983: 47–48, Caxaru mehande [m’ghandha]). 68 Ben Abdelkader et al. (1977: 21, 33); Zavadovskij (1979: 74). 69 Ibn Manẓūr (1981: 3487–3488); Wright (2005, I: 281). 70 de Soldanis (1750: 105 fuq, 134); Aquilina (1987: 368-369); Panzavecchia (1845: 247 fùⱪ); Vassalli (1796: 206e); Wettinger (2006: 52, 1543 fuc). 71 Cowell (2005: 485). 72 Ben Abdelkader et al. (1977: 2, 266); Zavadovskij (1979: 74). 73 Ibn Manẓūr (1981: 3516); Wright (2005, II: 186–187). 74 Aquilina (1990: 1100); Panzavecchia (1845: 248 ⱪabel); Vassalli (1796: 391d). 75 Cowell (2005: 487). 76 Ben Abdelkader et al. (1977: 45); Zavadovskij (1979: 74).

150 | Przemysław Turek

Table 8: Denominal prepositions V.

Classic/Modern Literary Arabic

Maltese

‫ ﻗﺒﺎﻟﺔ‬qubālata ‘opposite, face to face with, in front of’77 (AR43) waṣalū ʔilā tallati ʔAmmata. wa-taqaʕu tallatu ʔAmmata qubālata Ǧīḥin fī ṭ-ṭarīqi lmuʔaddiyati ʔilā ṣaḥrāʔi Ǧibʕūna. (Al-Kitāb almuqaddas 2009: 317) they were come to the hill of Ammah, that lieth before Giah by the way of the wilderness of Gibeon. (The Holy Bible 1953, I: 361, II Samuel, 2:24)

not used in Maltese (MT43) waslu ħdejn l-għolja ta’ Amma, li tinsab biswit Ġijaħ fuq it-triq li tieħu għad-deżert ta’ Gibgħon. (Il-Bibbja 2013: 342)

‫ ﻗﺒﻴﻞ‬qubayla ‘shortly before, prior to’78 (AR44) bi-l-ʔīmāni, bāraka Yaʕqūbu, qubayla mawti-hi, kulla wāḥidin min ibnay Yūsufa, AlKitāb al-muqaddas 1988, II: 339) By faith Jacob, when he was a dying, blessed both the sons of Joseph; (The Holy Bible 1953, II: 264, Hebrews, 11:21)

not used in Maltese (MT44) Permezz tal-fidi Ġakobb, hu u jmut, bierek lil kull wieħed minn ulied Ġużeppi (IlBibbja 2013: 1668)

‫ ﻗﺪﺍﻡ‬quddāma ‘before’ of place79 (AR45) wa-taġayyarat hayʔatu-hu quddāma-hum (Al-Kitāb al-muqaddas 1982, II: 31) And was transfigured before them: (The Holy Bible 1953, II: 23, Matthew 17:2)

quddiem ‘in front’80, cf. Tunisian Arabic quddǟm/quddām ‘in front of’81 (MT45) u tbiddel quddiemhom. (Il-Bibbja 2013: 1377)

‫ ﻗﺮﺏ‬qurba ‘near; close to; in the vicinity of’82 (AR46) wa-kāna Yūḥannā ʔayḍan yuʕammidu fī minṭaqati ʕAyni Nūna qurba qaryati Sālīma. (Al-Kitāb al-muqaddas 2009: 1114) And John also was baptizing in Aenon near to Salim, (The Holy Bible 1953, II: 112, John 3:23)

A variant qarib used in in Maltese; cf. Arabic qarībun ‘near’ (Adjective), usually as a part of phrasal preposition,83 (MT46) Ġwanni wkoll kien qiegħed jgħammed f’Għajnun, qrib Salim, (Il-Bibbja 2013: 1477)

‫ ﻣﺜﻞ‬miṯla ‘like; as’84

not used in Maltese, cf. Tunisian Arabic miṯl

|| 77 Ibn Manẓūr (1981: 3517); Ryding (2005: 396). 78 Ryding (2005: 396); Wright (2005, II: 186). 79 Ibn Manẓūr (1981: 3552); Wright (2005, II: 187). 80 de Soldanis (1750: 105 qoddièm); Aquilina (1990: 1163–1164); Vassalli (1796: 422b Φoddŷm). 81 Ben Abdelkader et al. (1977: 161); Zavadovskij (1979: 74). 82 Ibn Manẓūr (1981: 3566); Ryding (2005: 396). 83 Ryding (2005: 380). 84 Ibn Manẓūr (1981: 4132–4133); Wright (2005, II: 210).

Arabic prepositions and their Maltese equivalents | 151

Classic/Modern Literary Arabic

Maltese

(AR47) wa-ʔumūran kaṯiratan miṯla hāḏihi tafʕalūna (Al-Kitāb al-muqaddas 1982, II: 68) and many such like things do ye. (The Holy Bible 1953, II: 51, Mark, 7:13)

‘such as’85 (MT47) U bosta ħwejjeġ oħra bħal dawn tagħmlu. (Il-Bibbja 2013: 1409)

‫ ﻧﺤﻮ‬naḥwa ‘towards’86 not used in Maltese 87 (AR48) ṯumma madda yada-hu naḥwa (MT48) Mbagħad medd idu lejn id-dixxipli talāmīḏi-hi (Al-Kitāb al-muqaddas 1982, II: 68) tiegħu (Il-Bibbja 2013: 1371) And he stretched forth his hand toward his disciples, (The Holy Bible 1953, II: 17, Matthew 12:49) ‫ ﻭﺭﺍء‬warāʔa ‘behind, on the far side of’88 (AR49) fa-lā taḏhabū warāʔa-hum (Al-Kitāb almuqaddas 1982, II: 135) go ye not therefore after them. (The Holy Bible 1953, II: 102, Luke, 21:8)

wara ‘after, behind’89, cf. Tunisian Arabic wrā/urā ‘in back of, behind’90 (MT49) Tmorrux warajhom! (Il-Bibbja 2013: 1464)

‫ ﻭﺳﻂ‬wasṭa ‘in the middle of; in the midst of; among’91 (AR50) al-māšī wasaṭa l-manāʔiri ḏ-ḏahabiyyati s-sabʕi (Al-Kitāb al-muqaddas 2009: 1294) who walketh in the midst of the seven golden candlesticks; (The Holy Bible 1953, II: 286, Revelation, 2:1)

not used in Maltese, but see fost ‘amongst’ < fi + wast,92 cf. Tunisian Arabic wusṭ ‘in the middle of, among’ and fusṭ ‘inside’93 (MT50) Dak li jimxi fost is-seba’ kandelabri taddeheb: (Il-Bibbja 2013: 1704)

Only 11 of 32 Arabic prepositions belonging to this group are used in Maltese. Classical Arabic ʔamāma ‘before, opposite’ has been replaced in Maltese with its synonym quddiem (Arabic quddāma) and is also used as an equivalent for Arabic ladā, ladun ‘at, by, upon, having; with’ (MT12, MT18, MT45). Classical Arabic ḥawla disappeared in all the Arabic dialects west of Egypt (Procházka 1993: 130). It has been replaced e.g. by the compound preposition madwar, i.e. ma’ ‘with’ + dwar ‘on, about, concerning’94 (MT27) || 85 Ben Abdelkader et al. (1977: 502). 86 Ibn Manẓūr (1981: 4371); Wright (2005, II: 178). 87 Cf. Vassalli (1796: 530f). 88 Ibn Manẓūr (1981: 4807–4808); Penrice (1873: 158); Wright (2005, II: 187). 89 Aquilina (1990: 1518); Panzavecchia (1845: 248 uara); Vassalli (1796: 652k VARA). 90 Ben Abdelkader et al. (1977: 46); Zavadovskij (1979: 74). 91 Ibn Manẓūr (1981: 4831); Wright (2005, II: 188). 92 Aquilina (1987: 358); Aqulina (1990: 1521); Vassalli (1796: 204f). 93 Ben Abdelkader et al. (1977: 21). 94 Aquilina (1990: 767).

152 | Przemysław Turek

The Classical Arabic preposition ḥiḏāʔa ‘opposite’ occurs in both Tunisian Arabic and Maltese (MT25, MT43), which is another lexical testimony of the relationship between Maltese and the Maghreb dialects. It is not used in the Arabic translation of the Bible, not only because of its rarity, but most probably also for the reason that in these Arabic translations the very popular homograph and homophone ‫ ﺣﺬﺍء‬ḥiḏāʔ ‘shoe; sandal’ (Wehr 1994: 193) appears, in some cases (accusative in status constructus) identical with the preposition. Classical Arabic dāḫila ‘inside, within’ has been replaced in Maltese by common dialectal Arabic ġewwa ‘inside’ (AR32, M32, AR58, MT58). Classical Arabic ʕinda ‘with, at’ was and is pronounced ʕanda as well, hence Maltese għand (MT3, MT7, MT40), already attested in 15th century in Pietru Caxaru’s Il-Kantilena (Wettinger and Fsadni 1983: 47–48; Wright 2005, II: 179). Classical Arabic miṯla ‘like; as’, frequently used in Arabic dialects, including Tunisian, is not attested in Maltese. Its equivalent in this language is usually bhal (AR47, MT47). The Maltese equivalent of the Classical Arabic warāʔa ‘behind’, i.e wara, after the disappearance of the final glottal stop, unlike Arabic dialects, not only replaced its synonym ḫalfa, but also completely supplanted the preposition baʕda ‘after’ (MT19, MT20, MT30, MT49), remaining in everyday use in most Arabic dialects, cf. Libyan, Syrian, Egyptian and Tunisian Arabic baʕd ‘after’ (Cesàro 1939: 75; Cowell 2005: 487; Hinds and Badawi 1986: 86; Spiro 1895: 51; Zavadovskij 1979: 74). Therefore, the Maltese preposition wara covers the functional domain of temporal baʕd ‘after’ as well as that of locative ḫalf ‘behind’. This can be shown by Arabic translations of Maltese examples such as (MT51) ġie wara d-disgħa ta’ billejl ‘he came after 9 p.m.’ (Aquilina 1990: 1518) = (AR51) ǧāʔa baʕda s-sāʕati ttāsiʕati masāʔan [my translation] and (MT52) wara l-knisja ‘behind the church’ (Aquilina 1990: 1518) = (AR52) ḫalfa l-kanīsati [my translation]. Table 9 summarizes the results presented in the previous sections. In the upper part of Table 9, all those cases are mentioned where the Arabic preposition is the etymological source of a Maltese preposition. The lower part of the table hosts those cases in which no etymological successor of the Arabic preposition can be identified in Maltese. Table 9: Arabic prepositions and their Maltese successors.

Classical/Modern Standard Arabic

Maltese

bi- ‘in, by, with’

bi

li- ‘for, to, because of’

lil

ʔilā ‘to, unto, until’

’il

Arabic prepositions and their Maltese equivalents | 153

Classical/Modern Standard Arabic

Maltese

bi-lā ‘without’

bla

ʕalā ‘over, on, against’

għal

fī ‘in’

fi

maʕa ‘with’

ma’

min ‘from’

minn

munḏu ‘since’

mindu/0

bayna ‘between’

bejn

taḥta ‘under, below’

taħt

ḥiḏāʔa ‘opposite’

ħada, ħdej(n)

ʕinda ‘with, at’

għand

fawqa ‘on, over, above’

fuq

qabla ‘before’

qabel

quddāma ‘before’

quddiem

qurba ‘near; close to; in the vicinity of’

qarib

warāʔa ‘behind, on the far side of’

wara

ka- ‘as, like’

0

ḥattā ‘up to, as far as’

0

ʕan ‘from, about, concerning’

0

ladā, ladun at, by, upon, having; with’

0

ʔamāma ‘before, opposite’

0

baʕda ‘after’

0

tiǧāha, tuǧāha ‘opposite’

0

tilqāʔa ‘opposite; in front of’

0

ḥasaba ‘according to; in accordance with’

0

ḥawla ‘round about’

0

ḥawālay approximately’

0

ḫāriǧa ‘outside; outside of’

0

ḫalfa ‘behind’

0

ḫilāla ‘during’

0

dāḫila ‘inside, within’

0

dūna ‘on this side of, under, without’

0

siwā ‘other than; except’

0

ṣawba ‘in the direction of, toward, to’

0

ḍidda ‘against’

0

ḍimna ‘within, inside, among’

0

154 | Przemysław Turek

Classical/Modern Standard Arabic

Maltese

ṭawāla, ṭiwāla ‘during; for’

0

ʕabra ‘across, over’

0

qubālata ‘opposite, face to face with, in front of’

0

qubayla ‘shortly before, prior to’

0

miṯla ‘like; as’

0

naḥwa ‘towards’

0

wasṭa ‘in the middle of; among’

0

Of the 45 analyzed most popular prepositions found in Classical/Modern Standard Arabic, Maltese has retained 18, more or less phonologically modified. The missing forms, attested e.g. in Tunisian Arabic (5 prepositions), are replaced by other, multifunctional prepositions, such as bħal ‘like, similar to, as’, għal ‘for, at, on’, lejn ‘towards’, minn ‘from’, wara ‘after, behind’, etc.

6 Maltese prepositions of dialectal Arabic origin The structure of the following twelve Maltese prepositions and their use reflect the disappearance of many Classical Arabic prepositions and their replacement by the dialectal forms typical of the Maghreb dialects or those originating in Maltese itself. The data are presented in Tables 10–11. Table 10: Maltese prepositions of dialectal Arabic origin I.

Maltese preposition

Etymological commentary

barra ‘except; apart from’95 (MT53) U kienu madwar ħamest elef raġel dawk li kielu, barra n-nisa u t-tfal. (Il-Bibbja 2013: 1374) (AR53) wa-l-ʔākilūna kānū naḥwa ḫamsati ʔālāfi raǧulin mā ʕadā n-nisāʔi wa-l-ʔawlādi

dialectal Arabic barra ‘outside’, cf. Syrian Arabic barra ‘outside’, and Tunisian Arabic ilbarra ‘outside’, ḅaṛṛa96

|| 95 de Soldanis (1750: 122, 190); Aquilina (1987: 81); Panzavecchia (1845: 247); Vassalli (1796: 25c Barra opp. Гevvæ Foris, foras, extra, exterius). 96 Ben Abdelkader et al. (1977: 273); Cowell (2005: 485); Durand (2008: 237); Ibn Manẓūr (1981: 254); Zavadovskij (1979: 75).

Arabic prepositions and their Maltese equivalents | 155

Maltese preposition

Etymological commentary

(Al-Kitāb al-muqaddas 1982, II: 68) And they that had eaten were about five thousand men, beside women and children. (The Holy Bible 1953, II: 20, Matthew 14:21) bħal ‘like, similar to, as’97 From Arabic ‫ ﺑـ‬bi ‘in’ + ‫ ﺣﺎﻝ‬ḥāl ‘condition, (MT54) bħal ġebla niżlu fil-qiegħ. (Il-Bibbja way’98 2013: 76) (AR54) qad habaṭū fī-l-ʔaʕmāqi ka-l-ḥaǧari (AlKitāb al-muqaddas 1982, I: 111) they sank into the bottom as a stone. (The Holy Bible 1953, I: 64, Exodus 15:5) dwar ‘on, about, concerning’99 Arabic root dwr ‘revolve, rotate, turn’100 (MT55) Dan hu li jordna l-Mulej dwar l-ulied bniet ta’ Selofħad: (Il-Bibbja 2013: 192) (AR55) hāḏā mā ʔamara bi-hi r-rabbu ʕan banāti Ṣalufḥāda (Al-Kitāb al-muqaddas 1982, I: 276) This is the thing which the LORD doth command concerning the daughters of Zelophehad, (The Holy Bible 1953, I: 209, Numbers, 36:6) fost ‘amongst’101 fi + wast,102 Arabic ‫ ﻓﻲ‬fī ‘in’ + ‫ ﻭﺳﻂ‬wasṭ ‘middle, (MT56) Sellu għal Androniku u Ġunja, qraba center’103, cf. Tunisian Arabic fī wusṭ ‘in the center, in the middle’104 tiegħi u priġunieri miegħi, magħrufa fost lappostli, (Il-Bibbja 2013: 1572) (AR56) sallimū ʕalā ʔAndrūnikūsa wa-Yūnyāsa nasībay-ya l-maʔsūrīna maʕī l-laḏayni humā mašhūrāni bayna r-rusuli (Al-Kitāb al-muqaddas 1982, II: 266) Salute Andronicus and Junia, my kinsmen, and my fellowprisoners, who are of note among the apostles, (The Holy Bible 1953, II: 195, Romans 16:8)

|| 97 Aquilina (1987: 112); Vassalli (1796: 42); Wettinger and Fsadni (1983: 43, c. 1672 phal). 98 Wehr (1994: 252). 99 Aquilina (1987: 206); Vassalli (1796: 173a Dvâr (…) circum). 100 Wehr (1994: 348). 101 Aquilina (1987: 358); Panzavecchia (1845: 247); Vassalli (1796: 204f). 102 Aquilina (1987: 358), Aquilina (1990: 1521); Vassalli (1796: 204f). 103 Wehr (1994: 1250). 104 Ben Abdelkader et al. (1977: 21).

156 | Przemysław Turek

Maltese preposition

Etymological commentary

ġo, ġewwa ‘inside, in, within’105 Dialectal Arabic ǧuwwa ‘inside’, cf. Syrian (MT57) U jagħmluli santwarju, biex ngħammar Arabic žuwwa ‘inside’106 ġo nofshom; (Il-Bibbja 2013: 87) (AR 57) fa-yaṣnaʕūna l-ī maqdisan li-ʔaskuna fī wasṭi-him (Al-Kitāb al-muqaddas 1982, I: 126) And let them make me a sanctuary; that I may dwell among them. (The Holy Bible 1953, I: 96, Exodus, 25: 8) (MT58) Konna ġewwa l-għar ‘We were inside the cave’ (Borg and Azzopardi-Alexander 2012: 158) (AR58) kunnā dāḫila l-ġāri [my translation] għajr ‘except’107 (MT59) Aħna ma niktbulkomx għajr dak li tistgħu taqraw u tifhmu. (Il-Bibbja 2013: 1593) (AR59) fa-ʔinna-nā lā naktubu ʔilay-kum bišayʔin ʔāḫara siwā mā taqraʔūna ʔaw taʕrifūna (Al-Kitāb al-muqaddas 1982, II: 290–291) For we write none other things unto you, than what ye read or acknowledge; (The Holy Bible 1953, II: 212, II Corinthians 1:13)

Classical Arabic ‫ ﻏﻴﺮ‬ġayra (noun in Accusative used as preposition) ‘other than, different from, unlike’108, cf. Tunisian Arabic ġīr ‘without’109

Table 11: Maltese prepositions of dialectal Arabic origin II.

Maltese preposition

Etymological commentary

ħlief ‘except’110 Arabic ‫ ﺧﻼﻑ‬ḫilāf ‘difference, contrast’, ḫilāfa (MT60) Min hu agħma ħlief il-qaddej tiegħi? preposition ‘besides, apart’111, cf. Tunisian Arabic bi-ḫlǟf ‘except for’112 (Il-Bibbja 2013: 815) (AR60) man huwa ʔaʕmā siwā ʕabd-ī (Al-Kitāb al-muqaddas 1988, I: 865)

|| 105 de Soldanis (1750: 105 geuà, 135 geua); Aquilina (1987: 393–394); Panzavecchia (1845: 247 geuua); Vassalli (1796: 215n Γev ev гevvæ (...) Adverbialiter Intus intra). 106 Cowell (2005: 485); Durand (2008: 237). 107 Aquilina (1990: 948); Panzavecchia (1845: 248 gɧair); Vassalli (1796: 317b). 108 Wright (2005: II 208). 109 Zavadovskij (1979: 74). 110 Aquilina (1987: 546); Panzavecchia (1845: 248 ԧlièf); Vassalli (1796: 286, 325k). 111 Wehr (1994: 298). 112 Ben Abdelkader et al. (1977: 60, 134).

Arabic prepositions and their Maltese equivalents | 157

Maltese preposition

Etymological commentary

Who is blind, but my servant? (The Holy Bible 1953, I: 768, Isaiah, 42:19) lejn, lej ‘towards’113 (MT61) Jekk tiġbed lejn ix-xellug, jien immur lejn il-lemin, u jekk tiġbed lejn il-lemin, jien immur lejn ix-xellug. (Il-Bibbja 2013: 13) (AR61) ʔin ḏahabta šimālan fa-ʔanā yamīnan wa-ʔin yamīnan fa-ʔanā šimālan (Al-Kitāb almuqaddas 1982, I: 20) if thou wilt take the left hand, then I will go to the right; or if thou depart to the right hand, then I will go to the left. (The Holy Bible 1953, I: 17, Genesis, 13:9)

From Arabic ‫ ﺇﻟﻰ‬ʔilā ‘to’ + ‫ ﺃﻳﻦ‬ʔayna ‘where’114

Cf. Tunisian Arabic bi-žnäb ‘beside’116 maġenb ‘close to, beside’115 (MT62) Għajnejh bħal ħamimiet maġenb nixxiegħa ilma, (Il-Bibbja 2013: 762) (AR62) ʕaynā-hu ka-l-ḥamāmi ʕalā maǧārī lmiyāhi (Al-Kitāb al-muqaddas 1982, I: 989) His eyes are as the eyes of doves by the rivers of waters, (The Holy Bible 1953, I: 729, The Song of Salomon, 5:12) Arabic ‫ ﻗﻠﺐ‬qalb ‘heart; middle, center; mind’118 qalb ‘between, among’117 (MT63) Ara, jiena qiegħed nibgħatkom bħal nagħaġ qalb l-ilpup. (Il-Bibbja 2013: 1366) (AR63) hā ʔanā ʔursilu-kum ka-ġanamin fī wasṭi ḏiʔābin (Al-Kitāb al-muqaddas 1982, II: 18) Behold, I send you forth as sheep in the midst of wolves: (The Holy Bible 1953, II: 13, Matthew, 10:16) sa ‘till, until; as far as’119 (MT64) U semma l-bir Siba; għalhekk il-belt

The etymology of Maltese sa remains dubious.120

|| 113 Aquilina (1987: 739); Panzavecchia (1845: 247 lein, lei); Vassalli (1796: 435f). 114 Aquilina (1987: 739); Wehr (1994: 30–31, 48). 115 Aquilina (1987: 386 – ma’ + ġenb ‘side, flank’). 116 Ben Abdelkader et al. (1977: 47). 117 Aquilina (1990: 1111); Panzavecchia (1845: 247 ⱪalb); Vassalli (1796: 399i). 118 Wehr (1994: 918). 119 Aquilina (1990: 1244–1245); Vassalli (1796: 583g). 120 Aquilina (1990: 1245) preferred to derive the Maltese form from a contraction of Italian sino a, variant of fino a ‘until; as far as’; some scholars are inclined to propose a Berber influence: cf. preposition s- ‘to’, expressing direction (Procházka 1993; Boukhris et al. 2008: 112).

158 | Przemysław Turek

Maltese preposition

Etymological commentary

għadu jisimha Birsaba sal-lum. (Il-Bibbja 2013: 28) (AR64) fa-daʕā-hā Šibʕata. li-ḏālika smu lmadīnati Biʔru Sabʕin ʔilā hāḏā l-yawmi (AlKitāb al-muqaddas 1982, I: 41) And he called it Shebah: therefore the name of the city is Beersheba unto this day. (The Holy Bible 1953, I: 33, Genesis, 26:33) (MT 65) U l-ilma baqa’ tajjeb sal-lum, (IlBibbja 2013: 412) (AR65) fa-ṣāra l-māʔu ʕaḏban ḥattā yawmi-nā hāḏā (Al-Kitāb al-muqaddas 2009: 383) So the waters were healed unto this day, (The Holy Bible 1953, I: 433, II Kings, 2:22) ta’ ‘of’ used to express the Genitive121 (MT66) Jiena r-ranġisa ta’ Saron, il-ġilju talwidien. (Il-Bibbja 2013: 758) (AR66) ʔanā narǧisu šārūna sawsanatu lʔawdiyati (Al-Kitāb al-muqaddas 1982, I: 986) I am the rose of Sharon, and the lily of the valleys. (The Holy Bible 1953, I: 727, The Song of Salomon 2:1)

Cf. Tunisian Arabic mtǟʕ ‘belong to, of’.122 According to Bergsträsser (1995: 196): “The most important innovation in the area of nominal inflection is the development of genitival locutions. The most widespread is the substantive matāᶜun ‘property,’ Eg betāᶜ, in Maltese shortened to ta (with pronominal suffix tīᶜ-)”

The prepositions bħal ‘like, similar to, as’ (MT3, MT47, MT54, MT62, MT63), fost ‘amongst’ (MT6, MT50, MT56), maġenb ‘close to, beside’ (MT62) are in fact fossilized compound prepositions, i.e. they “have been lexicalized as unanalyzable units with prepositional functions.” (Stolz 2020: 460). The preposition lejn, lej ‘towards’ is also a lexicalized prepositional phrase with prepositional function (MT48, MT61). The common Arabic dialectal prepositions barra ‘except; apart from’ (MT29, MT53) and ġo, ġewwa ‘inside, in, within’ (M32, MT57, MT58), are widely used in Maltese having replaced Classical Arabic prepositions ḫāriǧa ‘outside; outside of’ (AR29, AR53) and dāḫila ‘inside, within’ (AR32, AR57, AR58), respectively. The most interesting example of the development of dialectal Arabic prepositions in Maltese is the form qalb ‘between, among’, originally ‘heart; middle, center’ (MT63), used in prepositional phrases e.g. in Egyptian dialect (Hinds and Badawi 1986: 713).

|| 121 de Soldanis (1750: 79 ta); Aquilina (1990: 1395–1397); Vassalli (1796: 62d); Wettinger and Fsadni (1983: 43 ta). 122 Ben Abdelkader et al. (1977: 46); Zavadovskij (1979: 74).

Arabic prepositions and their Maltese equivalents | 159

7 Conclusions The aim of my paper was the analysis of Arabic prepositions in Classical/Modern Standard Arabic and Arabic dialects and their Maltese equivalents to show how many Classical Arabic and/or dialectal Arabic prepositions have survived in Maltese and how many have disappeared. Besides, I verified the etymology of the Maltese prepositions which are a continuation or a dialectal development of Classical or dialectal Arabic forms. The data shows that the number of prepositions in Maltese has been reduced in comparison to e.g. Tunisian or Syrian dialect of Arabic. Some prepositions, e.g. wara, quddiem and għal have become multifunctional to a much higher degree than their equivalents in other dialects. One could ask whether the above-mentioned reduction can be attributed to phonological or other factors (such as the avoidance of too many homophones), especially when it comes to the preposition baʕd, whose trace we find in the adverb and conjunction mbagħad ‘then, next, later on’ (Arabic min + baʕd).123 My analysis shows also that Maltese prepositions of dialectal origin, with one exception (preposition sa, probably of Berber origin), can be derived from simple or complex forms of Arabic prepositions or nouns used as prepositions.

Acknowledgments: I would like to thank Thomas Stolz for commenting on the draft version, and Julia Nintemann for her kind help with technical and editorial matters in the preparation of this paper. All remaining shortcomings – be they factual errors or stylistic slips of the pen – remain exclusively mine.

References Al-Kitāb al-muqaddas. 1982. Al-Kitāb al-muqaddas ʔay kutub al-ʕahd al-qadīm wa-l-ʕahd alǧadīd wa-qad turǧima min al-luġāt al-ʔaṣliyya. Al-Qāhira: Dār Al-Kitāb al-muqaddas. Al-Kitāb al-muqaddas. 1988. Al-Kitāb al-muqaddas ʔay kutub al-ʕahd al-qadīm wa-l-ʕahd alǧadīd wa-qad turǧima bi-luġa ʕarabiyya ḥadīṯa. Al-Qāhira: G. C. Center. Al-Kitāb al-muqaddas. 2009. Al-Kitāb al-muqaddas bi-ʕahday-hi al-qadīm wa-l-ǧadīd. Attarǧama al-ʕarabiyya al-mubassaṭa. Ford Worth, Texas: Al-Markaz al-ʕālamī li-tarǧamat al-Kitāb al-muqaddas. Aquilina, Joseph. 1987–1990. Maltese-English dictionary. Vol. 1: A–L. Vol. 2: M–Z and Addenda. Valletta, Malta: Midsea Books Ltd.

|| 123 Aquilina (1990: 796).

160 | Przemysław Turek

Ben Abdelkader, Rached, Abdeljelil Ayed & Aziza Naouar. 1977. Peace Corps English-Tunisian Arabic dictionary. Washington, D.C.: Peace Corps. Bergsträsser, Gotthelf. 1995. Introduction to the Semitic languages. Text Specimens and Grammatical Sketches. Translated with Notes and Bibliography and an Appendix on the Scripts by Peter T. Daniels. Winona Lake, Indiana: Eisenbrauns. Borg, Albert & Marie Azzopardi-Alexander. 2012. Maltese. London & New York: Routledge Taylor & Francis Group. Boukhris, Fatima, Abdallah Boumalk, El Houssaïn El Moujahid & Hamid Souifi. 2008. La nouvelle grammaire de l'amazighe. Rabat: Institut Royal de la Culture Amazighe. Brincat, Joseph M. 2011. Maltese and other languages. A linguistic history of Malta. Sta Venera, Malta: Midsea Books Ltd. Cesàro, Antonio. 1939. L’arabo parlato a Tripoli. Grammatica – esercizi – testi vari. Roma: Casa Editrice A. Mondadori. Corriente, Federico. 2006. Andalusi Arabic. In Kees Versteegh & Mushira Eid (eds.), Encyclopedia of Arabic language and linguistics, 101–111. Leiden: Brill. Cowell, Mark W. 2005. A reference grammar of Syrian Arabic with audio CD (based on the dialect of Damascus). Washington D. C.: Georgetown University Press. Durand, Olivier. 2008. Dialettologia araba. Introduzione. Roma: Edizioni Nuova Cultura. Haywood, John Alfred & Hayim Musa Nahmad. 1990. A new Arabic grammar of the written language. London: Lund Humphries Publishers Ltd. Hinds, Martin & El-Said Badawi. 1986. A dictionary of Egyptian Arabic. Arabic-English. Beirut: Librairie du Liban. Ibn Manẓūr. 1981. Lisān al-ʕArab. Al-Qāhira: Dār al-maʕārif. Il-Bibbja. 2013. Il-Bibbja. Il-Kotba Mqaddsa miġjuba bil-Malti mill-ilsna oriġinali l-Lhudi u lGrieg mill-Għaqda Biblika Maltija. Furjana. Malta: Għaqda Biblika Maltija. Levin, Aryeh. 2007. ʔImāla. In Kees Versteegh & Mushira Eid (eds.), Encyclopedia of Arabic language and linguistics, 311–315. Leiden: Brill. Megiser, Hieronymus. 1610. Propugnaculum Europę. Wahrhaffte, Eigentliche vnd Außführliche Beschreibung, der viel und weit berühmbten Insul MALTA. Leipzig: Verlag Henning Großn des jüngern. Mifsud, Manwel. 2008. Maltese. In Kees Versteegh & Mushira Eid (eds.), Encyclopedia of Arabic language and linguistics, 146–159. Leiden: Brill. Panzavecchia, Fortunato. 1845. Grammatica della lingua maltese. Malta: M. Weiss. Penrice, John. 1873. A dictionary and glossary of the Kor-ân, with copious grammatical references and explanations of the text. London: Henry S. King & Co. Procházka, Stephan. 1993. Die Präpositionen in den neuarabischen Dialekten. Wien: Verband der wissenschaftlichen Gesellschaften Österreichs. Ryding, Karen C. 2005. A reference grammar of Modern Standard Arabic. Cambridge: Cambridge University Press. Saari, Rami. 2003. Milot ha-yaḥas ha-malṭeziyot. Yerušalayim: Hoṣa’at Karmel. Schmidt, Emeli, Maike Vorholt & Nele Witt. 2020. Form and behavior of Maltese prepositions – A usage-based approach. In Slavomír Čeplő & Jaroslav Drobný (eds.), Maltese linguistics on the Danube, 241–270. Berlin & Boston: De Gruyter Mouton. Soldanis, Giovanni Pietro Francesco Agius de. 1750. Della lingua Punica presentemente usata da Maltesi &c. Roma: Per Generoso Salomoni alla Piazza di S. Ignazio. Spiro, Socrates. 1895. An Arabic-English vocabulary of the Colloquial Arabic of Egypt. Cairo & London: Al-Mokattam Printing Office.

Arabic prepositions and their Maltese equivalents | 161

Stolz, Thomas. 2020. A chapter on compound prepositions in Maltese: PREP-PREP combinations and related issues. In Benjamin Fagard, José Pinto de Lima, Dejan Stosic & Elena Smirnova (eds.), Complex adpositions in European languages. A micro-typological approach to complex nominal relators, 439–469. Berlin & Boston: De Gruyter Mouton. Stolz, Thomas & Nataliya Levkovych. 2020. From variation towards the grammar of Maltese prepositions – first steps. In Slavomír Čeplő & Jaroslav Drobný (eds.), Maltese linguistics on the Danube, 199–240. Berlin & Boston: De Gruyter Mouton. Stumme, Hans. 1904. Maltesiche Studien. Eine Sammlung prosaischer und poetischer Texte in maltesischer Sprache nebst Erläuterungen. Leipzig: Hinrichs’sche Buchhandlung. The Holy Bible. 1953. The Holy Bible containing the Old and New Testaments. Philadelphia: The National Bible Press. Vassalli, Mikiel Anton. 1796. Ktŷb yl klŷm Mâlti ’mfisser byl-Latín u byt-Taljân sive Liber dictionum Melitensium. Romae: Apud Antonium Fulgonium. Vella, Andrew P. 1974. Storja ta’ Malta. L-ewwel Volum. Valetta, Malta: Edizzjoni Klabb Kotba Maltin. Wehr, Hans. 1994. A dictionary of Modern written Arabic (Arabic–English). Edited by J. Milton Cowan. Urbana, IL: Spoken Languages Services, Inc. Wettinger, Godfrey. 2006. Kliem Malti Qadim. Malta: L-Istitut tal-Lingwistika u d-Dipartiment tal-Malti fl-Università ta’ Malta. Wettinger, Godfrey & Mikiel Fsadni. 1983. L-Għanja ta’ Pietru Caxaru. Poeżija bil-Malti Medjevali. Malta: Printwell. Wright, Willliam. 2005. Arabic grammar. Mineola, New York: Dover Publications, Inc. Zavadovskij, Jurij. 1979. Tunisskij dialekt arabskogo jazyka. Moskva: Izdatel’stvo ‘Nauka’.

Thomas Stolz and Maike Vorholt

On WH-PREPS A contrastive grammatical sketch of Maltese fejn and Spanish donde/dónde Abstract: Maltese and Spanish share the possibility of using their spatial interrogatives of Place (and Goal) in prepositional function. The paper starts with the identification of the structural and functional similarities and differences the spatial interrogatives have in the two languages. The main focus is on selected issues of the synchronic morphosyntax of Maltese fejn and Spanish donde/ dónde which are investigated with a view to determining whether we are dealing with genuine prepositions in the first place. Keywords: Maltese; prepositions; Spanish; spatial interrogatives

1 Introduction This paper investigates a property unexpectedly shared by Maltese and Spanish in the domain of prepositions. Both languages give evidence of the prepositional use of their basic spatial interrogative of Place (and Goal), i.e., Maltese fejn = Spanish dónde(/donde1) with both meaning ‘where’ – a practice which otherwise is exceptional in the European context. From the perspective of our native German, the possibility to employ one’s spatial interrogative of Place (and Goal) as a preposition is remarkable since a similar word-class flexibility is ruled out

|| 1 In Spanish orthography, a difference is made between dónde ‘where?’ as proper spatial interrogative and donde ‘where’ as spatial relativizer and preposition. The acute accent marks the interrogative like in other cases such as e.g. cuándo ‘when?’ vs. cuando ‘when’, cómo ‘how?’ vs. como ‘like, as’, quién ‘who?’ vs. quien ‘who’, etc. In what follows we respect this purely orthographic convention wherever reference is made explicitly to the spatial interrogative. Elsewhere in the text, we either use donde without diacritic alone or the pair of forms, i.e. dónde/donde. || Thomas Stolz: FB 10: Linguistics/Language Sciences, University of Bremen, UniversitätsBoulevard 13, 28359 Bremen, Germany, E-mail: [email protected] Maike Vorholt: FB 10: Linguistics/Language Sciences, University of Bremen, UniversitätsBoulevard 13, 28359 Bremen, Germany, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-007

164 | Thomas Stolz and Maike Vorholt

not only for German wo ‘where’ but also for its functional counterparts in many languages such as English where, French où, Czech kde, Welsh ble, Finnish missä, etc.2 If so many languages disallow the prepositional use of the spatial interrogative, the question arises as to what Maltese and Spanish have in common structurally which licenses the prepositional use of the spatial interrogative. The necessity to scrutinize this parallel between two genetically unrelated languages receives further support ex negativo in the sense that the prepositional use of the items under review is mentioned – neither in Procházka’s (1993) comparative study of the prepositional systems of neo-Arabic varieties, – nor in the description of the system of Spanish prepositions by López (1999), – nor in Saari’s (2003) dissertation on Maltese prepositions, – nor in Hagège’s (2010) account of adpositions in cross-linguistic perspective. In contrast, other studies (to be reviewed in Section 2) assume that fejn and donde are prepositions at least in certain morphosyntactic contexts. This controversy calls for a dedicated inquiry into the problem. There is thus a descriptive gap to be filled. We understand our own study as an incentive for follow-up studies which investigate the intricacies of the problem in more detail. For ease of reference, we attach the label WH-PREP to the elements under scrutiny whenever they are used prepositionally in the examples and the ensuing discussion. This study tries to establish the synchronic similarities and differences of the grammar of the two WH-PREPS in order to determine whether their classification as prepositions is justified in the first place. Diachronic issues are therefore not addressed systematically. In addition to data from the extant descriptive linguistic literature (grammars, dictionaries, dedicated studies), the phenomena are illustrated with sentential examples drawn from a) a random collection of conventionally published modern Maltese and Spanish fiction (both original and translated prose), the composition of which is spelled out in the primary sources section below, and b) two electronic corpora which cover a wide range of different genres. For Maltese, we have consulted the Korpus Malti (3.0) at http://mlrs.research. um.edu.mt/ (~ 250 million words), whereas El Corpus del Español at https://www.corpusdelespanol.org/ (~ 5.1 billion words) serves as our database for Spanish.

|| 2 Norbert Boretzky (p.c.) directs our attention to possible parallels in some languages of the Balkan Sprachbund such as Macedonian. Monaro and Poletto (2020) show that the phenomenon is also common across the Italo-Romance varieties in southern Italy (Sicily, Calabria, Naples, and Corsica). These cases deserve to be investigated separately in a follow-up study.

On WH-PREPS | 165

It is worth noting that Maltese fejn is not tagged as a preposition in the Korpus Malti (3.0) (where it is always registered as interrogative pronoun). The same holds for donde in El Corpus del Español. Moreover, the corpora are not two of a kind since El Corpus del Español contains data from varieties of twenty-one Spanish-speaking countries world-wide so that variation is very likely to occur. Lectal variation is an aspect that is important also for the topic of this study as will become clear in due course. Our methodology is overwhelmingly qualitative. We look at the data from the vantage point of functional typology. Moreover, we adopt a framework-free approach as advocated by Haspelmath (2010). The paper is organized as follows. In Section 2, we look at how the multifunctionality of fejn and donde/dónde is presented in major reference works representative of the current state of the art in Maltese and Spanish philology and linguistics. The non-prepositional usages of the two elements are illustrated in Section 3 in order to show that there are enough similarities between them to render it likely that we find further shared properties outside the domain covered in Section 3. Section 4 zooms in on the main issue of this study. In four subsections, Maltese and Spanish are compared directly to each other as to the similarities and dissimilarities of fejn and donde in certain structural contexts. The conclusions are presented in Section 5.

2 Background According to Aquilina (1987: 316–317), Maltese fejn displays multiple word-class membership.3 Its primary English translation as – in Aquilina’s parlance – an interrogative adverb equivalent to where and whither4 notwithstanding, fejn

|| 3 In spoken Maltese, the final nasal of fejn is often dropped to yield the apocopated form fej. This practice is reflected in examples like (i) from (Bartolo and Vella 2009: 264) where the dropping of the word-final /n/ is emblematic of the colloquial speech-style of one of the protagonists of the story: (i) Imm’ inti taf fej qegħdin, u fej sejrin? where stay:AP:PL and where go:AP:PL but you 2SG.IMPERF:know ‘But do you know where they are and where they are going?’ 4 For practical reasons, we make use of the somewhat anachronistic spatial interrogatives whither (for Goal) and whence (for Source) of Early Modern English when we translate the Maltese and Spanish examples in which the elements under review are unanalyzable monoword constructions. In contrast, transparent morphologically complex constructions including

166 | Thomas Stolz and Maike Vorholt

may also come as a conjunction, a (relative) pronoun, a noun, or a preposition. In the latter function, the English translations include simple and complex prepositions such as near, beside, and as compared to/with as given in Aquilina’s dictionary. The reference grammar of Maltese mentions fejn not only as member of the category of question words and relatives (Borg and AzzopardiAlexander 1997: 211–212) but also as one of the prepositions which express proximity (Borg and Azzopardi-Alexander 1997: 157). The inclusion of fejn in the inventory of Maltese prepositions is also argued for by Stolz and Levkovych (2020). Schmidt et al. (2020) conduct their investigation of various aspects of the grammar of Maltese prepositions on the same basis. Similarly, the 2019 edition of the authoritative dictionary of the Real Academia Española (https://dle.rae.es/?w=donde&m=form) enumerates twelve different usages for donde5 ranging from those of a relative pronoun to three cases which are labeled as prepositional with a socio-linguistically informed addition according to which the use of donde as preposition belongs to the colloquial register. We also have to take account of what is said in the same dictionary with reference to dónde ‘where?’ (https://dle.rae.es/?w=d%C3%B3 nde&m=form). The list of functions thus can be complemented with those of the spatial interrogative of Place and Goal, adverbs employed in emphatic contexts (exclamations), and the nominalization el dónde ‘(lit.) the where’. This profile largely corresponds to the picture painted of donde/dónde in the monumental Nueva Gramática de la Lengua Española where the items under review are registered as spatial relative (“adverbio relativo locativo”), subordinating conjunction (“conjunción subordinante”), interrogative, and preposition (Real Academia Española 2009). Note that the prepositional status of donde is accepted only with certain reservations by the Real Academia Española (2009: 1605). According to Kany (1994: 422–423), the WH-PREP is especially common in Latin American varieties of Spanish. Quilis and Casado-Fresnillo (1995: 243) provide ample evidence for this phenomenon in the Spanish of Equatorial Guinea. The same authors state that donde as a WH-PREP is firmly established also in the Spanish variety of the Philippines (Quilis and Casado-Fresnillo 2008: 126).

|| PPs receive phrasal translations, namely to where (for Goal) and from where (for Source) with prepositional stranding wherever this strategy is called for pragmatically. 5 There are several diatopic variants of (a)donde in the Spanish varieties of Europe, the Americas, Africa, and the Philippines. Kany (1994: 423) mentions (a)onde and ande, the Real Academia Española (2009: 1597) additionally registers the nowadays obsolete forms onde, do, ado, and adolo.

On WH-PREPS | 167

We will come back to the issue of the use of donde in the extra-European varieties of Spanish in Section 4. The employment of fejn and donde as prepositions is thus an acknowledged fact for many scholars. Moreover, the functional domains of fejn and donde/ dónde resemble each other closely. What distinguishes the one from the other is the different socio-linguistic classification. As to the Spanish case, the prepositional use of donde is explicitly marked as colloquial whereas no such specification is given for Maltese fejn as a preposition so that it is legitimate to assume that the latter is admissible across different styles and registers. It cannot be ruled out therefore that the socio-linguistic restrictions observed for Spanish donde have an impact on the token frequency of the WH-PREP in the corpora.

3 Non-prepositional usages compared In this section, we review a selection of the major functions other than those of prepositions with which the two multifunctional free morphemes are associated. We assume that if the elements can be shown to have properties in common in this domain, it is the more probable that they also behave similarly to each other when they are used as prepositions. We start with a look at their use as spatial interrogatives in direct questions in Section 3.1 because we assume that this is the original nucleus of their functional domain from which the other uses can be derived (cf. Figure 1). INTERROGATIVEDIRECT > INTERROGATIVEINDIRECT > RELATIVIZER > PREPOSITION

Figure 1: Chain of functional domain expansion.

Section 3.2 focuses on the same elements in indirect questions. This leads us directly to the subsequent Section 3.3 where the employment of fejn and donde as spatial relativizers is addressed. The phenomena are illustrated by a sequence of examples for each of the three basic spatial relations of general location, namely Place, Goal, and Source (Goldap 1991: 22). The choice of topics of Sections 3.1–3.3 is rooted in the primarily spatial character of the constructions which they share with the WH-PREPS.6 For reasons of space, further functions of fejn and donde, no matter how interesting they might be, cannot be investigated in this study. || 6 Following Lehmann (1984: 330), we assume that the three sentence types to be discussed in Sections 3.1–3.3 form a functionally-based network.

168 | Thomas Stolz and Maike Vorholt

3.1 Spatial interrogative – direct questions As to the organization of their paradigms of spatial interrogatives, Maltese and Spanish belong to two different classes of languages – at least superficially. The languages differ insofar as Maltese gives evidence of Place-Goal syncretism, i.e. fejn is used for both (static) Place and (dynamic) Goal (Stolz et al. 2017b: 385), whereas in many (but by no means all) varieties of Spanish, the spatial interrogatives of Place and Goal are formally distinct. The Spanish spatial interrogative of Place is dónde as opposed to adónde which is used for Goal (Stolz et al. 2017b: 95–96). Table 1 summarizes the above aspects which are illustrated in the subsequent paragraphs with sentential examples. Table 1: Syncretistic patterns with spatial interrogatives.

Relation

Maltese

Place

fejn

Goal

fejn

Source

minn fejn

Spanish dónde dónde mnejn

adónde

de dónde

Grey shading indicates that two cells of one and the same paradigm host identical forms so that syncretism applies. As will transpire from the discussion in Sections 3.2–3.3, the patterns presented in Table 1 hold good for spatial interrogatives in indirect questions and for spatial relativizers, too. Examples (1)–(2)7 give us an idea of how the spatial interrogatives of Place behave in Maltese and Spanish. (1)

Place – Maltese Fejn hi darek? where she house:POR.2SG ‘Where is your home?’

(2)

Place – Spanish ¿Harry, dónde estabas? Harry where be.TEMP:IMPERF:2SG Harry, where have you been?’

[Saint-Exupéry 2000: 8]

[Rowling 1999: 189]

|| 7 In the sample sentences, glosses, and translations, boldface marks out those items which are of interest for the ensuing discussion. Single underlining identifies prepositions which accompany the items under review. Except otherwise stated, the English translations are ours. With a view to avoid overburdening the presentation with too much superfluous information, we employ morphological segmentation only very sparingly.

On WH-PREPS | 169

The matter becomes more interesting as soon as we take account also of their behavior in examples (3)–(4). (3)

Goal – Maltese [Saint-Exupéry 2000: 8] Fejn trid tmur biha n-nagħġa tiegħi? where 2SG.IMPERF:want 2SG.IMPERF:go with:3SG.F DEF-sheep of:1SG ‘Where do you want to go with my sheep?’

(4)

Goal – Spanish ¿(A-)dónde vas? (to-)where go:2SG ‘Where are you going (to)?’

[Real Academia Española 2009: 1652]

As mentioned in the first paragraph of this section, the Maltese spatial interrogatives are not sensitive to the distinction of Place and Goal with fejn being attested in both (1) and (3). This is a property Maltese has in common with the bulk of the Romance languages like French and Italian – except those of the Ibero-Romance branch to which Spanish belongs. However, the brackets in example (4) are indicative of variation in the Spanish-speaking world. The formerly mandatory use of the univerbated PP a dónde ‘to where’ > adónde has become optional so that dónde, the spatial interrogative of Place, can also function as spatial interrogative of Goal. This means that at least optionally Spanish displays the same pattern of Place-Goal syncretism as Maltese. The spatial relation of Source shows variation on the Maltese side where mnejn ‘whence’8 and the PP minn fejn ‘from where’ compete with each other (Borg and Azzopardi-Alexander 1997: 2011). The competition is illustrated in examples (5)–(6). In contrast, Spanish consistently makes use of the PP de dónde ‘from where’ as exemplified in (7). (5)

Source – Maltese Mn’ejn int ġej ħabib ċkejken whence you come.AP friend little ‘Where do you come from, my little friend?’

[Saint-Exupéry 2000: 8] tiegħi? of:1SG

(6)

Source – Maltese [Bartolo and Vella 2009: 267] Minn fejn kienet ġejja t-twerżiqa? from where be:3SG.F.PERF come:AP:F DEF-scream ‘Where was the scream coming from?’

|| 8 In example (5), the form mn‘ejn ‘whence’ appears in lieu of mnejn. Aquilina (1991: 847) registers only the form without internal apostrophe. We have checked the Korpus Malti (3.0) for the token frequency of both versions. The results are uncontroversial. There are 1,949 hits for mnejn whereas the turnout for mn’ejn is nil.

170 | Thomas Stolz and Maike Vorholt

(7)

Source – Spanish ¿De dónde sacas todo ese from where take_out:2SG all this ‘Where do you take all this money from?’

[Tristante 2013: 199] dinero? money

In Stolz et al. (2017b: 437–438), it is argued that Maltese mnejn and minn fejn are not necessarily historically related in the sense that the former is a univerbated and contracted form of the PP.9 Interestingly, Spanish dónde is the result of a diachronic process during which Latin unde ‘whence’ was generalized over the entire paradigm of spatial interrogatives so that it became necessary to distinguish the different spatial relations by way of adding prepositions. In the case of Source, the ablative preposition de ‘from’ came into play which subsequently fused with the successor of unde to yield (de + unde >) dónde. History repeated itself when the newly created spatial interrogative of Source was again generalized over the entire paradigm. Thus, the need arose to distinguish the spatial relation once more by way of employing distinctive prepositions, namely (a + donde >) adónde for Goal and (de + donde >) de dónde for Source (Stolz et al. 2017a).

3.2 Spatial interrogative – indirect questions It is no surprise to see that in indirect questions, the spatial interrogatives display the same properties as in direct questions. Examples (8)–(9) are perfectly in line with what we have learned from examples (1)–(2). (8)

Place – Maltese [Bartolo and Vella 2009: 332] Dil-vuċi sabiħa qed tistaqsini DEM.F.PROX:DEF-voice beautiful:F PROG 3SG.F.IMPERF:ask:DO.1SG fejn qegħdin. where stay:AP:PL ‘This beautiful voice is asking me where they are.’

(9)

Place – Spanish [Tristante 2013: 141] descubrimos dónde mataron de verdad al pobre discover:1PL.PAST where kill:3PL.PAST from truth IO:DEF poor chaval guy ‘…we discovered where they really killed the poor guy…’

|| 9 We assume that mnejn forms part of the original Semitic heritage of Maltese whereas minn fejn is an innovation (Brockelmann 1908: 328).

On WH-PREPS | 171

Much the same holds for the next triplet of sentences. Examples (8) and (10) corroborate the above Place-Goal syncretism as documented in examples (1) and (3). (10)

Goal – Maltese [Bartolo and Vella 2009: 322] ma kienx jaf fejn kien NEG be:NEG 3SG.M.IMPERF:know where be se jwasal dan l-interrogatorju FUT 3SG.M.IMPERF:arrive DEM.M.PROX DEF-interrogation ‘…he did not know where this interrogation would take him …’

(11)

Goal – Spanish [Tristante 2015: 160] pero no sé a-dónde lleva esto. but NEG know.1SG to-where take.3SG this ‘…but I do not know where this will lead to.’

(12)

Goal – Spanish [https://www.corpusdelespanol.org/now/] No sé dónde va el dinero NEG know.1SG where go.3SG DEF money ‘I do not know where the money goes …’

Similarly, the Spanish examples in (11)–(12) reflect the pattern of variation we have identified in example (4) above in the sense that dónde can replace adónde relatively freely. According to the Real Academia Española (2009: 1652), there is a residual use of adónde in static contexts too which is said to have been much more wide-spread in earlier periods of the language. In examples (13)–(14), we have further evidence of the competition of minn fejn and mnejn in Maltese which perfectly matches the pattern illustrated in (5)– (6) above. The Spanish example (15) underlines the fact that the expression of Source is the monopoly of de dónde as referred to already in connection to example (7). (13)

Source – Maltese [Saint-Exupéry 2000: 7] Domt ħafna biex fhimt take_time:1SG.PERF much in_order_to understand:1SG.PERF minn fejn kien ġej. from where be come.AP ‘It took me a very long time to understand where he was coming from.’

172 | Thomas Stolz and Maike Vorholt

(14)

Source – Maltese10 Bħalissa għandi nagħti elf ewro u presently at:1SG 1SG.IMPERF:give 1000 euro and ma nafx mnejn se nġibhom. NEG 1SG.IMPERF:know:NEG whence FUT 1SG.IMPERF:bring:DO.3PL ‘Presently I have to give 1,000 € and I do not know whence I will take them.’

(15)

Source – Spanish [https://www.corpusdelespanol.org/now/] No sé de dónde viene esa imaginación NEG know.1SG from where come.3SG DEM.PROX.F imagination ‘I do not know where this imagination comes from.’

Direct questions and indirect questions paint identical pictures as to the behavior of the spatial interrogatives. On this basis, it is only logical to assume that we will be confronted with the same patterns also in the domain of relativization.

3.3 Spatial relativizers In point of fact, the data prove this hypothesis right. In (16)–(17), fejn and donde function as relativizers of Place,11 which introduce a spatial attributive relative clause to semantically identical definite common nouns, namely Maltese post and Spanish lugar both meaning ‘place’. (16)

Place – Maltese [Bartolo and Vella 2009: 196] Viċin tal-post fejn trabbejt jien. near of:DEF-place where grow_up:1SG.PERF I ‘Near the place where I grew up.’

(17)

Place – Spanish [Tristante 2013: 131] Ella no puede escapar del lugar donde vive she NEG can:3SG escape:INF of:DEF place where live.3SG ‘She cannot escape from the place where she lives.’

Since Place-Goal syncretism applies in Maltese, the spatial relativizer of Goal in (18) is the same as that of Place in (16). As to Spanish, examples (19)–(20) illus-

|| 10 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=ma+nafx+ mnejn&qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y]. 11 For further properties of fejn as relativizer, we refer the reader to Camilleri (2014: 180–187).

On WH-PREPS | 173

trate the pattern of variation between adonde and donde which is already familiar from the foregoing sections. (18)

Goal – Maltese [Żahra 2008: 100] kien irnexxielu jasal be succeed:IO:3SG.M 3SG.M.IMPERF:arrive fejn l-ebda Malti qatt ma wasal where DEF-no Maltese ever NEG arrive ‘… he managed to arrive where no Maltese has ever arrived …’

(19)

Goal – Spanish [Real Academia Española 2009: 1602] veía ahora el jardín, a-donde daba see:3SG.IMPERF now DEF garden to-where give:3SG.IMPERF el ventanal de la alcoba DEF window of DEF.F alcove ‘… he now saw the garden which the window of the alcove looked out onto …’

(20)

Goal – Spanish [https://www.corpusdelespanol.org/now/] No es el presidente de la República NEG be.3SG DEF president of DEF.F republic el que decide el lugar DEF REL decide.3SG DEF place donde va a ir el Papa. where go.3SG to go:INF DEF Pope ‘The President of the Republic is not the one to decide upon the place where the Pope will go.’

Like in the case of the Spanish spatial interrogatives of Goal mentioned in Section 3.2, one can find remnants of an older practice according to which adonde could be used with a static interpretation, i.e. as spatial relativizer of Place. The Real Academia Española (2009: 1603) claims that this usage has become productive in some Latin American varieties of Spanish such as that of Peru whereas it is considered archaic for instance in Peninsular Spanish. Examples (21)–(23) fail to tell us anything new. The alternation between Maltese mnejn and minn fejn is illustrated once again in the first pair of sentences whereas the uncontested position of de donde in Spanish comes to the fore in example (23).

174 | Thomas Stolz and Maike Vorholt

(21)

Source – Maltese12 Fi ftit tal-ħin waslu fil-post in little of:DEF-time arrive:3PL.PERF in:DEF-place minn fejn Mikiel u l-oħrajn qabdu neżlin from where Mikiel and DEF-other:PL catch:3PL.PERF descend:AP:PL ʼl isfel fl-irdum to below in:DEF-cliff ‘In a short time they arrived at the place from where Mikiel and the others had started to descend downwards on the cliff.’

(22)

Source – Maltese13 Kif wasal fil-post mnejn kienu telqu how arrive in:DEF-place whence be:3PL.PERF depart:3PL.PERF ‘As he arrived at the place from where they had departed …’

(23)

Source – Spanish [Tristante 2013: 407] sale corriendo hacia el coche leave.3SG run:GER towards DEF coach de donde vuelve con el maletín de su marido from where return.3SG with DEF suitcase of POR.3SG husband ‘She ran towards the coach from where she returned with her husband’s suitcase.’

The hypothesis that Maltese fejn and Spanish donde/dónde resemble each other on the morphosyntactic level receives additional support from their possibility to combine relatively freely with various prepositions to cover further spatial relations outside the domain of general location. The grammar of the Real Academia Española (2009: 1601–1602) refrains from providing an exhaustive list of combinations but mentions explicitly desde ‘since, from’ + donde = ‘from where’; hacia ‘towards’ + donde = ‘towards where’; hasta ‘until’ + donde = ‘up to where’; para ‘for’ + donde = ‘to where’; por ‘for, through’ + donde = ‘through where’; en ‘in’ + donde = ‘where’, to which the already familiar; de ‘from’ + donde = ‘from where’ and a ‘to’ + donde = ‘to where’ have to be added.

|| 12 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=post+minn +fejn&qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y]. 13 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=post+mnejn &qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y].

On WH-PREPS | 175

As to the combinability of Maltese fejn with prepositions, Borg and AzzopardiAlexander (1997: 157 and 211) enumerate the following three patterns: minn ‘from’ + fejn = ‘from where’; lejn ‘towards’ + fejn = ‘to where’; sa ‘until’ + fejn = ‘up to where’.14 Independently of the possibility that there might be more (but infrequent) combinations in Maltese, this triplet forms a proper subset of the spatial relations involved in the Spanish case. We interpret this as another piece of evidence of the similarity of the two languages in the domain under scrutiny.

4

WH-PREPS

This section is composed of four parts. In Section 4.1, we adduce evidence in support of the hypothesis that we are dealing with proper prepositions. In Sections 4.2–4.3, we focus first on the definiteness of the complement NP and in the next step on pronominal complements. In Section 4.4, we discuss the possibility of analyzing the PPs as cases of ellipsis. Owing to limitations of time and space, further questions can neither be asked nor answered in this study. They have to be relegated to future research on this subject matter.

4.1 Proof To prove that we are indeed dealing with prepositions, we open this section with sentences (24)–(25) which form a very typical pair of examples. The PPs are highlighted in boldface. Indexed square brackets are added to identify (the most interesting parts of) the internal structure of the PP. (24)

WH-PREP – Maltese

[Cauchi 1990: 52] Fost l-għaġeb ta’ kulħadd wasal among DEF-wondering of everybody arrive [fejnPREP [ilDEF-ħarruba]NP]PP f’ kemm trodd salib WH-PREP DEF-locust_tree:F in how_much 2SG.IMPERF:return cross ‘To everybody’s surprise he arrived at the locust tree in no time.’ [with trodd is-salib being an idiomatic expression corresponding to English You make the sign of the Cross]

|| 14 Kirsty Azzopardi (p.c.) mentions that Maltese native speakers also use ‘il fejn ‘to where’.

176 | Thomas Stolz and Maike Vorholt

(25)

WH-PREP – Spanish

[Tristante 2013: 296] ¿Y decís que está [dondePREP [laDEF fuente de Rogez]NP]PP? and say:2PL SUBORD be.TEMP:3SG WH-PREP DEF:F fountain of Rogez ‘And you say that it is near the Rogez fountain?’

In (24), Maltese fejn is the head of a PP whose complement is the definite NP ilħarruba ‘the locust tree’. This NP forms the Ground of the spatial situation in which the Figure (in this case, the story’s protagonist Mastru Gerfex) moves in space to arrive at the Goal, namely at the locust tree. The Spanish example (25) involves donde as head of a PP whose complement is the complex NP la fuente de Rogez ‘the Rogez fountain’ which could also be classified as a place name in lieu of a common noun since in other examples Fuente is written with upper case initial . This NP is again the Ground of the spatial situation. In contrast to il-ħarruba in (24) however, la fuente de Rogez forms part of a static situation of Place since the Figure – a certain journalist – is described as presently being in the vicinity of the fountain. The whereabouts of the Figure are inquired into by means of the temporary (or spatial) copula estar ‘to be (somewhere/in a given non-permanent state)’. As examples (26)–(27) taken from the same primary sources as (24) and (25) show, fejn and donde occupy the same slots as bona fide prepositions which take the very same complements. (26)

Ordinary preposition – Maltese [Cauchi 1990: 51] Fiehmu li jtellqu mit-tarf understand:3PL SUBORD 3.IMPERF:race:PL from:DEF-end tat-triq [saPREP-[lDEF-ħarruba]NP]PP of:DEF-street until-DEF-locust_tree ‘They agreed upon running from the end of the road up to the locust tree …’

(27)

Ordinary preposition – Spanish [Tristante 2013: 300] está a unos metros [dePREP [laDEF Fuente de Rogez]NP]PP be.TEMP:3SG to some:M:PL meter:PL from DEF:F fountain of Rogez ‘… he is a few meters from the Rogez Fountain.’

In (26), the position of the prepositional head is occupied by sa ‘until’ which is thus in the same slot as fejn in (24). Sa takes the same complement as fejn. Similarly, in (27), the preposition de ‘of, from’ is attested in the same slot as donde in (25). In these examples, de and donde take identical complement NPs. Diagram 2 serves as visualization of the syntactic equivalence of sa and fejn on the one hand, and donde and de on the other.

On WH-PREPS | 177

Maltese

Spanish

fejn

donde (i)l-ħarruba

sa [PREP

la Fuente de Rogez de

NP]PP

[PREP

NP]PP

Figure 2: Syntactic equivalence.

These structural parallels between established prepositions and those derived from the spatial interrogative clearly speak in favor of the prepositional status of the latter two. On the basis of the above proof we can proceed by way of zooming in on the complements of the WH-PREPS and the relations within the PPs.

4.2 Definiteness What strikes the eye immediately is the preponderance of definite complement NPs in both languages. Examples like (28)–(29) are typical of this preference. (28)

[Bartolo and Vella 2009: 90] lesta biex toħroġ ready:F in_order_to 3SG.F.IMPERF:exit [fejnPREP [ilDEF-pjanti u lDEF-fjuri tagħha]NP]PP WH-PREP DEF-plant:PL and DEF-flower:PL of:3SG.F ‘[Estella was] ready to emerge close to her plants and flowers.’

(29)

WH-PREP – Spanish

WH-PREP – Maltese

[Real Academia Española 2009: 1605] Luego iré contigo soon go:1SG.FUT with:2SG [dondePREP [tu cuñado y su mujer]NP]PP WH-PREP POR.2SG brother_in_law and POR.3SG woman ‘Soon I will go with you to your brother-in-law and his wife.’

In both examples, we have coordinated NPs as complement of the WH-PREP. In (28), the definiteness is marked overtly on both NPs by the proclitic il-. There is no definite article in (29). Nevertheless, the coordinated NPs are definite because they are accompanied by possessive pronouns. That (28)–(29) are indeed

178 | Thomas Stolz and Maike Vorholt

representative of the bulk of our data becomes clear when we check the occurrences of WH-PREPS in selected works of two writers. Table 2 contains all instances of the Maltese WH-PREP fejn in (Vella 2014) which is the primary source with the highest type and token frequency of WHPREPS in our Maltese sub-corpus. There are altogether thirteen sentences which host fourteen WH-PREP tokens. The definiteness proclitic on the leftmost component of the complement NP is indexed additionally. None of the cases involves an indefinite complement NP. Table 2: Examples taken from (Vella 2014).

#

Page Example + translation

(a)

53

kienu qed jiġru [fejnPREP [itDEF-toilets]NP]PP ‘…they were running in the vicinity of the toilets…’

(b)

73

Trid jew tidħol [fejnPREP [ilDEF-klassijiet]NP]PP. ‘You either enter where the classes are.’

(c)

90

Lanqas Mr Mallia [fejnPREP [ilDEF-bieb ta’ barra]NP]PP jħares lejn is-subien minn fuq s’isfel. ‘Not even Mr Mallia at the main entrance looking up the boys from head to toe.’

(d)

99

Il-mekkanik [ta’ [fejnPREP [lDEF-iskola]NP]PP]PP taqagħlu xi ħaġa tal-ħadid mal-art. ‘The mechanic from near the school let some iron thing drop to the ground.’

(e)

105

kien tqatel miegħu dik l-ewwel darba [fejnPREP [itDEF-toilets]NP]PP ‘…he had fought with him this first time near the toilets…’

(f)

135

f’ħames minuti jkun [fejnPREP [ilDEF-baħar]NP]PP ‘…in five minutes he will be at the sea…’

(g)

135

Jieqaf biss għal sekonda twila ħafna [fejnPREP [ilDEF-ħajt tal-Muddie’s]NP]PP ‘He only stops for a very long second near the wall of Muddie’s…’

(h)–(i)

135

jimxi [minn [fejnPREP [Walter]NP]PP]PP li qed idaħħal il-kelb wara l-mixja [fejnPREP [ilDEF-baħar]NP]PP. ‘…he walks from Walter who is taking in the dog after the walk at the sea.’

(j)

143

imma Ms Muscat tkun qed iddur [fejnPREP [ilDEF-line]NP]PP ‘…but Ms Muscat will be lingering near the line…’

(k)

162

jitlajja [fejnPREP [Bieb il-Belt]NP]PP ‘…he lingers near the City Gate…’

(l)

180

huwa u jitnikker mal-ħadid [fejnPREP [Bieb il-Belt]NP]PP ‘…while he is loitering at the iron fence near the City Gate…’

(m)

241

reġa’ ġie [fejnPREP [lDEF-iskrivanija]NP]PP. ‘…he had come back near the desk.’

(n)

259

T.A. u C.A. [fejnPREP [iċDEF-Ċimiterju]NP]PP. ‘T.A. and C.A. near the cemetery.’

Only three out of fourteen complement NPs are not equipped with an initial definiteness marker. The three putative exceptions can be found in (h), (k), and (l) in Table 2. The absence of the definiteness marker is easily explained. The

On WH-PREPS | 179

complement NPs Walter and Bieb il-Belt (twice) are proper names and thus inherently definite. Since names are monoreferential in principle (Nübling et al. 2015: 17–20), there is no need to establish their definiteness by formal means. Accordingly, the feature [+definite] applies in 100% of the cases of PPs with the WH-PREP as head given in Table 2. What is more, we have not found a single example of an indefinite complement NP in the assorted exemplars of conventionally published Maltese prose texts. Even the Korpus Malti (3.0) does not feature many uncontroversial examples of PPs with the WH-PREP as head in combination with an indefinite complement NP. Kirsty Azzopardi (p.c.) emphasizes that combinations of this kind are nevertheless possible. The examples we have come by overwhelmingly stem from the sports press and are mostly representative of a non-spatial function of fejn as shown in (30). (30)

Non-spatial function – Maltese15 hemmhekk tinduna li għadna lura there_exactly 2SG.IMPERF:notice SUBORD still:1PL back [fejnPREP [timijiet oħra]NP]PP compared_to team:PL other ‘…exactly there you notice that we are still lagging behind in comparison to other teams.’

Apart from non-spatial cases like (30), the WH-PREP tends to take a definite complement in the Korpus Malti (3.0).16 That prepositions are specialized for combinations with complements which are either definite or indefinite is not exceptional for Maltese. Of the two prepositions which correspond to English without, only mingħajr takes definite as well as indefinite complement-NPs whereas the synonymous bla is restricted to combinations with indefinite complements (Saari 2003: 139–140).

|| 15 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=lura+fejn+ timijiet+&qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y]. 16 Superficially, there seems to be at least one exception in the Korpus Malti (3.0), namely: (ii) Irnexxilna aslu [fejnPREP [stazzjon tal-Protestanti]NP]PP succeed:IO:1PL 1IMPERF:arrive:PL WH-PREP station of:DEF-Protestant:PL ‘We managed to arrive at a station of the Protestants.’ This sentence is taken from a short religious text (religion310). However, the sentence occurs in a paragraph that is fragmentary, i.e. some parts seem to be missing or are placed in the wrong order. It is therefore possible that (ii), too is incomplete (for whatever reasons). The definiteness proclitic preceding the noun stazzjon might have been omitted accidentally. Kirsty Azzopardi (p.c.) claims that (ii) makes perfect sense in Maltese when referring to a Protestant station which could be any one and not just one specific station.

180 | Thomas Stolz and Maike Vorholt

Table 3 contains the same number of WH-PREP tokens as Table 2. The Spanish examples are taken from the Víctor-Ros series authored by Jerónimo Tristante. Whereas one book was sufficient to produce fourteen examples for Maltese fejn, we need the entire pentalogy to reach the same number of examples for Spanish donde. Table 3: Examples taken from Tristante (2008, 2010, 2013, 2015, 2016).

#

Page

Example + translation

(a) 2008: 11

los agentes llegaran [dondePREP [elDEF suceso]NP]PP ‘…the policemen arrived at the scene of crime…’

(b) 2008: 25

Y ahora vaya [dondePREP [su familia]NP]PP ‘And now go to your family…’

(c) 2008: 40

se llegó [dondePREP [laDEF tumba]NP]PP ‘…he arrived at the grave…’

(d) 2008: 81

[DondePREP [losDEF Chisperos]NP]PP ‘At the nail smiths.’17

(e) 2008: 187

El hombre se acercó [dondePREP [losDEF detectives]NP]PP ‘The man got closer to the detectives…’

(f)

Lo enterraron [dondePREP [losDEF indigentes]NP]PP ‘They buried him near the poor people…’

2010: 80

(g) 2010: 145

y al final, [dondePREP [losDEF iluminadores]NP]PP en el gallinero, las clases más populares, la gente de la calle. ‘…and finally, close to the lighting, in the chicken coop, the most Plebeian classes, the street people.’

(h) 2013: 71

Cuando llegan [dondePREP [losDEF cocheros]NP]PP pregunta a viva voz. ‘When they arrive where the coachmen are he asks loudly…’

(i)

2013: 296

¿Y decís que está [dondePREP [laDEF fuente de Rogez]NP]PP? ‘And you say that it is near the Rogez fountain?’ = (25)

(j)

2013: 418

Entonces se acerca [dondePREP [elDEF caballo]NP]PP ‘Thus he gets closer to the horse…’

(k) 2015: 47 (l)

Sí, la conozco, justo [dondePREP [laDEF fuente de la Red de San Luís]NP]PP. ‘Yes, I know it, right at the Fountain of San Luís’s Net.’

2015: 102–103 tenemos que acercarnos [dondePREP [Martin]NP]PP. ‘…we have to get closer to Martin.’

(m) 2016: 61

Fueron a parar [dondePREP [losDEF suicidas y los vagabundos]NP]PP ‘They went to stop at the suicides and vagabonds…’

(n) 2016: 168

El subinspector llegó [dondePREP [elDEF forcejeo]NP]PP ‘The deputy inspector arrived where the struggle took place…’

|| 17 Chispero is the colloquial nickname for the inhabitants of Maravillas, a famous quarter of Madrid.

On WH-PREPS | 181

Twelve of the examples in Table 3 involve complement NPs which are accompanied by the Spanish definite article. This is different with examples (b) and (l) which are definite nevertheless. In the case of (b), the possessive pronoun su ‘(polite) your’ renders the complement definite whereas in (l) we are dealing with the person name Martin in complement function. This means that all of the complement NPs of the WH-PREP in Table 3 are definite, be it overtly or inherently. The Spanish situation thus does not differ in any way from that pictured for Maltese above. The preference for definite complements does not preclude the possibility that there is also the occasional indefinite complement NP which combines with donde in Spanish. In their account of the use of donde in the Spanish of Equatorial Guinea, Quilis and Casado-Fresnillo (1995: 243) mention a handful of pertinent cases of which (31) is representative. (31)

WH-PREP + indefinite complement – Spanish, Equatorial Guinea

se

recurre [dondePREP resort.3SG WH-PREP ‘One resorts to a medicine man.’

REFL.3SG

[Quilis and Casado-Fresnillo 1995: 243] [unINDEF curandero]NP]PP INDEF medicine_man

It remains doubtful however, whether this example and others of its kind mentioned in the same source reflect the use of Spanish as a native language. It cannot be ruled out that we are facing L2-data. This is different with example (32) which is representative of the colloquial speech style of Columbian Spanish. (32)

WH-PREP + indefinite complement – Spanish, Columbia

[https://www.corpusdelespanol.org/now/] Llegué [dondePREP [unosINDEF tíos]NP]PP arrive:1SG.PAST WH-PREP some:PL uncle:PL y me puse a trabajar fuertemente and REFL.1SG put:1SG.PAST to work:INF strongly ‘I came to some guys and started to work hard.’ In point of fact, every hit for the string ir donde un NP [{go:INF} {WH-PREP} {INDEF} {NP}] ‘to go to a NP’ in El Corpus del Español stems from Latin America (from Mexico down to Chile). The same picture emerges if ir is replaced with other verbs such as quedarse ‘stay’, llegar ‘arrive’, etc. Peninsular Spanish – the variety of which the data in Table 3 are specimens – does not seem to favor the combination of WH-PREP and indefinite complements. What the foregoing paragraphs teach us is that the WH-PREPS of Maltese and Spanish are not absolutely excluded from combinations with indefinite com-

182 | Thomas Stolz and Maike Vorholt

plements. This possibility of combining the WH-PREPS with indefinite NPs is overshadowed by a plethora of cases in which the complement-NPs are definite though. It is therefore justified to postulate a more or less pronounced preference for the pattern [WH-PREP [COMPLEMENTDEF]NP]PP in both Maltese and Spanish. Since personal pronouns (and their bound equivalents) are by definition definite, the phenomena addressed in Section 4.3 fit nicely into the above construction pattern.

4.3 Pronominal complements Under the looking glass, the interaction of the WH-PREP and pronominal elements turns out to be not as straightforward as it might seem superficially. In this section, our point of departure is marked by examples (33)–(34). (33)

WH-PREP + PRO – Maltese

Il-maid

[Camilleri 2013: 197] u supperva and haughty:F

Rumena qsajra, irqiqa Romanian:F short:F slim:F daħlet [fejnPREP-homPRO]PP enter:3SG.F.PERF WH-PREP-3PL ‘The Romanian maid – short, slim, and haughty – entered where they were…’ DEF-maid

(34)

WH-PREP + PRO – Spanish, Philippines

[Quilis and Casado-Fresnillo 2008: 126] vas a esperar que se vengan NEG go:2SG to hope:INF SUBORD REF.3 come:3PL.SUBJ [dondePREP [tiPRO]NP]PP WH-PREP you.OBL ‘…you will not expect that they come to you.’

no

In both sentences, the WH-PREP takes a pronominal complement. Maltese and Spanish differ as to the morpheme class to which the complement belongs. In (33), it is shown that the PP consists of a single word, namely the WH-PREP fejn which hosts the pronominal affix -hom for the 3rd person plural. In contrast, example (34) involves the free personal pronoun of the 2nd person singular in the oblique form (labeled “pronombre tonico” by the Real Academia Española 2009: 1179) ti as complement of donde. Fejn belongs to the large class of Maltese prepositions which inflect for person and thus do not combine with a free personal pronoun (Borg and Azzopardi-Alexander 1997: 196). With the notable exception of

On WH-PREPS | 183

con ‘with’,18 Spanish prepositions do not have a paradigm of person inflection but take free pronouns in their oblique form as complement (cf. below). As examples (35)–(36) suggest, there is variation as to which of the forms of the personal pronoun are appropriate in Spanish PPs with a WH-PREP head. (35)

WH-PREP + PRO – Spanish, Peru

[Kany 1994: 423]

el

alcalde vino [dondePREP [míPRO]NP]PP DEF mayor come.3SG.PAST WH-PREP me ‘The mayor came to me.’

(36)

WH-PREP + PRO – Spanish, Guatemala

espero que vengas [a hope:1SG SUBORD come:2SG.SUBJ to ‘…I hope that you come to me.’

[dondePREP WH-PREP

[Kany 1994: 424] [yoPRO]NP]PP]PP I

In (35), donde takes the pronominal complement of the 1st person singular in the oblique form mí which is what we expect of a regular Spanish preposition. However, in (36), the pronoun of the same person in complement function has the shape of the subject form, namely yo. In El Corpus del Español [https://www.corpusdelespanol.org/now/] the PP donde yo is a hapax (from a Mexican journal) as opposed to nine instances of donde mí.19 Interestingly, all of these nine tokens combine with motion verbs, i.e. the spatial relation of Goal is expressed. The sole example of donde yo in El Corpus Español is involved in the expression of Place. In the same corpus, there are only two hits for donde ti but none for *donde tú patterned on the model of donde yo. Donde ti is attested exclusively in Goal-constructions. This privilege of Goal holds also in the case of the pronominal complements of the 2nd person singular polite usted (2 hits); 1st person plural nosotros (12 hits); and 3rd person plural feminine ellas (3 hits). For the other person categories, Goal claims the majority of tokens, namely for the

|| 18 Con ‘with’ is unique among the Spanish prepositions insofar as it has special forms for the 1st person singular conmigo ‘with me’, 2nd person singular contigo ‘with you’, and 3rd person reflexive consigo ‘with himself/herself/themselves’ whereas in all other slots of the paradigm PPs with the structure [PREP [PRO]NP]PP, i.e. with a free pronoun as complement as in con nosotros ‘with us’ are the rule (Iggesen 2005: 16). 19 For practical reasons, we have searched El Corpus Español exclusively for strings of the kind donde PRO##, i.e. for PPs in sentence-final position. This decision most probably limits the turnout of hits.

184 | Thomas Stolz and Maike Vorholt

3rd person singular masculine él (30 hits out of 35 attestations); 3rd person singular feminine ella (23 hits out of 27 attestations); 3rd person plural masculine ellos (13 hits out of 16 attestations); and 2nd person plural (polite) ustedes (2 hits out of 3 attestations). Note that Spanish has distinct subject and oblique forms only for the 1st (yo vs. mí) and 2nd person singular (tú vs. tí) as well as for the 3rd person (reflexive) of both numbers (él/ella/ellos/ellas vs. sí), this means that the above frequency counts involve a majority of pronouns which do not formally distinguish subject from oblique forms. More than 90% of all cases of the WH-PREP with pronominal complements have a Latin American background. The variation illustrated in examples (35)–(36) demands an explanation. It is exactly this explanation which might challenge the prepositional status of donde – and in analogy to the Spanish case, to some extent also that of Maltese fejn. Why this is so can be gathered from the discussion of ellipsis in Section 4.4.

4.4 Ellipsis 4.4.1 Spanish According to the Real Academia Española (2009: 1605), the WH-PREP arises in elliptical constructions. In connection to donde-clauses, two types of ellipsis are assumed, namely a) Equi-V deletion – which involves the deletion of the finite verb in a spatial relative clause introduced by donde if the deleted verb is lexically identical to the verb in the main or preceding clause (as shown in (37)), b) copula deletion – which removes estar ‘to be (somewhere/in a given nonpermanent state)’ from a spatial relative clause introduced by donde if the meaning of Place is already sufficiently conveyed by donde itself (as shown in (39)). (37)

Ellipsis – Spanish [Real Academia Española 2009: 1605] unos decían que no deberían dejar some:PL say:IMPERF:3PL SUBORD NEG shall:POT:3PL leave:INF que SE BAÑARAN los perros SUBORD 3.REFL bathe:SUBJ.PAST:3PL DEF:PL dog:PL donde SE BAÑABAN las personas where 3.REFL bathe:IMPERF:3PL DEF:F:PL person:PL

On WH-PREPS | 185

‘Some said that they should not allow that the dogs the people TOOK THEIR BATH.’

TAKE A BATH

where

In (37), the reflexive verb bañarse ‘take a bath’ occurs twice, once as se bañaran with the subject-NP los perros and once in the form se bañaban with the subjectNP las personas. These verb forms are marked out in boldface and small caps. In the adverbial clause introduced by donde, the verb under scrutiny is subject to Equi-V deletion (together with the reflexive pronoun) so that the sentence is reshaped to (38) without change of meaning in comparison to (37). (38)

Deletion applied – Spanish [Real Academia Española 2009: 1605] unos decían que no deberían dejar que se bañaran los perros donde Ø las personas

As to the internal structure of the donde-clause, we doubt that the NP las personas can be analyzed as an argument or complement of donde since there is no direct relation of government. The NP las personas remains the subject argument of the deleted verb not the least because the meaning of the action verb is too specific to be transferrable to donde. It is therefore ruled out that in cases like (38) donde has prepositional function at all. In point of fact, there is no evidence of the deletion type (a) in Table 3. The situation is different with deletion type (b) as illustrated in (39). All examples of the WH-PREP in Table 3 reflect deletion type (b). (39)

Ellipsis – Spanish [Real Academia Española 2009: 1605] Yo no iré nunca donde ESTÁ mi familia I NEG go:1SG.FUT never where be.TEMP.3SG POR.1SG family ‘I will never go where my family IS.’

Equi-V deletion does not apply in the case of (39) because the motion verb ir ‘go’ in the initial clause is different from the supposed copula verb estar ‘to be (somewhere/in a given non-permanent state)’ in the donde-clause. According to the Real Academia Española (2009: 1605), estar can be deleted nevertheless because “el relativo aporta la idea de ubicación que el verbo reitera”20, meaning: estar is superfluous semantically. In a way, donde inherits the subject-NP from the deleted copula. After deleting estar, the sentence is fully grammatical and retains its meaning also in the form given in (40) (cf. (39)). (40)

Deletion applied – Spanish [Real Academia Española 2009: 1605] Yo no iré nunca donde Ø mi familia.

|| 20 Our translation: ‘the relativizer conveys the idea of location which the verb reiterates.’

186 | Thomas Stolz and Maike Vorholt

Also with reference to deletion type (b), the Real Academia Española (2009: 1605) puts forward the constraint that the deletion of estar is not licit if donde has an overt antecedent, i.e. an NP which is relativized spatially as in (41) (cf. (17)). (41)

Deletion blocked – Spanish [Tristante 2010: 286] En la primera occasión ustedes miraban hacía in DEF:F first:F occasion you.POLITE:PL look:IMPERF:3PL towards allí, a la farola donde ESTÁ Eduardo there to DEF:F streetlight where be.TEMP:3SG Eduardo ‘On the first occasion, you looked in that direction, towards the streetlight where Eduardo IS…’

The copula estar cannot be deleted from the donde-clause because donde relativizes the NP la farola ‘the streetlight’. As a matter of fact, none of our Spanish sources – be it primary or secondary – produces examples to the contrary. The WH-PREP donde is never attested with an overt antecedent. Therefore, donde is a relativizer in cases like (41) but has prepositional properties in cases like (40). This functional split is captured in Figure 3.

donde

relativizer

/

NPANTECEDENT ___ V

preposition

/

V [ ___ NP]STATIC ##

→

Figure 3: Functional split/Spanish.

The Real Academia Española (2009: 1605) assumes further that the deletion type (b) derives from (a). Moreover, it is argued that type (b) is restricted to inanimate referents although the same source concedes that person names, etc. can be complements of the WH-PREP provided reference is made to their home or place of residence. They are thus considered constructions which are functionally equivalent to Maltese [għandPREP [PERSON NAME]NP]PP ‘at someone’s place’ (Borg and Azzopardi-Alexander 1997: 156). Moreover, the Real Academia Española (2009: 1605) concedes that the above explanations are challenged by the fact that donde takes pronominal complements of the 1st/2nd person singular in the oblique as in (34)–(35) since the explanations given in (a)–(b) would favor the subject form as in (36). The Real Academia Española (2009: 2253) claims that the oblique forms of the pronominal complements are infrequent and belong entirely to the colloquial style.

On WH-PREPS | 187

The supposed infrequency and dependency on style are, however, not an explanation of the use of the oblique in these contexts. What is more, no example is provided for the use of the subject form of the pronominal complements either. To come at least a tiny bit closer to a potential solution we have to take a detour via Maltese interrogative clauses in Section 4.4.2.

4.4.2

Maltese

The Maltese examples in Table 2 are fully in line with the restrictions formulated for the Spanish WH-PREP in the foregoing section. In none of the cases is there an antecedent for fejn. If fejn is preceded by an antecedent as in example (16) or (42), the fejn-clause contains a verb or verb-like predicate nucleus. (42)

Predicate – Maltese [Bartolo and Vella 2009: 268] ftit ‘il fuq mill-passaġġ fejn KIENU huma little to above from:DEF-passage where be:3PL.PERF they ‘… a little bit upwards from the passage where they WERE.’

The perfective copula is present in (42) because fejn acts as relativizer of its antecedent il-passaġġ. In the absence of a relativizer, the copula is not compulsory so that a situation similar to that depicted for Spanish emerges, cf. Figure 4 to Figure 3 above. relativizer fejn

/ NPANTECEDENT ___ V

→ preposition

/

V [ ___ NP]STATIC ##

Figure 4: Functional split/Maltese.

This resemblance notwithstanding, the two languages do not correspond to each other in each and every structural detail as the subsequent paragraphs are going to show. As it will turn out, Figure 4 cannot be taken at face value. In Maltese interrogative clauses, fejn can function as a fully blown predicate. In this case, there are two options. Either fejn hosts the pronominal affixes as subject agreement markers as in (43)–(44) or the personal pronoun is employed as ersatz-copula as in (45) since, in stark contrast to Spanish, Maltese has no genuine copula in the present tense. According to Borg (1988: 271), the zero-copula is blocked in interrogative clauses of this kind (cf. (8) above).

188 | Thomas Stolz and Maike Vorholt

(43)

Predicate – Maltese Rob, Kristjan fejn-u? Rob Kristjan where-3SG.M ‘Rob, Christian, where is he?’

[Bartolo and Vella 2009: 273]

(44)

Predicate – Maltese [Bartolo and Vella 2009: 360] Fejn-hom uliedek u wlied uliedek? where-3PL children:POR.2SG and children children:POR.2SG ‘Where are your children and grand-children?’

(45)

Predicate – Maltese [Bartolo and Vella 2009: 272] Fejn hu Kris? Fejn huma l-oħrajn? where he Chris where they DEF-other:PL ‘Where is Chris? Where are the others?’

In (43)–(44), the inflected forms of fejn are identical to those of the WH-PREP as host of the pronominal suffixes as illustrated in (33). The set of suffixes is not only identical with the one employed for inflected prepositions in general but is also common for many of the so-called pseudo-verbs of Maltese (Fabri 1993: 198–201). With examples (46)–(47), we leave the domain of direct questions. In these examples, we contrast the WH-PREP and the predicative fejn – both with an argument in the 1st person singular. (46)

WH-PREP – Maltese21

Ġiet [fejnPREP-iPRO]PP persuna Għawdxija serjissima come:3SG.F.PERF WH-PREP-1SG person:F Gozitan:F serious:INTENS:F ‘A very serious person from Gozo came to me …’ (47)

Predicate – Maltese22 Bqajt skantat bir-reazzjoni tan-nies remain:1SG.PERF amaze:PASSP with:DEF-reaction of:DEF-people li litteralment bkewli biex SUBORD literally weep:3PL.PERF:IO:1SG in_order_to nibqa’ fejn jiena 1SG.IMPERF:remain where I

|| 21 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=fejni& qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y]. 22 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=fejn+jiena& qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y].

On WH-PREPS | 189

‘I was amazed by the reaction of the people who literally cried for me to stay where I was.’ It is possible to interpret the fejn-clause in (46) as an instance of the zero-copula whose use is ruled out for direct question as mentioned above. Accordingly, we assume an additional structural constellation of the type [fejn Ø PRO]CLAUSE, i.e. the pronoun functions as subject of fejn much the same as yo functions as subject of donde in (36). In contrast to Spanish, Maltese has several possibilities for the subject-fejn-relationship, namely agreement marking by person inflection on fejn in the presence of a co-referential lexical subject-NP as in (43)–(44), the ersatz-copula as link between fejn and its lexical subject-NP as in (45), and the asyndetic juxtaposition of fejn and the subject pronoun as in (46). The question arises how the person inflection on fejn has to be understood if no lexical subject–NP is co-present in the clause. For the sake of the argument, let us assume that example (47) fits the description of deletion type (a) as sketched for Spanish in the previous Section 4.4.1. In the clause preceding the donde-clause the verb baqa’ ‘remain’ is used with a 1st person singular subject. The donde-clause also involves a participant in the 1st person singular. The donde-clause is devoid of any verb. Is it the case that the empty slot of the verb in the fejn-clause is virtually occupied by baqa’, too? If this is true, the second occurrence of baqa’ can be considered superfluous and thus dispensable. We illustrate this possibility with a parallel example from the Korpus Malti (3.0) in (48). Note that in the original, there is no lexical verb in the fejn-clause. (48)

Deletion type (a) – Maltese23 inti tista’ TIBQA’ you 2SG.IMPERF:can 2SG.IMPERF:remain fejn inti TIBQA’ kemm trid where you 2SG.IMPERF:remain how 2SG.IMPERF:want ‘You may STAY where you are ARE STAYING as long as you want…’

However, we doubt that deletion type (a) is operative in Maltese at all. Everything points in a different direction. This means that examples like (47) and (48) involve the zero-copula. There is thus no need for postulating the double occurrence of one and the same lexical verb prior to the deletion of the verb which

|| 23 [http://mlrs.research.um.edu.mt/CQPweb/malti03/concordance.php?theData=fejn+inti& qmode=sq_nocase&pp=50&del=begin&t=&del=end&uT=y].

190 | Thomas Stolz and Maike Vorholt

forms part of the fejn-clause. Accordingly, in (49), we provide an alternative analysis for the example given in (48). (49)

Zero-copula – Maltese (cf. (48)) inti tista’ tibqa’ you 2SG.IMPERF:can 2SG.IMPERF:remain fejn Ø inti kemm trid where Ø you how 2SG.IMPERF:want ‘You may stay where you ARE as long as you want…’

It depends entirely on the syntactic model the descriptive linguists adhere to whether we interpret the absence of a material copula as direct zero-marking or as the product of a deletion process. In the latter case, an underlying copula is subject to deletion of type (b) as in Spanish. This scenario receives support from the possibility to employ an ersatz-copula in constructions of this kind because there is a syntactic slot available as landing-site for the ersatz-copula in the first place. In contrast to the Spanish case, however, the application of deletion type (b) in Maltese does not automatically create PPs with a WH-PREP head. Given that the examples in Table 2 are bona fide cases of WH-PREPS, we have to face the following problem. The combination of fejn + NPLEXICAL counts as a PP. If the same NP is replaced with a referentially identical pronoun however, the resulting combination fejn + PRO is interpreted as a clause involving the zero-copula. The PP-analysis is possible only if the pronominal participant is represented by bound morphology on fejn in lieu of a free personal pronoun. In (50), it is shown that even the pronominally inflected fejn sometimes requires a stylistically adequate English translation which involves the verb to be. (50)

Genitive case assignment – Maltese [Borg and Azzopardi-Alexander 1997: 157]24 Ġriet [minnPREP [fejnPREP-[iPRO]NP]PP]PP s’ għandek. run:3SG.F.PERF from where-1SG until at:2SG ‘She ran all the way from where I was up to your place.’

Since the pronominal affixes are identical to those used to mark inalienable possession in Maltese, we follow Stolz and Levkovych (2020) who postulate that the preposition requires the complement to appear in the genitive so that fejni in (50) could be interpreted along the lines of a possessive construction like my

|| 24 Original translation, the glosses are ours.

On WH-PREPS | 191

place/location in space.25 This solution entails that the relation of fejn and lexical complements receives a genitival interpretation, too. Figure 5 reflects the distinction of non-prepositional uses of fejn in combination with free pronouns and NPs in the nominative on the one hand and bound pronominal affixes and NPs in the genitive on the other.

fejn

~PREP

/

V ___ (COP) PRONOM/NPNOM

WH-PREP

/

V ___-PROGEN/NPGEN

→

Figure 5: Nominative vs. genitive patterns.

Only in the latter case do we have evidence of a genuine WH-PREP. The tenability of our hypothesis hinges crucially on two factors which need to be studied in the months to come. First of all, there is the necessity of finding tangible proof that the distinction of nominative and genitive with lexical NPs makes sense linguistically. Secondly, the supposed pseudo-verb status of pronominally inflected fejn requires further inspection. There is thus still a lot of work to be carried out in the future.

5 Conclusions The previous sections have shown that the structural and functional resemblance of Spanish and Maltese in the domain of the WH-PREPS is straightforward only on the surface. On the one hand, it cannot be denied that fejn and donde share many properties and that the common traits are particularly intriguing wherever the elements are put to prepositional use. These similarities notwithstanding, we have noticed that it is by no means possible to describe the WHPREPS of both languages by way of postulating only one set of rules. An attempt to describe the Spanish facts with reference to the Maltese rules and vice versa is bound to failure. On top of this, fejn and donde seem to differ from the prototypical Maltese and Spanish preposition, respectively, so that there might re|| 25 In the model defended in Stolz and Levkovych (2020), the free pronoun represents the nominative.

192 | Thomas Stolz and Maike Vorholt

main doubts as to the appropriateness of classifying them as prepositions in the first place. A task for the future consists in determining the location of the WHPREPS on the continuum of “prepositionhood”. Another issue that promises interesting insights is diachrony. With reference to prior (but largely unsystematic) work in Spanish philology, Kany (1994: 422–424) argues that the Spanish WH-PREP is possibly of long standing since it is attested as early as the 16th century. The author further claims that it is not only widely diffused in several regions of Spain but also in Sephardic which means that it must have been a structural fact of Spanish already by 1492 when the Jews were expelled from Spain. Since the Iberian Peninsula used to be the arena of Semitic-Romance language contacts for some 700 years, chances are that there is a historical connection between the genesis of the WH-PREP in Spanish and the widely attested spatial interrogative constructions without overt copula in many neo-Arabic varieties such as those of eastern Libya (Owens 1984: 93) and Syria (Cowell 1964: 574). It is thus clear that this study has only scratched the surface. It is hoped that follow-up studies will shed more light on the fascinating topic of the WH-PREPS – perhaps going (far) beyond the comparison of the two languages featured here.

Abbreviations 1, 2, 3 AP CAUS COP DEF DEM DO F FUT GEN GER IMPERF INDEF INF INTENS IO M NEG NOM

NP

1st, 2nd, 3rd person active participle causative copula definite demonstrative direct object feminine future tense genitive gerund imperfective indefinite infinitive intensive indirect object masculine negation nominative nominal phrase

On WH-PREPS | 193

OBL PASSP PAST PERF PL POR POT

PP PREP PRO PROG PROX REFL REL SG SUBJ SUBORD TEMP WH-PREP

oblique passive participle past tense perfective plural possessor potential prepositional phrase preposition pronoun progressive proximate reflexive relativizer singular subjunctive subordinator temporary where-preposition

Acknowledgments: This study forms part of the project Präpositionen und ihre Grammatik im Maltesischen / Prepositions and their grammar in Maltese (Grant STO 186/22-1) funded by the Deutsche Forschungsgemeinschaft. We owe the inspiration to conduct this investigation on Maltese fejn and Spanish donde to Norbert Boretzky (Bochum) who about thirty years ago came back from his holidays in Andalusia to tell his then PhD-candidate Thomas Stolz in a sideways remark that Spanish donde is used like the corresponding interrogative in Arabic – a casual remark that re-emerged from the addressee’s memory only in 2019 when he started to think seriously about a suitable topic for this edited volume. Kirsty Azzopardi and Julia Nintemann kindly commented upon the draft version of this paper. Their comments helped us very much to improve the quality of our study. We gratefully acknowledge the technical support by the members of our research-team in Bremen: Sonja Kettler, Nataliya Levkovych and Beke Seefried have helped to make this publication possible. For whatever is said in this paper we assume the exclusive and full responsibility.

194 | Thomas Stolz and Maike Vorholt

Primary Sources Bartolo, Simon & Loranne Vella. 2009. Il-ġnien tad-dmugħ. Blata l-Bajda: Merlin. Camilleri, Mark. 2013. Volens. Blata l-Bajda: Merlin. Cauchi, Carmel G. 1990. Il-praspar ta’ Mastru Gerfex. Blata l-Bajda: Merlin. Rowling, Joanne K. 1999. Harry Potter y la piedra filosofal. [translated by Dolores Avendaño] Barcelona: Salamandra. Saint-Exupéry, Antoine de. 2000. Iċ-Ċkejken Prinċep. [translated by Toni Aquilina] Msida: Mireva. Tristante, Jerónimo. 2008. El caso de la viuda negra. Madrid: Maeva. Tristante, Jerónimo. 2010. El enigma de la Calle Calabria. Barvelona: Debolsillo. Tristante, Jerónimo. 2013. La última noche de Víctor Ros. Barcelona: Debolsillo. Tristante, Jerónimo. 2015. Víctor Ros y el gran robo del oro español. Barcelona: Plaza Janés. Tristante, Jerónimo. 2016. Víctor Ros: El misterio de la casa Aranda. Barcelona: Debolsillo. Vella, Mark. 2014. X’seta’ ġralu lil Kevin Cacciattolo? Blata l-Bajda: Merlin. Żahra, Trevor. 2008. Il-Ġenn li jżommni f’sikti. Blata l-Bajda: Merlin. Korpus Malti (3.0): http://mlrs.research.um.edu.mt/ El Corpus del Español: https://www.corpusdelespanol.org/

References Aquilina, Joseph. 1987. Maltese-English dictionary. Vol. 1: A–L. Malta: Midsea. Aquilina, Joseph. 1991. Maltese-English dictionary. Vol. 2: M–Z and Addenda. Malta: Midsea. Borg, Albert. 1988. Ilsienna. Studju grammatikali. Ħas-Sajjied: s.l. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London & New York: Routledge. Brockelmann, Carl. 1908. Grundriss der vergleichenden Grammatik der semitischen Sprachen. 2 Bände. Berlin: Reuther & Reichart. Camilleri, Maris. 2014. The Maltese restrictive relative clause. In Albert Borg, Sandro Caruana & Alexandra Vella (eds.), Perspectives on Maltese linguistics, 161–200. Berlin: Akademie Verlag. Čéplö, Slavomír & Jaroslav Drobný (eds.). 2020. Maltese linguistics on the Danube. Papers from the 6th International Conference on Maltese Linguistics (Lingwistika Maltija), Bratislava, Slovakia, June 8–9, 2017. Berlin & Boston: De Gruyter Mouton. Cowell, Mark W. 1964. A reference grammar of Syrian Arabic (based on the dialect of Damascus). Washington/DC: Georgetown University Press. Fabri, Ray. 1993. Kongruenz und die Grammatik des Maltesischen. Tübingen: Niemeyer. Goldap, Christel. 1991. Lokale Relationen im Yukatekischen. Frankfurt a.M.: Lang. Hagège, Claude. 2010. Adpositions. Oxford: Oxford University Press. Haspelmath, Martin. 2010. Framework-free grammatical theory. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of linguistic analysis, 287–310. Oxford: Oxford University Press.

On WH-PREPS | 195

Iggesen, Oliver. 2005. Case-asymmetry. A world-wide typological study on lexeme-classdependent deviations in morphological case inventories. München & Newcastle: LINCOM Europa. Kany, Charles E. 1994. Sintaxis hispanoamericana. Madrid: Gredos. Lehmann, Christian. 1984. Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. López, María del Carmen Fernández. 1999. Las preposiciones en español. Valores y usos. Construcciones preposicionales. Salamanca: Colegio de España. Munaro, Nicola & Cecilia Poletto. 2020. Prepositional ‘where’ in Southern Italian dialects. In Jacopo Garzonio & Silvia Rossi (eds.), Variation in P. Comparative approaches to adpositional phrases, 114–131. Oxford: Oxford University Press. Nübling, Damaris, Fabian Fahlbusch & Rita Heuser. 2015. Namen. Eine Einführung in die Onomastik. Tübingen: Narr. Owens, Jonathan. 1984. A short reference grammar of Eastern Libyan Arabic. Wiesbaden: Harrassowitz. Procházka, Stephan. 1993. Die Präpositionen in den neuarabischen Dialekten. Wien: VWGÖ. Quilis, Antonio & Celia Casado-Fresnillo. 1995. La lengua Española en Guinea Ecuatorial. Madrid: Universidad Nacional de Educación a Distancia. Quilis, Antonio & Celia Casado-Fresnillo. 2008. La lengua Española en Filipinas. Historia. Situación actual. El chabacano. Antología de textos. Madrid: Consejo Superior de Investigaciones Científicas. Real Academia Española. 2009. Nueva gramática de la lengua Española. Tomos I–II. Madrid: Asociación de Academias de la Lengua Española. Real Academia Española. 2019. Diccionario de la lengua española. Edición del Tricentenario. Actualización 2019. Madrid: Asociación de Academias de la Lengua Española. https://dle.rae.es/ Saari, Rami. 2003. Milot Hayaxas Hamalteziyot. Jerusalem: Carmel. Schmidt, Emeli, Maike Vorholt & Nele Witt. 2020. Form and behavior of Maltese prepositions – A usage based approach. In Slavomír Čéplö, & Jaroslav Drobný (eds.), Maltese linguistics on the Danube. Papers from the 6th International Conference on Maltese Linguistics (Lingwistika Maltija), Bratislava, Slovakia, June 8–9, 2017, 241–270. Berlin & Boston: De Gruyter Mouton. Stolz, Thomas & Nataliya Levkovych. 2020. From variation towards the grammar of Maltese prepositions – first steps. In Slavomír Čéplö, & Jaroslav Drobný (eds.), Maltese linguistics on the Danube. Papers from the 6th International Conference on Maltese Linguistics (Lingwistika Maltija), Bratislava, Slovakia, June 8–9, 2017, 199–240. Berlin & Boston: De Gruyter Mouton. Stolz, Thomas, Nataliya Levkovych & Aina Urdze. 2017a. Spatial interrogatives: Typology and dynamics (with special focus on the development from Latin to Romance). In Silvia Luraghi, Tatiana Nikitina & Chiara Zanchi (eds.), Space in diachrony, 207–240. Amsterdam & Philadelphia: John Benjamins. Stolz, Thomas, Nataliya Levkovych, Aina Urdze, Julia Nintemann & Maja Robbers. 2017b. Spatial interrogatives in Europe and beyond: Where, Whither, Whence. Berlin & Boston: De Gruyter Mouton.

Maike Vorholt

Can frequency predict length? A crosslinguistic investigation of Zipf’s law for European adpositions Abstract: The aim of this crosslinguistic investigation is to shed light on European adpositions. The proportion of adpositions in press corpora is compared for the sample languages. Additionally, a relationship between length and frequency as stated by Zipf’s law of abbreviation (Zipf 1935) will be examined. Languages were chosen from different language families in Europe, which led to a convenience sample of 16 languages, including Maltese. The comparison reveals that adposition inventories are quite divergent, both in terms of size and composition. Moreover, it will be shown that adposition proportions vary significantly across corpora. The results show that length can be predicted by frequency for the majority of languages in the sample, Maltese being one of them. Keywords: adpositions; frequency; word length; Zipf’s law

1 Introduction Is there a correlation between a word’s frequency and its length? This question has been present in the field of linguistics for at least 90 years and was first posed by George K. Zipf. He put forward the assumption that words become shorter the more they are used (Zipf 1929). For example, the Maltese preposition fi ‘in’ occurs more than 25 thousand times per one million words while the phonologically longer preposition ħdejn ‘beside’ appears just over one hundred times in a million words. Is this a pattern that can be observed for all words and does it apply to all languages? Many scholars (Hatzigeorgiu et al. 2001; Sigurd et al. 2004; Strauss et al. 2007) have tried to answer this question for several languages (including Greek, English, Swedish, Russian, Hungarian, German, Croatian, Czech, and Indonesian), with some researchers making slight adjustments or modifications to

|| Maike Vorholt: FB 10: Linguistics/Language Sciences, University of Bremen, UniversitätsBoulevard 13, 28359 Bremen, Germany, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-008

198 | Maike Vorholt

Zipf’s proposal. Nevertheless, many languages and text types have not been tested. If there is a relationship between the two phenomena, what exactly does it look like and can one be predicted on the basis of the other? The question of whether Zipf’s law can be applied equally to specific word classes is still unanswered. In the present study, the word class of adpositions will be scrutinized. There have been numerous studies on different aspects of adpositions from a crosslinguistic perspective: on adpositions in general (cf. Hagège 2010), their semantics (cf. Levinson et al. 2003), or syntactic relations (Saint-Dizier 2006). However, as König and Nekula (2013: 26) note, there has not yet been a satisfactory comparison of adposition inventories across languages. Svenonius (2007) gives a crosslinguistic overview of how different languages behave in respect to their adposition inventories. For most languages investigated in this study, no thorough inquiry of adpositions has been previously conducted. Although there have been some works aimed at comparing adpositions of and between different languages, adposition frequency is an area that has not been investigated intensively and, in the rare instances it has been studied, limited to language acquisition, error prevention (cf. Lowie and Verspoor 2004), or language change (cf. Bybee 2001; Diessel 2007). What makes this gap in research particularly interesting is that several studies deal with adpositions exactly for their high frequency of occurrence. Lowie and Verspoor, for example, chose to focus on prepositions for their study on the role of frequency and familiarity in L2 acquisition since they “occur relatively frequently” (Lowie and Verspoor 2004: 83). Additionally, in the 1980s, Schröder (1983: 22) had already acknowledged that preposition frequency in German and other Indo-Euopean languages is so high that at least a group of them belongs to the most frequently used words. Other word classes have been examined for their frequency or proportion in texts or corpora. Hudson (1994: 338) observes that “there seem to be regularities in language of which most of us have been unaware – regularities which involve the statistical probability of any randomly selected word belonging to a particular word-class”. Liang and Liu (2013) compared the proportion of nouns in English, Russian, Chinese, French, German, Italian, and Japanese. There are, of course, frequency dictionaries for several languages. The Leipziger Universitätsverlag has published them for twelve languages so far (Leipziger Universitätsverlag 2019). Kaeding (1898) made the first effort to count words in German in his Häufigkeitswörterbuch der deutschen Sprache. In this frequency dictionary, he did not group the words according to parts of speech but, still, among the ten most frequent words there are four prepositions, namely, zu ‘to’, in ‘in’, auf ‘on’, and an ‘at’. In a frequency dictionary for French by Juilland et al. (1970), 100 high frequency

Can frequency predict length? | 199

words are listed, of which eleven are defined as prepositions: à ‘at, to, on’, avec ‘with’, dans ‘in, during’, de ‘from’, du ‘from’, ne ‘not’, par ‘by’, pour ‘for’, sans ‘without’, sur ‘on, upon’; Juilland et al. (1970: L). Some studies of word class proportions mention adpositions as well. Best (1997) compared ten German texts for the proportions of word classes and found that for prepositions they ranged from 3.5% to 10.25% (Best 1997: 284). Mikosch (1987) analyzed a corpus of free conversations that contains 520,000 words, 28,958 of which were prepositions (5.57%). She found that ‘in’ is the most frequently used preposition in German. The only study that deals with frequencies of Maltese adpositions was conducted by Schmidt et al. (2020). For Turkish, Kabak (2006) uses the frequency of postpositions to analyze their morphologization process. Although adpositions are often mentioned in studies, they hardly ever constitute the main research focus. Zipf (1935: 20) stated that “[p]robably the most striking feature of words is difference in length”. He tried to explain this difference by looking at the relationship between word length and frequency, and found that commonly used words tend to be shorter than less frequent ones (Zipf 1935). Subsequently, Köhler (1986: 69) observes that the connection between length and frequency of words seems to be the best-known quantitative relation between two linguistic features. The aim of this empirical study is to provide insight into the similarities and differences in the size and usage of European adposition inventories with special focus on Maltese. As the literature review has shown, there still exists a need for crosslinguistic investigation of adpositions and especially their frequencies. This study will take a closer look at adpositions and their frequencies and length in 16 European languages. I will also investigate if length is influenced by frequency in the sense that one can be predicted by the other. Corpus linguistic methods are applied in this study in order to discover typological differences in the word class of adpositions among European languages. This study is synchronic and will, therefore, not deal with the development of adposition categories in different languages. Moreover, only written data from the press genre will be considered. The paper is divided into five sections. Section 2 deals with frequency and Zipf’s law. Section 3 provides definitions and describes the methodology used. Section 4 presents the results obtained. Section 5 provides a discussion and outlines the implications of the findings. Section 6 concludes the findings of this study and outlines future research possibilities.

200 | Maike Vorholt

2 Frequency effects in language Frequency of use is a topic that has often been investigated in linguistics. It has been shown that it can influence language processing and acquisition (cf. Ellis 2002). This is relevant to the present study because frequency can have various effects on language, among them language change (cf. Diessel 2007), especially in terms of phonological reduction of high frequency words (cf. Bybee 2001). The theory behind this usage-based view is that “language use shapes grammar” (Bybee and Thompson 2007: 269). Mańczak (2010: 56) observes the balance between the length of words and their frequency and the resulting consequences when this balance is not maintained: [T]he size of linguistic elements is not stable. […] Since the frequency of words is not stable either, it may happen that the balance between the size of a word or of a morpheme and its frequency is disturbed. If a word or a morpheme becomes too short in relation to its frequency, it is replaced by a longer one. But if a linguistic element (i.e. a morpheme, word, or group of words) becomes too long in relation to its frequency, it must be shortened.

Bybee and Thompson (2007: 270) state that the ‘reducing effect’ that goes along with high token frequency is “particularly noticeable in grammaticizing elements or phrases that undergo drastic reduction as they increase in frequency”.

2.1 Zipf’s law of abbreviation George Kingsley Zipf’s most famous law concerns the relationship between the rank and frequency of a word (Zipf 1949). There is another law he formulated, however, that will constitute the focus of this study: the assumption that there is a connection between the length of a word and its frequency. Although this law was first formulated more than 70 years ago, it remains relevant and is frequently referenced in contemporary linguistics. It marks a first attempt at an explanation of the effect that frequency can have on words. Zipf (1929: 3) stated that “whenever a word, hitherto uncommon, suddenly comes to be frequently used, it is likely to undergo a weakening of form to make it more easily pronounceable”. Zipf (1929: 3) provides certain specific examples, such as gas for gasoline, auto for automobile or movie for moving-pictures. If there is a causal relationship between length and frequency, it would be either that length is the cause of a word’s frequency of usage, or vice versa (Zipf 1935: 29). Zipf states that the first theory cannot be true, as speakers do not choose their words according to their length. In Zipf’s opinion, the second possibility is more likely

Can frequency predict length? | 201

because there are examples in languages in which word length decreases as a result of increased frequency (Zipf 1935: 29). He tested this assumption with Chinese, English, and Latin data and found “that the magnitude of words tends, on the whole, to stand in an inverse (not necessarily proportionate) relationship to the number of occurrences” (Zipf 1935: 25). In other words, the more frequent a word, the shorter it will be. He referred to this phenomenon as the ‘law of abbreviation’ (Zipf 1935: 28). Since then, many scholars have examined Zipf’s law of abbreviation across various languages and text types (cf. Hammerl 1990; Piantadosi et al. 2011; Sigurd et al. 2004; Strauss et al. 2007). In general, scholars have verified its accuracy by employing different adjustments to the design and calculation. Hammerl (1990), for example, compared different models that explained the relationship between length and frequency based on either actual frequency or rank order, while a study conducted by Piantadosi et al. revealed that “information content is a considerably more important predictor of word length than frequency” (Piantadosi et al. 2011: 3528). Zipf’s law was even applied to artificial communication by Kanwal et al. (2017). They conducted a study on the pressure for accurate communication and the pressure for less effort and found that only when both pressures were present did the law of abbreviation emerge (Kanwal et al. 2017: 51). Bentz and Ferrer-i-Cancho (2016) tested Zipf’s law of abbreviation as a language universal with a sample of 986 languages. They found a negative correlation between frequency and length for almost all languages in the sample (Bentz and Ferrer-i-Cancho 2016: 1). More recently, Zipf’s ideas have been adapted by Haspelmath and Karjus to study morphosyntactic asymmetries. They discovered that, due to economy, less frequent grammatical patterns tend to be coded with more coding material than more frequent ones (Haspelmath 2008; Haspelmath and Karjus 2017). Köhler (1986) assumes that the manner of shortening a word that becomes more frequent depends on the length it has in the moment the process of shortening begins. Consequently, he reasoned that words that are already short will not be reduced in the same way that longer words are (Köhler 1986: 71f.). Zipf does not formulate how exactly he expects the relationship between length and frequency to be manifest, but rather simply that it is not necessarily proportionate, as mentioned above. However, recent studies focusing on Zipf’s law of abbreviation found that exponential functions are fitting to model the relationship between frequency and length (cf. Sigurd et al. 2004; Strauss et al. 2007).

202 | Maike Vorholt

3 Definitions and methodology This chapter will describe the subject and methodology of this study. The subsections will describe research questions, composition of the sample, definitions, data collection, analysis, and statistical tests that were used.

3.1 Research questions This paper will provide answers to the following three research questions: 1. How are adposition inventories of different languages composed? 2. Are adposition proportions in corpora significantly different across languages? 3. Can the length of adpositions be predicted on the basis of their frequency? The first research question is purely descriptive. More specifically, I will show how many members and what types of adpositions the sample languages display. It is expected that there is variation as to the number of adpositions across languages of different families. Second, I will analyze frequency of occurrence. Frequencies of all adpositions will be collected and the percentage of adpositions in the corpora will be compared between languages. This will be done only in the context of written language from the press genre, as previously stated. It is expected that there is considerable variance in the number of adpositions in corpora of different languages because various strategies are employed for marking the relationship between words in different languages. Additionally, semantic differences are an important factor for variation. The third question will take a closer look at the connection between length and frequency. The question is based upon Zipf’s law of abbreviation but will go further than simply verifying a correlation between the two variables, as this has already been verified by previous studies. Instead, it will evaluate whether length can be predicted by frequency and whether Zipf’s law applies to the word class of adpositions. This question will be analyzed for all languages in the sample separately. As stated in Section 2.1, Zipf’s law has been established for numerous languages and text types. Accordingly, it is expected that a connection between length and frequency, in the sense that frequency influences length, in the case of adpositions will be found for at least some of the languages studied. Thus, it is expected that the null-hypothesis – that there is no significant prediction of length by frequency – will be rejected.

Can frequency predict length? | 203

3.2 Sample Only languages spoken in Europe were chosen for this study. This led to a convenience sample of 16 languages. The languages chosen had to have available resources and a corpus containing a press section. The sample comprises the following languages: Albanian, Armenian, Basque, Finnish, Georgian, German, Greek, Hungarian, Irish, Italian, Lezgian, Lithuanian, Maltese, Ossetic, Polish, and Turkish.

3.2.1 Adposition inventories For each distinct language in this study, a list of adpositions was compiled using a grammar or scholarly paper focused on adpositions. The list of Maltese prepositions is provided in Appendix 2. Only one source per language was used (listed in Appendix 1). This restriction had to be made so that all of the languages were treated equally, as the available literature varied greatly between languages. The source had to be written either in German or English and contain information about adpositions in the particular language. In order to achieve a thorough description of the adpositions, more than one source was used for some languages. Given the limits of this study, it would be impossible to check whether the definition in Section 3.3 below applies to all of the adpositions mentioned in the respective literature on each language in the sample while also performing a crosslinguistic investigation. Most grammars do not mention the definition by which adpositions were selected nor do they provide enough examples to decide whether the criteria are met by every item. As a result, elements will be accepted as adpositions of a language if they are listed as such in the literature and match the restrictions mentioned in Sections 3.3.1 and 3.3.2. As stated in Section 3.3.2, some forms of inflected adpositions are counted seperately. Additionally, it is likely there are more items in the languages that fit the definition but are not listed in the source, and these will thus not be taken into consideration here.

3.3 Definition of adposition Some difficulties arise when attempting to define the category of adpositions for different languages, as DeLancey (2005: 187) observes: “[i]t is impossible to provide a comprehensive and clear semantic definition of the category, as adpo-

204 | Maike Vorholt

sitions within a given language often show a range of rather distinct semantic functions”. Nevertheless, Hagège (2010: 8) suggests a definition for adpositions that will be used in the present study: An adposition […] is an unanalysable or analysable grammatical word constituting an adpositional phrase […] with a term that it puts in relationship, like case affixes1, with another linguistic unit, by marking the grammatical and semantic links between them.

There are three types of adpositions. A ‘preposition’ is located in front of its governed term, a ‘postposition’ after it (Hagège 2010: 8). The term ‘ambiposition’ will be used for adpositions that can either be used as prepositions or as postpositions (Hagège 2010: 114).

3.3.1 Compound adpositions What Hagège (2010: 130) refers to as compound adpositions includes adpositions that are written separately (like the English ahead of, apart from etc.) and adpositions that are written as one word but can be analyzed in more parts (like the Albanian për ‘for’ in përballë ‘opposite’). Only one-word adpositions will be taken into consideration to limit the scope of the present study. A word is defined orthographically here and not morphologically, so compound adpositions in Hagège’s sense (which can be analyzed in multiple parts but are written as one word) will be nonetheless included. This limitation does not prevent parts of multi-word adpositions from being counted as single-word adpositions because multi-word adpositions are often comprised of single-word adpositions. For example, the in and of in the multiword adposition in front of will be counted as an instance of the adposition in and of, respectively.

3.3.2 Inflection In this study, the term ‘inflection’2 will be used to describe the fusion of adpositions with personal pronouns, definite articles, or case markers. In Lezgian (1) and in Finnish (2), for instance, adpositions can bear case markers.

|| 1 Case affixes are not considered adpositions in this study. 2 See Stolz (1990) for a detailed discussion of the term inflection used for adpositions.

Can frequency predict length? | 205

Each of these adpositions with different case markers will be treated as a separate adposition since they can express different semantic functions. (1)

къен-ез q̃en-ez inside-DAT ‘(to) inside’

(2)

al-la under-ADE ‘(at) under’

3

къен-яй q̃en-äj inside-INEL ‘(from) inside’

[Haspelmath 1993: 205]

al-ta al-le al-itse under-ABL under-ALL under-PRO ‘(from) under’ ‘(to) under’ ‘(via) under’ [Sulkala and Karjalainen 1992: 342]

In German, fused forms of adposition and definite article containing an exponent of case will not be counted as separate adpositions, since the system is not considered as developed as in other languages (cf. Nübling 1998: 275). An example is provided in (3). (3)

in ‘in, into, inside’ + dem ‘the (DAT)’ → im ‘in the (DAT)’

3.4 Case affixes Case and adpositions are both considered to be function marking and thereby syntactically share the same role (Hagège 2010: 37). There are, however, differences between the two, and many languages have both case and adpositions at their disposal (Hagège 2010: 37). In this study, no case affix will be counted as belonging to a language’s adposition inventory.4 Thus, in order to treat all languages equally, instances of case will not be treated as belonging to adpositions, even if the reference grammar treats them as such. Furthermore, the relationship between length and frequency could not be investigated in a comprehensive way if case suffixes were counted as well, as they are usually shorter than adpositions (Hagège 2010: 25).

3.5 Etymology In some adposition inventories, there are members that are loanwords from other languages. They will be treated as any other adposition since this study || 3 A transliteration will be given for languages that do not use Latin script. 4 It would certainly be interesting to see whether Hagège’s (2010: 29) theory that case affixes are generally shorter and more frequent than adpositions can be validated by corpus data.

206 | Maike Vorholt

does not concern etymology and is purely synchronic. Of course, it is likely that there is a connection between frequency of occurrence and etymology for some of the sample languages, as observed by Schmidt et al. (2020) with Maltese prepositions.

3.6 Frequency data Corpora were used to collect token frequency for all adpositions. Frequency was normalized for occurrence per one million words in order to make the data comparable. The term ‘frequency’ will refer to usage in the respective corpora. The available corpora for the languages investigated in this study show variation in their size, composition, representative character, and time frame. For this reason, the study is limited to written press texts. This register was chosen since it reflects everyday use of language and has the advantage of being available in many corpora. The corpora5 used are listed in Appendix 1. Although some corpora used here are tagged for parts of speech, this does not always apply to all items on the adposition lists. As a result, not all of the adpositions would be found in the corpora when merely searching for words tagged as adpositions. Thus, in order to treat all of the languages and adpositions equally, no part of speech tags were considered.6 Adpositions that inflect for person, as in Turkish, Hungarian, or Finnish, and those that fuse with the definite article, as in Maltese, Italian, or German, were regarded in all of their different inflected forms.

3.7 Length Word length can be defined in respect to different units (i.e. phonemes, graphemes, morphemes, syllables, and so on). This study will focus on phonological length. The segmental value attributed to different phonemes is based on Stolz et al. (2017). According to their method, affricates and consonants with secondary articulation are treated as monosegmental and thus have a value of one

|| 5 When a corpus is mentioned in the following sections, it refers to the press section of each corpus and not the entire corpus in case it is not a genre specific corpus. 6 This restriction can be problematic since many nouns or other word classes can have the same surface form as adpositions (cf. Svorou 1986: 517) but is a necessary methodological restriction for this study.

Can frequency predict length? | 207

(Stolz et al. 2017: 57). Geminates, diphthongs, long vowels, and nasal vowels are bisegmental with a value of two. The programming environment RStudio (RStudio Team 2015) was used to determine the length of each adposition in this study, using the function nchar(). Digraphs, trigraphs, and the items mentioned above were manually adjusted to their segmental value beforehand. Phonological information was taken from grammars and dictionaries of the sample languages. Optional longer forms and variants of the same adposition were analyzed separately. This does not include all possible forms containing person suffixes of fusion with definite articles. Thus, Turkish postpositions with a compulsory person infix were counted in their usual form of reference (i.e. 3rd person singular).

3.8 Statistical tests Statistical tests applied in this study were carried out with the programming environment RStudio (RStudio Team 2015). An Analysis of Variance (ANOVA) was used to determine if there is variation in length. A chi-square test was applied to show differences in adposition proportions in the corpora of different languages. A linear regression analysis was conducted to see if frequency, in fact, determines length.

3.8.1 ANOVA An Analysis of Variance (ANOVA) was used to test variation in the variable length between languages. Levene’s test (cf. Field et al. 2012: 186) was utilized to test for homogeneity of variance because a normal ANOVA assumes normality and equal variance. As variance in adposition length between languages was significant (F(15, 924) = 3.83, p < 0.001), a robust ANOVA was chosen with 20% trimmed means. Correspondingly, a robust measure of effect size was conducted (cf. Wilcox 2012: 166ff.). Robust post hoc tests that compare all of the languages with each other were conducted with 20% trimmed means as well (cf. Wilcox 2012: 319).

3.8.2 Chi-square test A chi-square test compares observed and expected frequencies. It was used to reveal statistically significant differences in the frequency of adpositions in the

208 | Maike Vorholt

sample languages. Pearson’s chi-square test was chosen for this study. The data are categorical (language) and continuous (frequency) so the test is an adequate and reliable measure.7 Cramér’s V was calculated to measure the effect size. A graphic representation was created in order to see which cells contribute the most to the chi-square score as instructed in Chi-Square Test of Independence in R (Kassambara 2018).8

3.8.3 Linear regression In regression analysis, a statistical model is fitted to the data in order to predict one variable from one or more explanatory variables. A simple regression will be carried out, as only one explanatory variable (i.e. frequency) will be fitted here. Linear regression is applicable to the data in this study because both variables (length and frequency) are ratio-scaled. A logarithmic transformation was applied to the variable frequency in order to reduce positive skew (Baayen 2008; Field et al. 2012). As there is no logarithm of a zero value, a constant of 1 was added to all frequencies beforehand (Field et al. 2012: 197). The variable length did not exhibit such skew, so no logarithmic transformation was applied. As mentioned in 2.1, the relationship between the independent variable and the dependent variable can be expected to follow an exponential function. Thus, a logarithmic transformation generates a linear relationship between the variables. In regression analysis, R2 is calculated, which specifies how much variance the predictor variable can account for in the outcome variable (Field et al. 2012: 258). F-statistic shows if the null-hypothesis can be rejected as well as if there is a relationship between the two variables. The Shapiro-Wilk test was used to test whether the residuals were normally distributed. One regression coefficient represents the intercept the other one specifies the gradient of change the response variable exhibits when one unit of the explanatory variable is added. Applied to this study, it indicates the change in length if one unit of logarithmic frequency is added.

|| 7 For criticism of the chi-square test for comparing frequencies across corpora see Lijffijt et al. (2016). 8 Many thanks to Benjamin Saade for help with the script for Rstudio.

Can frequency predict length? | 209

4 Results In this section, the results for Maltese will be looked at in more detail. Following the Maltese results, the variation among adposition inventories will be discussed and the different languages will be compared with regard to the proportion of adpositions in the respective corpora, their frequency, and length.

4.1 Maltese Maltese adpositions vary in length between two and eight phonemes. The distribution of length can be seen in Figure 1. The most common length is four. The mean length is 4.80 segments. 18 16

Number of adpositions

14 12 10 8 6 4 2 0

2

3

4

5

6

7

8

Length (in phonemes)

Figure 1: Length distribution of Maltese adpositions.

The most frequent adposition is ta’ ‘of’ with 45,559.55 occurrences per million words. Two adpositions that are listed for Maltese were not represented in the

210 | Maike Vorholt

corpus: oltri ‘except’ and sotta ‘under’.9 Maltese adpositions show a mean frequency of 2,685.50 per million words. Figure 2 visualizes a negative relationship between length and logarithmic frequency. Linear regression analysis shows a significant model fit for the two variables (F(1, 63) = 10.92, p < .01). Frequency can explain almost 15% of the variation in length (R² = .1477). Length decreases by .28 with every additional unit of logarithmic frequency.

Figure 2: Relationship between length and logarithmic frequency of Maltese adpositions.

4.2 Variation among adposition inventories Figure 3 shows the distribution of types of adpositions (pre-, post-, and ambipositions) in the sample’s languages. It can be observed that all three types are present in Armenian, Finnish, and German. Ossetic employs both pre- and postpositions. Georgian uses post- and ambipositions, whereas Albanian uses prepositions and one ambiposition. Basque, Lezgian, Turkish, and Hungarian

|| 9 It is not surprising that these two prepositions are not attested in the corpus since they are very rare in spoken language (p.c. Ray Fabri 17.12.2020) or restricted to specific registers (Aquilina 1991: 1014).

Can frequency predict length? | 211

only use postpositions, while Irish, Lithuanian, Greek, Italian, Maltese, and Polish only use prepositions. 100 90 Number of adpositions

80 70

postpositions

60

ambipositions

50

prepositions

40 30 20 10 Irish

Greek

Albanian

Italian

Ossetic

Georgian

Lezgian

Lituanian

Polish

Maltese

Finnish

Hungarian

Turkish

Basque

German

Armenian

0

Figure 3: Distribution of types of adpositions.

German has the largest adposition inventory in this sample with 99 members. Greek and Irish are located at the other end of the scale with the smallest inventories of just 18 adpositions. This nicely matches DeLancey’s observation that “[t]here is substantial variation in the size of the adpositional class across languages which manifest such a category” (DeLancey 2005: 187).

4.3 Frequency and length In this section, adposition proportion, frequency, and length will be compared across languages.

4.3.1 Frequency Table 1 lists the highest ranking adposition of each language in the sample. The highest frequencies that adpositions show are quite divergent across the sample. The Italian adposition di ‘of, by, from’ has a frequency of occurrence of

212 | Maike Vorholt

almost 60 thousand per one million words. The most frequent Hungarian adposition szerint ‘according to, in the opinion of’ on the other hand occurs fewer than 80 times in one million words. Table 1: Adpositions with highest frequency.

Language

Adposition with highest frequency

Highest frequency (per 1 mio words)

Italian

di ‘of, by, from’

59,858.55

Maltese

ta’ ‘of’

45,559.55

Greek

σε se ‘to, into; at, in’

40,957.92

Polish

w ‘in, on’

35,392.88

Irish

i ‘in’

29,508.13

Albanian

në/ndë ‘in, on’

27,931.07

Turkish

ile ‘with’

6,949.80

German

in ‘in, into, inside’

5,788.29

Lezgian

гaлaз galaz ‘with’

4420.45

Armenian

հետ het ‘with’

3,614.51

Georgian

შემდეგ šemdeg ‘after’

2,826.16

Basque

artean ‘among’

1,430.35

Ossetic

фӕдыл fædyl ‘behind, after’

1,081.55

Finnish

ennen ‘before’

610.07

Lithuanian

į ‘into, to’

343.79

Hungarian

szerint ‘according to, in the opinion of’

77.00

4.3.2 Adposition proportion The proportion of adpositions that a corpus contains varies in the languages of this study. Figure 4 shows that Maltese has the largest proportion of adpositions with 15.27%, followed by Italian with 14.89%. Ossetic, Finnish, and Lithuanian display the smallest proportion of adpositions in the sample with .64%, .44%, and .15%, respectively.

Can frequency predict length? | 213

16

Proportion in corpus (in %)

14

15.27 14.89 13.17 12.35

12 10 8

9.09 7.74

6 4.20 3.84 3.84

4 2

2.91 2.04 1.89 1.73 0.64 0.44 0.15

0

Figure 4: Proportion of adpositions in corpus of each language.

A chi-square test revealed a significant relationship between language and adposition frequency (χ2(15) = 795,990, p < .001), though with a small effect size (Cramér’s V = .22). Figure 5 visualizes the Pearson residuals for the test. For each cell, the size of the circle is proportional to the extent of its contribution. White indicates a positive association and black a negative one. Maltese, Italian, Irish, and Polish show much higher adposition frequencies than expected. Ossetic, Finnish, and Lithuanian, on the other hand, use fewer adpositions than expected. All percentages of contribution are listed in Appendix 3. Maltese contributes the most to the chi-square value with 18.82%, followed by Italian with 17.20%.

Figure 5: Residuals of frequency of occurrence. White = positive residual, black = negative residual. Not adposition = all tokens, except adpositions.

214 | Maike Vorholt

4.3.3 Length The range of length was described for each language separately in Section 4. A comparison across all languages of the sample will be given here. Figure 6 visualizes the variance in adposition length across languages. The boxplot reveals that Basque, German, and Italian show a great range in adposition length, while Finnish, Lezgian, and Maltese exhibit less deviation. Some languages have a small range for the middle 50% but the distribution includes outliers represented by black dots (e.g. as in Hungarian and Ossetic). White dots indicate the mean length of the adpositions of each language. Irish adpositions possess the lowest mean length with 2.83 segments. Turkish adpositions, on the other hand, show the highest mean length with 7.11 segments. Armenian, Irish, Italian, Lithuanian, and Polish adpositions have a minimum length of one segment. In contrast, Finnish adpositions are at least four segments long. The longest adposition in the sample, következtében ‘as a consequence of’, belongs to Hungarian and has 14 segments. Across all of the languages in the sample, the mean length is 5.68.

Figure 6: Adposition length across languages.

Can frequency predict length? | 215

The ANOVA shows that there is significant variance in mean adposition length across the whole sample (F(15, 155.23) = 18.11, p < .001) with an effect size of .61, which is considered a large effect (cf. Wilcox 2012: 169). Post hoc tests show that 80 of 120 different language combinations are significantly different from each other (p < .05). The mean length of Irish adpositions is significantly different compared to all of the other languages in the sample. This makes Irish the language that differs most among the sample languages. Italian adposition length, on the other hand, is only significantly different to six of the languages in the sample: Albanian, Basque, Finnish, Greek, Irish, and Turkish.

4.3.4 Correlation between frequency and length A significant correlation between length and frequency was found for twelve of the 16 languages (detailed results are listed in Appendix 4). Results highlighting the combination with language family and types of adpositions present in the languages can be found in Table 2. For most Indo-European languages in the sample, length can be predicted by frequency. However, languages from other language families show a significant relationship between the two variables, as well. Both languages from the Caucasian region (Georgian and Lezgian) did not yield significant results. Seven of the twelve languages that show correlation between length and frequency only use prepositions, while three only use postpositions. One language uses both prepositions and ambipositions and two languages have all three types of adpositions in their inventory. Table 2: Correlation between frequency and length combined with language families and types of adpositions.

Language

Language family/ Significant Linguistic area correlation

Types of adposition preposition

Maltese

Afroasiatic

X

X

Albanian

Indo-European

X

X

German

Indo-European

X

X

Greek

Indo-European

X

X

Irish

Indo-European

X

X

Italian

Indo-European

X

X

postposition

ambiposition

X X

X

216 | Maike Vorholt

Language

Language family/ Significant Linguistic area correlation

Types of adposition preposition

Lithuanian

Indo-European

X

X

Polish

Indo-European

X

X

Basque

Isolate

X

Turkish

Turkic

X

Finnish

Uralic

X

Hungarian

Uralic

X

Georgian

Caucasian region

postposition

ambiposition

X X X

X

X

X X

Lezgian

Caucasian region

Armenian

Indo-European

X

X

Ossetic

Indo-European

X

X

X

X X

5 Discussion The adposition inventories of the languages in this study differ greatly from one another. The number of items considered adpositions varies between 18 and 99. Additionally, the types of adpositions the languages possess are different. While all three types are used in some languages, other languages exclusively use one type. The distribution of types is different across languages as well. Thus, as expected, the composition of adposition inventories in Europe is quite diverse. This finding conflicts with Hagège’s assessment that most languages that have both [prepositions] and [postpositions] do not have only one accidental [preposition] or [postposition]. The relative distribution of [prepositions] and [postpositions], in these languages, is of two types. First, we find languages in which the distribution is uneven. […] Second, there are languages in which a fairly good number of each of the two positional types of [adpositions] is found, even if we note a statistically dominant type. (Hagège 2010: 113)

In this sample, some languages do have only one or two members in one adposition category, while the majority of the members are of the other type. For instance, Ossetic has two prepositions while all of the others are postpositions, and Albanian only uses prepositions with the exception of one ambiposition. Although the aim was to list all of the adpositions of a language, there are undoubtedly items missing for most languages in this sample. First, this is due to the fact that only one source was used per language, which was a necessary

Can frequency predict length? | 217

constraint in order to treat all languages equally. Secondly, for most languages no complete list of adpositions is available in the extant literature. Additionally, authors do not mention whether or not the adposition list they provide is exhaustive. While ideally avoided, these limitations were unavoidable given the limited scope of this study. Adpositions are not used equally often across different languages. The proportion of adpositions in corpora of the sample languages varies greatly. A variation across languages was expected because some languages use other strategies for expressing relationships between words as well. It is nonetheless interesting to see which languages show the greatest divergence. Languages of the same language family do not necessarily show the same behavior when it comes to their adpositions. At the same time, a smaller adposition inventory does not lead to a smaller adposition proportion in the corpus. Irish, with less than 20 adpositions, has the third highest adposition proportion in the sample with over 13%. In contrast, German has almost 100 adpositions that cover less than 3% of the corpus. Across all languages in the sample, the proportion of adpositions varies between 15.27% (Maltese) to .15% (Lithuanian). It seems that the proportion of adpositions is not as consistent across languages as for other word classes, as has been found for nouns in several languages (cf. Hudson 1994; Liang and Liu 2013). Additionally, the highest frequency of occurrence is very different across the sample. Italian has the highest frequency of an individual adposition with almost 60 thousand, followed by Maltese and Greek while Finnish, Lithuanian, and Hungarian show a much smaller highest frequency of 600 per million words and less. The counted frequencies correspond only to a certain corpus and genre. Usage in different genres or spoken languages could thus behave differently. Furthermore, homophonous words belonging to other word classes might have been counted due to methodological restrictions mentioned in 3.6. Adposition length shows significant variance within and across different languages as well. Adpositions contain between one to 14 segments. The findings suggest that length can be predicted from frequency in twelve of the 16 languages in the sample. Thus, Zipf’s law of abbreviation does apply to the word class of adpositions. That said, this does not mean that frequency is the only relevant factor in predicting adposition length. For Italian, for example, 65% of the variation in length can be explained by frequency, which is the highest rate in the sample. These findings show that there is not just a connection between frequency and length but that one variable is enough to predict part of the other at least for some languages. For Hungarian, on the other hand,

218 | Maike Vorholt

frequency can only explain about 10% of variation in length. There are therefore other factors that need to be considered as well. The results do not rule out that frequency influences length for Armenian, Georgian, Lezgian, and Ossetic as well. They simply show that the regression model did not have a significant fit to the data. One way to yield higher rates to predict length could be to exclude outliers from the inventories. Should one do so, however, the data would not include all adpositions, so this method was not selected in the present study. All in all, there seems to be a tendency for languages that have prepositions in their inventory to yield a significant correlation between length and frequency. The sample is, of course, heavily biased for Indo-European languages as well as languages which use prepositions, so no generalizations can be made. A change in frequency leads to different changes in length, depending on the language. The highest rate of change can be observed for Italian with a decrease of .71, while Maltese adpositions exhibit a decrease of .28. This study has shown that Maltese takes on a special role when it comes to adposition proportion, with the highest proportion within the sample and contributing to the chi-square test with a higher frequency than expected. It also shows the smallest rate of change in length due to frequency and, after Italian, the second highest frequency for an individual adposition. However, it behaves just like most of the languages in the sample with regard to the size of its adposition inventory, length, and predictability of length.

6 Conclusion and outlook This study took a closer look at the word class of adpositions in a number of European languages using quantitative methods. Additionally, the variable length and frequency were observed in detail and the relationship between the two was investigated. The findings of this study expanded the work of previous researchers concerned with Zipf’s law of abbreviation, but also incorporated a novel focus specifically on adpositions. The study has shown that the word class of adpositions is quite diverse, in terms of size, frequency, and length, among the languages in the sample. Nevertheless, length can be predicted by frequency for most of the languages in the present study. Thus, though the languages are divergent, the same relationship between frequency and length can be observed. Although the results came out as expected, a pressing question for future researchers remains, namely, why a correlation between length and frequency does not apply to all languages is the sample. Surprisingly, these languages

Can frequency predict length? | 219

(Armenian, Georgian, Lezgian, and Ossetic) are all located in the Caucasian linguistic area. Future studies will need to test whether this applies to other languages spoken in that region as well. The findings need to be tested on other languages and genres to achieve wider coverage. It would be particularly interesting to see whether other units of length like syllables or morphemes lead to similar results. Furthermore, it is worth investigating whether the size of adposition inventories and their frequency is connected to having a rich case system. Basque, Hungarian, and Turkish use locative cases despite still having fairly large adposition inventories. To better understand the implications of these results, future studies need to address the question of whether languages that use prepositions, as opposed to post- or ambipositions, are more likely to show a correlation between frequency and length and, furthermore, why that may be the case. The results of this study illuminate important aspects of adpositions across a set of European languages and reveal correlations between them. More is known now about the frequency of adpositions as well as the fact that they do not take up the same proportion in corpora of different languages. Although adposition inventories might vary in size and composure, Zipf’s law is valid for the word class of adpositions. Though it does not apply to all inventories, it is nonetheless possible to predict the length of an adposition based on its frequency. These findings confirm that frequency of usage and length are strongly correlated, which makes it even more intriguing that some languages seem to lack this connection. The findings suggest a possible connection between the types of adpositions a language uses and the emergence of the law of abbreviation.

Abbreviations ABL ADE ALL DAT INEL PRO

ablative adessive allative dative inelative prolative

Acknowledgments: This paper is a concise version of my master’s thesis (Vorholt 2019) supervised by Thomas Stolz. Additionally to Thomas Stolz I am grateful to Julia Nintemann and Ray Fabri for their valuable and constructive comments on the draft of this paper.

220 | Maike Vorholt

References Aquilina, Joseph. 1991. Maltese-English dictionary. Vol. 2: M–Z and Addenda. Malta: Midsea. Baayen, Harald. 2008. Analyzing linguistic data. A practical introduction to statistics using R. Cambridge: Cambridge University Press. Bentz, Christian & Ramon Ferrer-i-Cancho. 2016. Zipf’s law of abbreviation as a language universal. In Christian Bentz, Gerhard Jäger & Igor Yanovich (eds.), Proceedings of the Leiden Workshop on Capturing Phylogenetic Algorithms for Linguistics. Tübingen: University of Tübingen, online publication system. Best, Karl-Heinz. 1997. Zur Wortartenhäufigkeit in Texten deutscher Kurzprosa der Gegenwart. In Karl-Heinz Best (ed.), Glottometrika 16 The distribution of word and sentence length, 276–285. Trier: Wissenschaftlicher Verlag Trier. Bybee, Joan. 2001. Phonology and language use. Cambridge: Cambridge University Press. Bybee, Joan & Sandra Thompson. 2007. Three frequency effects in syntax. In Joan Bybee (ed.), Frequency of use and the organization of language, 269–278. Oxford: Oxford University Press DeLancey, Scott. 2005. Adpositions as a non-universal category. In Zygmunt Frajzyngier, Adam Hodges & David S. Rood (eds.), Linguistic diversity and language theories, 185–202. Amsterdam & Philadelphia: John Benjamins. Diessel, Holger. 2007. Frequency effects in language aquisition, language use, and diachronic change. New Ideas in Psychology 25. 108–127. Ellis, Nick C. 2002. Frequency effects in language processing: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition 24(2). 143–188. Field, Andy P., Jeremy Miles & Zoë Field. 2012. Discovering statistics using R. London: Sage. Hagège, Claude. 2010. Adpositions. Oxford: Oxford University Press. Hammerl, Rolf. 1990. Länge-Frequenz, Länge-Rangnummer. Überprüfung von zwei lexikalischen Modellen. In Rolf Hammerl (ed.), Glottometrika 12, 1–24. Bochum: Brockmeyer. Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive Linguistics 19(1). 1–33. Haspelmath, Martin & Andres Karjus. 2017. Explaining asymmetries in number marking: Singulatives, pluratives, and usage frequency. Linguistics 55(6). 1213–1235. Hatzigeorgiu, Nick, George Mikros & George Carayannis. 2001. Word length, word frequencies, and Zipf’s law in the Greek language. Joural of quantitative linguistics 8(3). 175–185. Hudson, Richard. 1994. About 37% of word-tokens are nouns. Language 70(2). 331–339. Juilland, Alphonse, Dorothy Brodin & Catherine Davidovitch. 1970. Frequency dictionary for French words. The Hauge: Mouton. Kabak, Barış. 2006. An obstacle to the morphologization of postpositions. Studies in Language 30(1). 33–68. Kaeding, Friedrich Wilhelm. 1898. Häufigkeitswörterbuch der Deutschen Sprache. Steglitz bei Berlin: self edited. Kanwal, Jasmeen, Kenny Smith, Jennifer Culbertson & Simon Kirby. 2017. Zipf’s law of abbreviation and the principle of least effort: Language users optimise a miniature lexicon for efficient communication. Cognition 165. 45–52.

Can frequency predict length? | 221

Kassambara, Alboukadel. 2018. Chi-square test of independence in R. Statistical tools for highthroughput data analysis. http://www.sthda.com/english/wiki/chi-square-testofindependence-in-r#infos. Köhler, Reinhard. 1986. Zur linguistischen Synergetik: Struktur und Dynamik der Lexik. Bochum: Brockmeyer. König, Ekkehard & Marek Nekula. 2013. Zum Verhältnis von Kontrastiver Linguistik und Sprachtypologie: Präpositionen im Vergleich. In Marek Nekula, Katerina Sichová & Jana Valdrová (eds.), Bilingualer Sprachvergleich und Typologie: Deutsch – Tschechisch, 15–46. Tübingen: Stauffenberg/Julius Groos. Leipziger Universitätsverlag. 2019. Frequency dictionaries – Häufigkeitswörterbücher. https://www.univerlag-leipzig.de/catalog/bookstore/category/142Frequency_Dictionaries_Haeufigkeitswoerterbuecher. Levinson, Stephen, Sérgio Meira & The language and cognition group. 2003. “Natural concepts” in the spacial topological domain – Adpositional meaning in crosslinguistic perspective. An exercise in semantic typology. Language 79(3). 485–516. Liang, Junying & Haitao Liu. 2013. Noun distribution in natural languages. Poznan Studies in contemporary linguistics 49(4). 509–529. Lijffijt, Jefrey, Terttu Nevalainen, Tanja Säily, Panagiotis Papapetrou, Kai Poulamäki & Heikki Mannila. 2016. Significance testing of word frequencies in corpora. Digital Scholarship in the Humanities 31(2). 374–397. Lowie, Wander & Marjolijn Verspoor. 2004. Input versus transfer? – The role of frequency and similarity in the aquisition of L2 prepositions. In Michel Achard & Susanne Niemeier (eds.), Cognitive linguistics, second language acquisition, and foreign language teaching, 77–94. Berlin & New York: De Gruyter. Mańczak, Witold. 2010. Irregular sound change due to frequency in Germanic languages. Lingua Posnaniensis 52(1). 55–65. Mikosch, Ingrid. 1987. Die Präpositionen in gesprochener Sprache. Vorkommen und Funktion: untersucht an Tonbandaufnahmen aus Baden-Württemberg, Bayrisch-Schwaben und Vorarlberg. Tübingen: Max Niemeyer. Nübling, Damaris. 1998. Wann werden die deutschen Präpositionen flektieren? Grammatisierungswege zur Flexion. In Ray Fabri, Albert Ortmann & Teresa Parodi (eds.), Models of inflection, 266–289. Tübingen: Max Niemeyer. Piantadosi, Steven T., Harry Tily & Edward Gibson. 2011. Word lengths are optimized for efficient communication. PNAS 108(9). 3526–3529. RStudio Team. 2015. RStudio: Integrated Development for R. Boston: RStudio, Inc. http://www.rstudio.com/. Saint-Dizier, Patrick (ed.). 2006. Syntax and semantics of prepositions. Dordrecht: Springer. Schmidt, Emeli, Maike Vorholt & Nele Witt. 2020. Form and behavior of Maltese prepositions – A usage-based approach. In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese Linguistics on the Danube, 241–270. Berlin & Boston: De Gruyter Mouton. Schröder, Jochen. 1983. Überlegungen zu Status der Präpositionen. In Gabriele Schieb, Wolfgang Fleischer, Rudolf Große & Gotthard Lerchner (eds.), Beiträge zur Erforschung der deutschen Sprache, vol. 3, 22–47. Leipzig: VEB Bibliographisches Institut Leipzig. Sigurd, Bengt, Mats Eeg-Olofsson & Joost ven de Weijer. 2004. Word length, sentence length and frequency – Zipf revisited. Studia Linguistica 58(1). 37–52. Stolz, Thomas. 1990. Flexion und Adpositionen, flektierte Adpositionen, adpositionelle Flexion. ZPSK 43(3). 334–354.

222 | Maike Vorholt

Stolz, Thomas, Nataliya Levkovych, Aina Urdze, Julia Nintemann & Maja Robbers. 2017. Spatial interrogatives in Europe and beyond: where, whither, whence. Berlin & Boston: De Gruyter Mouton. Strauss, Udo, Peter Grzybek & Gabriel Altmann. 2007. Word length and word frequency. In Peter Grzybek (ed.), Contributions to the science of test and language, 277–294. Dodrecht: Springer. Sulkala, Helena & Merja Karjalainen. 1992. Finnish. London: Routledge. Svenonius, Peter. 2007. Adpositions, particles, and the arguments they introduce. In Eric J. Reuland, Tanmoy Bhattacharya & Giorgos Spathas (eds.), Argument structure, 63–103. Amsterdam & Philadelphia: John Benjamins. Svorou, Soteria. 1986. On the evolutionary paths of locative expressions. Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, 515–527. Vorholt, Maike. 2019. Can frequency predict length? A crosslinguistic investigation of Zipf's law for European languages. Unpublished master’s thesis. University of Bremen. Wilcox, Rand R. 2012. Introduction to robust estimation and hypothesis testing. Amsterdam: Academic Press. Zipf, George Kingsley. 1929. Relative frequency as a determinant of phonetic change. Harvard Studies in Classical Philology 40. 1–95. Zipf, George Kingsley. 1935. The psycho-biology of language. An introduction to dynamic philology. Boston: Houghton Mifflin Company. Zipf, George Kingsley. 1949. Human behavior and the principle of least effort. An introduction to human ecology. Cambridge: Addison-Wesley.

Literature for adposition lists Abaev, Vasilij I. 1964. A grammatical sketch of Ossetic. Bloomington: Indiana University, translated and edited by Herbert Paper. Ambrazas, Vytautas (ed.). 1997. Lithuanian grammar. Vilnius: Baltos Lankos. Bammesberger, Alfred. 1982. A handbook of Irish. 1. Essentials of Modern Irish. Sprachwissenschaftliche Studienbücher. Heidelberg: Carl Winter. Camaj, Martin. 1984. Albanian grammar. With exercises, chrestomathy, and glossaries. Wiesbaden: Otto Harrassowitz. Dékány, Éva. 2009. The nanosyntax of Hungarian postpositions. Nordlyd 36(1). 41–76. Dum-Tragut, Jasmine. 2009. Armenian. Modern Eastern Armenian. Amsterdam & Philadelphia: John Benjamins. Haspelmath, Martin. 1993. A grammar of Lezgian. Berlin & New York: De Gruyter. Hewitt, B. George. 1995. Georgian. A structural reference grammar. Amsterdam & Philadelphia: John Benjamins. Hualde, José Ignacio. 2003. Postpositions. In José Ignacio Hualde & Jon Ortiz de Urbina (eds.), A grammar of Basque, 187–90. Berlin & New York: De Gruyter. Holton, David, Peter Mackridge & Irene Philippaki-Warburton. 2004. Greek. An essential grammar of the modern language. London: Routledge. Kornfilt, Jaklin. 1997. Turkish. London: Routledge. Lehmann, Christian & Christel Stolz. 1992. Bildung von Adpositionen im Deutschen. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 6. Schwarze, Christoph. 1995. Grammatik der italienischen Sprache. Tübingen: Niemeyer.

Can frequency predict length? | 223

Skibicki, Monika. 2007. Polnische Grammatik. Hamburg: Buske. Stolz, Thomas & Nataliya Levkovych. 2020. From variation towards the grammar of Maltese prepositions – first steps. In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese Linguistics on the Danube, 199-240. Berlin & Boston: De Gruyter Mouton. Sulkala, Helena & Merja Karjalainen. 1992. Finnish. London: Routledge.

Corpora Albanian

Armenian Basque Finnish Georgian German Greek Hungarian Irish Italian

Lezgian Lithuanian Maltese Ossetic Polish Turkish

Albanian National Corpus, http://webcorpora.net/AlbanianCorpus/search/index.php?interface_language=en (19.09.2018) Eastern Armenian national corpus, version: 3.0, http://eanc.net/ (19.09.2018) Egungo Testuen Corpusa, version: 06.05.2016, https://www.ehu.eus/etc/ (22.11.2018) Kansalliskirjaston lehtikokoelman suomenkieliset lehdet, https://korp.csc. fi/#?stats_reduce=word&cqp=%5B%5D (22.11.2018) Leipzig Corpora Collection Georgian, http://corpora.unileipzig.de/en?Corpus Id=kat_newscrawl_2011 (22.11.2018) DWDS Kernkorpus 21, https://www.dwds.de/r?corpus=korpus21 (19.09.2018) Corpus of Modern Greek, http://web-corpora.net/GreekCorpus/search/ (19.09.2018) Hungarian National Corpus, version: v2.0.5, http://corpus.nytud.hu/mnsz/index_eng.html (22.11.2018) The new corpus for Ireland, version: ver:SkE-2.60.1-focloir-2.58.17, http://corpas.focloir.ie/ (19.09.2018) Repubblica, https://corpora.dipintra.it/public/run.cgi/first_form?corpname=repubblica;a lign= (22.11.2018) Corpus of Standard Lezgian, http://www.daglanguages. org/LezgianCorpus/search/index.php?interface_language=en Corpus of the contemporary Lithuanian language, http://tekstynas.vdu.lt/tekstynas/ (22.11.2018) Korpus Malti, version: 3.0, http://mlrs.research.um.edu.mt/CQPweb/malti03/ (19.09.2018) Ossetic National Corpus, http://corpus.osseticstudies. org/search/index.php?interface_language=en (19.09.2018) Leipzig Corpora Collection Polish, http://corpora.unileipzig. de/en?corpusId=pol_newscrawl_2011 (19.09.2018) Leipzig Corpora Collection Turkish, http://corpora.unileipzig. de/en?corpusId=tur_news_2005 (22.11.2018)

Computer programs R: RStudio:

version 3.5.2, available at: https://ftp.gwdg.de/pub/misc/cran/ version 1.1.463, available at: https://www.rstudio.com/products/rstudio/download/#download

224 | Maike Vorholt

Appendix I: Corpora Language

Corpus

Number of Tokens (in news section)

Albanian

Albanian National Corpus

12,230,215

Armenian

Eastern Armenian national corpus

47,264,999

Basque

Egungo Testuen Corpusa

17,000,000

Finnish

Kansalliskirjaston lehtikokoelman suomenkieliset lehdet

5,246,334,710

Georgian

Leipzig Corpora Collection Georgian

10,074,787

German

DWDS Kernkorpus 21

107,54,643

Greek

Corpus of Modern Greek

31,206,417

Hungarian

Hungarian National Corpus

430,303,070

Irish

The new corpus for Ireland

5,688,534

Italian

Repubblica

380,823,725

Lezgian

Corpus of Standard Lezgian

1,591,240

Lithuanian

Corpus of the contemporary Lithuanian language

86,497,837

Maltese

Korpus Malti

89,235,823

Ossetic

Ossetic National Corpus

3,192,632

Polish

Leipzig Corpora Collection Polish

96,476,260

Turkish

Leipzig Corpora Collection Turkish

13,506,869

Can frequency predict length? | 225

Appendix II: Bremen List of Maltese Prepositions (BLOMP, as described in Stolz and Levkovych 2020) Preposition

English translation

apparti

apart from

(min)barra

except

bejn

between

bħal

like

bħala

as

bi

with

biswit

facing

bla

without

daqs

equal to

dwar

about

faċċata

opposite

favur

on the side of

fejn

near

(min(n))flok

instead of

fi

in

fost

between

fuq

on

ġew(wa)

in

ġo

in

għajr

except

għal(a)

for

għand

at one’s house

ħdejn

beside

ħlief

except

‘il

to

kif

as, when, while

kontra

against

lejn

towards

lil

OBJECT MARKER

ma

with

madwar

around

maġenb

close to

226 | Maike Vorholt

Preposition

English translation

m(in)ħabba

on account of

mingħajr

without

mingħala

is someone’s opinion, under the impression of

mingħand

from s.o.

minkejja

in spite of

minn

from

mnejn

from where

oltri

except

permezz

by means of

qabel

before

qalb

amongst

qrib

near

quddiem

in front of

rigward

concerning

sa

till

sforz

thanks to

skont

according to

sotta

under

ta'

of

taħt

under

(ma)tul

during

versu

towards

viċin

near

waqt

instance

wara

after

Appendix III: Chi-square contribution in % Language

Adposition

Not adposition

Maltese

18.82

1.18

Italian

17.20

1.08

Irish

11.89

0.75

Polish

8.82

0.55

Greek

2.44

0.15

Can frequency predict length? | 227

Language

Adposition

Not adposition

Albanian

0.71

0.04

Hungarian

0.63

0.04

Armenian

0.92

0.06

Turkish

0.92

0.06

German

1.93

0.12

Lezgian

3.21

0.20

Basque

3.46

0.22

Georgian

3.74

0.24

Ossetic

5.93

0.37

Finnish

6.39

0.40

Lithuanian

7.08

0.44

Appendix IV: Results of regression analysis Multiple R2

P-value

logfrequency

F-ratio

df

Albanian

0.2553

1.42400E-03

-0.33783

12

35

Armenian

0.01947

0.1274

-0.13061

2.365

95

Language

Basque

0.3454

1.74E-09

-0.4675

45.38

86

Finnish

0.2694

1.54700E-05

-0.35647

22.12

60

Georgian

0.04832

1.25100E-01

-0.1874

2.437

48

German

0.3789

7.39500E-12

-0.668

60.39

99

Greek

0.3215

0.01413

-0.4804

7.582

17

Hungarian

0.09917

5.27800E-03

-0.2899

8.257

75

Irish

0.5454

0.0003053

-0.5348

20.39

17

Italian

0.65

5.05400E-09

-0.70944

61.29

33

Lezgian

0.01667

0.3715

-0.09222

0.8138

48

Lithuanian

0.2358

0.0001117

-0.4802

17.28

44

Maltese

0.1477

1.57500E-03

-0.28238

10.92

63

Ossetic

0.04986

0.2833

-0.09704

1.207

23

Polish

0.3639

1.62E-08

-0.47554

40.61

71

Turkish

0.1812

3.54700E-05

-0.38723

19.04

86

| Part IV: Phonology

Andrei A. Avram

Regressive voicing assimilation in Maltese: Diachrony and typology Abstract: In Modern Maltese obstruents in clusters agree in voicing. Earlier Maltese, however, did not have a rule of voicing assimilation. The present paper reconstructs the diachrony of voicing assimilation on the basis of the spelling used in records of earlier Maltese. It is shown that voicing assimilation appears to have become a rule of Maltese towards the end of the 18th century. Diachronically, voicing assimilation is therefore a case of rule addition and illustrative of lexical diffusion. The combined effect of the rules of voicing assimilation in obstruent clusters and of word-final obstruent devoicing is a typological shift. Keywords: lexical diffusion; obstruents; rule addition; typological shift; voicing assimilation

1 Introduction In Modern Maltese obstruents in clusters must agree in voicing. As put by Borg (1975: 15), “when pairs of voiced and voiceless sounds are brought together […], there often occurs the phenomenon of voicing assimilation (usually regressive)”. Regressive voicing assimilation1 triggers both obstruent devoicing and obstruent voicing, i.e. voicing neutralization is pre-obstruent neutralization. Borg (1975: 15) illustrates regressive devoicing with examples such as the following: (1) a. b. c.

[Borg 1975: 15] [rapta] ‘bond’ vs. [rabat] ‘he tied’ [ʔatfa] ‘rowing’ vs. [ʔadef] ‘he rowed’ [sfi:n] ‘dancing’ vs. [zifen] ‘he danced’

The next set of examples illustrates regressive voicing:

|| 1 Also known as “voicing harmony” (Borg and Azzopardi-Alexander 1997; Galea 2016; Galea and Ussishkin 2018). || Andrei A. Avram: Department of English, University of Bucharest, 7–13 Pitar Moş Str., 010451 Bucharest, Romania, E-mail: [email protected]

https://doi.org/10.1515/9783110783834-009

232 | Andrei A. Avram

(2) a. b.

[Borg 1975: 16] [vdejtna] ‘you redeemed us’ vs. [feda] ‘he redeemed’ [zdi:ri] ‘waistcoats’ vs. [sidrijja] ‘waistcoat’

To account for the cases of regressive devoicing exemplified in (1), Borg (1975: 15) proposes the following rule: (3)

cns [+voice]

→

cns [−voice]

/__ cns [−voice]

As for regressive voicing, illustrated in the examples under (2), Borg (1975: 16) formulates the rule below: (4)

cns [−voice]

→

cns [+voice, +obstruent]

/__ cns [+voice]

The rules in (3) and (4) are inadequate on three grounds. Firstly, they fail to capture the fact that regressive voicing assimilation affects obstruents only, to the exclusion of sonorants. Secondly, they do not show that the phonological context for the operation of the rules is restricted to obstruent clusters. Thirdly, the two separate rules for regressive voicing assimilation can be conflated into a single one, using the so-called “α notation”. Parawahera (1994–1995) proposes the rule in (5) to account for both regressive devoicing and voicing: (5)

C → [−son]

C _______ C [β voice] [β voice]

Formulated as such, the rule correctly indicates that it only targets obstruents, but its context of application is incorrectly stated. A reformulation is given below: (6)

[+obstruent] → [+obstruent, α voice] / __ [+obstruent, α voice]

Finally, as shown by Galea (2016: 27) and Galea and Ussishkin (2018: 66–67), “the voicing harmony rule is not strictly respected in clusters”.2 As is well known, Modern Maltese also exhibits word-final obstruent devoicing (Cohen 1966: 13; Borg 1975: 19; Krier 1975; Parawahera 1994–1995: 174; Borg 1997: 223; Borg and Azzopardi-Alexander 1997: 307). The relevant rule can be formulated as follows (Avram 2020: 28): (7)

[+obstruent] → [−voice] / __#

|| 2 Exceptions include e.g. dħul [dħu:l] ‘entrances’ (Galea 2016: 27), bqajt [bʔɐjt] ‘[I] stayed’ (Galea and Ussishkin 2018: 67).

Regressive voicing assimilation in Maltese: Diachrony and typology | 233

As noted by Parawahera (1994–1995: 174), word-final obstruent devoicing precedes regressive voicing assimilation (8a), whereas the reverse ordering yields an incorrect surface form (8b): (8)

a.

b.

/ħobz/ ħobs ħops [ħops] /ħobz/ ħobz ħobs *[ħobs]

UR word-final devoicing regressive voicing assimilation SR UR regressive voicing assimilation word-final devoicing SR

Parawahera (1994–1995: 174) also states that “Voicing Assimilation is a rule that spreads over to the neighboring consonants in a cluster” and illustrates spreading with the following example: “in [níktbu] the phonetic representation is [nígdbu] showing that [t] first assimilated with the voice quality of [b] and afterwards [k] became [g]”.3 This is shown in the derivation below: (9)

/niktbu/ nikdbu nigdbu [nigdbu]

UR regressive voicing assimilation of t regressive voicing assimilation of k SR

Finally, as noted by Fabri (2010: 792), in Modern Maltese “the letters generally retain their sound value”, however, “there are some surface phonetic/phonological rules that bring about changes in pronunciation that are not reflected in the spelling”. This holds for both regressive voicing assimilation of obstruents and word-final obstruent devoicing. The aim of the paper is to attempt to reconstruct the diachrony of word-final obstruent devoicing in Maltese. Together with word-final obstruent devoicing, regressive voicing assimilation has led to “a notable historical restructuring of the function of voicing in obstruents” (Borg 1997: 223). As will be shown, the combined effect of the two phonological processes is also a typological shift. The paper is organized as follows. Section 2 presents the corpus and the methodology. Section 3 illustrates regressive voicing assimilation in earlier stages of Maltese. Section 4 discusses the findings.

|| 3 Where [níktbu] should read /níktbu/ since this is the underlying representation.

234 | Andrei A. Avram

2 Corpus and methodology The corpus of Maltese covers a period ranging from the 15th century to the beginning of the 19th century. The evidence examined is from texts, word lists, vocabularies, dictionaries, and lists of place-names. The texts are Caxaro’s Cantilena (Wettinger and Fsadni 1968); Buonamico’s Sonnet (Cachia 2000); the sermons of Ignazio Saverio Mifsud (Ġabra tal-Malti Qadim n.d.); de Soldanis’ grammar (de Soldanis 1750); the Christian Catechism (Wzzino 1752), de Soldanis’ dialogues (Id-Djalogi ta’ de Soldanis n.d.); Cannolo’s translation of La Via Sagra (Cannolo 1796), and anonymous poems (Marshall and Vella Bonavita 1975). The lexicographical works consist of Megiser’s word list (Megiser 1610); Thezan’s dictionary (Cassola 1992); Skippon’s word list (Skippon 1732); de Sentmenat’s vocabulary (Queraltó Bartrés 2003); de Soldanis’ short dictionary (de Soldanis 1750); Il Mezzo Vocabolario Maltese-Italiano del’700 (Cassola 1996). The place names are from Abela (1647). The timeline of the regressive assimilation of voicing in Maltese is inferred from the orthography used in the sources. Reference is also made to metalinguistic comments by late 18th-century authors (de Soldanis 1750; Vassalli 1791, 1796) as well as by others writing in the first half of the 19th century (Vassalli 1827; Vella 1831 and 1838; Falzon 1845; Panzavecchia 1845). All examples appear in the orthography or transcription system used in the sources. The entries include the translation, the date of the attestations, the source, and comparative Arabic or Modern Maltese forms. The original glosses in Catalan, German, Italian or Latin in the sources have all been translated into English. When an exact year cannot be safely established, the following system is used: a year preceded by a hyphen reads ‘in or before’ and if followed by a hyphen ‘in or after’. For Arabic forms the following dictionaries have been consulted: Ben Sedira (1910); Barbera (1939a–b, 1940a–b); Harrell (1963); Ben Abdelkader (1977); Wehr (1976); Cherbonneau (1981); Aquilina (1987a–b).

3 Regressive voicing assimilation in early Maltese 3.1 The 15th century Caxaro’s Cantilena (c. 1450), the earliest text in Maltese, provides evidence that regressive voicing assimilation of obstruents was not yet a phonological process

Regressive voicing assimilation in Maltese: Diachrony and typology | 235

in the language. The relevant form is sib ‘[I] found’, which occurs in the verse reproduced below: (10)

fen timayt insib il gebel sib where hope.PFV.1SG 1SG.find.IPFV DEF stone.PL find.PFV.1SG tafal morchi c. 1450 [Wettinger and Fsadni 1968: 36] clay soft ‘I found loose clay where I had hoped to find rock.’

In their comments, Wettinger and Fsadni (1968: 45) only provide the Modern Maltese equivalent and the English translation: “sib tafal morchi, i.e. sibt tafal merħi: I found loose clay”. However, in their detailed linguistic analysis of Caxaro’s Cantilena, Cohen and Vanhove (1986: 190) state that “in sib tafal (“I found clay”) […] one must suppose that the suffix -t, […] of the 1st person of the perfective, was omitted at the end of sib by the copyist, because of the connection with the following word”. According to Cohen and Vanhove (1986: 190), this is “a frequent phenomenon with copyists in the Middle Ages”. Therefore, Caxaro’s ‘[I] found’ should read . On this analysis, in the obstruent cluster /bt/, the voiced /b/ does not undergo regressive voicing assimilation to [p].

3.2 The 16th century Regressive voicing assimilation is attested for the first time in Megiser’s list of Maltese words and phrases4, collected in 1588 (Megiser 1610). Two relevant forms occur in onset clusters: (11) a. b.

Guir ‘big.M’ 1588 cf. Ar. kabīr Sbe ‘morning’ 1588 cf. Ar. ṣabāḥ

[Megiser 1610: 9] [Megiser 1610: 12]

The in (11a) is an error of transcription or a misprint and should read . The form given by Megiser (1610: 10) is therefore gbir, identical to its Modern Maltese counterpart kbir [gbi:r]. However, the case of sbe, corresponding to Modern Maltese sbiħ [zbi:ħ], is less straightforward since Megiser (1610) uses to render either [s] or [z]5. There is also a third form, in a coda cluster:

|| 4 See also the detailed analysis in Cowan (1964). 5 As in Sigura ‘tree’ and Siemel ‘horse’, respectively (Megiser 1610: 11), cf. Modern Maltese siġra and żiemel.

236 | Andrei A. Avram

(12)

Chops ‘bread’ 1588 cf. Ar. ḫubz

[Megiser 1610: 10]

Note that the form in (12) illustrates both word-final devoicing of z to s and regressive voicing assimilation of b to p. Here again the form in Megiser (1610: 10) is identical to Modern Maltese ħobs [ħops].

3.3 The 17th century The first, and most revealing, source is Thezan’s (–1647) dictionary. As shown below, the forms6 listed in this dictionary illustrate regressive voicing assimilation of all types of obstruents, i.e. stops, fricatives and affricates as well as in all the types of clusters, i.e. onset, word-medial and coda clusters. The occurrence of regressive voicing assimilation is particularly well represented in onset clusters, as in the examples below: (13) a. b. c. d. e. f. g. h. i.

ptala ‘holiday’ –1647 [Cassola 1992: 171r] cf. Ar. baṭāla/biṭāla ptelt ‘[I] bathed’ –1647 [Cassola 1992: 134v] cf. Ar. ʔibtalalt pkeit ‘[I] cried’ –1647 [Cassola 1992: 134v] cf. Ar. bakayt pkait ‘[I] remained’ –1647 [Cassola 1992: 134v] cf. Ar. baqayt t‫ﺥ‬alt ‘[I] entered’ –1647 [Cassola 1992: 137v] cf. Ar. daḫalt p‫ ﺵ‬ein ‘for nothing’ –1647 [Cassola 1992: 171r] cf. Mod. Mal. b’xejn p‫ﺡ‬aira ‘melon field’ –1647 [Cassola 1992: 171r] cf. Alg. Ar. bḥayra7, Tun. Ar. buḥayra ‘vegetable garden’8 p‫ﺡ‬al ‘like’ –1647 [Cassola 1992: 171r] cf. Mod. Mal. bħal p‫ ﺵ‬ara ‘tip’ –1647 [Cassola 1992: 171r] Mor. Ar. bšaṛa9

|| 6 Thezan (–1647) uses a number of Arabic letters. The following appear in the examples taken from his dictionary: ﺫ‬ð], ﺵ‬ʃ], ﺥ‬x], ﺡ‬ħ], ﻫ‬h], and چ‬ʧ]. 7 See Behnstedt and Woidich (2012: 307). 8 See Wehr (1976: 43).

Regressive voicing assimilation in Maltese: Diachrony and typology | 237

Regressive voicing assimilation is also attested in a Romance loanword: (14)

p‫چ‬ei‫چ‬a ‘small piece’ –1647 [Cassola 1992: 121v] cf. Sic. piccia10, Mod. Mal. bċejċa (diminutive of biċċa ‘piece’)

Regressive voicing assimilation also occurs in several word-medial clusters (15) a. b. c. d. e. f.

‫ﺡ‬apka ‘basil’ –1647 [Cassola 1992: 145v] cf. Ar. habaqa ‫ ﺵ‬epka ‘net’ –1647 [Cassola 1992: 178r] cf. Ar. šabaka opsor ‘predict!’ –1647 [Cassola 1992: 170r] cf. Ar. ʔubṣur ‘see clearly, understand, comprehend’ mep‫ﺥ‬ara ‘censer’ –1647 [Cassola 1992: 163r] cf. Ar. mibḫara etfen ‘bury!’ –1647 [Cassola 1992: 137r] cf. Ar. ʔidfin esfen ‘dance!’ –1647 [Cassola 1992: 135v] cf. Ar. ʔizfin

Coda clusters also illustrate the occurrence of regressive voicing assimilation: (16) a. b. c. d. e. f. g.

sept ‘Saturday’ –1647 cf. Ar. sabt nep‫‘ ﻫ‬barking’ –1647 cf. Ar. nabḥ sep‫‘ ﻫ‬daybreak’ –1647 cf. Ar. ṣubḥ ‫ﺥ‬ame‫ ﺵ‬t ‘[I] spoiled’ –1647 cf. Ar. ḫamiǧt ‫ﺡ‬ege‫ ﺵ‬t ‘set ablaze!’ –1647 cf. Ar. haǧǧaǧt teie‫ ﺵ‬t ‘[I] married’ –1647 cf. Ar. tawwaǧ ‘to crown’ zeuüe‫ ﺵ‬t ‘[I] got married’ –1647 cf. Ar. tazawwaǧt, Mod. Mal. iżżewweġ

[Cassola 1992: 174v] [Cassola 1992: 167v] [Cassola 1992: 174v] [Cassola 1992: 155v] [Cassola 1992: 146v] [Cassola 1992: 181r] [Cassola 1992: 185v]

|| 9 The Moroccan Arabic word is polysemous. It is glossed “1. gift given to the mother of a newborn baby; 2. reward (e.g., for finding s.o.’ lost object); 3. good news” in Harrell (1963: 13). The second meaning is related to that given by Thezan (–1647). Modern Maltese bxara means ‘announcement, good news’. 10 See Barbera (1939a: 190), Aquilina (1987a: 116).

238 | Andrei A. Avram

Note that in examples (16d–g) the etymological voiced affricate ʤ is phonetically realized as the voiceless fricative [ʃ], an issue which is discussed in Section 4. There is also evidence of variation in the phonetic realization of obstruents in pre-obstruent position: (17)

lipsa vs. libsa ‘garment’ –1647 cf. Ar. libsa

[Cassola 1992: 7r and 157v, respectively]

On the other hand, the majority of forms do not exhibit regressive assimilation of voicing. These include the following: (18) a. b. c. d. e.

arkobtein ‘kneeling’ –1647 cf. Mod. Mal. għarkobbtejn ‫ﺥ‬obz ‘bread’ –1647 cf. Ar. ḫubz keb‫‘ ﺵ‬ram’ –1647 cf. Ar. kabš me‫ ﺵ‬eb‫ ﻫ‬a ‘likened’ –1647 cf. Mod. Mal. imxebbha e‫ﺡ ﺫ‬er ‘appear!’ –1647 cf. Ar. ʔiẓhar

[Cassola 1992: 119v] [Cassola 1992: 156v] [Cassola 1992: 152v] [Cassola 1992: 164r] [Cassola 1992: 129v]

The lists of place names in Abela (1647) include only two relevant forms, neither of which exhibits regressive voicing assimilation: (19)

Ta Tborsa 1647 cf. ?11

[Abela 1647: 105]

(20)

Vyed el Besbies ‘Valley of the Fennel’ 1647 cf. Ar. basbās12

[Abela 1647: 112]

Consider next Skippon’s word list (Skippon 1732: 624–626). Although dated at least 14 years after Thezan’s (–1647) dictionary, there is no evidence of regressive voicing assimilation in the only four relevant forms. Two of these are onset clusters: (21) a. b.

kbir ‘big.M’ 1664 cf. Ar. kabīr sbiacha ‘beauty’ 1664 cf. Ar. √ṣbḥ13

[Skippon 1732: 626] [Skippon 1732: 625]

|| 11 Tborsa may be a forgotten surname (Aquilina 1987b: 1412). 12 See Aquilina (1987a: 156). The Modern Maltese forms are busbies, biżbież or bużbież (Aquilina 1987a: 156).

Regressive voicing assimilation in Maltese: Diachrony and typology | 239

Note that Skippon consistently uses for [k] and for [s]. To the forms under (21) the following should be added: (22)

muscbahalu ‘unlike’ 1664 cf. Mod. Mal. mux bħalu ‘unlike him’

[Skippon 1732: 626]

The first in bahalu is not etymological, given that Maltese b is the contracted form of the preposition bi, cf. Arabic and Modern Maltese bi. Therefore, Skippon’s form bahalu is an error of transcription and should read bhalu, i.e. without regressive voicing assimilation of etymological b. Similarly, b fails to be devoiced in the following coda cluster: (23)

rekobt ‘knee’ 1664 cf. Mod. Mal. rkobbt- ‘knee-’

[Skippon 1732: 625]

Consider finally Buonamico’s Sonnet (Cachia 2000: 18), dated 1672. All three relevant forms exhibit regressive voicing assimilation, in one onset cluster (24) and two coda clusters (25), respectively: (24)

phal ‘like’ 1672 cf. Mod. Mal. bħal

(25) a. b.

[Cachia 2000: 18]

[Cachia 2000: 18] neptet ‘sprouted-F’ 1672 cf. Ar. nabatat tepki ‘[she] cries’ 1672 cf. Ar. tabkī

3.4 The 18th century Maltese is far better documented in the 18th century. Not only are there more numerous texts, but these also cover a wide range of genres: sermons, a translation of the Christian Catechism, dialogues, poems, a grammar, and several lexicographical works. These texts provide further evidence for the occurrence of regressive voicing assimilation. It is also in this century that explicit mention is made of this phonological process.

|| 13 The Modern Maltese equivalent is sbuħija ‘beauty’. The form recorded by Skippon is presumably an archaic one, not found in any of the dictionaries of Maltese consulted, derived from the comparative isbaḥ ‘more beautiful’. This derivational pattern is attested in Maghrebian Arabic: cf. Algerian Arabic biyāḍa ‘blancheur’ (Cherbonneau 1981: 58), from abyaḍ ‘whiter’.

240 | Andrei A. Avram

In chronological order, the first textual evidence comes from the sermons of Ignazio Saverio Mifsud, covering the period 1739–1746. This is the first Maltese text in which forms exhibiting regressive voicing assimilation outnumber those which do not. The following illustrate the occurrence of regressive voicing assimilation in onset clusters in words of Semitic origin: (26) a. b. c. d. e. f. g.

[Ġabra tal-Malti Qdim] pcait ‘[I] remained’ 1739 cf. Ar. baqayt p’caua ‘with force’ 1746 cf. Ar. bi qūwa p’cheit ‘[I] cried’ 1741 cf. Ar. bakayt psacaic ‘with your feet’ 1740 cf. Mod. Mal. b’saqajk psci ‘with some’ 1746 cf. Mod. Mal. b’xi thalt ‘[I] entered’ 1746 cf. Ar. daḫalt phal ‘like’ 1739 cf. Mod. Mal. bħal

Consider also the Romance loanword below: (27)

p’tant ‘with so much’ 1741 cf. Sic. tantu, It. tanto, Mod. Mal. b’tant

[Ġabra tal-Malti Qdim]

Regressive voicing assimilation is also attested in word-medial clusters: (28) a. b. c. d. e.

[Ġabra tal-Malti Qdim] darptein ‘twice’ 1741 cf. Mod. Mal. darbtejn harcupteina ‘on our knees’ 1741 cf. Mod. Mal. għarkobbtejna Nipchi ‘[I] cry’ 1741 cf. Mod. Mal. nibki ithac ‘[he] laughs’ 1741 cf. Ar. yaḍḥak imscepha ‘likened’1740 cf. Mod. Mal. imxebbha

Regressive voicing assimilation in Maltese: Diachrony and typology | 241

Again, this also holds for the following loanwords of Romance origin, in which syncope, i.e. the loss of the unstressed etymological vowel, brings together v and t: (29) a. b.

[Ġabra tal-Malti Qdim] sefturi ‘servants’ 1739 cf. Sic. servituri Issiftura ‘the maid’ 1741 cf. Sic. servitura

Regressive voicing assimilation also occurs in coda clusters: (30) a. b. c. d.

[Ġabra tal-Malti Qdim] hrapt ‘[I] ran away’ 1741 cf. Ar. harabt hsipt ‘[I] thought’ 1746 cf. Ar. ḥasibt hops ‘bread’ 1746 cf. Ar. ḫubz insapsc ‘was not found’ 1739 cf. Mod. Mal. insabx

Variation is also attested: (31)

tipca vs. tibca ‘you remain’ 1740 cf. Ar. tabqā

[Ġabra tal-Malti Qdim]

A few forms do not exhibit regressive voicing assimilation: (32) a. b. c. d.

[Ġabra tal-Malti Qdim] B’cant ‘with a chant’ 1740 cf. Mod. Mal. b’kant b’fidi14 ‘with faith’ 1741 cf. Mod. Mal. b’fidi bsebhu ‘with his finger’ 1740 cf. Mod. Mal. bsebgħu tifda15 ‘[you] trust’ 1740 cf. Mod. Mal. tafda

|| 14 From Sicilian fidi or Italian fede (Aquilina 1987a: 334). 15 From Sicilian fidari or Italian fidare (Aquilina 1987a: 295).

242 | Andrei A. Avram

In the Catalan-Maltese vocabulary attributed to Marquis de Sentmenat, and believed to have been written at the beginning of the second half of the 18th century (Queraltó Bartrés 2003: 7), forms which exhibit regressive voicing assimilation outnumber by a ratio of 2 to 1 those which do not. The forms below exemplify regressive voicing assimilation in onset clusters: (33) a. b. c.

Pkaila ‘spinach’ 1750– 16 cf. Ar. buqayla ptaiel ‘holidays’ 1750– cf. Mod. Mal. btajjel phal ‘like’ 1750– cf. Mod. Mal. bħal

[Queraltó Bartrés 2003: 28] [Queraltó Bartrés 2003: 103] [Queraltó Bartrés 2003: 102]

Regressive voicing assimilation in word-medial clusters occurs in examples such as the following: (34) a. b. c. d. e.

ar kupteik ‘on your knees’ 1750– cf. Mod. Mal. għarkobbtejk Ipki ‘[he] cries’ 1750– cf. Ar. yabkī Lipsa ‘garment’ 1750– cf. Ar. libsa 17 Sipsi ‘pipe’ 1750– Alg. Ar. sebsi18, Mor. Ar. sebsi19, Ethhhól ‘enter!’ 1750– cf. Ar. ʔudḫul

[Queraltó Bartrés 2003: 67] [Queraltó Bartrés 2003: 37] [Queraltó Bartrés 2003: 28] [Queraltó Bartrés 2003: 36] [Queraltó Bartrés 2003: 28]

Consider also the form in (35). On the assumption that the Maltese reflex of uvular q had shifted to a velar articulation, i.e. that of [k], this is a rare case of obstruent voicing due to regressive assimilation:20 (35)

togba ‘hole’ 1750– cf. Ar. ṯuqba

[Queraltó Bartrés 2003: 29]

|| 16 Diminutive of baqla ‘greens’. 17 A word apparently no longer in use in Modern Maltese, not recorded in any of the dictionaries consulted. 18 See Ben Sedira (1910: 474). 19 See Harrell (1963: 134). 20 The very few examples of regressive voicing of obstruents may well be an artefact of the sources.

Regressive voicing assimilation in Maltese: Diachrony and typology | 243

Regressive voicing assimilation in coda clusters is illustrated by the examples below: (36) a. b.

[Queraltó Bartrés 2003: 30] Gipps ‘plaster’ 1750– cf. Ar. ǧibs Hhops ‘bread’ 1750– cf. Ar. ḫubz

Forms which do not display regressive assimilation of voicing include the following: (37) a. b. c.

kbira ‘big-F’ 1750– cf. Ar. kabīra darbtein ‘twice’ 1750– cf. Mod. Mal. darbtejn tilhhhabx ‘don’t play’ 1750– cf. Mod. Mal. tilgħabx

[Queraltó Bartrés 2003: 30] [Queraltó Bartrés 2003: 41] [Queraltó Bartrés 2003: 31]

In Wzzino’s (1752) translation of Dottrina Cristiana, the first book printed in Maltese, forms which display regressive voicing assimilation outnumber by a ratio of more than 2 to 1 those which do not. Regressive voicing assimilation is attested in a number of onset and word-medial clusters, as shown in the examples under (38) and (39), respectively: (38) a. b. c. d. (39)

ptajel ‘holidays’ 1752 cf. Mod btajjel p-chul ‘with every’ 1752 cf. Ar. bi kull pceiec ‘pieces’ 1752 cf. Mod. Mal. bċejjeċ phhal ‘like’ 1752 cf. Mod. Mal. bħal

jipk̇a ‘[he] remains’ 1752 cf. A yabqā

[Wzzino 1752: 49] [Wzzino 1752: 113] [Wzzino 1752: 32] [Wzzino 1752: 45] [Wzzino 1752: 78]

The following forms illustrate the non-occurrence of regressive voicing assimilation: (40) a. b.

jibtell ‘[he] bathes’ 1752 cf. Mod. Mal. jibtell nibchu ‘[we] cry’ 1752 cf. Mod. Mal. nibku

[Wzzino 1752: 98] [Wzzino 1752: 89]

244 | Andrei A. Avram

c. d.

ḣhobs ‘bread’ 1752 cf. Ar. ḫubz ma nidnibsċ ‘[I] do not sin’ 1752 cf. Mod. Mal. ma nidnibx

[Wzzino 1752: 111] [Wzzino 1752: 108]

The next texts examined, with a slight violation of chronological order, are two works by de Soldanis, i.e. his grammar and dictionary of Maltese (de Soldanis 1750) and his dialogues (Id-Djalogi ta’ de Soldanis n.d.), written sometime after 1760. In de Soldanis (1750) forms exhibiting regressive voicing assimilation are attested in onset and coda clusters only. Consider the examples under (41) and (42), respectively: (41) a. b. c. (42)

psciara ‘happy announcement’ 1750 21 cf. Ar. bišāra ‘good news’ phhalu ‘like him’ 1750 cf. Mod. Mal. bħalu sfin ‘dancing’ 1750 cf. Mod. Mal. żfin

hhaps ‘prison’ 1750 cf. Ar. ḥabs

[de Soldanis 1750: 165] [de Soldanis 1750: 187] [de Soldanis 1750: 93] [de Soldanis 1750: 144]

In one case, variation in pronunciation is explicitly mentioned: (43)

hhobs or chops ‘bread’ 1750 cf. Ar. ḫubz

[de Soldanis 1750: 193]

Regressive voicing assimilation does not occur in forms such those under (44): (44) a. b. c. d. e.

Qbir ‘big.M’ 1750 cf. Ar. kabīr ibqa ‘[he] remains’ 1750 cf. Ar. yabqā gebt ‘[I] brought’ 1750 cf. Mod. Mal. ġibt ma gebsc ‘[I] didn’t bring’ 1750 cf. Mod. Mal. ma ġibtx Nebhh ‘barking’ 1750 cf. Ar. nabḥ

[de Soldanis 1750: 167] [de Soldanis 1750: 115] [de Soldanis 1750: 113] [de Soldanis 1750: 113] [de Soldanis 1750: 162]

|| 21 For the earlier and current meanings of the Modern Maltese counterpart bxara see fn. 9.

Regressive voicing assimilation in Maltese: Diachrony and typology | 245

Regressive voicing assimilation is also documented in de Soldanis’ dialogues (Id-Djalogi ta’ de Soldanis. n.d.), in a majority of forms. The examples below illustrate the occurrence of regressive voicing assimilation in onset clusters: (45) a. b. c. d. e.

[Id-Djalogi ta’ de Soldanis n.d.] ptan ‘breed’ 1760– cf. Mod. Mal. btan pkaina ‘[we] remained’ 1760– cf. Ar. baqaynā dbeddel ‘[it] changed’ 1760– cf. Mod. Mal. tbiddel phhaga ‘with a thing’ 1760– cf. Ar. bi ḥāǧa phhal ‘like’ 1760– cf. Mod. Mal. bħal

It is worth drawing attention to the form in (45c) since it is one of the extremely rare instances of regressive voicing of obstruents found in pre-19th-century sources.22 Word-medial clusters also include several forms displaying regressive voicing assimilation: (46) a. b. c.

[Id-Djalogi ta’ de Soldanis n.d.] i chaptulek ‘[they] knock on your’ 1760– cf. Mod. Mal. iħabbtulek iepka ‘[he] remains’ 1760– cf. Ar. yabqā ietchol ‘[he] enters’ 1760– cf. Ar. yadḫul

Finally, regressive voicing assimilation is attested in coda clusters as well. Note in (47d) that the fricative [ʃ] is the devoiced counterpart of the affricate ʤ. (47) a. b. c.

[Id-Djalogi ta’ de Soldanis n.d.] Hhsept ‘[I] thought’ 1760– cf. Ar. ḥasibt Ghagest ‘[I] have grown old’ 1760– cf. Mod. Mal. għaġiżt ma essipsc ‘[it] is not found’ 1760– cf. Mod. Mal. ma ssibx

|| 22 See also fn. 20.

246 | Andrei A. Avram

d.

ma żeuesctc ‘[I] did not get married’ 1760– cf. Mod. Mal. ma żżewwiġtx

As shown by e.g. Marshall (1971), de Soldanis’ dialogues reflect variation in Maltese, as spoken in the second half of the 18th century. Three forms illustrate intra-speaker variation (48) and, respectively, inter-speaker variation (48b–c) in the occurrence of regressive voicing assimilation, where D = domanda ‘question’ and R = risposta ‘answer’: (48) a. b. c.

[Id-Djalogi ta’ de Soldanis n.d.] D tebka vs. D tepka‘[she] remains’ 1760– cf. Ar. tabqā D sehhebtek ‘your companion’ vs. R sehhepti ‘my companion’ 1760– cf. Mod. Mal. seħibti, seħibtek R n-elbsu ‘[we] get dressed’ D lepsin ‘dressed-PL’ 1760– cf. Mod. Mal. lebsin, nilbsu

Finally, a few forms do not exhibit regressive assimilations of voicing: (49) a. b. c. d.

[Id-Djalogi ta’ de Soldanis n.d.] gebt ‘[I] brought’ 1760– cf. Mod. Mal. ġibt klubkom ‘your hearts’ 1760– cf. Ar. qulūbukum chobs ‘bread’ 1760– cf. Ar. ḫubz ma en hhobsc ‘[I] don’t love’ 1760– cf. Mod. Mal. ma nħobbx

The next set of examples is from a partially preserved dictionary, Il Mezzo Vocabulario Maltese-Italiano del ‘700 (Cassola 1996), believed to have been written before 1775. Regressive voicing assimilation is best reflected in onset clusters, including a Sicilian-derived loanword (50i): (50) a. b. c.

[Cassola 1996: 132] ptàla ‘holiday’ –1775 cf. Ar. baṭāla/biṭāla pkia ‘rest’ –1775 cf. Ar. baqīya psisa ‘dust of small stones’ –1775 cf. Ar. basīsa ‘flour mixed with butter and oil’23

|| 23 See Aquilina (1987a: 157). See also the discussion of Modern Maltese bsisa in Cassola (1996: 132–133).

Regressive voicing assimilation in Maltese: Diachrony and typology | 247

d. e. f. g. h. i.

pxara ‘tip’ –1775 Mor. Ar. bšaṛa pxein ‘for nothing’ –1775 cf. Mod. Mal. b’xejn pxima ‘placenta’ cf. Leb. Ar., Syr. Ar. bašīma24 phaira ‘melon field’ –1775 cf. Alg. Ar. bḥaira, Tun. Ar. buḥaira ‘vegetable garden’ phal ‘like’ –1775 cf. Mod. Mal. bħal pceicia ‘little piece’ cf. Sic. piccia, Mod. Mal. bċejċa (diminutive of biċċa ‘piece’)

Forms such as those reproduced below illustrate the occurrence of regressive voicing assimilation in word-medial clusters: (51) a. b. c. d.

hapka ‘basil’ –1775 cf. Ar. habaqa mepchara ‘censer’ –1775 cf. Ar. mibḫara tesfen ‘you dance’ –1775 cf. Ar. tazfin zeuextu ‘you.PL married’ –1775 cf. Mod. Mal. żewweġtu

[Cassola 1996: 62] [Cassola 1996: 106] [Cassola 1996: 82] [Cassola 1996: 175]

Finally, regressive voicing assimilation is also attested in coda clusters. Consider the following examples: (52) a. b.

cherext ‘[he] got out’ –1775 cf. Ar. ḫaraǧt zeuext ‘[I] married’ –1775 cf. Ar. zawwaǧt

[Cassola 1996: 125] [Cassola 1996: 175]

However, the majority of relevant forms, in onset, word-medial and coda clusters, do not evince regressive assimilation of voicing, as seen in the examples below: (53) a. b.

rcobtein ‘knees’ –1775 cf. Mod. Mal. rkobbtejn cbir ‘big.M’ –1775 cf. Ar. Kabīr

[Cassola 1996: 136] [Cassola 1996: 89]

|| 24 According to Barbera (1939a: 257), bašīma is used especially in Lebanon.

248 | Andrei A. Avram

c. d. e. f. g.

liebsa ‘dressed.F’ –1775 cf. Ar. lābisa gibt ‘[I] brought’ –1775 cf. Mod. Mal. ġibt habs ‘prison’ –1775 cf. Ar. ḥabs nebh ‘barking’ –1775 cf. Ar. nabḥ sebh ‘daybreak’ –1775 cf. Ar. ṣubḥ

[Cassola 1996: 94] [Cassola 1996: 55] [Cassola 1996: 58] [Cassola 1996: 120] [Cassola 1996: 142]

As mentioned at the beginning of this subsection, it is in 18th century works that the occurrence in Maltese of regressive voicing assimilation is explicitly mentioned. In the first edition of his grammar of Maltese, Vassalli (1791: 80) writes about the letter “Be Bb” that “before T K Ф S &  […] it sounds P”.25 Vassalli (1791: 80) illustrates this case of regressive assimilation of voicing with btâla ‘holiday’, bkejt ‘[I] cried’, bɸajt ‘[I] remained’, lybsæ ‘garment’ and beuɸa ‘with a wish’, about which he writes that “these words are read and heard btâla pkejt pɸajt lypsæ p euqa”.26 A similar statement is made with respect to the letter “Zajn Zz”, which “before Ф, T […] sounds S” (Vassalli 1791: 90). The examples given by Vassalli (1791: 90) include zɸūɸ ‘bellies’ and hmyzt ‘[I] pinned”.27 Regressive voicing assimilation is again briefly discussed in Vassalli’s (1796) dictionary, this time with reference to two obstruents only. One case mentioned is “the letter B [which] when it meets with one of these five letters T K Ф S  is however pronounced P” (Vassalli 1796: XXXII). The other case is the “letter D” which “before the letters F  K & Ф is read T” (Vassalli 1796: 140). Some variation appears to be still occurring, however, as suggested by examples found in the last 18th-century source examined, Cannolo’s (1796) translation of La Via Sagra. On the one hand, regressive voicing assimilation occurs in several onset and word-medial clusters, as shown in the examples under (54) and (55), respectively: (54) a.

psabar ‘with patience’ 1796 cf. Ar. bi ṣabr

[Cannolo 1796: 11]

|| 25 In Vassalli’s (1791) Maltese alphabet, stand for Modern Maltese , < ,  > for . 26 In btâla is an error of transcription and should read

. 27 The third example given by Vassalli (1791: 80) is irrelevant: obz ‘bread’, where stands for etymological ḫ.

Regressive voicing assimilation in Maltese: Diachrony and typology | 249

b. (55) a. b.

Psalibech ‘with your cross’ 1796 cf. Mod. Mal. b’salibek

[Cannolo 1796: 7]

[Cannolo 1796: 11] jipchu ‘they cry’ 1796 cf. Mod. Mal. jibku ulietchom cf. Mod. Mal. uliedkom

On the other hand, variation between forms with and without regressive voicing assimilation is attested in the same sentence: (56)

la ti-bch-u-sc fuk-i […], ima i-pch-u fuk-om 1796 no 2-cry.IMP-PL-NEG over-1SG but 2-cry.IMP-PL over-2PL ‘don’t cry over me, but cry over yourselves’ [Cannolo 1796: 11]

Moreover, three forms do not exhibit regressive voicing assimilation: (57) a. b. c.

chbir ‘big.M’ 1796 cf. Ar. kabīr chbira ‘big.F’ 1796 cf. Ar. kabīra l’ibsa ‘garment’ 1796 cf. Ar. libsa

[Cannolo 1796: 11] [Cannolo 1796: 15] [Cannolo 1796: 13]

In light of the evidence surveyed in this section, it can be concluded that towards the end of the 18th century regressive voicing assimilation appears to have essentially become a phonological process in Maltese.

3.5 The 19th century Two types of evidence are relevant to the issue of regressive voicing assimilation in early 19th-century Maltese. The first type consists of evidence from a number of texts whose authors employ what appears to be, in spite of inconsistencies, a more phonetically-oriented orthography. Marshall and Vella Bonavita (1975) published four anonymous poems which appear to date from the first decades of the 19th century. The orthography employed is phonetic and therefore reflects pronunciation. Regressive voicing assimilation is attested in examples such as those given below: (58) a.

p’feugia ‘with a breeze’ 1800– [Marshall and Vella Bonavita 1975: 120] cf. Mod. Mal. b’fewġa

250 | Andrei A. Avram

b. c. d. e. f.

phal ‘like’ 1800– cf. Mod. Mal. bħal thalt ‘[I] entered’ 1800– cf. Ar. daḫalt imhaptec ‘your love’ 1800– cf. Mod. Mal. imħabbtek ipchi ‘[he] will cry’ 1800– cf. Ar. yabkī iipca ‘[it] will stay’ 1800– cf. Ar. yabqā

[Marshall and Vella Bonavita 1975: 122] [Marshall and Vella Bonavita 1975: 124] [Marshall and Vella Bonavita 1975: 112] [Marshall and Vella Bonavita 1975: 106] [Marshall and Vella Bonavita 1975: 118]

Regressive voicing assimilation is amply represented in book primer (spelling book) Anon. (1824). The following are cases of regressive devoicing of obstruents, in onset, word-medial and coda clusters: (59) a. b. c. d. e.

phal ‘like’ 1824 cf. Mod. Mal. bħal tip-ka [you] remain’ 1824 cf. Ar. tabqā ithol ‘[he] enters’ 1824 cf. Ar. yadḫul ithak ‘[he] laughs’ 1824 cf. Ar. yaḍḥak tigdipx ‘don’t lie’ 1824 cf. Mod. Mal. tigdibx

[Anon. 1824: 6] [Anon. 1824: 10] [Anon. 1824: 19] [Anon. 1824: 11] [Anon. 1824: 19]

Note, in particular, example (59b), in which the word is divided into syllables, with a devoiced reflex of etymological b in the coda of the first syllable. Since in Maltese regressive voicing assimilation is not syllable-final, the occurrence of [p] in this form can only be due to the influence of the following k. It is also worth mentioning that cases of regressive voicing of obstruents are well represented. The examples under (60) are forms of adjectives and verbs from the root √kbr: (60) a. b. c. d.

gbira ‘big.F’ 1824 cf. Ar. kbīra gbar ‘big.PL’ 1824 cf. Ar. kibār agbar ‘bigger’ 1824 cf. Ar. akbar tigber ‘[you] grow’ 1824 cf. Ar. takbir

[Anon. 1824: 3] [Anon. 1824: 30] [Anon. 1824: 13] [Anon. 1824: 27]

Regressive voicing assimilation in Maltese: Diachrony and typology | 251

e. f.

igber ‘[he] grows’ 1824 cf. Ar. yakbir nigbru ‘[we] grow’ 1824 cf. Mod. Mal. nikbru

[Anon. 1824: 20] [Anon. 1824: 27]

Deverbal nouns (61a) and derived forms of verbs (61b–d) with a voiced reflex of an etymological t are also represented: (61) a. b. c. d.

dbatia ‘suffering’ 1824 cf. Mod. Mal. tbatija dbi‫ﻉ‬ed ‘[he] went away’ 1824 cf. Mod. Mal. tbiegħed dbixclet ‘[she] was embroiled’ 1824 cf. Mod. Mal. tbixklet dbelleh ‘[he] lost his wits’ 1824 cf. Mod. Mal. tbellah

[Anon. 1824: 27] [Anon. 1824: 18] [Anon. 1824: 15] [Anon. 1824: 22]

The second type of evidence consists of metalinguistic comments by various Maltese authors from the first half of the 19th century. For instance, the issue of the regressive assimilation of voicing is again addressed, in more detail, in the second edition of Vassalli’s grammar (Vassalli 1827). About “the B” it is mentioned that “whenever it immediately meets with one of the following letters, c, k, q, ɦ, ɧ, s, , t, will be heard pronounced […] as if it were p” (Vassalli 1827: 6); this is illustrated with seven words: bcejcæ ‘small piece’, bkejna ‘[we] cried’, bɦâli ‘like me’, bɧǔr ‘incense’, bsarna ‘[we] predicted’, bâra ‘announcement’, and btâla ‘holiday’.28 Next, Vassalli (1827: 6–7) states that “the D […] is heard however pronounced t, whenever it is immediately followed by f, ɦ, ɧ, k, q”, and that this happens in “for instance, nydfen, [I] will bury; nydɦak, [I] will laugh; nydɧol, [I] will enter”. It is also noted that “the G […] is sometimes heard as ; e.g. ɦrigt, [I] went out’ (Vassalli 1827: 7). The last case mentioned is that of “the Z [which] before q […] is heard pronounced as if it were an s”, as in “Alla tâhom yr-ryzq […]; God gave them fortune” (Vassalli 1827: 9). Consider next Vella’s (1831) grammar. Regressive voicing assimilation is only mentioned and illustrated with reference to the letters and , respectively. It is stated about the former that “when it is followed by c,  , ḣ, S, T, X, [it] is pronounced P” (Vella 1831: 12). The pronunciation is indicated in an English-like transcription: “Bciecen Pigeons pronounce Pchiechen, Bcheina We

|| 28 In Vassalli’s (1827) alphabet, = [ʧ], and represent etymological and respectively.

252 | Andrei A. Avram

cryed [sic] Pkeina, Ob ra Seas Op [sic], Tibḣir Perfume Tipḣir, Libsa Address [sic]29 Lipsa, Btie i Cloisters Ptie i” (Vella 1831: 12). As for , the author writes that it “assumes the sound of T when it is followed by the letters , ḣ, k” (Vella 1831: 13), and the following examples illustrate its regressive voicing assimilation: “Dḣaḣen Smoks [sic] pronounce Tḣaḣen, Midfur Weaved Mitfur, Midkuk Beaten Mitkúk, Nid ac I laugh Nit ak” (Vella 1831: 14). Essentially the same account is found in Vella’s (1838) school primer (spelling book). The author writes about that “this letter followed by c,  , ḣ, j, t, x is pronounced like the p” (Vella 1838: 6). Similarly, it is stated with respect to that “this letter takes the sound of T when it is followed by the letters  , ḣ, k” (Vella 1838: 6). A more comprehensive description of regressive voicing assimilation in Maltese is also found in Falzon (1845), the first edition of his dictionary. Falzon (1845: viii) writes with respect to that “being however followed by the letters c, f, ħ, ḣ, k, q, s, t, x, z it is pronounced as p” and provides the following examples:30 “Bciecèn, pigeons; B’fer∩a, with a branch; Bħajra, mellon field, pumpkin field; Bḣùr, perfume, incense; Bkièt, [she] cried; Bqajla, spinach; Bsarna, [we] predicted; Btàla, holiday, festive day; Bxàra, good news; Bzièzen, roll of bread” (Falzon 1845: viii). Similarly, “when it is immediately followed by the letters f, ħ, ḣ, k, q, it is heard pronounced like t” as in “Nidfru, [we] braid; Nidħku, [we] will laugh; Nidḣlu, [we] will enter; Tidkik, pounding; Dqìq, flour” (Falzon 1845: viii). The “is sometimes heard x; as, ḣrigt ‘[I] went out’” (Falzon 1845: ix). Finally, “before q […] is heard like an s” (Falzon 1845: x). Rather surprisingly, although textual evidence documents it, voicing of stops in consonant clusters, e.g. /kb/ realized as [gb] or /tb/ realized as [db], due to regressive assimilation is not mentioned by any of these above authors. Note also that regressive voicing of fricatives, as in the examples given in section 1, is not mentioned either. Summing up, both the empirical and the metalinguistic evidence discussed in this section converge on the conclusion that in the first decades of the 19th century the rule of regressive voicing assimilation of obstruents is firmly entrenched in the phonology of Maltese. Regressive voicing assimilation is therefore a late development in the historical phonology of Maltese. As seen in section 1, regressive voicing assimilation continues to operate as a synchronic rule of Modern Maltese. That is, for some two centuries now, regressive voicing assimilation has been a persistent rule (in the sense of Cser 2015: 197) of Maltese phonology. || 29 The gloss should read “Dress”. 30 In Falzon (1845), = [ʧ], and represent etymological ḫ and ḥ respectively.

Regressive voicing assimilation in Maltese: Diachrony and typology | 253

4 Discussion and conclusion Before discussing the findings, two preliminary issues need to be addressed. The first one is the reliability of the sources and, by way of consequence, of the data. As seen in Section 3, four of the sources are works by non-Maltese authors: Megiser’s word list, Thezan’s dictionary, Skippon’s word list, and de Sentmenat’s vocabulary. Two of these, Hieronymus Megiser and Philip Skippon, were speakers of languages which do not exhibit regressive voicing assimilation of obstruents: German does have obstruent devoicing, but it is syllable-final (Lombardi 1999: 268; Wetzels and Mascaró 2001: 208); in English, there is voicing assimilation of obstruents, but it is rather progressive. Thezan and de Sentmenat, however, were speakers of French and Catalan respectively, both displaying regressive voicing assimilation of obstruents (Lombardi 1999: 269). A naturally arising question is whether Thezan and de Sentmenat may have been influenced by the phonologies of their mother tongues when transcribing Maltese. In the case of Thezan, the answer is safely in the negative. In the entries of his dictionary, nouns are given in their singular and plural forms, as in (62a), and for verbs the imperative masculine singular, the perfective of the 1st and 3rd person singular are listed. The entries provide ample evidence that the author was generally very much aware of the occurrence of alternants: (62) a. b.

be‫چ چ‬a ‘piece’ vs. p‫چ‬ei‫‘ چ‬pieces’ –1647 cf. Mod. Mal. biċċa [bitʧa] vs. bċejjeċ [pʧejjeʧ] t‫ﺥ‬alt ‘[I] entered’ vs. da‫ﺥ‬al ‘[he] entered’ -1647 cf. Mod. Mal. dħalt [tħalt] vs. daħal [daħal]

[Cassola 1992: 123] [Cassola 1992: 137v]

As for de Sentmenat, as already mentioned in Section 3.4, forms which exhibit regressive voicing assimilation outnumber those which do not by a ratio of 2 to 1, whereas in the translation of the Christian Catechism by Wzzino, who was a native speaker of Maltese writing in approximately the same period, the ratio is of more than 2 to 1. The fate of /ʤ/, when undergoing devoicing due to regressive assimilation, is also of some interest. As shown in Sections 3.3 and 3.4 respectively, in all the relevant examples from Thezan, de Soldanis and Il Mezzo Vocabolario MalteseItaliano del ’700, the devoiced realization of the affricate /ʤ/ is the fricative [ʃ], rather than the expected affricate [ʧ]. The same holds for the descriptions of regressive voicing assimilation by Vassalli (1827), Vella (1838) and Falzon (1845). On the other hand, Panzavecchia

254 | Andrei A. Avram

(1845: 29) explicitly mentions variation:31 “the G changes sometimes into Sc, therefore, so instead of ‘ƕrigt’ one says ‘ƕrisct’; other times it changes into  ; so instead of ‘ƕrigt’ one also says ‘ƕri t’”.32 After almost a century, Sutcliffe (1936: 2–3) states in his grammar of Maltese that the letter “assumes the sharp sound of ċ […] when followed by […] t, e.g. ħriġt I went out (pronounced also ‘ħrixt’*)”. Some thirty years later, Aquilina (1965: 19) gives [ʧ] as what he calls the “breathed” or “unvoiced” counterpart of voiced /ʤ/. More recently, Cardona (1997: 104) writes that “in consonant clusters, the last consonant which is voiced or voiceless affects the preceding one” and one of the examples given is “ħriġt → hriʧt”. This suggests that the devoiced realization of /ʤ/ in consonant clusters went through the following stages: [ʃ] in pre-1800 Maltese; a transitional stage, up to the first half of the 20th century, in which either [ʃ] or [ʧ] occurred; finally, [ʧ] from the second half of the 20th century onwards. Turning to the findings, consider first the approximate date when regressive voicing assimilation became a phonological process in Maltese. As mentioned in Section 3.5, it is suggested that this must have occurred by the end of the 18th century. In addition to the empirical and metalinguistic evidence surveyed in Section 3, this conclusion finds support in typological findings. In her crosslinguistic study of voicing neutralization, Lombardi (1994: 54) writes that, in principle, “there could be preobstruent neutralization, which would apply in clusters; spreading; and a separate rule of word-final neutralization”. However, she further states that this “predicts that a language could have word-final devoicing, but no [voicing] neutralization in clusters” and that she has “found no languages of this type” (Lombardi 1994: 54). In other words, as noted by among others Lombardi (1994: 54), Wetzels and Mascaró (2001: 209), there is an implicational universal accounting for obstruent voicing assimilation and devoicing: word-final devoicing ⊃ regressive voicing assimilation. Diachronically, this means that cross-linguistically word-final obstruent devoicing of necessity precedes or coincides with the generalization of regressive voicing assimilation as a phonological rule. This is also the case in Maltese, in which word-final obstruent devoicing is first attested in the 15th century and becomes a phonological rule by the end of the 18th century, as shown in Avram (2017, 2020). The generalization of regressive voicing assimilation appears to have been a slow process, lasting more than two centuries, from its first attestation in 1588,

|| 31 In the alphabet used by Panzavecchia (1845) corresponds to etymological ḫ or ḥ, and represents [ʧ]. 32 Incidentally, this is one of the words which consistently exhibits [ʃ] as the devoiced realization of /ʤ/ in pre-19th-century sources.

Regressive voicing assimilation in Maltese: Diachrony and typology | 255

in the word list compiled by Megiser (1610), to the end of the 18th century, when it was first mentioned by Vassalli (1791, 1796). Consider in Table 1 a selection of forms illustrative of regressive voicing assimilation in 10 of the sources examined in Section 3: Table 1: Regressive voicing assimilation in onset clusters.

Megiser Thezan

Skippon Mifsud

Sentmenat Wzzino de

de

Mezzo

1588

1664

1750–

Soldanis

Soldanis

Vocabolario 1796

1750

1760–

–1775

1647

1739–

1752

1746

ptala

ptaiel

pkeit

p’cheit

pkait

pcait

kbira

Qbir

p‫ ﺵ‬ara

pschara psciara

p‫ﺡ‬aira p‫چ‬ei‫چ‬a

pkia

jibtel kbir

t‫ﺥ‬alt

ptàla

pkaina

eptel Gir

ptajel

Cannolo

kbir

chbir pxara phaira

thalt pceiec

pceicia

As can be seen, while several forms consistently exhibit regressive voicing assimilation in all the sources, others continue to contain unassimilated obstruents. A particularly striking case is that of ‘big’: although Gir is one of two earliest examples of regressive voicing assimilation, the word still displays a cluster consisting of a voiceless obstruent and a voiced one as late as 1796, in chbir. Consider next the forms in Table 2 which illustrates cases of regressive assimilation in word-medial clusters, attested in 8 sources: Here again some forms, e.g. rcobtein ‘knees’ display unassimilated obstruents, even though regressive voicing assimilation is attested in earlier records. Also, there is variation in several forms in the same paradigm, e.g. tibchusc ‘don’t cry!’ ~ ipchu ‘cry!’, the latter dating from more than 150 years after the attestation of epki ‘cry!’ with regressive voicing assimilation.

256 | Andrei A. Avram

Table 2: Regressive voicing assimilation in word-medial clusters.

Thezan

Buonamico Mifsud

Sentmenat Wzzino

de Soldanis Mezzo Vocabolario Cannolo

1647

1672

1750–

1760–

1739–1746

arkobtein

harcupteina ar kupteik

darbtein

darptein

Darbtein

tipchi

ipki

epki

tepki

epka

1752

–1775

1796

rcobtein

tibchusc ipchu

nibchu nipġhu

tibca tipca

tebka tepka

‫ﺡ‬apka

hapka

libsa lipsa

lipsa

lipsa

me‫ ﺵ‬eb‫ ﻫ‬a

imscepha

n-elbsu lepsin

mep‫ﺥ‬ara

liebsa

mepchara

et‫ﺥ‬ol

ithol

Ethhhól

ietchol

esfen

tesfen

A similar picture emerges from the comparison of selected forms with coda clusters, occurring in 8 sources: Table 3: Regressive voicing assimilation in coda clusters.

Megiser

Thezan

Mifsud

Sentmenat

Wzzino

de Soldanis

de Soldanis

Mezzo Vocabolario

1588

1647

1739–1746

1750–

1752

1750

1760–

–1775

hsipt sept

Hhsept Issabt

hhaps Chops

‫ﺥ‬obz nep‫ﻫ‬ sep‫ﻫ‬

habs hops

Hhops

ḣhobs chops hhobs

chobs Nebhh

nebh sebh

Note, for instance, the forms for ‘bread’: it is first attested in 1588 with the obstruent cluster [ps] which agrees in voicing, but forms with [bs] still occur in the second half of the 18th century. Consider also the forms habs ‘prison’ and sebh ‘daybreak’, both recorded by 1775, with the unassimilated voiced obstruent [b],

Regressive voicing assimilation in Maltese: Diachrony and typology | 257

compared to hhaps and respectively sep‫ ﻫ‬with the assimilated voiceless obstruent [p], attested some 125 years earlier. The data in the comparative Tables 1–3 suggest that regressive voicing assimilation in Maltese is a case of lexical diffusion. The central claim of lexical diffusion theory is that sound change is phonetically abrupt, but lexically gradual (Phillips 2015; Bybee 2016: 39). On this theory, a sound change does not affect all relevant lexical items, but occurs only in a small subset of the potential targets. Also, in some words there will be both intra- and inter-speaker variation for a certain period of time, characterized by the coexistence of two competing phonetic realizations – the initial one and the new one resulting from the sound change. At a still later stage, the sound change spreads both to other lexical items and to other speakers. Finally, the sound change may extend to all relevant words. For various reasons, it is impossible to follow in detail the diffusion of regressive voicing assimilation in Maltese. One such reason is the paucity of 15th- and 16th-century records of Maltese. Also, the available records are certainly representative of several dialectal varieties of Maltese. Thezan’s dictionary reflects the variety of 17th-century Maltese spoken in the area of Valletta (Hull 1994: 394), whereas de Soldanis’ works represent the dialect of Gozo (Vella 2013). On the other hand, as put by Cardona (1997: 22), “we do not know whether Megiser or Skippon took those words from the city or from the countryside”. Finally, the task is certainly not facilitated by the fact that transcriptions are marred by inconsistencies. Diachronically, regressive voicing assimilation in Maltese is a case of rule addition, one of the four major types of rule change within the framework of generative historical phonology (Dresher 2015). This type of rule change does not trigger a modification of underlying representations (Dresher 2015: 511). Indeed, the current underlying representations of most of the forms examined in Section 3 contain clusters of obstruents which do not agree in voicing. There are only very few exceptions. Consider, for instance, two of the forms listed in Section 3.4, psisa ‘dust of small stones’ and pxima ‘placenta’. Their Modern Maltese counterparts are or 33 [psi:sɐ], and respectively or 34 [pʃi:mɐ]. The first word, a singulare tantum, is not integrated in a paradigm, whereas in the case of the second one, its plural or [pʃejjem] contains the same word-initial cluster. Etymologically, both words are derived from forms with word-initial b; synchronically, however, their underlying representations are /psi:s/ and /pʃi:mɐ/ respectively. Hence, || 33 Aquilina (1987a: 145). 34 See Schembri (1998: 27).

258 | Andrei A. Avram

while the spelling with reflects the etymology, the one with

is justified phonologically. Finally, the emergence in Maltese of regressive voicing assimilation together with word-final obstruent devoicing also has typological consequences. Work by Lombardi (1994: 53–54, 1999: 268–269), Wetzels and Mascaró (2001: 208– 209), Kenstowicz et al. (2003: 260) on obstruent voicing assimilation and wordfinal neutralization of voicing has identified the types of languages attested. Modern Maltese belongs to the type “voicing assimilation in obstruent clusters with word-final neutralization” (Lombardi 1999: 269). However, Arabic – its putative ancestor – belongs to the type called “voice unrestricted” by Lombardi (1999: 268), in which there is “contrast of voiced and voiceless obstruents initially, medially, finally and in clusters” (Kenstowicz et al. 2003: 260). The combined effect of obstruent regressive voicing assimilation and word-final obstruent devoicing has thus triggered a typological shift.

Abbreviations 1, 2 Alg. Ar. DEF IMP IPFV F

It. Leb. M

Mod. Mal. Mor. NEG PFV PL SG

Sic. SR Syr. Tun. UR

1st, 2nd person Algerian Arabic definite imperative imperfective feminine Italian Lebanese masculine Maltese Moroccan negator perfective plural singular Sicilian surface representation Syrian Tunisian underlying representation

Regressive voicing assimilation in Maltese: Diachrony and typology | 259

References Abela, Giovanni Francesco. 1647. Della descrittione di Malta isola nel mare siciliano. Con le sue antichita, ed altre notitie. Malta: Paolo Bonacota. Anon. 1824. Chtieb ilkari yau Dahla ‫ﻉ‬al ilsien Malti [The Book of Reading or Introduction to the Maltese Language]. Ligorni. Aquilina, Joseph. 1965. Maltese. A complete course for beginners. Sevenoaks: Hodder and Stoughton. Aquilina, Joseph. 1987a. Maltese – English dictionary, vol. 1, A–L. Valletta: Midsea Books. Aquilina, Joseph. 1987b. Maltese – English dictionary, vol. 2, M–Z. Valletta: Midsea Books. Avram, Andrei A. 2017. Word-final obstruent devoicing in Maltese: Inherited, internal development or contact-induced?. Academic Journal of Modern Philology 6. 23–37. Avram, Andrei A. 2020. The diachrony of word-final obstruent devoicing in Maltese. In Slavomír Čéplö & Jaroslav Drobný (eds.), Maltese Linguistics on the Danube, 27–57. Berlin & Boston: De Gruyter Mouton. Barbera, Giuseppe. 1939a. Dizionario maltese-arabo-italiano. Con una grammatica comparata arabo-maltese, vol. I, A–E. Beirut: Imprimerie Catholique. Barbera, Giuseppe. 1939b. Dizionario maltese-arabo-italiano. Con una grammatica comparata arabo-maltese, vol. II, F–L. Beirut: Imprimerie Catholique. Barbera, Giuseppe. 1940a. Dizionario maltese-arabo-italiano. Con una grammatica comparata arabo-maltese, vol. III, M–S. Beirut: Imprimerie Catholique. Barbera, Giuseppe. 1940b. Dizionario maltese-arabo-italiano. Con una grammatical comparata arabo-maltese, vol. IV, T–Ġ. Beirut: Imprimerie Catholique. Behnstedt, Peter & Manfred Woidich. 2012. Wortatlas der arabischen Dialekte, vol. II, Materielle Kultur. Leiden & Boston: Brill. Ben Abdelkader, Rached. 1977. Peace Corps English-Tunisian Arabic dictionary. Washington, D.C.: Peace Corps. Ben Sedira, Bel Kassem. 1910. Dictionnaire Français-Arabe de la language parlée en Algérie. Algiers: Adolphe Jourdan. Borg, Alexander. 1975. Maltese morphophonemics. Journal of Maltese Studies 10. 11–28. Borg, Alexander. 1997. Maltese phonology. In Alan S. Kaye (ed.), Phonologies of Asia and Africa, vol. 1, 245–285. Winona Lake, IN: Eisenbrauns. Borg, Albert & Marie Azzopardi-Alexander. 1997. Maltese. London & New York: Routledge. Bybee, Joan. 2016. Language change. Cambridge: Cambridge University Press. Cachia, Lawrenz. 2000. Ħabbew l-ilsien Malti [They loved the Maltese language]. Zabbar: Veritas Press. Cannolo, Gioseppe Martino. 1796. La Via Sagra, tradotta in Maltese da Gioseppe Martino Cannolo. Malta. Cardona, Tony. 1997. Introduzzjoni għal-lingwistika Maltija [Introduction to Maltese linguistics]. L-Msida: Mireva Publications. Cassola, Arnold. 1992. The Biblioteca Vallicelliana Regole per la lingua Maltese. The earliest extant grammar and dictionary of the Maltese language. Valletta: Said International. Cassola, Arnold (ed.). 1996. Il Mezzo Vocabolario Maltese-Italiano del ’700. Valletta: Said International. Cherbonneau, Auguste. 1981. Dictionnaire français-arabe. Beyrouth: Librairie du Liban.

260 | Andrei A. Avram

Cohen, David. 1966. Le système phonologique du maltais. Aspects synchroniques et diachroniques. Journal of Maltese Studies 3. 1–26. Cohen, David & Martine Vanhove. 1986. La Cantilène maltaise du 15ème siècle: remarques linguistiques. Comptes-rendus du Groupe linguistique d’études chamito-sémitiques XXIX–XXX. 177–200. Cowan, William. 1964. An early Maltese word-list. Journal of Maltese Studies 2. 217–225. Cser, András. 2015. Basic types of phonological change. In Patrick Honeybone & Joseph C. Salmons (eds.), The Oxford handbook of historical phonology, 193–204. Oxford: Oxford University Press. Id-Djalogi ta’ de Soldanis [de Soldanis’s Dialogues]. n.d. http://malti.skola.edu.mt/ kotba/djalogi_desoldanis.pdf. Dresher, Elan. 2015. Rule-based generative historical phonology. In Patrick Honeybone & Joseph C. Salmons (eds.), The Oxford handbook of historical phonology, 501–521. Oxford: Oxford University Press. Fabri, Ray. 2010. Maltese. Revue Belge de Philologie et d’Histoire 88(3). 791–816. Falzon, Giovanni Battista. 1845. Dizionario Maltese-Italiano-Inglese arrichito di nomi di botanica, ittiologia, ornitologia, e marineria come pure di varie frasi, modi di dire, e proverbi, preceduto da una breve esposizione grammaticale della lingua maltese. Malta. Galea, Luke. 2016. Syllable structure and gemination in Maltese. Cologne: University of Cologne PhD dissertation. Galea, Luke & Adam Ussishkin. 2018. Onset clusters, syllable structure and syllabification in Maltese. In Patrizia Paggio & Albert Gatt (eds.), The languages of Malta, 55–79. Berlin: Language Science Press. Ġabra tal-Malti Qdim [Corpus of Old Maltese]. n.d. http://mlrs.research.um.edu.mt/ resources/gabrastorika. Harrell, Richard S. (ed.). 1963. A dictionary of Moroccan Arabic:Moroccan-English. Washington, D.C.: Georgetown University Press. Hull, Geoffrey. 1994. Review of Arnold Cassola, The Biblioteca Vallicelliana: Regole per la lingua maltese. The earliest Extant Grammar and Dictionary of the Maltese Language. Journal of Semitic Studies 39(2). 391–395. Kenstowicz, Michael, Mahasen Abu-Mansour & Miklós Törkenczy. 2003. Two notes on laryngeal licensing. In Stefan Ploch (ed.), Living on the edge: 28 Papers in honour of Jonathan Kaye, 260–282. Berlin & New York: De Gruyter. Krier, Fernande. 1975. Analyse phonologique du maltais. Phonetica 32. 103–129. Lombardi, Linda. 1994. Laryngeal features and laryngeal neutralization. London: Routledge. Lombardi, Linda. 1999. Positional faithfulness and voicing assimilation in Optimality Theory. Natural Language and Linguistic Theory 17. 267–302. Marshall, David R. 1971. A social and historical commentary on ‘The Dialogues’ of de Soldanis. Journal of Maltese Studies 6. 1–33 Marshall, David R. & Roger Vella Bonavita. 1975. Four anonymous Old Maltese Poems. Journal of Maltese Studies 10. 89–127. Megiser, Hieronymus. 1610. Propugnaculum Europe. Wahrhafte/Eigentliche und Ausführliche Beschreibung/der viel und weitberühmten Afrikanischen Insul Malta. Leipzig: Henning Großn. Panzavecchia, Fortunato. 1845. Grammatica della lingua maltese, spiegata secondo I principi delle lingue orientali e della lingua italiana. Malta: M. Weiss.

Regressive voicing assimilation in Maltese: Diachrony and typology | 261

Parawahera, N. P. 1994–1995. Lexical strata in Maltese. Journal of Maltese Studies 25–25. 166– 178. Phillips, Betty S. 2015. Lexical diffusion in historical phonology. In Patrick Honeybone & Joseph C. Salmons (eds.), The Oxford handbook of historical phonology, 359–373. Oxford: Oxford University Press. Queraltó Bartrés, Alexandre. 2003. Un vocabulari català – maltès manuscript del segle XVIII. Edició, transcripció i estudi. Barcelona: PPU. Schembri, Ludovik. 1998. Dictionary Dizzjunarju Maltese – English English – Maltese. Mġarr: Colour Image. Skippon, Philip. 1732. An Account of a Journey Made thro’ Part of the Low-Countries, Germany, Italy, and France. London. 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. Roma: Salomoni. Sutcliffe, Edmund F. 1936. A grammar of the Maltese language with chrestomathy and vocabulary. Oxford: Oxford University Press. Vassalli, Mikiel Anton. 1791. Mylsen Phoenico-Punicum sive Grammatica Melitensis. Rome: Antonio Fulgoni Vassalli, Mikiel Anton. 1796. Ktŷb yl Klŷm Mâlti. Mfysser byl-Latin u byt-Taljan. Lexicon Melitense-Latin-Italum. Vocabolario maltese. Recato nelle lingue Latina e italiana. Rome: Antonio Fulgoni. Vassalli, Michelantonio. 1827. Grammatica della lingua maltese. Rome: The Author. Vella, Francis. 1831. Maltese grammar for the use of the English. Leghorn: laucus Masi. Vella, Francesco. 1838. Abbecedario gal kadi tas schejel ta taglm xilxjeni jau Senga ḣafifa biex tarreġ fil kari it tfal [Primer for service in schools for mutual teaching or easy steps to teach children how to read]. Livorno: Matbag Volpi. Vella, Olvin. 2013. A Gozitan dialect for Standard Maltese. 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 & Mikiel Fsadni. 1968. Peter Caxaro’s Cantilena. A Poem in Medieval Maltese. Malta. Wetzels, W. Leo. & Joan Mascaró. 2001. The typology of voicing and devoicing. Language 77. 207–244. Wzzino, Francesco. 1752. Taġhlim nisrani. Migiub fil Ġharbi bl’ordini ta sant. mem. ta Paulu V. Min gdid stampat fit-Talian mis Sacra Congregazioni tal Propaganda Fidi, u’ migiub issa fil Malti [Christian Catechism. Translated into Arabic on Orders of the Sacred Memory of Paul V. Reprinted in Italian by the Sacred Congregation de Propaganda Fide, and Translated Now into Maltese]. Rome: Arcangelo Casaletti.

Index of Authors Á Kempis, Thomas 11, 13 Abela, Giovanni Francesco 234, 238 Aikhenvald, Alexandra Y. 24 Albright, Adam 65 Al-Jallad, Ahmad 60 Amaira, Mark 87 Amberber, Mengitsu 114f. Ambros, Arne Amadeus 61 Aoun, Joseph 105, 115f. Aquilina, Joseph 3f., 7, 24, 47f., 50, 57ff., 70, 72, 74f., 91f., 94ff., 99, 101, 107, 136ff., 145f., 149ff., 154, 165f., 169, 210, 234, 237f., 241, 246, 254, 257 Arkadiev, Peter 110 Avram, Andrei A. 232, 254 Azzopardi, Kirsty 83f. Azzopardi-Alexander, Marie 105, 107ff., 119f., 146, 156, 166, 169, 175, 182, 186, 190, 231f. Baayen, Harald 124, 208 Badawi, El-Said 152, 158 Baker, Mark S. 5f. Baldwin, Timothy 24 Barbera, Giuseppe Maria 7, 56, 72, 234, 237, 247 Bartolo, Simon 165, 169ff., 177, 187f. Baumann, Stefan 120 Behnstedt, Peter 59, 61, 236 Béjar, Susana 110 Bellarmino, Roberto 10ff. Ben Abdelkader, Rached 139ff., 144f., 147ff., 154ff., 234 Ben Sedira, Bel Kassem 234, 242 Bentz, Christian 201 Bergsträsser, Gotthelf 138, 140f., 158 Best, Karl-Heinz 199 Bickel, Balthasar 106 Black, Jeremy 9 Blanc, Haim 53 Blevins, Juliette 52 Booij, Geert 24

https://doi.org/10.1515/9783110783834-010

Borg, Albert 83ff., 89f., 105, 107ff., 119f., 146, 156, 166, 169, 175, 182, 186f., 190, 231f. Borg, Alexander 231ff. Borsley, Robert D. 58 Bossong, Georg 106 Boukhris, Fatima 157 Brincat, Joseph M. 4 Brockelmann, Carl 170 Brustad, Kirsten 116 Bybee, Joan L. 65f., 198, 200, 257 Cachia, Lawrenz 234, 239 Calzolari, Nicoletta 24, 44 Camilleri, Maris 84f., 172, 182 Cannolo, Gioseppe Martino 234, 248f., 255f. Cardona, Tony 254, 257 Casado-Fresnillo, Celia 166, 181f. Cassola, Arnold 71, 234, 236ff., 246ff., 253 Cauchi, Carmel G. 175f. Čéplö, Slavomír 69, 74, 105, 108, 114, 117, 120f. Cesàro, Antonio 152 Cherbonneau, Auguste 234, 239 Cohen, David 232, 235 Cojocaru, Dana 114 Comrie, Bernard 110 Corriente, Federico 59 Cowan, William 235 Cowell, Mark W. 116, 139, 141f., 145, 149, 152, 154, 156, 192 Cser, András 252 Culbertson, Jennifer 109f. Dalrymple, Mary 109 Danecki, Janusz 4 De Cat, Cécile 105, 110 De Liguori, Alfonso Maria 14ff. Deguara, Abigail 32 DeLancey, Scott 203, 211 Demuth, Katherine 105 Diessel, Holger 198, 200 Dixon, R. M. W. 24 Downing, Laura J. 110

264 | Index of Authors

Dresher, Elan 257 Du Bois, John W. 110f. Durand, Olivier 154, 156 Edzard, Lutz 6 Ellis, Nick C. 200 Fabri, Ray 7, 36, 42, 44, 83ff., 89f., 105ff., 188, 210, 233 Falzon, Giovanni Battista 234, 252f. Fanselow, Gisbert 111 Fenech, Edward 53 Ferrer-i-Cancho, Ramon 201 Field, Andy P. 207f. Finlay, W. M. 50 Fischer, Wolfdietrich 72f. Flask, Wayne 97 Fodor, Jerry A. 5 Forker, Diana 106 Franks, Steven 113f. Friedman, Victor A. 110, 113 Friggieri, Olivier 16 Fsadni, Mikiel 139, 141f., 146, 149, 152, 155, 158, 234f.

Ibn Manẓūr 137, 139ff., 154 Iemmolo, Giorgio 110, 114, 117 Iggesen, Oliver 183 Ingham, Bruce 59 Jakobson, Roman 66 Johnstone, Thomas M. 60 Juilland, Alphonse 198f. Just, Erika 109, 112ff.

Gai, Amikam 6 Galea, Luke 231f. Givón, Talmy 6 Goldap, Christel 167 Guillaume, Jean-Patrick 6

Kabak, Barıš 199 Kaeding, Friedrich Wilhelm 198 Kallulli, Dalina 105, 113f. Kanwal, Jasmeen 201 Kany, Charles E. 166, 183, 192 Karjalainen, Merja 205 Karjus, Andres 201 Kassambara, Alboukadel 208 Katz, Jerrold 5 Kenstowicz, Michael 258 Khan, Geoffrey 8 King, Tracy H. 113f. Klamer, Marian 110f. Köhler, Reinhard 199, 201 König, Ekkehard 198 Kontzi, Reinhold 71 Kratchovíl, František 110f. Krer, Mohamed 58 Krier, Fernande 3, 232 Krzeszowski, Tomasz P. 5

Hagège, Claude 164, 198, 204f., 216 Haig, Geoffrey 106 Hammerl, Rolf 201 Harrell, Richard S. 234, 237, 242 Hartmann, Reinhard R. 5 Haspelmath, Martin 165, 201, 205 Hasselbach, Rebecca 72 Hatzigeorgiu, Nick 197 Haywood, John Alfred 137 Hinds, Martin 152, 158 Holes, Clive 56 Hooper, Joan 65 Hothorn, Torsten 125 Hudson, Richard 198, 217, 220 Huehnergard, John 60f. Hull, Geoffrey 257

Lachowski, Joseph 9 Lane, Edward 74 Latin 201, 205, 234 Lehmann, Christian 167 Levin, Aryeh 116, 135 Levinson, Stephen 198 Levkovych, Nataliya 166, 190f. Levshina, Natalia 124 Liang, Junying 198, 217 Lieber, Rochelle 24 Lijffijt, Jefrey 208 Liu, Haitao 198, 217 Lombardi, Linda 253f., 258 Loop, Jan 8 López, María del Carmen Fernández 164 Lowie, Wander 198

Index of Authors | 265

Lucas, Christopher 69, 74 Lyons, Christopher 119 Maas, Utz 84f., 88 Magidow, Alexander 72 Mańczak, Witold 200 Markantonatou, Stella 24 Marshall, David R. 234, 246, 249f. Mascaró, Joan 253f., 258 Matić, Dejan 110 Megiser, Hieronymus 136, 234ff., 253, 255ff. Michalski, Marcin 6 Mifsud, Manwel 69, 136 Mikosch, Ingrid 199 Muxí, Isabel 110 Nahmad, Hayim Musa 137 Nekula, Marek 198 Nichols, Johanna 106 Nikolaeva, Irina 109 Nübling, Damaris 179, 205 O’Leary, De Lacy Evans 72 Owens, Jonathan 6, 61, 192 Panzavecchia, Fortunato 7, 15, 139ff., 145f., 149, 151, 154ff., 234, 253f. Parawahera, N. P. 232f. Penrice, John 137, 139ff., 144ff., 149, 151 Pereira, Christophe 58 Peterson, John 84 Phillips, Betty S. 257 Piamenta, Moshe 58 Piantadosi, Steven T. 201 Prémare, André-Louis de 60 Priestley, Carol 111 Procházka, Stephan 136ff., 140, 151, 157, 164 Queraltó Bartrés, Alexandre 234, 242 Quilis, Antonio 166, 181f. Retsö, Jan 61 Riedel, Kristina 110f. Rohlfs, Friedrich Gerhard 72 Rowling, Joanne K. 168 Ryding, Karen C. 137, 142f., 146ff., 150

Saari, Rami 136f., 164, 179 Saint-Dizier, Patrick 198 Saint-Exupéry, Antoine de 168f., 171 Saydon, Peter P. 47, 51, 72 Schembri, Ludovik 257 Schikowski, Robert 114 Schmidt, Emeli 137, 166, 199, 206 Schnell, Stefan 114 Schröder, Jochen 198 Scicluna, Karl 7 Seržant, Ilja A. 106, 109, 128 Sigurd, Bengt 197, 201 Singer, Hans-Rudolf 58, 64, 71, 73 Skippon, Philip 234, 238f., 253, 255, 257 Skopeteas, Stavros 111 Snelling, Thomas 50 Soldanis, Giovanni Pietro Francesco Agius de 7, 15, 60, 62, 138f., 141f., 145, 149f., 154, 156, 158, 234, 244ff., 253, 255ff. Souag, Lameen 60, 77 Spiro, Socrates 152 Štekauer, Pavol 24 Stokes, Phillip 59 Stolz, Thomas 24, 84f., 87f., 91, 137, 144, 158, 163, 166, 168, 170, 190f., 204, 206f. Strauss, Udo 197, 201 Stumme, Hans 70, 136 Su Nam, Kim 24 Sulkala, Helena 205 Sutcliffe, Edmund F. 7, 56, 254 Svenonius, Peter 198 Svensson, Ann-Marie 65 Svorou, Soteria 206 Tagliamonte, Sali 124 Tasmowski, Liliane 105 Taylor, Charles V. 112 Taylor, Richard 8, 13f., 17 Thompson, Sandra 200 Trask, Robert L. 88 Trimble, Louis P. 3 Tristante, Jerónimo 170ff., 174, 176, 180, 186 Ussishkin, Adam 231f. VanDam, Mark 51f. Vanhove, Martine 235

266 | Index of Authors

Vassali, Michael Antonio 7 Vassalli, Michelantonio 84, 86 Vassalli, Mikiel Anton 60, 62, 74, 136, 138ff., 145, 149ff., 154ff., 234, 248, 251, 253, 255 Vassallo, Ġan Anton 13 Vella Bonavita, Roger 234, 249 Vella, Alexandra 105 Vella, Andrew P. 136 Vella, Francesco 234, 251ff. Vella, Francis 7 Vella, Loranne 165, 169ff., 177, 187f. Vella, Mark 178 Vella, Olvin 257 Verspoor, Marjolijn 198 Versteegh, Kees 4, 6 Vicente, Ángeles 59 Vorholt, Maike 24

Wedgwood, Daniel 110 Wehr, Hans 137, 148, 152, 155ff., 234, 236 Wettinger, Godfrey 139, 141f., 146, 149, 152, 155, 158, 234f. Wetzels, W. Leo 253f., 258 Wilcox, Rand R. 207, 215 Witzlack-Makarevich, Alena 106, 109, 112ff., 128 Woidich, Manfred 61, 236 Wright, William 137, 139ff., 149ff., 156 Wzzino, Francesco 234, 243f., 253, 255f. Yoda, Sumikazu 58, 72 Żahra, Trevor 173 Zammit Ciantar, Joe 14ff. Zammit, Martin R. 57 Zavadovskij, Jurij 139, 141f., 144ff., 149ff., 154, 156, 158 Zipf, George Kingsley 197ff., 217ff., 222

Index of Languages Albanian 113, 203f., 210ff., 215f., 222 Amharic 114f. Arabic 3ff., 12ff., 17ff., 48f., 51ff., 56ff., 69ff., 105, 115f., 135ff., 143ff., 152ff., 164, 192, 234, 236, 239, 258 – Algerian ~ 60, 236, 239, 242, 247 – Andalusian ~ 4, 135 – Classical ~ 4, 6, 8, 56f., 59ff., 64, 71, 74, 76, 136ff., 151ff., 156, 158f. – Egyptian ~ 152, 158 – Ethiopic ~ 72 – Lebanese ~ 115f., 247 – Libyan ~ 62, 73, 152 – Moroccan ~ 60, 23gf., 242, 247 – Najdi ~ 59 – Old ~ 60, 135 – Proto ~ 60f. – Siculo ~ 4, 135f. – Standard ~ 52f., 59, 136ff., 144, 152ff., 159 – Syrian ~ 116, 138ff., 141f., 144, 149, 152, 154, 156, 159, 247 – Tunis/ian ~ 59, 62, 64, 71, 73f., 135, 138ff., 144f., 147ff., 154ff., 236, 247 Aramaic 72 Armenian 203, 210, 211f., 214, 216, 218f. Basque 203, 210, 211f., 214ff., 219 Catalan 110, 234, 242, 253 Chichewa 110 Chinese 198, 201 Croatian 197 Czech 164, 197

Georgian 203, 210, 212, 215f., 218f. German 163f., 197ff., 203, 205f., 210ff., 214f., 217, 234, 253 Greek 7, 113f., 197, 203, 211f., 215, 217 Hungarian 197, 203, 206, 210ff., 214, 216f., 219 Indonesian 197 Irish 203, 211ff., 217 Italian 3f., 8f., 15f., 18, 49ff., 69f., 74ff., 157, 169, 198, 203, 206, 211ff., 217f., 234, 241 Japanese 198 Kesawai 111 Latin 8, 142, 170, 201, 205, 234 Lezgian 203f., 210, 211f., 214ff., 218f. Lithuanian 203, 211ff., 216f. Macedonian 113f., 164 Maltese 3ff., 13ff., 23f., 32, 36, 38f., 42, 44, 47f., 51ff., 64f., 68ff., 83ff., 87ff., 105ff., 114ff., 120f., 124, 126ff., 135ff., 163ff., 167ff., 184, 186ff., 194, 197, 199, 203, 206, 209ff., 217f., 231, 233ff., 239f., 242ff., 246ff., 257ff. Nepali 114 Nkore-Kiga 111ff. Ossetic 203, 210ff., 216, 218f.

English 4, 7, 9, 24ff., 30, 36, 47, 49f., 52, 58, 65f., 68ff., 94, 138, 164ff., 168, 179, 190, 197f., 201, 203f., 234f., 251, 253 Finnish 164, 203f., 206, 210, 212ff. French 49ff., 65f., 75, 109f., 119, 164, 169, 198, 253

Polish 203, 211ff., 216 Romanian 113f. Russian 197f. Ruuli 112ff. Sambaa 111, 113, 119 Selayarese 110

https://doi.org/10.1515/9783110783834-011

268 | Index of Languages

Sicilian 4, 15f., 49f., 70, 72, 74, 76, 116, 237, 240f., 247, 259 Spanish 163ff., 180ff., 189ff. Swedish 197

Teiwa 111 Turkish 199, 203, 206f., 210ff., 214ff., 219 Vera'a 114 Welsh 164

Index of Subjects adjectives 3, 5ff., 26, 39ff., 52, 59, 118, 150, 250 adposition 106, 164, 198f., 202ff., 209ff. agreement 6, 26, 35, 86ff., 205f., 187, 189

indexing 105ff., 121f., 124ff. inflection 8, 48, 53, 63ff., 69, 87, 158, 183, 189, 204 information structure 107f., 122 Islam 4f., 61

Bible 9, 19, 137ff., 155ff.

lexicon 3, 18f., 44, 51, 62, 64f., 69, 74, 136 loanword 15f., 47f., 51, 64ff., 69, 76, 205, 237, 240f., 246

case affixes 204f. clitic doubling 105, 108, 116 comparative 13ff., 239 comparison 3, 5, 19, 106, 116, 120, 192, 198, 214, 256 compounds 7, 17, 24, 27f., 31, 44, 54, 56, 137, 141, 144, 151, 158, 204 corpus 7, 24, 43, 68, 108, 113f., 116f., 120f., 124, 127f., 164f., 178, 181, 183, 199, 203, 205f., 210, 212f., 217, 234 definiteness 35, 37f., 108, 110, 114, 127, 175, 177ff. demonstrative 38f., 49, 67, 72f., 118 derivation 6f., 10f., 14, 19, 233, 239 dialect 4, 50, 53, 56ff., 70, 72ff., 135ff., 141, 143f., 151f., 154, 156, 158f., 257 diminutive 7, 15, 18, 70, 237, 242, 247 elative 13ff., 17 etymology 9f., 13, 19, 47f., 51, 56ff., 60ff., 64, 68, 71ff., 137, 152, 154ff., 205f., 238f., 241, 248, 250ff., 254, 257f. frequency 51, 57, 64ff., 68f., 72f., 76, 167, 169, 178, 184, 197ff., 205ff gender 6, 35, 42f., 86, 90, 118 glottal stop 75, 136, 152 Goal 163, 165ff., 176, 183 grammarians 4, 6f., 52 holiness 3, 5, 7, 9ff., 15, 17ff.

https://doi.org/10.1515/9783110783834-012

morphology 6, 8, 24, 44, 52, 54, 61, 111, 190 morphosyntax 23, 31, 35, 84, 88f., 91ff., 97f., 100ff., 164, 174, 201 multi-word expression 24, 27, 44, 204 number 6, 35, 42, 50, 52, 70, 86f., 90, 105, 118 numerals 36, 41, 118 obstruents 231ff., 235f., 238, 242, 245, 248, 250, 252ff. orthography phonology 8, 65f., 70, 252, 257 place 32, 163, 166ff., 176, 183f. possession 24, 139, 190 preposition 13, 23f., 27ff., 31, 33, 40f., 44, 55f., 88, 107, 109, 114ff., 119, 121, 136ff., 163ff., 170, 174ff., 179, 182ff., 190ff., 197ff., 203f., 206, 210f., 215f., 218f., 239 quantifier 40f. quantitative/qualitative methods 113, 118, 121, 128, 165, 218 reduplication 66, 105 relativizer 163, 167f., 172f., 185ff. rule addition

270 | Index of Subjects

sample 51, 84, 89ff., 117, 120f., 124, 127, 168, 201ff., 206ff., 210ff., 214ff. semantic marker 10, 17 semantics 7, 84, 114, 118, 120f., 124, 128, 198 Source 27, 165ff., 174 spatial interrogatives 165, 167ff., 172f. superlative 13f., 16f., 39 syllable 47, 57, 63, 65, 67ff., 71f., 74ff., 250, 263

verb chain/sequence 83ff., 90f., 93, 95ff., 101 voicing assimilation 231ff. word formation 7, 24 word length 199, 201 word order 107f., 111ff., 118, 122, 125ff. Zipf’s law 198ff., 217ff.