<<

A Tale of Two Morphologies

Verb structure and argument alternations in Maltese

Dissertation zur Erlangung des akademischen Grades eines Doktors der Philosophie

vorgelegt von Spagnol, Michael

an der

Geisteswissenschaftliche Sektion

Sprachwissenschaft

1. Referent: Prof. Dr. Frans Plank 2. Referent: Prof. Dr. Christoph Schwarze 3. Referent: Prof. Dr. Albert Borg

To my late Nannu Kieli, a great story teller

Contents

Acknowledgments ...... iii Notational conventions ...... v Abstract ...... viii

Ch. 1. Introduction ...... 1 1.1. A tale to be told ...... 2 1.2 Three sides to every tale ...... 4

Ch. 2. Setting the stage ...... 9 2.1. No is an island ...... 10 2.2. -and-pattern ...... 13 2.3. Concatenative morphology ...... 41 2.4 Summary ...... 47

Ch. 3. The protagonists: roots and patterns ...... 52 3.1. The ingredients of roots ...... 54 3.2. The functions of patterns ...... 63 3.3. Root derivation and derivation ...... 71

Ch. 4. When patterns put down roots ...... 87 4.1. Database of roots and patterns ...... 88 4.2. The routes patterns take ...... 103 4.3. A(n i)rregular binyan system ...... 125

Ch. 5. Towards a unified account ...... 140 5.1. Getting to the root of the alternation ...... 142 5.2. Maltese: between (anti) and labile ...... 152 5.3. Summary ...... 171 ii

Ch. 6. Bringing down the curtain ...... 175 6.1. Contrasting the two morphologies ...... 175 6.2. Comparing the two morphologies ...... 177 6.3. Tales in search of an author ...... 182

References ...... 184

Appendix I ...... 193 Appendix II ...... 283

iii

Acknowledgments

This study builds on a number of studies on the morphology and lexical semantics of Maltese verbs. It owes a great deal to three works in particular, Edmund Sutcliffe’s (1936) pioneering grammar, Albert Borg’s (1981, 1988) influential study on templatic verbs, and Manwel Mifsud’s (1995) detailed survey of loan verbs. It has greatly bene- fited from Arad’s (2005) illuminating study of the verbs in Hebrew, Frans Plank’s thought-provoking seminars on the direction of derivation, and Martin Haspelmath (1993) and Bernard Comrie’s (2006) typological work on the causative-inchoative al- ternation. If I argued against parts of their work, it was always out of deep admiration. Many people have helped me put this work together. I am particularly grateful to Frans Plank for having taught me to think outside the box and to rethink what is inside the box. I am also indebted to Albert Borg, Manwel Mifsud and Alexandra Vella for hav- ing instilled in me the passion for Maltese linguistics during my undergraduate studies. Another important person in that period was Alina Twist. Assisting her in her doctoral work on Maltese verbs was the spark that ignited a string of events which led to the completion of the present work. I would like to thank Josef Bayer, Albert Borg, Miriam Butt, Eleanor Coghill, Bariş Kabak, Paul Kiparsky, Chris Lucas, Fatemeh Nemati, Frans Plank, Christoph Schwarze, and Adam Ussishkin for having discussed many ideas with me and for their detailed comments on earlier versions of the manuscript. Ingrid Kauf- mann deserves special mention for her comments and suggestions, and for having translated the abstract. I am very grateful to Mark Amaira whose help with the appen- dices has been invaluable, and to Albert Gatt for his interest in my research and for his help at various stages of my work. A special thanks to my friends, Elif Bamyacı, Laura Bonelli, Maialen Iraola Azpiroz, Thomas Mayer, Roxanne Mifsud, Fatemeh Nemati, Florian Schönhuber, Laura Sghendo Loredana Theuma, Joeaby Vassallo, Jonathan Xuereb, for their constant support. I also would like to thank my family. You are all very special to me, in particular my niblings, Nolan, Martina, and little Kieran, who was born only a few hours before this work went to print.

iv

This project was supported by the Deutsche Forschungsgemeinschaft through the Sonderforschungsbereich 471, “Variation und Entwicklung im Lexikon”, by the Compu- tational Analysis of Linguistic Development, CALD Project, and by the Zukunftskolleg at the University of Konstanz.

v

Notational conventions

1. Phonetic symbols

1.1

Orthographic Phonetic

b [b] ċ [tʃ] d [d] f [f] ġ [dʒ] g [g] ħ [ħ] [j] k [k] l [l] m [m] n [n] p [p] [ʔ] r [r] s [s] t [t] w [w] [ʃ] ż [z] z [ts]

vi

1.2 Vowels

Orthographic Phonetic, short Phonetic, long

a [ɐ] [ɐː] e [ɛ] [ɛː] i [ɪ] [iː] ie [ɪː] o [ɔ] [ɔː] u [ʊ] [uː, ʊː]

Notes

1. In Maltese, there is a close correspondence between orthographic and phonological representation. For this reason, throughout this study examples in Maltese are generally given in standard . Note that orthography does not usually represent such phonological processes as final devoicing and voicing . However, it does represent other processes such as vowel syncope and . 2. Word stress is not indicated in the orthography (other than a few cases where the final, vowel-ending is stressed, e.g., verità ‘truth’), but in many cases it is predictable from the syllable structure. Word stress is very often on the penulti- mate syllable, unless there is a final superheavy syllable, which is always stressed. 3. There are two other orthographic symbols, għ and h, which have been described as virtual or ghost phonemes. Depending on the environment, they may have no pho- netic realization at all, e.g., għasfur ‘bird’ /ɐsˈfuːr/ and hawn ‘here’ /ɐwn/, or may mark an increase in duration of an adjacent vowel, e.g., għamel ‘do, make’ /ˈɐːmɛl/ and hemeż ‘attach’ /ˈɛːmɛz/. In definable morphophonemic contexts, they are real- ized as [ħ], usually in word final position, e.g., dmugħ ‘tears’ /dmʊːħ/ and ikrah ‘ugly’ /ˈɪkrɐħ/ or as [w], when occurring between two vowels, e.g., żagħżugħa ‘young girl’ /zɐˈzʊwɐ/and kruha ‘ugliness’ /ˈkrʊwɐ/.1 4. Vowel quantity is not indicated, except for ie. Where necessary, I represent long vowels with repeated sequences, such as aa for [ɐː].

1 When għ occurs word finally and has no phonetic realization, it is orthographically rendered as an apos- trophe, e.g., sema’ ‘hear’.

vii

2. Morphemic glossing

1 first person M masculine

2 second person PFV perfective

3 third person PL

DEF definite PROG progressive

F feminine PST past

FUT future PTCP

INCH inchoative REFL reflexive

IPFV imperfective SG singular

Notes

1. As far as is reasonably practical, the conventions followed are those given in The Leipzig Glossing Rules (Bickel et al. 2008). 2. It is standard to represent verbs in Maltese by the third person singular masculine perfective since it is morphologically unmarked. 3. Verbal, nominal and adjectival patterns are indicated by a string of C’s for conso-

nants and v’s for vowels, such as C1vC2C2vvC3. Subscript numbers represent the root consonants, doubled v’s mark long vowels, while doubled consonants indicate .

4. Superscript i at the beginning of a word, e.g., issoda ‘strengthen’ and inkiteb ‘be written’, marks a prosthetic vowel, that is always inserted before initial gemination and certain consonants clusters, unless the preceding word in the same

phonological phrase ends in a vowel.

3. Other abbreviations

R. regular (root) Red. reduplicative (root) v.i intransitive v.t W-F weak-final (root) W-I weak-initial (root) W-M weak-medial (root)

viii

Abstract

Diese Dissertation behandelt die morphologische Struktur und die Semantik von Ver- ben im Maltesischen. Das Maltesische gehört zu den semitischen Sprachen, ist aber durch den fast 1000-jährigen Kontakt zu den romanischen Sprachen sprachtypologisch zwischen den semitischen und den romanischen Sprachen anzusiedeln. Der Kontakt mit typologisch so unterschiedlichen Sprachen wie Arabisch, Italienisch und in jüngerer Zeit Englisch hat dazu geführt, dass das Maltesische über zwei unterschiedliche mor- phologische Strategien verfügt: die für semitische Sprachen charakteristische “root- and-pattern”-Morphologie und die konkatenative Morphologie. Im maltesischen Verb- System finden sich daher einerseits Verben, die durch die Kombination von konsonan- tischen Wurzeln mit festgelegten Konsonant-Vokal-Schemata (Binyanim, engl. templa- tes) gebildet werden, und andererseits Verben, die durch Suffigierung gebildet werden. Diese Arbeit gibt einen umfassenden Überblick über die beiden morpholo- gischen Prozesse und ihrer Koexistenz im Maltesischen. Die Eigenschaften der beiden Verbklassen werden systematisch beschrieben und die wichtigsten Unterschiede in Be- zug auf die morphologische Struktur der Stämme, die Derivations- und Flexionsmor- pheme, ihre Produktivität und ihre Etymologie herausgearbeitet. Auf dieser Basis wird dann eine einheitliche morphologische und semantische Analyse der Verben und Ar- gumentstrukturalternationen im Maltesischen entwickelt. Der Schwerpunkt liegt dabei auf der Kausativ-Inchoativ-Alternation. Anhand einer umfassenden Datenbasis werden insbesondere die komplexen Beziehungen zwischen den konsonantischen Wurzeln und den Verbal-Schemata detailliert analysiert. Dabei werden die Beobachtungen und Be- deutungsanalysen früherer Grammatiken des Maltesischen präzisiert und die morpho- syntaktische Struktur der schema-basierten Verben im Rahmen der Distributed Morphology modelliert. In einer Gegenüberstellung der semantischen und morphologi- schen Eigenschaften von Kausativ-Inchoativ-Verbpaaren aus beiden morphologischen Klassen werden schließlich charakteristische Unterschiede zwischen den beiden mor- phologischen Systemen innerhalb des Maltesischen aufgezeigt.

Chapter 1 Introduction

This work is an investigation on the morphology and lexical semantics of verbs in Mal- tese, a language which typologically stands between Semitic and Romance , as it is genealogically related to , on the one hand, and has been in close contact with Italian and Sicilian for about one millennium, on the other. A long history of contact be- tween typologically diverse languages gave rise to two different word formation strate- gies in modern Maltese, root-and-pattern morphology and concatenative morphology. Word formation in Maltese can be compared to a checkerboard on which language us- ers may play chess and draughts, two different games with a different set of rules. For any given lexical item, a speaker of Maltese must know whether it is a piece for chess or a piece for draughts, and play the morphosyntactic game accordingly. In the present analysis of the verbal system of Maltese, this translates as a distinc- tion between templatic verbs, formed by the association of consonantal roots with a set of verbal patterns, and concatenative verbs, built by the combination of syllabic roots with verbal suffixes. In this study, I seek to provide a description of the two morpho- logical processes and the way they coexist in one language. I begin with a contrastive examination, outlining the most important differences between the two classes of verbs in terms of stem structure, derivational and inflectional morphology, productivity, and etymology. Following that, I advance a unified morphological and lexical semantic ac- of Maltese verbs, by focusing on argument structure alternations, in particular the causative-inchoative alternation. In so doing, I give a detailed analysis of templatic verbs and the complexity of relations that hold between consonantal roots and verbal patterns. Thus, the aim of this work is threefold. First, it describes the main characteristics of templatic and concatenative verbs in Maltese. Second, by examining a comprehensive database of consonantal roots and the verbal patterns they occur in, it refines some of the intuitions of previous grammars and develops an analysis of the morphosyntactic 2 A Tale of Two Morphologies

makeup of templatic verbs based on the model of Distributed Morphology. Finally, it studies the lexical semantics and formal encoding of the causative-inchoative alterna- tion in both templatic and concatenative verbs, providing a comparative treatment of the two verb classes in Maltese. The rest of this chapter, which serves as a curtain raiser for my study of Maltese verbs, is structured as follows. In Sect 1.1, I briefly review previous research on the verbal system of Maltese, motivating the need for the present work. Next, in Sect. 1.2, I introduce the general assumptions underlying the analysis and give an outline of the main argument of the study as a whole.

1.1 A tale to be told

Parts of this tale have already been told. The verbal system of Maltese has been the sub- ject of several studies. Two important works are Albert Borg’s A Study of Aspect in Mal- tese (1981), an elaboration of which was published in Ilsienna (1988), and Manwel Mif- sud’s Loan Verbs in Maltese (1995).1 As the title suggests, Borg’s work is primarily con- cerned with the grammatical aspect of verbs in Maltese. However, a large part of his book is devoted to what defines as a “tentative (and at times speculative)” semantic classification of verbs, in particular of templatic verbs of movement. The main focus of Mifsud’s work, on the other hand, is the integration of Romance and English loan verbs into Maltese. In describing the phenomenon, he provides an in-depth analysis of con- catenative verbs, chiefly from a morphological point of view. Let us give a brief over- view of these two studies. Couched in a localist framework, Borg’s (1981, 1988) work gives a detailed exami- nation of the morphosyntactic and semantic relations that hold among templatic verbs. His main contributions are two. First, unlike other grammars, which focus on the se-

mantic and syntactic contribution of individual verbal patterns or binyanim (SG. binyan, of system a constituting as verbs templatic at looks Borg building’),2‘ בנין Hebrew from relations among morphologically related verbs. Second, by examining the associations

1 For an overview of other grammatical works of Maltese, see Mifsud (1995a), Brincat (2004), and various other references made in the present study. 2 They are also referred to as themes, forms, conjugations, verbal stems, and awzaan ‘weights, measures’ in Arabic. Throughout this work, the two terms verbal patterns and binyanim are used interchangeably.

Introduction 3

between verbs sharing the same root consonants, he identifies certain pockets of regu- larity in the binyan system in terms of and directionality of derivation. Some tales need to be retold, in some more detail. While Borg’s study draws very im- portant conclusions, most of which are reconfirmed in the present study, its main drawback is perhaps the fact that it focuses on a handful of templatic verbs, mostly verbs of motion. In this work, I build on Borg’s contribution by carrying out a quantita- tive and qualitative analysis on a comprehensive database of templatic verbs in Maltese. Another part of this tale has been told by Mifsud (1995a), who describes in great detail the assimilation of loan verbs, mostly from Romance and English, into Maltese. He proposes a four-way classification of loan verbs, based on the morphology and, in some cases, on the etymology of the verbs, which may be reduced to two main types. The first group consists of loans that, in terms of stem structure and derivational potential, are identical to templatic verbs (e.g., pitter ‘paint’, cf. Sicilian pitturi ‘painter’). The second group includes loans which are integrated into the verbal system of Maltese through concatenative morphology, typically involving initial gemination and some verbal suffix (e.g., ipparkja ‘park’, issimplifika ‘simplify’). Some tales beg a sequel. The current tale arises from the need to account for the di- chotomous nature of the verbal system in Maltese. With Borg’s morphosyntactic and semantic examination of templatic verbs and Mifsud’s exhaustive study of the morphol- ogy of concatenative verbs in hand, I set out to compare and contrast these two verb classes in Maltese with the aim of giving a unified account of some aspects of both sys- tems. After describing the formal properties of the two verb formation strategies indi- vidually, I seek to (i) establish the nature of the relation between the two morphologies, and (ii) determine whether the differences between templatic and concatenative verbs are purely morphological or whether there are deeper semantic and syntactic differ- ences between the two verb classes. In order to address these issues, I first examine an exhaustive list of templatic verbs from a quantitative and qualitative point of view, providing a morphosyntactic and lexi- cal semantic analysis of the ways consonantal roots interact with the verbal patterns. In light of the finding that the main function of the binyan system is to mark argument al- ternations, in the last part of this work, I take the causative-inchoative alternation as a starting point for a unified treatment of the formal and lexical semantic characteristics of templatic and concatenative verbs.

4 A Tale of Two Morphologies

1.2 Three sides to every tale

The structure of this work is as follows. Ch. 2 sets the stage for the rest of the study by providing an initial description of the templatic and concatenative verb formation strategies, and highlighting the points of convergence and divergence between the two verb classes. Following that, the tale takes two intersecting paths. The first one leads to an in-depth analysis of templatic verbs. In Ch. 3, I describe the regular and irregular sides of the binyan system in Maltese, and account for these conflicting aspects by in- troducing two assumptions, one concerning the semantic contents of roots, the other dealing with a structural distinction in word formation. With this as background, in Ch. 4, I then move on to spell out and quantify the regular and irregular share in the mor- phological realization and lexical semantics of templatic verbs. The second path leads to an attempt at a unified approach of root-and-pattern and concatenative morphology. This task is the of Ch. 5, where the formal encoding of the causative-inchoative alternation by templatic and concatenative verbs is studied in light of the lexical seman- tics of verbs in terms of internal and external causation.

Preliminaries The plot of this tale unfolds in two main parts, one dealing with the binyan system, and another with the causative-inchoative alternation. They are preceded by Ch. 2, whose purpose is to pave the way for the main discussion by introducing a number of impor- tant issues concerning the double-sided morphology of Maltese verbs. The chapter be- gins with some background on the long, multiethnic history of , which brought about the hybrid linguistic structures of Maltese. After briefly surveying the stratigra- phy of the language, I introduce the two verb formation strategies. It is important to note that throughout this study, I follow recent work in Distributed Morphology (cf. Marantz 2000, 2007; Arad 2005; inter alia) in adopting two interrelated assumptions. One concerns the existence of the root as an underlying core, which is not fully specified in terms of form, meaning and lexical category. The second assumes that all morpho- logical composition takes place in by merging a root with some syntactic head. Accordingly, root-and-pattern morphology is described from a rather traditional perspective. Templatic verbs, such as kiteb ‘write’, are viewed as composed of two in- terdigitated segmental units, a tri- or quadri-consonantal root (√ktb) and a binyan

(C1vC2vC3). The verbal pattern determines the syllabic structure of the stem, including the number of , vowel length, and gemination of root consonants. As the name

Introduction 5

suggests, concatenative verbs like intensifika ‘intensify’ are constructed in the more fa- miliar structure involving morpheme concatenation of a syllabic root (√INTENS) and some verbal suffix (-ifika). The rest of Ch. 2 spells out in detail the main characteristics of the two verb classes. To begin with, I describe the two components of templatic verbs, including the formal properties of the consonantal root, co-occurrence restrictions in terms of place of ar- ticulation, the controversy surrounding root-based analyses, the syllabic configuration of the binyanim and their vocalic sequences, and the debate concerning the inflectional or derivational status of the binyanim. After arguing that binyan morphology is essen- tially derivational, I discuss the inflectional morphology of templatic verbs. Following that, I review some of the reasons why the templatic verb formation strategy was inter- rupted and a new, concatenative system contrived. Finally, I describe the formation of concatenative verbs and their inflectional morphology, showing how they differ from templatic verbs with regard to their structural makeup, morphological productivity, derivational and inflectional morphology, and etymology.

The binyan system in Maltese Drawing on these detailed descriptions, I move on to study the morphological and lexi- cal semantic aspects of templatic verbs. The point of departure for the discussion in this section are the observations (a) that all templatic verbs must be in the form of a binyan, and (b) that one consonantal root typically occurs in more than one verbal pattern, even though the system is full of gaps, i.e. binyan slots that are left empty. Verbs derived from a single consonantal root may:

(i) mark a regular argument structure alternation (e.g., √ktb, kiteb ‘write’ – inkiteb ‘be written’) or have two semantically distant interpretations (e.g., √xrb, xorob ‘drink’ – xarrab ‘wet’); (ii) have the same vowel sequence, as in kiteb and inkiteb, or display vocalic varia- tion, as in xorob and xarrab; (iii) be fully productive (e.g., passives in binyan V, tħassar ‘be cancelled’, always have an active counterpart, typically in binyan II, ħassar ‘cancel’) or exhibit gaps in the system (e.g., inchoatives in IX, smar ‘get tanned’, do not always have a corre- sponding causative).

The binyan system, therefore, has two sides, a regular one (it marks regular argument structure alternations, exhibits no gaps, and retains the same vowel sequences) and an

6 A Tale of Two Morphologies

irregular one (allows for gaps, different vowel sequences, and multiple, unpredictable interpretations). However, traditional grammars of Maltese generally view templatic verbs as a transparent combination of a root, which carries the core lexical meaning of the word, and a binyan, which has a fixed lexical semantic role such as causative, passive, reflex- ive. On the basis of an examination of templatic verbs, I argue that:

(i) roots have minimal semantic content; they are assumed to be semantically un- derspecified, and it is only when they are inserted in a verbal or nominal pat- tern that they take on a specific interpretation; (ii) there is no exclusive mapping between the semantic and syntactic properties of verbs and their morphological form (i.e. binyan): , for instance, may appear in more than one binyan (e.g., binyan I, II and III); and verbs in the same verbal pattern (e.g., binyan V), may express different properties, such as in- choative, passive, and reflexive.

The exact meaning of a verb cannot be unequivocally predicted from the core lexical meaning and grammatical features of its constituent morphemes, i.e. the root and the binyan, independently. I argue that the binyan system in Maltese is neither completely transparent nor totally opaque. In the rest of Ch. 3 and Ch. 4, I try to answer two related questions. First, why is the binyan system not entirely ir/regular? And second, how regular and how irregular is it? In order to address the first question, in Ch. 3, I introduce an assumption related to the formation of , which neatly explains the behavior of templatic verbs. In line with Marantz (1997, 2001, 2007), Arad (2003a, 2003b, 2005) and other works couched in the framework of Distributed Morphology, I assume word-formation to be a syntactic process where roots, category-less, non-decomposable elements, merge with a syntac- tic head, which turns the root into a , a verb or an . However, words can be formed not only by combining a root with a category-bearing head, but also by com- bining an already formed word with a new syntactic head. This structural distinction between words derived from roots and word derived from previously formed words translates into irregular, non-productive processes (root derivation), on the one hand, and regular, productive processes (word derivation), on the other. Turning to the second question, after accounting for the inherent regularity and ir- regularity in the binyan system by means of this distinction in word formation, the next step is to look at a comprehensive database of templatic verbs to quantify the regular

Introduction 7

and irregular portions in the system. This task is taken up in Ch. 4, which, to begin with, examines the distribution of roots among patterns, and the morphological productivity of the binyanim. Following that, it draws a number of generalizations on the relations that hold between verbs containing the same root consonants. This qualitative analysis reveals that roots fall into four main categories:

(i) argument alternations, roots that mark alternations such as active-passive, causative-inchoative, when they occur in more than one verbal pattern (61%); (ii) synonyms, roots that create two or more verbs with identical meaning (18%); (iii) singletons, roots that appear in one binyan only (12%); (iv) multiple interpretations, roots that take on semantically distant interpretations when cast in different verbal patterns (9%).

In the remaining part of the chapter, I describe in detail the morphological realization of each one of these four categories. The regular and irregular aspects observed in each category is then connected to the structural distinction in word formation from roots and word formation from existing words, presenting a full-fledged account of the lexical semantics and morphological marking of templatic verbs in Maltese.

The causative-inchoative alternation in Maltese In the final phase of this study, I provide a comparative analysis of templatic and con- catenative verbs in Maltese by focusing on the causative-inchoative alternation, which is one of the basic argument alternations in Maltese. The chapter begins with a descrip- tion of this verb alternation from a typological and a lexical semantic perspective. It first illustrates the cross-linguistic variation found in the formal encoding of the alter- nation. Following that, it reviews some of the main theoretical views proposed in the literature on the underlying structures of causatives and inchoatives. After arguing that the morphological irregularities of the causative-inchoative al- ternation are best captured by a root-based analysis, I show how the alternation is un- dergoing considerable change in Maltese. While templatic verbs generally mark the al- ternation by means of two different morphologically related verbs (with the being generally morphologically more complex than the causative one), concate- native verbs do not express the alternation overtly. Rather, labile verbs, i.e. ambitransi- tive verbal forms, are used to mark both the causative and the inchoative alternant. In spite of the differences in the formal encoding of the alternation by templatic and concatenative verbs, a unified analysis of the causative-inchoative alternation in Mal-

8 A Tale of Two Morphologies

tese can be provided if we assume that all verbs taking part in the alternation are cases of root derivation. In addition, a sentence creation task and a corpus study on a set of around 40 labile verbs in Maltese reveal that, even though they fail to show any mor- phological mark that would distinguish the causative or inchoative alternant as for- mally more marked than the other, labile verbs nonetheless evince a bias in transitivity. In other words, labile verbs do not constitute a homogeneous class: some of them are more likely to occur in transitive frames, while others tend to pattern with intransitive constructions. This difference among alternating labile verbs is then accounted for in lexical se- mantic terms. Assuming that the morphosyntactic structure of verbs is directly related to their lexical semantic structure, I argue that the variation found in labile verbs in Maltese is sensitive to the kind of event denoted by the verbs in question, whether it is internally or externally caused. Externally caused events, like iċċarġja ‘recharge’, are conceptualized as coming about due to a force external to the entity undergoing the change of state. By contrast, in internally caused events, like sparixxa ‘vanish’, the means for bringing about the change of state is conceptualized as residing in the entity undergoing the change itself. It follows from this that externally caused labile verbs are used more frequently in transitive constructions, whereas internally caused labile verbs occur more often in intransitive frames. Finally, a general conclusion is found in Ch. 6. It is organized in two main parts. First, it brings together the principal arguments and the main findings presented throughout this study. Second, it suggests a number of possible avenues for further re- search in this area, with the aim of achieving an all-inclusive unified account of tem- platic and concatenative verbs in Maltese.

Chapter 2 Setting the stage

2.1 No language is an island ...... 10 2.2 Root-and-pattern morphology ...... 13 2.2.1 The root ...... 18 2.2.2 The pattern...... 26 2.2.3 of templatic verbs...... 37 2.3 Concatenative morphology ...... 41 2.3.1 Templatic loan verbs ...... 41 2.3.2 Concatenative verbs...... 44 2.4 Summary...... 47

The aim of this chapter is to give an overview of Maltese verbal morphology as one di- chotomous system where root-and-pattern morphology, typical of , rubs shoulders with concatenative morphology, which is in general the result of rather intense language contact with Italian varieties of Romance and, more recently, English. In characterizing the templatic verbal morphology of Maltese, I occasionally draw on some insights from other . The description of concatenatively formed verbs is followed by a comparison of the general properties of the two types of verbal morphologies in Maltese, laying the foundation for the rest of the work, which seeks to advance a unified morphological and lexical semantic account of Maltese verbs. The structure of this introductory chapter is as follows. Sect. 2.1 starts off with a short flashback into the history of the language and a brief description of the linguistic stratigraphy of Maltese. Sect. 2.2 and 2.3 provide a detailed analysis of root-and-pattern and concatenative morphology respectively. Finally, in Sect. 2.4, I show how, in terms of stem structure, inflectional morphology and derivational potential, Maltese verbs can be divided synchronically into two classes, templatic and concatenative.

10 A Tale of Two Morphologies

2.1 No language is an island

The hybrid nature of Maltese and its shaping into a language which, as we shall see, is elusive to rigorous typological classifications is best understood in the light of the his- torical and socio-cultural events that affected the Maltese islands. Due to their geo- graphical position in the centre of the Mediterranean Sea, at the periphery between southern and , the history of the islands is characterized by a suc- cession of rulers and colonizers, stretching from at least the eighth century BCE to the latter half of the 21st century when the country became independent and subsequently a republic. The Phoenicians, Romans, Byzantines, Normans, and the Aragonese are among some of the cultural and political powers that made their linguistic mark on the inhabitants. A deeper imprint was, however, left by the who conquered the small archipelago in 870 and most probably imposed their language on the islanders. The - riod of the Arabic dominance indeed marked the beginning of the , more specifically in 1048 when the rulers brought to Malta a new community that spoke a Siculo-Arabic linguistic variety. Though in origin Maltese was a dialect of Arabic, it almost immediately underwent a process of Latinization in 1090 when the Normans captured the islands and annexed them to the Kingdom of . By the second half of the 13th century, there was very little contact with the and the archipelago was completely Christianized. Contacts with Sicily and the south of became stronger and borrowings from Italian varieties must have begun to penetrate the native dialect. The Romance element branched further out into the Semitic structures of the language in the 16th century when the islands passed under the of St. John for over two centu- ries. At that time, Malta grew into a cosmopolitan centre, especially the Grand Harbor area, which teemed with intercultural activity. Even if the great majority of the Knights were French and Spanish, at the official level, Italian was used as the written language of the administration and culture, replacing Sicilian and . Hoberman & Aronoff (2003: 64) rightly point out that “[i]t is striking and somewhat puzzling that a little over two hundred years of Muslim rule was sufficient to replace whatever language or lan- guages had previously been spoken in Malta with Arabic, while the last seven hundred years have not seen that language supplanted by Sicilian or Italian.” The role of Italian as the language of culture in Malta was threatened by the arrival of the British Empire in the 19th century. A process of Anglicization was set in motion, which led up to the Language Question, a linguistic and cultural conflict between the

Setting the stage 11

attempts of the British to oust the long established Italo-Maltese culture and the by using Maltese as a lever. In 1934, English, the language of the ruler, and Maltese, which by then had been standardized with a literature in early blossom, be- came official languages, with Maltese as the at the expense of Italian.1 The intricate history of the Maltese islands is mirrored in the stratigraphy of the language, which may be represented in a simplified manner in terms of three major strata. The main is Semitic, forming the basis of the , morphology, and the basic lexicon. Maltese shares many characteristics with Western Arabic, in par- ticular Maghrebi dialects (Aquilina 1961a, 1979; inter alia). However, “[i]t is likely that successive waves of impact reached the Maltese shores from different Arab stations and at different points in the island’s history” (Mifsud 2008: 146). Maltese, in fact, displays a number of curious Levantine traits (Alex. Borg 1994, 1997; Borg & Mifsud 2002) and some researchers suggest that the main stratum of Maltese is actually Siculo-Arabic (Brincat 2004; cf. also Agius 1996). The Romance superstratum is chiefly represented by lexical, syntactic, and, to a lesser extent, phonological and morphological accretions from Sicilian and southern Italian dialects (roughly from 1090 to 1530) and from Italian (from 1530 on). The influ- ence of Italian comes down to the present day mainly through television programs, geographical and cultural proximity with Italy, as well as tourism and commerce.2 For this reason, Italian can still be regarded as an adstratum in the linguistic stratigraphy of Maltese. Other than modern Italian, English constitutes a strong adstratum. It mainly consists of lexical material from the period of the British rule (1800 up to World War II) and, most importantly, to the postwar period and later, when English grew into a widely used language on an international level.3 Inter-language contact between morphosyntactic systems of languages as typologi- cally diverse as Arabic (Semitic), Italian (Romance), and English (Germanic) gives rise to a heterogeneous language structure in Maltese. As a result of extensive contact with European, mostly Romance, languages, Maltese typologically occupies an intermediate position between Semitic and Romance, though in several respects it is closer to the former. From a comparative quantitative analysis it comes to light that “while Romance

1 For detailed accounts of the history of Maltese, see Brincat (2000, 2004, 2008), Borg (forthcoming), Mif- sud (1995: Ch.2), Mifsud & Borg (1994), among others. 2 Brincat (1998, 1999) and Caruana (2006) study the acquisition of Italian through television among Mal- tese speakers. 3 Since the development of Maltese is characterized by the presence of other languages, usually of greater prestige, it can be also described in terms of , as a low variety interacting with various high varie- ties such as Latin, Sicilian, Italian, French and English (cf. Brincat 2004: 25-27, 249ff.).

12 A Tale of Two Morphologies

influence has moved Maltese typologically in the direction of Romance, Maltese still remains typologically closer to Arabic than to Romance” (Comrie 2009: 9-10). In other words, due to foreign pressure exerted on its Semitic structures, Maltese has undergone grammatical reorganization according to the patterns it interacted with only to a cer- tain extent, which did not drastically change its Semitic character (cf. Mifsud 1995b). A statistical analysis of the general lexicon reveals that the number of lexical items of Italian and Sicilian origin is rather high. According to Brincat (1996, 2000, 2004), the etymological sources of Aquilina’s (1987-1990) dictionary entries are distributed as follows: Semitic 32.41%, Siculo-Italian 52.46%, and English 6.12%. Local formations and items with obscure etymology make up the remaining percentage. Type vs. token frequency in journalistic texts reveal that items of Semitic origin (72.92%) are more frequent in use because prepositions, , and other grammatical words are mostly derived from Arabic (Fenech 1978: 140; Brincat 2004: 363-366). It comes therefore as no surprise that Maltese is often described as a “mixed lan- guage” (Aquilina 1958, 1959; Kontzi 1981; Farrugia 1998), a term in need of a clearer definition. The status of Maltese with respect to the typology of language contacts, how- ever, is still an unsettled matter. Stolz (2003) argues that Maltese is not an ideal case of a “”, since no sharp distinction can be made between the lexical compo- nent derived from one language and grammar belonging to another. Nor is Maltese an ordinary case of “massive borrowing”, because the non-Semitic share in the lexicon goes significantly beyond the 45 per cent borderline suggested in the language contact literature (cf. Bakker & Mous’s 1994 language intertwining approach). In a continuum of language typology, Maltese would therefore occupy an in-between position, pre- sumably unfilled by any language, between clear cases of language mixing and massive borrowing (Stolz 2003: 306-308).4 From this brief sketch of the main linguistic phases in the , it is ap- parent that a variety of threads are woven into the linguistic fabric of Maltese. The con- stant interaction between Semitic and Romance structures results in a double-sided morphology. For practically any lexical category, language learners of Maltese are faced with two word formation strategies. By way of illustration, verbs, agent , and ver- bal nouns may be formed both by merging a consonantal root with verbal or nominal patterns (cf. Table 2.1a) and by concatenating verbal or nominal to a root or stem (cf. Table 2.1b).

4 Other studies on contact linguistics in Maltese include Drewes (1994) and Mori (2009).

Setting the stage 13

Table 2.1a Root-and-pattern word formation strategy

Root Word pattern Verb/Noun Meaning

√ktb C1vC2vC3 kiteb to write

C1vC2C2vvC3 kittieb writer

C1vC2C3a kitba writing

Table 2.1b Concatenative word formation strategy

Root Verb/Noun Meaning

i 5 √TAJP -ja ttajpja to type -ista tajpista typist -ar ittajpjar typing

In what follows, I lay stress on the verbal morphology of Maltese, which is like a fork in the road where root-and-pattern and concatenative morphology come together. The systems are first dealt with separately (Sect. 2.2, 2.3), highlighting instances of fu- sion between the two morphologies (Sect. 2.4). A unified, contrastive analysis that seeks to determine the nature of the relation between the two morphologies is then presented in Ch. 5.

2.2 Root-and-pattern morphology

Roots and patterns are basic concepts every student of Maltese is confronted with, es- pecially when studying the verbal system of the language. The phenomenon is not char- acteristic of Maltese in particular, but of Semitic languages generally. In Maltese root- and-pattern morphology, verbs, as well as a large number of nouns and , are composed of two basic derivational morphemes, a consonantal root and a word pattern, interwoven within each other in a non-concatenative or discontinuous manner. The root is an ordered set of typically three or four consonants. The consonants of the root,

5 As discussed in Sect. 2.3, initial gemination operates as a general for concatenative verbs and nouns derived from them (cf. the discussion on word formation from roots and words in Ch. 3).

14 A Tale of Two Morphologies

also known as radicals, form a skeleton on which the vowels are intercalated and affixes are added. The pattern, also referred to as template or scheme, represents the structure of the word itself, which is then filled in by the root consonants and the vocalization. Table 2.2 illustrates the structure of some Maltese words formed by inserting the root √sbħ in different word patterns.

Table 2.2 An illustration of some of the derivations of the root √sbħ

Orthographic form Meaning Word pattern

sebbaħ to beautify, embellish C1vC2C2vC3

sbieħ to become beautiful C1C2vvC3

sbejjaħ pretty C1C2vjjvC3

tisbiħ embellishment tvC1C2iiC3

sbuħija beauty C1C2uC3ija

This non-concatenative or non-linear process of word formation is seen by most researchers as the distinctive feature of Semitic morphology, whereby the morphemic structure of words is characterized by the interdigitation of roots and patterns, rather than by morphemes being placed linearly to the left or right of the word stem, as pre- fixes and suffixes. A Semitic stem may, however, be flanked by linearly attached affixes to produce inflected forms as well as to form derivatives, such as the adjective-forming suffix -i, e.g., suf ‘hair’ – sufi ‘hairy’. Thus, Semitic words can be viewed as having a mixed morphology: non-concatenative with regard to the stem of the word, and concatenative with regard to the affixes attached to it (cf. Sect. 2.2.3). According to most grammars of Maltese, the root usually carries the core lexical meaning of the word. It is claimed that a root such as √ktb means ‘to write’ or at any rate identifies the basic lexical meaning of ‘writing’ (cf. Table 2.1). The word pattern, then, creates variation on its meaning, determines the word class and is responsible for other grammatical characteristics. Such claims rest uncomfortably with the evidence in at least two respects. First, the lexical core attributed to the root is generally the pri- mary meaning it acquires in one specific verbal environment, namely the first verbal

pattern, C1vC2vC3 (kiteb), glossing over secondary meanings of the same pattern and meanings it may acquire in other verbal and/or nominal environments. Second, al- though word patterns shape the actual meaning of words, the semantic characteristics

Setting the stage 15

of patterns, especially verbal patterns, are often inconsistent, so much so that the exact meaning of a verb – and in many cases of nouns too – cannot be unequivocally pre- dicted from the core lexical meaning and grammatical features of its constituent mor- phemes, i.e. the root and the pattern, independently. This point is discussed in detail in Ch. 3. One and the same root may appear in different patterns, and a single pattern ac- commodates different roots. The formation of Maltese templatic verbs and nouns can be viewed as an operation that consists of two lists: a list of no more than 150 patterns and a much longer list of around 2,000 consonantal roots, which is in principle an open list (cf. Ch. 4; Appendix I & II).6 The Periodic Table in chemistry has been used as a metaphor to describe the system of interrelations between roots and patterns in Se- mitic languages (cf. Ornan 1990). This is how Shimron (2003b: 6) puts it:

The two lists can be imagined as two axes of a huge matrix, where words are placed in cells that correspond to a certain template on the templatic axis and a root on the root axis. The intersec- tion between the two axes may be seen as a cell containing a word or a potential word. The fact that not every cell is actually occupied leaves room for the construction of new words without adding roots or templates to the matrix as a whole.

This latter point on the (large) number of gaps in the system, i.e. possible pattern slots that are not filled by actual verbs or nouns, is taken up in Ch. 4. What has been delineated above may be called the traditional view of Semitic mor- phology, backed by a long history of linguistic research (cf. Goldenberg 1994; Owens 1997), where the root-pattern construct appears as the central morphological structure of this . There are nonetheless reservations about the centrality of roots and patterns. As discussed below, a number of researchers have recently argued against the traditional root-based approach to Semitic languages, in favor of a word- or stem- based approach (cf. Bat-El 2001, 2003a, 2003b; Ussishkin 2000, 2005; inter alia). In the remaining part of this section, I take a closer look at roots (Sect. 2.2.1) and patterns (Sect. 2.2.2) from a somewhat traditional point of view, bearing the root-based vs. word-based controversy in mind. Finally, I give an overview of the inflection of Mal- tese templatic verbs in Sect. 2.2.3. Before zooming in on the major features of roots and patterns, it is important to note that I am here concerned only with verb-creating roots and verbal patterns in Mal-

6 Comprehensive lists of verbal, nominal, and adjectival patterns in Maltese are given by Aquilina (1959: 142-286) and Borg & Azzopardi-Alexander (1997: 280ff.) who count many more than 150 patterns be- cause they take different vocalisms as constituting different patterns.

16 A Tale of Two Morphologies

tese. There are a number of differences between verbs and nouns. Not only are nominal patterns much more numerous in Maltese (as in other Semitic languages) than verbal ones, but psycholinguistic evidence from other Semitic languages shows that in non- concatenative morphology nouns and verbs are organized and accessed differently in the mental lexicon (Deutsch, Frost & Forster 1998; Deutsch & Frost 2003). Further noun-verb asymmetries in Semitic languages are pointed out in Aronoff (1994: Ch. 5), Buccellati (1997), Owens (1997), Arad (2005), and others. These studies suggest that roots and patterns are sensitive to grammatical categories and that their roles may not be the same for nouns and verbs because of structural differences between the two classes. A fuller investigation of nouns and verbs is, however, required to determine if such differences hold also true for Maltese.

Table 2.3 Examples of derivational nominal patterns in Maltese

Nominal pattern Root Noun Meaning

(1) C1C2vvC3 √ftħ ftuħ opening √qdf qdif rowing √ħls ħlas payment

(2) C1C2uuC3a √qrs qrusa sourness √blh bluha foolishness √mlħ imluħa saltiness

(3) C1C2uC3ija √qlb qlubija bravery √bżl bżulija industriousness √tjb tjubija goodness

(4) mvC1C2vvC3 √ftħ muftieħ key √qdf moqdief oar √smr musmar nail

(5) tC1vC2C2iiC3 √ftħ tfettiħ widening √ksr tkissir breaking √fdl tfaddil saving

(6) tvC1C2iiC3a √fsr tifsira meaning √qsm taqsima section √fkr tifkira memory

Setting the stage 17

One key difference between nouns and verbs of Semitic origin in Maltese concerns the obligatoriness of pattern morphology. While roots and patterns are central to tem- platic verbs, as they may not be formed outside the pattern system, nouns fall into two different groups: those that are formed by the combination of roots and nominal pat- terns (known as mishqalim ‘meters’ in Hebrew) and those that are not. Table 2.3 lists some nominal patterns in Maltese and nouns that appear in them. In the above list, which is by no means exhaustive, nominal patterns are distin- guished on the basis of , suffixes, and inherent vowels (i.e. underlying specified vowels). The patterns are ordered as follows: (1) no affixes, no inherent vowels; (2) no affixes, inherent vowels; (3) suffix and inherent vowel; (4) , no inherent vowel;7 (5) prefix and inherent vowel; (6) prefix and suffix.8 Note that the same root may ap- pear in several nominal patterns, creating numerous nouns, e.g., √ftħ and √qdf. Words belonging to one nominal pattern may share some meaning component, e.g., items with the nominal pattern in (4) generally denote nouns of instrument. Other nouns, by contrast, do not readily fit into any of the nominal patterns. Nouns like alla ‘god’, bniedem ‘human’, żnuber ‘Aleppo pine’, omm ‘mother’, lack the prefixes, suffixes, vowels and syllabic structure typical of nouns like those in Table 2.3, and may not be further decomposed into a root and a pattern. If, however, we take all Maltese nouns and verbs into account, not only those of Se- mitic origin, then, the difference between the two grammatical categories is not so large. As a result of heavy borrowing from Romance and English, there are (i) many nouns, such as vjolin ‘violin’, pubblikazzjoni ‘publication’, whose structure does not in- volve a fixed nominal pattern, and (ii) numerous verbs that are formed concatenatively, rather than by root and pattern association (cf. Sect. 2.3). In relation to this, Twist (2006: Ch. 5) concludes, on the basis of a visual masked priming experiment, that tem- platic verbs in Maltese are comparable to nouns in Hebrew: while consonantal roots provide organizational structure, verbal patterns do not seem to bear psychological sa- lience. I am inclined to think this is due to the presence of many non-templatic verbs in Maltese, just as in Hebrew there are many non-mishqalim nouns (cf. Arad 2005: Ch. 2).

7 One might argue, on historical grounds, that the second vowel of the nominal pattern (4) is inherent, as it can either be ie or aa, which historically converge in one vocalic segment. In this study, vowels are taken to be inherent only when the same vowel is retained in the various manifestations of a given pattern. 8 I consider the final vowel in (2) different from the final vowel in (6). While the former is an inherent vowel, which, by definition, is always present (hence the non-occurrence of forms like *qrus and *bluh – not to be confused with nominal pattern (1), where the vowel u is not inherent), the latter is a suffix, which creates (verbal) nouns of unity, i.e. the singulative of collectives such as tifsir ‘meaning’ and taqsim ‘dividing up, separation’.

18 A Tale of Two Morphologies

2.2.1 The root

In the linguistic literature, the definition of a root is a complicated matter (cf. Aronoff 1994 for a discussion). I here take the view of the root as an unanalyzable morphologi- cal unit obtained by stripping away all morphological structure from a word form (Aro- noff 1994). The goal of this subsection is to shed light on the properties of roots, ad- dressing two main questions. The first one concerns the types of verb-creating roots in Maltese and their phonological structure, specifically in terms of root co-occurrence restrictions (and preferences). The second question deals with the centrality of the root, taking into account studies that downplay or eliminate the role of the root, on the one hand, and evidence, mostly from experimental psycholinguistic data, that roots have independent status in the mental lexicon, on the other.

Types of roots in Maltese Verb-creating roots in Maltese typically consist of three or four consonants in fixed se- quence, e.g., √ktb, √sbħ, √ħrbt, √tktk.9,10 Tri-consonantal roots may be divided into two major groups, strong and weak. Roots are taken to be strong unless they contain one or more weak radicals, namely the phonemes /j/ or /w/, glides that are silent in some

words and audible in others (e.g., √wżn > t-i_żen ‘IPFV.3SG.F-weigh’, but wiżn-et ‘weigh-

PFV.3SG.F’). Strong roots are further subdivided into regular, having the structure C1C2C3,

and reduplicative, with the structure C1C2C2. Weak radicals may occupy initial, medial or final position.11 There are a small number of doubly-weak roots, such as √xwj (xewa ‘grill’) and √ħjj (ħejja ‘prepare’), which behave in every respect like weak-final roots. As we shall see in the rest of this chapter, these root types and subtypes determine the syllabic structure of verbs and have an effect on their inflectional morphology. To get a complete picture of the inflectional system of Maltese, it is useful to have a third group for liquid/nasal-medial roots, i.e. roots that have l, m, n or r for second radical,

9 This characterization of verb-creating roots in Maltese implies that they are taken to be consonantal. Ar- guments for the presence of vowels in Semitic roots are found in Lederman (1982), Bat-El (1994), Ussish- kin (1999), Dell & Elmedlaoui (2002), inter alia. 10 Although the sequence of radicals is fixed, metathesis does sometimes take place within the root: (i) either in entire paradigms, e.g., √fġr, faġar > faraġ, as in ilqattu u fraġtu (for fġartu) ‘I hit him and made his nose bleed’; √sfq, seffaq > saqqaf, as in saqqaft (for seffaqt) wiċċi ‘face up’, even though both √frġ and √sqf are already found in the language, appearing in patterns such as faraġ ‘comfort’, farraġ ‘to console’, saqaf ‘ceiling, roof’, saqqaf ‘to roof’; (ii) or in a few cells in the paradigm, e.g., √bkj > bk-ejt ‘cry-1SG.PFV’ > kbejt; √dħl > n-idħl-u ‘IPFV.1-enter-PL’ > niħdlu. 11 Other names given to these root types include doubled and geminates for Red., assimilative and primae infirmae for W-I, hollow and media infirmae for W-M, defective and tertiae infirmae for W-F.

Setting the stage 19

and a fourth group for silent-final roots, i.e. roots whose third radical is the silent or, as Hoberman (2007) calls it, the “ghost” consonant, orthographically represented as a di- graph għ or an apostrophe (cf. Sect. 2.2.3; Notational conventions).12 Finally, there is a small group of irregular roots, with at least one of the C-slots un- filled (cf. af ‘know’, ħa ‘take, idda ‘shine’, ra ‘see’, ta ‘give’), through historical loss of some consonant (cf. Sutcliffe 1936: 132-140).

Table 2.4 The main types of verb-creating roots in Maltese

Root type Example(s) Verb(s) Meaning(s)

Tri-consonantal

Strong regular √ktb kiteb write reduplicative √xmm xamm smell

Weak initial √wħl, √jsr weħel, jassar stick, enslave medial √dwr, √tjr dar, tar turn, fly final √dgħw, √qrj dagħa, qara swear, read

Quadri-consonantal

Strong √ħrbt, √tktk ħarbat, tektek confuse, tap Weak final √pnġj pinġa draw

Quadri-consonantal roots may also be divided into strong and weak(-final). A quadri-consonantal root is weak only if it has a /j/ in final position. In all other cases, i.e.

• when all radicals are strong; • when /j/ or /w/ occupy first, second and/or third position, e.g., √wrwr, werwer ‘scare’, √ċjpr, ċajpar ‘blur’;

• when one or two radicals are reduplicated, e.g., C1C2C1C2 √tktk, tektek ‘tap’, C1C2C1C3

√brbq, berbaq ‘squander’, C1C2C3C2 √ċqlq, ċaqlaq ‘move’, C1C2C3C3 √qnċċ > qanċeċ ‘scrimp on’;

12 Roots with a silent radical, għ or h, in initial (e.g., √għml, √hmż), medial (e.g., √lgħb, √fhm) and final posi- tion (e.g., √smgħ), are taken to be tri-consonantal nonetheless. The same holds for quadri-consonantal roots with silent radicals (e.g., √kgħbr, √hwdn). For discussion on the status of għ and h and the implica- tions and assumptions on their possible underlying forms, see Brame (1972), Fenech (1980: 68-73), Hume et al. (2009), inter alia.

20 A Tale of Two Morphologies

roots are considered strong.13 Verbs derived from these roots behave inflectionally in the same way (cf. Sect. 2.2.2). The main root types are summarized in Table 2.4 with relevant examples.

Co-occurrence restrictions on roots One issue that has received much attention in the literature on Semitic roots concerns the phonological structure of the root. There is a discrepancy, in Maltese as in other Semitic languages, between the number of verbal roots that are theoretically possible and their actual numbers as attested in the lexicon. The question arises naturally whether the unattested forms are mere accidental gaps or whether there are systematic constraints for the proper shape of a root. Greenberg (1950) was among the first to observe that not any sequence of conso- nants correspond to a potential root. Examining 3775 tri-consonantal roots in Arabic, he arrived at the conclusion that radicals are subject to co-occurrence restrictions of

two main kinds: homogeneous identity (identical consonants in one root, such as C1C2C2

or C1C1C3) and homorganic identity (phonemes with the same ) re- strictions. A complete identity of all root consonants such as √ttt never occurs. How- ever, identity of two consonants does occur. For tri-consonantal roots, there can be three types of such identities:

(a) C1 and C2 - for the first two radicals, both complete identity and homorganic similarity (i.e. root such as √ttk and √tdk) are prohibited;

(b) C2 and C3 - the last two radicals can be identical but not homorganic (i.e. roots such as √tkk are possible, but roots such as √tkg are not);

(c) C1 and C3 - the first and last radical are rarely identical or homorganic (i.e. roots such as √tkt and √tkd are disfavored).

According to McCarthy (1981, 1986), the restriction in (a), follows from the Obligatory Contour Principle (OCP), a universal phonological constraint which disfavors adjacent identical phonological elements, including features, segments and tones. Co-occurrence restrictions (a) and (b) point to an asymmetry. While root initial gemination is extremely rare, root final gemination, i.e. cases where the second and

13 Serracino-Inglott (1966, 1975-1989) argues that quadrilaterals are in fact “inflated triliterals” (1966: 50). A segment such as x or b – which he considers to be reduced forms of xi ‘some’ and bi ‘with’ – is added to the left or right of a tri-consonantal root such as √qlb, qaleb ‘turn’ and √grf, giref ‘scratch’, forming a quadri-consonantal root, √xqlb, xaqleb ‘incline’, √grfx, gerfex ‘disorder’. Such an analysis inevitably runs into many theoretical as well as practical problems (cf. Ellul 2010 for a discussion).

Setting the stage 21

third radical are identical, but not homorganic, is relatively frequent. Now, the OCP ex- cludes identical adjacent consonants at the left edge of the root, but does not do so at the right edge of the root, cf. co-occurrence restriction in (b). To account for this asym- metry in the location of geminates in Semitic roots, McCarthy assumes that C1C2C2 roots, in which the last two consonants are identical, are underlyingly bi-consonantal C1C2 roots. The geminates, then, appear on the surface during the process of word formation. A Maltese verb like xamm ‘smell’ is thus formed by aligning the bi-consonantal root √xm with the three slots for root consonants in the word pattern CvCC, leaving the third consonant slot empty, which is then filled by the spreading of the second radical to the rightmost root-consonant position in the word pattern (Figure 2.1; spreading is indi- cated by a broken line).

C V C C | | x m √xmm Figure 2.1 Left-to-right spreading

Since alignment, according to a general convention in autosegmental phonology (Gold- smith 1990), proceeds from left to right, it is always the third C-slot that remains empty. Therefore, √xxm and √xmx are ill-formed because the former violates rightward spreading (Figure 2.2), and the latter requires the crossing of alignment, which is not possible (Figure 2.3).

C V C C C V C C | | | | x m x m √xxm √xmx Figure 2.2 Violation of left-to-right Figure 2.3 Violation of prohibition spreading on crossing alignments

Having said that, it is interesting to note that the root in Figure 2.3 happens to be an ac- tual root in Maltese, √xmx, xemx ‘sun’, ixxemmex ‘sunbathe’, etc. The OCP, however, fails to explain the restriction on first-third radical identity in (c) above. The occurrence, al-

22 A Tale of Two Morphologies

beit infrequent, of such roots suggests that the non-adjacency of segments also plays a role in the constraint. More recent work in Optimality Theory (e.g., Gafos 1998, 2001; Rose 2000) has ar- gued that reduplicative roots may be explained not as autosegmental spreading, but rather as cases of . In Rose (2000), reduplicative roots are still subject to the OCP but, in line with Optimality Theory, they are allowed to violate OCP in order to satisfy higher ranked constraints. According to Gafos (2001), the underlying represen- tation of verb stems created from roots like √xmm is not a bi-consonantal (√xm) or tri- consonantal (√xmm) form, but rather a form containing a final geminate. Reduplicative stems are, therefore, represented as geminates, that is single elements on the melodic tier (see, among others, Frisch 2004; Frisch et al. 2004 for further discussion). An assessment of the co-occurrence restriction on root consonants in Maltese was carried out by examining the distribution of root instances in the lexicon (cf. Spagnol & Mayer 2011). The analysis was based on an exhaustive list of around 2,000 verb- creating roots in Maltese, which was compiled using Serracino-Inglott (1975-1989) and Aquilina’s (1987-1990) dictionaries as well as Mifsud’s (1995a: 272-295) corpus of loan verbs fully assimilated into the root-and-pattern system (cf. Appendix I). Note that, with respect to Arabic, Maltese has lost eleven phonemes, seven of which are recorded in old Maltese documents, and added seven new phonemes, /p, v, g, tʃ, ts, dz, ʒ/, from contact with Romance and English (cf. Aquilina 1961b; Mifsud 2008: 146- 9). Despite such a reshuffle in the consonantal inventory of Maltese, and the constant contact with non-Semitic patterns, root consonants follow the constraints found in other Semitic languages as Arabic and Hebrew (cf. Greenberg 1950; Bachra 2001; inter alia), disfavoring:

(i) complete identity of all radicals; (ii) first-second identity and homorganic similarity;14

and restricting:

(iii) second-third homorganically, but not for complete identity (reduplicative roots add up to 158, that is 11% of all tri-consonantal verb-creating roots); (iv) first-third identity partially on both counts.

14 Tri-consonantal roots in which the first two radicals are identical, such as √kkj kakka ‘defecate’, √nnj ninna ‘sleep’, and √ppj pappa ‘eat’, increased through language contact with Romance. Note that the words derived from these roots all belong to the nursery vocabulary, which is not surprising as cross-linguistically nursery words tend to include the same sound in repeated sequence (cf. Mifsud 1995a: 89).

Setting the stage 23

According to Mifsud (2008: 156), Maltese dictionaries record “373 fully integrated loan verbs which hark back to 247 different roots”. Frisch et al. (2004) examined roots of Romance origin incorporated in Maltese in terms of place of articulation and found out that Italian words that violate OCP are less likely to be incorporated into the root- and-pattern morphology of Maltese. These results were compared to a sample of Italian verbs to ensure that the co-occurrence patterns observed in Maltese loans are not ex- plained entirely by co-occurrence constraints that are (already) found in Italian. Com- parative analyses of this sort are supported by the fact that a tendency for the so-called “repulsion of likes” or “similar place avoidance” is observed in other, non-Semitic lan- guages (cf. Padgett 1991; Pozdniakov & Segerer 2007; inter alia). On examining a large crosslinguistic lexical database with data on more than 4,500 languages, Mayer et al. (2010) claim that this phenomenon constitutes a universal tendency.

How radical is the root? The view of the root described here is by no means uncontroversial. The notion of the consonantal root as a central morphological unit is axiomatic in most Semitic research, but is questioned by a number of studies that consider the word as the basic unit of morphology. It has recently been proposed that the consonantal root is irrelevant for some or all root-and-pattern word formation processes in Arabic (McCarthy 1993; Ben- mamoun 2003), Hebrew (Bat-El 1994, 2001, 2003a, 2003b; Ussishkin 2000, 2005), and other Semitic languages (Rose 2003; Buckley 2003; inter alia). In these word-based approaches, word formation operates in an output-to-output fashion, positing morpho- logical rules that turn words into other words. The base for word formation is not a consonantal root, but an output form that has already been derived. Such word-based analyses of Semitic reject the consonantal root, often seeing it as nothing more than a convenient way to list words in dictionaries. This study argues in favor of a root-based analysis. The meaning similarity across derived forms, the semantic as well as phonological regularity and irregularity found in word formation, and other empirical phenomena discussed in the following chapters would have been difficult to explain without considering the root as a central element in the structure of templatic verbs in Maltese. It is important to note that it is not the scope of this study to undermine the arguments for word-based approaches to Semitic morphology. A composite model which admits both root-based and word-based deriva- tion is most likely correct for Semitic, as argued by Arad (2003a, 2003b, 2005) for He- brew, by Tucker (2009a) for Arabic, and by myself in this work for Maltese. In the rest

24 A Tale of Two Morphologies

of this subsection, I motivate the existence of the consonantal root in root-and-pattern morphology, surveying some of the evidence put forward in various studies. To start with, the Greenbergian restrictions introduced above are taken by several researchers as an argument in favor of the root morpheme. Crucially, the OCP does not apply to any domain larger than the root. The fact that deverbal nominalizing mor- phemes such as m- and t- can appear with /m/- and /t/-initial roots (e.g., √mlj mimli ‘full’, √tjb titjib ‘improvement’), demonstrates that the OCP targets as domain the con- sonantal root not the prosodic word (but cf. the discussion in Bat-El 2003a). Moreover, evidence has been accumulating for the psychological reality of the Se- mitic root, especially in Arabic and Hebrew. Ephratt (1997) tested native speakers of Arabic and Hebrew on root awareness by means of a coloring task. She found that they consistently colored root letters inside the words rather than resorting to other strate- gies such as coloring initial or final letter sequences. In the same vein, Frost, Forster & Deutsch (1997) studied the lexical representation of Hebrew words by testing adults on reading non-vowelled words. They found that previous exposure to the root consonants facilitated lexical access and naming of targets that were derivations of the root. A vis- ual masked priming experiment conducted by Twist (2006) on Maltese also reveals that consonantal roots bear psychological salience. The orthography of Maltese, unlike that of Arabic, Hebrew, and other Semitic languages, is not consonant-based but also in- cludes vowels, and so the Maltese case provides evidence for the consonantal root that is not biased by a consonantal . Other experimental studies of Semitic morphology and speakers’ sensitivity to roots, including Berent & Shimron (2003), Deutsch & Frost (2003), and Goral & Obler (2003), all point to the independent psycho- logical reality of the root. A survey of priming experiments and other so-called external evidence from Arabic and Hebrew that reach the same conclusion is found in Prunet (2006).15 Compelling external evidence for the root in Arabic comes from a case study of roots in aphasic speech. Metathesis in the speech of an Arabic-French bilingual aphasic speaker takes place within the root domain, while the pattern, including affixal or epen- thetic consonants, remains unchanged (e.g., √ntq > minṭaqatun ‘region’ metathesized as *miṭnaqatun). Moreover, the consonants of Arabic roots frequently undergo metathesis while those of French roots do not: the aphasic subject produces 25 times more conso-

15 Note, however, that Berent et al. (2007) present evidence from rating and lexical decision experiments with both visual and auditory stimuli in Hebrew, which support the stem hypothesis, though not necessar- ily falsifying root-based approaches. Thanks to Adam Ussishkin for bringing this study to my attention.

Setting the stage 25

nantal metatheses, affecting only root consonants, in Arabic than in French. Prunet, Bé- land & Idrissi (2000) interpret this empirical result as evidence for the morphemic status of the consonantal root in Arabic.16 In a follow-up study, Idrissi, Prunet & Béland (2008) test weak roots, i.e. roots con- taining a glide that is silent in some words and audible in others, in Arabic aphasic er- rors, and found out that glides that are inaudible in a number of words resurface in some errors. On the basis of this data and other results obtained from hypocoristic for- mation, they argue that errors operate not on fully formed words, but on underlying root morphemes. Such findings suggest a representation of word structure in which the root consonants of Arabic occupy a separate autosegmental tier, as in McCarthy (1981). Semitic roots, it has been shown, have a prominent role not only in the representa- tion and organization of Hebrew and Arabic speakers’ mental lexicons, but also in the acquisition of root-and-pattern morphology. Language acquisition studies demonstrate that - and Hebrew-speaking children from a young age are able to identify and isolate the consonantal skeleton and to produce new words from the same root (cf. Berman 2003; Ravid 2003). More time, however, is required to master the of- ten idiosyncratic links between roots and patterns (cf. Borer 2003). Morphological evidence supports the view that the consonantal root plays a crucial role in the formation of the stative in Coptic, an Afroasiatic language which, like Semitic, has a root-and-pattern morphology. Kramer (2006) argues that the form of the stative of any given verb in Coptic can only be reliably predicted from the root, whether it is bi- or tri- or quadri-consonantal. Statives are insensitive to the phonological properties of the , which would be taken as the base form in an output-output approach. Tucker (2009b) brings to the table further arguments for the root morpheme. He shows that in Iraqi Arabic voicing assimilation proceeds regressively at the prosodic word level, but progressively at the root plus affix level, namely in binyan VIII, i.e. the verbal pattern C1taC2aC3, formed by infixation of -t- after the first root consonant. The same phenomenon takes place in Maltese: the direction of voicing assimilation, which is typically regressive, is progressive when -t- is infixed after the first radical in the forma- tion of binyan VIII (cf. Sect. 2.2.2): compare √sjd stad ‘fish’ with √żjd żdied ‘increase’

16 This pattern of metathesis is recreated in a language game in the Hijazi dialect of Arabic, which is played by freely metathesizing (only) root consonants around the vocalic pattern (McCarthy 1981: 379- 380).

26 A Tale of Two Morphologies

(for *żtied). These facts demonstrate that root consonants are treated differently from other segments in root-and-pattern morphology.17 It appears that, while in the realm of psycholinguistics, researchers agree almost unanimously on the need for consonantal roots in the representation of Semitic lan- guages, formal linguists are divided in their views on consonantal roots.18 Engaging in a proper discussion of this issue would take this work too far afield. For the present pur- poses it suffices to say that, on the basis of the arguments outlined above and the evi- dence presented in Ch. 3 and Ch. 4, I take all derivation to begin with the root. Throughout this section I introduced the root as a notion that is not without its problems, and surveyed the formal properties of verb-creating roots: the types of roots found in Maltese and their co-occurrence restrictions. The issue of the semantic con- tents of roots is taken up in Sect. 3.1. I now turn to a description of the morphemes roots are bound to, the verbal patterns in non-concatenative morphology.

2.2.2 The pattern

The Semitic pattern, roughly speaking, is whatever is left of the (uninflected) word after the consonantal root is extracted. It is the unit with which a root, which cannot be pro- nounced as a phonological unit on its own, may combine to form a verbal or nominal lexeme. Pattern morphology contains slots for the root consonants to fit it in, vowels, and a specific syllabic structure. Some patterns also include other consonants that may or may not be considered affixes. Nominal patterns may include inherent vowels too. The pattern determines the lexical category: it is only when the category-neutral roots are embedded in a pattern, that an actual verb or noun is formed. Patterns therefore serve at least two functions: they make the root consonants pronounceable and provide the category feature, turning the root into a verb or a noun.19 Much of the work done so far on Maltese patterns, especially verbal ones, departs from the view of the pattern having a specific interpretation or function, such as causa- tivity, , and reflexivity. In Ch. 3, I show that, in spite of certain tendencies, trying to

17 See Faust & Hever (2010) for further morphological arguments in favor of the root in and Chaha. 18 Some studies, like several of the contributions to Shimron (2003a), try to reconcile formal theories that deny the existence of the consonantal root with the experimental research such as that reviewed above. 19 In a model called Root and Prosody, Tucker (2009b, 2010) argues that patterns are unnecessary and proposes a morphemic analysis which breaks the Arabic verb into roots, vowels and verbal affixes, but not patterns. Inasmuch as the question whether this model is applicable to templatic verbs in Maltese has not yet been answered, I take patterns as relevant elements in the formal description of word formation in Maltese.

Setting the stage 27

match verbal patterns with particular interpretations such as passive, causative and reflexive leads nowhere, as a given pattern may serve more than a single role, according to the various relations it bears to other patterns in the verbal system. I now take a closer look at the characteristics of verbal patterns in Maltese to make up an inventory of the verbal patterns or binyanim that occur in Maltese. Next, I exam- ine the vowel sequences of each binyan. Finally, I consider the issue of whether bin- yanim are inflectional or derivational.

The binyan system in Maltese The Maltese verbal system has eleven major binyanim, referred to as forom ‘moulds, forms’: nine for tri-consonantal roots, two for quadri-consonantal roots. Because most grammars generally take the descriptions for developed by Western scholars as a model for classifying the templatic verbs of Maltese, there is a discrepancy between the number of categories proposed by grammars and the actual number of binyanim found in the language. Roman numerals from I to X are conventionally as- signed to the verbal patterns consisting of tri-consonantal roots, even though only nine patterns are attested in Maltese. One category, namely binyan IV, is vacant.20 Although, in general, grammars agree that binyan IV is a disused class in modern Maltese, they retain the tenfold classification based on the Western model for Arabic (cf. Aquilina 1979 and Marshall 1968 for a comparison of the verbal patterns in Arabic and Mal- tese).21 Note that binyan IV has disappeared from almost all (cf. Kaye & Rosenhouse 1997; Erwin 2004; Versteegh 2004; inter alia). On the basis of an analysis of an extensive database of verb-creating roots and the verbal patterns they occur in, I argue in Sect. 4.3 that tri-consonantal roots make up a four-way system, as certain couples of binyanim can be said to be in complementary distribution. This leaves us with four major patterns: I, II, V, VII, and their ‘shadow’

20 This class traditionally accounts for causative verbs like wera ‘show’, morphologically related to ra ‘see’ (cf. Sutcliffe 1936: 108). The CV structure and paradigm of wera are of a weak-final binyan I verb. Mifsud (1995a: 312-315) discusses the loss of binyan IV and its redistribution in other binyanim, mostly binyan I, because of similarity in the paradigms of the two verb classes. Cachia (1984: 21-28; 1994: 209-211) argues, not quite convincingly, that binyan IV includes: (i) no fewer than eight verbs, namely ambi ‘need’, arra ‘re- gret’, emmen ‘believe’, idda ‘sparkle’, idden ‘crow’, iżża ‘render (thanks)’, ibber ‘shine’; (ii) a handful of verbs with a st- prefix, such as staħa ‘be shy’, stenna ‘wait’, which were at some point ‘misclassified’ as binyan X verbs; and (iii) another two, agħma and agħna (somehow distinct from standard orthography forms għama ‘blind, grow blind’ and għana ‘enrich’), whose membership in binyan IV is open to question. 21 Earlier grammars, such as Vassalli (1827) and Vella (1831), present a syllabic classification, listing the binyanim as monosyllabic or disyllabic. In his Grammatika Umanizzata tal-Ilsien Malti, Serracino-Inglott (1975-1989, IX: 228-235) proposes an eleventh class of verbs to accommodate verbs that fall outside the binyan system, namely concatenative verbs, which are the topic of Sect. 2.3.

28 A Tale of Two Morphologies

Table 2.5a Traditional classification of tri-consonantal binyanim

Binyan Structure(s) Root Verb Meaning

I C1vC2vC3 √tlf tilef lose

C1vC2C2 √rss rass press

WvC2vC3 √wżn wiżen weigh

C1vvC3 √swq saq drive

C1vC2v √nsj nesa forget

II C1vC2C2vC3 √tlf tellef disturb

C1vC2C2vC2 √rss rassas press repeatedly

C1vC2C2v √nsj nessa make forget

III C1vvC2vC3 √wġb wieġeb answer

C1vvC2vC2 √qrr qarar confess, v.t

C1vvC2v √ndj nieda launch

i V tC1vC2C2vC3 √tlf ttellef be disturbed

tC1vC2C2vC2 √rss trassas be pressed

tC1vC2C2v √nsj tnessa be forgotten

VI tC1vvC2vC3 √wġb twieġeb be answered

tC1vvC2vC2 √qrr tqarar be confessed

tC1vvC2v √ndj tnieda be launched

i VII nC1(t)vC2vC3 √tlf ntilef get lost

i n(t)C1vC2C2 √rss ntrass press oneself

i ntvC2vC3 √wżn ntiżen weigh oneself

i nC1(t)vvC3 √swq nsaq be driven

i n(t)C1vC2vC3 √ħbj nħeba be hidden

VIII C1tvC2vC3 √fkr ftakar remember

i C1tvC2C2 √mdd mtedd lie down

C1tvvC3 √sjd stad fish

i C1tvC2v √nsj ntesa be forgotten

IX C1C2vvC3 √krh krieh become ugly

X stvC1C2vC3 √nbħ stenbaħ wake up, v.i

stC1vC2C2vC3 √krh stkerrah abhor

stC1vC2C2 √qrr stqarr declare

stvC1C2v √ħbj staħba hide oneself

Setting the stage 29

binyanim, IX, III, VI, VIII respectively. There is arguably a fifth marginal pattern, binyan X, with only 24 representatives. In this chapter, which is meant as a general introduc- tion, I give an overview of the traditional classification. The main verbal patterns for Maltese, including their allomorphic variations, are laid out in Table 2.5a and 2.5b. Since no single tri-consonantal root appears in all of the nine binyanim, the examples are taken from different verb-creating roots.

Table 2.5b Traditional classification of quadri-consonantal binyanim

Binyan Structures Root Verb Meaning

QI C1vC2C3vC4 √ħrbt ħarbat confuse

C1vC2C3v √lstj lesta finish, carry out

QII tC1vC2C3vC4 √ħrbt tħarbat get confused

tC1vC2C3v √lstj tlesta be carried out

The verbal patterns listed in the tables above contain consonant and vowel slots, and (optionally) prefixes. They are associated with a fixed prosody, constraining the syllabic structure of verbs: all binyanim are disyllabic, except for binyan IX, which is monosyllabic. Some disyllabic binyanim have monosyllabic variants, in particular when verbs are derived from reduplicative or weak-medial roots (see, for example, the struc- tures for binyan I). The allomorphic variations for each binyan are mainly a result of different root types, strong (regular, reduplicative) and weak (initial, medial, final). However, the di- visions outlined in Table 2.4 for weak roots matter mostly for binyan I, VII and VIII verbs. For instance, binyan II and V have a slightly different structure only for weak- final roots, but other weak root types have the same structure as binyanim with strong roots. Similarly, the structure of binyan IX verbs does not vary, regardless of the kind of root that is embedded in the pattern. Thus, regular √krh, reduplicative √ħff, and weak √twl, all share the structure CCvvC when cast in binyan IX (cf. krieh ‘become ugly’, ħfief ‘become light’, twal ‘become tall’). A number of binyanim may be analyzed as involving a prefix (t- for V, VI, QII; n- for VII; and st- for X) or an (-t- for VIII). Two of these affixes, prefixes t- and n-, have phonologically conditioned allomorphs, as discussed next.

30 A Tale of Two Morphologies

Complete assimilation of the t- prefix takes place when the first radical of the root is coronal, that is /d/, the alveolar or postalevolar /s, z, ʃ/ or the /tʃ, dʒ, ts/. Compare, for example, V t-kisser ‘be broken’ with is-sakkar ‘lock oneself’; VI t-biegħed ‘distant oneself’ with iġ-ġieled ‘fight’; and QII t-farfar ‘dust oneself off’ with id-dardar ‘get dizzy’. The n- prefix has an nt- allomorph conditioned by liquids and nasals in the first C- slot of the root, e.g., int-libes ‘be worn’, int-mesaħ ‘be wiped’, int- ‘bind oneself’. Weak-initial roots, such as √wżn, are also prefixed by nt-, as the weak radical does not surface at any point in the paradigm: int-iżen ‘weigh oneself’. Both n- and nt- prefixes attach to roots whose initial radical is /b/ (in-bela’ ‘be swallowed’, int-bagħat ‘be sent’), silent għ (in-għaqad ‘join’, int-għażel ‘be chosen’), or /ʔ/ (in-qabad ‘get caught’, int-qal ‘be said’). When roots start with /s/ or /ʃ/, these sibilants undergo metathesis with pre- fixal t- as in in-s-teraq ‘be stolen’, in-x-t-eħet ‘throw oneself’, and not *int-seraq, *int- xeħet. In such cases, the verb forms seem to be hybrids of binyan VII (prefix n-) and VIII (infix -t-). And, in fact, some dictionaries classify them as VII+VIII verbs (cf. Aquilina 1987-1990). Binyan VIII involves a -t- infixed between the first C- and V-slot. For phonotactic reasons, roots are not cast in this pattern when their first consonant is a stop /b, p, t, d, k, g, ʔ/,22 an /tʃ, dʒ, ts/, a glide /j, w/, or silent għ and h. By contrast, roots with liquids, nasals, fricatives, and sibilants as first radical do appear in binyan VIII.23 Unlike binyanim I to IX, the variation within binyan X verbs is neither phonologi- cally conditioned, nor is it entirely determined by the root types. Verbs traditionally listed under binyan X do not share a common CV structure. It appears that these verbs are rather grouped together on the basis of the prefix st-, which is common to all forms. There are two main kinds of structures, with the latter being further subdivided into five groups, depending on different types of roots, as summarized in Table 2.6. The verbs listed in this table, including another four that consist of roots with some missing consonants (staħa ‘be shy’, stenna ‘wait’, stieden ‘invite’, †stienes ‘become socia- ble’), are the only verbs that fall under binyan X. Verbs marked by a cross are no longer

22 One exception seems to be btaram ‘be twisted or twined’ recorded in an old dictionary of Maltese, Dizi- onario Maltese-Italiano-Inglese compiled by G.B. Falzon in 1845 (second edition, 1882), which has fallen out of use, with its semantics redistributed to binyan VII verb inbaram. 23 There is only one ż-initial root under binyan VIII, √żjd. When cast in binyan VIII, the infix undergoes par- tial progressive assimilation in terms of voicing, ż-d-ied ‘increase’ for *ż-t-ied (cf. Sect. 2.2.1).

Setting the stage 31

current in the standard language.24 Phonotactic constraints must be part of the explana- tion for this extremely small number of verbs in one class. Such consonant clusters as std, stġ, stb are not permissible in Maltese. Note, however, that str, for instance, is a pos- sible cluster in onset position (e.g. strofa ‘stanza’, stramb ‘strange’), but no /r/-initial root appears in binyan X.25 See Sect. 4.3 for further combinatorial constraints.

Table 2.6 CV structures of binyan X verbs

Structures Subtypes Root Verb Meaning

stC1vC2C2vC3 strong √ħrġ stħarreġ investigate √krh stkerrah abhor weak √ħjl stħajjel imagine √hwj †sthewwa lust for s.o.

stvC1C2vC3 R √għdr stagħdar stagnate √għġb stagħġeb be surprised √nbħ stenbaħ wake up Red. √għll †stgħall enjoy the fruit √ħqq stħaqq deserve √knn stkenn take shelter √krr †stkerr be repulsive √mrr stmerr detest √qrr stqarr state W-I √jqr stejqer come to senses W-M √għjn †stgħan strive √għjt †stgħat wish for food √għjr †stgħar arouse envy W-F √għnj stagħna become rich √ħbj staħba hide oneself √ħjj †staħja refresh oneself

24 Aquilina (1987-1990: 1590) lists the no longer used verb stebagħ ‘be painted’ (√żbgħ) under binyan X. This is clearly a binyan VIII verb, with a -t- infixed after the first radical (cf. żebagħ ‘paint’). In fact, in pg. 1628 the same verb is listed as VIII żteba’. It is not clear whether progressive (as one would suppose from the orthographic rendering stebagħ) or regressive assimilation (as the form żteba’ would indicate) took place. 25 Similar morphonological constraints on binyan morphology have been reported for Modern Hebrew (cf. Laks 2009). Thanks to Adam Ussishkin for bringing this to my attention.

32 A Tale of Two Morphologies

Vowel patterns of the binyanim Most of the discussion so far was centered on root consonants and affixes. In this sub- section, I take a closer look at the vowels in the verbal patterns. Unlike Classical Arabic, Modern Hebrew and other Semitic languages, vowels in the verbal templates of Maltese are not fixed. Each binyan allows several different vowel patterns. For instance, for strong (regular) roots, binyan I & VII have six different vowel sequences, while binyan III & VI have only three. Also, vowels can be lexically distinct in Maltese, as in rema ‘throw away’ vs. rama ‘set up’, għala ‘boil’ vs. għola ‘rise’, and salva ‘save’ vs. solva ‘solve’. This large variety of vowel patterns has developed in part as a consequence of the loss of the emphatic consonants /ṣ, ṭ, ḍ, ẓ/. In Arabic, vowels adjacent to emphatics have back allophones, while vowels adjacent to non-emphatics have front allophones. Since the consonantal opposition is no longer found in Maltese, the associated vowel differ- ences have become phonemic. Also, the strong influence from Romance varieties has introduced novel vowel patterns (e.g., i-a as in pinġa ‘draw’ and o-a as in solva ‘solve’), and possibly loosened restrictions on older ones (see discussion in Cowan 1966; Trim- ble 1971; Alex. Borg 1978: 26ff.; Hoberman & Aronoff 2003; Walter 2006; Hoberman 2007). The vowel sequences for binyan I to IX, including their allomorphic variations, are given in Table 2.7a. Some binyanim, such as I & VII and II & V, are listed together as their V-slots are occupied by the same vocalisms. Single vowels a, e, i, o correspond to the short vowels /ɐ, ɛ, ɪ, ɔ/, while aa, ee, and ie correspond to the long vowels /ɐː, ɛː, ɪː/. For the vowel sequences of binyan X verbs, refer to Table 2.6, which lists exhaustively the verbs that belong to this pattern. Table 2.7b then lists the vowel sequences for quadri-consonantal binyanim.

Table 2.7a Vowel sequences for binyan I to IX

Binyan Root types Vowels Verb(s) Meaning(s)

26 i I, VII R, W-I a-a qasam, nqasam split (v.t), split (v.i) a-e qaleb, inqaleb turn (v.t), turn (v.i) e-e ħeles, inħeles release, be released

26 Since silent għ and h generally result in vowel lengthening, the first vowel of some binyan I (CvCvC) verbs is long, as in għaraf ‘recognize’, għażel ‘choose’, hemeż ‘attach’. For this reason, some grammars (e.g., Cachia 1984: 13) confuse them with binyan III verbs which have the CV structure CvvCvC.

Setting the stage 33

e-a rebaħ, intrebaħ win, be won i-e tilef, intilef lose, get lost o-o xorob, inxtorob drink, shrink Red. a xamm, inxtamm smell, be smelled e kedd, inkedd vex, get vexed W-M aa sab, instab find, be found ie biegħ, inbiegħ sell, be sold

27 i W-F a-a qara, nqara read, be read e-a kera, inkera hire, be hired

II, V R, Red., W-I a-a qassar, tqassar shorten, be shortened a-e qalleb, tqalleb turn, be turned e-e ġedded, iġġedded renew, be renewed e-a dejjaq, iddejjaq upset, get upset i-e kisser, tkisser break, be broken W-F a-a baxxa, tbaxxa lower (v.t), lower (v.i) e-a tebba’, ittebba’ stain, get stained o-a poġġa, tpoġġa place, be placed

III, VI R, Red., W-I aa-a qarar, tqarar confess, be confessed aa-e ħares, tħares protect, be protected ie-e wieġeb, twieġeb answer, be answered W-F aa-a gara, tgara hurl, be hurled ee-a fera, tfera injure, get injured ie-a rieġa, trieġa support, be supported

VIII R a-a ftakar remember e-e xteħet throw oneself e-a stelaħ get bruised i-e intiseġ be woven Red. a stadd clog up e imtedd lie down W-M aa stad fish ie ħtieġ need W-F a-a xtara buy e-a intesa be forgotten

IX R, Red., W-M aa twal get tall, long ie ħxien get fat

27 Other vocalisms for W-F binyan I verbs include i-a, e.g., siqa, xila, which are considered as dialectal vari- ants of seqa ‘water’, and xela ‘accuse’, and o-a found in only two verbs, għola ‘rise’ and għoxa ‘swoon’.

34 A Tale of Two Morphologies

Table 2.7b Vowel sequences for quadri-consonantal binyanim

Binyan Root types Vowels Verb(s) Meaning(s)

28 i QI, QII Strong a-a xandar, xxandar broadcast, be broadcast a-e ħanżer, tħanżer make filthy, pig out on e-e kexkex, tkexkex terrify, get terrified e-a berbaq, tberbaq squander, be squandered Weak a-a kanta, tkanta sing, be sung e-a lesta, tlesta finish, be carried out i-a pinġa, tpinġa draw, be drawn o-a solva, issolva solve, be solved

Note that these are the vowel patterns in the perfective, specifically for the third-person singular masculine, which is the citation form conventionally used in grammars and dictionaries. In the imperfective, then, some vowel patterns remain unchanged (qasam,

jaqsam ‘split.PFV, IPFV’), while others change (rebaħ, jirbaħ ‘win.PFV, IPFV’). The range of vowel sequences is usually higher in the imperfective, with such sequences as i-o

(daħal, jidħol ‘enter.PFV, IPFV’) and i-a (feraħ, jifraħ ‘rejoice.PFV, IPFV’), which are not available in the perfective of binyan I verbs. For some studies on the correspondences between vowel patterns in the perfective and imperfective, see Aquilina (1959: 254- 255), Brame (1972), Puech (1978), Hume (1991, 1996). Finally, some linguists observe that vowel quality may be conditioned by the adja- cent root consonants (cf. Sutcliffe 1936: 71-74; Cremona 1973: 45). Hoberman (2007: 268) claims that the vowel sequences a-e, e-e, and e-a in binyan I “occur only in the neighborhood of the gutturals għ, ħ, h, q”, while o-o sequence “occurs almost exclusively with gutturals.” The other two sequences, a-a and i-e are “freely available for verbs that contain none of the guttural consonants”. He concludes that, with a few exceptions, a regular (i.e. not reduplicative or weak) binyan I verb without guttural consonants has a perfective stem with either of two structures, CaCaC or CiCeC. Further research in this direction is required to establish correspondences between radicals and vowel pat- terns.

28 Quadri-consonantals derived from strong roots can also have the sequences i-e fixkel ‘obstruct’, ie-e miegħrek ‘scrub’ and o-o għosfor ‘disappear’, which are very rare (cf. Ellul 2010: 98-107).

Setting the stage 35

Inflectional or derivational? Before moving on to an examination of the inflection of Maltese templatic verbs, let us discuss the status of the binyanim with regard to inflection and word formation. The binyanim, together with the category-free roots, are understood here as an integral part of templatic verbs. Put differently, a templatic verb of Maltese such as fetaħ ‘open’ and inħasel ‘wash oneself’ is the end result of the combination of one particular root (√ftħ,

√ħsl) and one particular binyan (C1vC2vC3, nC1vC2vC3). A number of important points need to be made. First, paraphrasing Aronoff (1994: 125), no Maltese templatic verb exists outside the binyan system. That is, all templatic verbs must be in the form of a binyan, possibly with the exception of a handful of verbs derived from roots missing some consonant, such as idda ‘sparkle’, idden ‘crow’, iżża ‘render (thanks)’, which do not readily fit into any of the verbal patterns (cf. irregular roots in Appendix I). Second, a root can in principle combine with several binyanim to form different verbal lexemes. The system is however full of gaps, as roots do not combine with cer- tain patterns. For instance, √bsm appears only in one binyan, V tbissem ‘smile’, √fsr oc- curs only twice, in II fisser ‘explain’ and V tfisser ‘be explained’, while √ksħ appears only in three, I kesaħ ‘cool down, v.i.’, kessaħ ‘cool down, v.t.’ and tkessaħ ‘be cooled down’.

This is, to a certain extent, comparable to English roots such as √RELIG and √STICK, which take the adjectival affixes -ious and -y, creating religious and sticky, and are not assigned an adjectival interpretation in the environment of the other suffix, *relig-y, *stick-ious (cf. Arad 2005: 194). Similarly, tri- and quadri-consonantal roots select only for some of the verbal patterns. Third, even if there are some tendencies for the binyanim to serve particular func- tions, roots generally acquire highly specialized, non-transparent meanings in different patterns. A single root may be assigned numerous interpretations in the environment of different verbal patterns (e.g., √ħrġ > ħareġ ‘take/go out’ in I, ħarreġ ‘train’ in II, stħarreġ ‘investigate’ in X). Unlike what is suggested by several grammars of Maltese, in this study I argue that each binyan does not have a well-defined function through which it relates to the other binyanim, and the meaning of a given verb is not necessarily asso- ciated with the binyan to which it belongs in a systematic and transparent way (cf. Sect. 3.2). Finally, one morphological assumption underlying this work is that the binyanim belong to the domain of word formation or derivation, not to that of inflection and the marking of grammatical categories such as person, number, tense-aspect, which are

36 A Tale of Two Morphologies

discussed in the following section. The binyanim exhibit two main characteristics of word formation. First, they are not completely productive. As a matter of fact, there is not one tri-consonantal root which creates nine different verbs. It is also interesting to note that, while there are quadri-consonantal roots which do create both QI and QII verbs, about half of them appear in only one pattern, e.g., √ħbrk creates ħabrek ‘QI, strive’ only and √gxtr creates tgexter ‘QII, become unsociable’ only (cf. Ch. 4 for the sta- tistics). Second, word formation, by definition, creates words with lexicalized meanings, and in the environment of different binyanim, roots may be assigned lexicalized, non- transparent interpretations. This view is challenged in Aronoff (1994: Ch. 5), where it is maintained that the bin- yanim in Hebrew (as in other Semitic languages) are similar in all respects to inflec- tional classes in other languages, such as verb conjugations in Latin. In a research pro- gram that advocates an autonomous status of morphology, independent of other grammatical components such as syntax and phonology, Aronoff assumes that binyan morphology realizes both inflectional and derivational processes. The binyanim are derivational in that they are associated with idiosyncrasies: gaps and non-transparent, non-predictable meanings assigned to the verbs that are embedded in them. They are inflectional in that they are obligatory for all (templatic) verbs. Aronoff untangles this apparent paradox by considering inflection and derivation not as two types of mor- phology, but rather as two uses of morphology: “inflection is the morphological realiza- tion of syntax, while derivation is the morphological realization of word-formation” (Aronoff 1994: 126). As elegantly as it might explain the regularity (obligatoriness) and irregularity (gaps and lexicalized, unpredictable meanings) in the binyan system, Aronoff’s account of “Morphology by Itself” fails to capture the strong connection between the binyanim and syntax. One important way how the binyanim interact with syntax is in terms of transitivity: some are inherently intransitive, some are predominantly transitive or in- transitive, and others may host either transitive or intransitive verbs (see Ch. 4 for other generalizations between the binyanim and argument alternations). Any theory has to account for a significant fact like this one: there must be some mapping rule from syntax to morphology that, for instance, prohibits verbs with direct from being mapped into the inherently intransitive patterns. As Arad (2005: 154) observes:

Setting the stage 37

If we treat morphology by itself, the interaction between syntactic structures and binyan mor- phology remains arbitrary. On the other hand, if we take morphology to be a reflection of syntac- tic processes, the relation between binyanim and the syntactic properties of verbs appearing in them is not only accidental, but expected.

In addition, in a study centered on Maltese verbs, Borg & Mifsud (1999) raise a number of problems regarding Aronoff’s treatment of the binyanim as inflectional classes. They argue for the derivational status of the Maltese binyanim on the basis of a number of criteria such as the gaps in the system, the semantic non-transparency of the binyanim, the relative order of morphemes (derivational ones are central in the word, inflectional ones are added laterally in the form of prefixes and suffixes), and the change in word class (as when a verb, e.g., tħanżer ‘make a pig of oneself’, xemmex ‘ex- pose to the sun’, is derived from a nominal, ħanżir ‘pig’, xemx ‘sun’). As shown in Ch. 4, verbs have some degree of morphological freedom in their choice of binyan morpheme. Inchoative verbs, for instance, may appear in different binyanim: I għala ‘boil, v.i’, V tgħawweġ ‘bend, v.i’, VII inkiser ‘break, v.i’, VIII ftaqar ‘become poor’, IX krieh ‘become ugly’, X stgħana ‘become rich’. The derivational nature of the binyanim is also compatible with this variation in morphological marking. In conclusion, on all accounts it is safe to say that the Maltese binyanim belong to the domain of derivation.29 The derivational relations between verbal patterns are taken up in Ch. 3 and Ch. 4, where the binyanim are not analyzed in isolation, but rather in their interaction with roots and other binyanim.

2.2.3 Inflection of templatic verbs

Each binyan may be inflected for tense-aspect (perfective, imperfective), mood (indica- tive, imperative), person (first, second, third), number (singular, plural for each of the three persons), and, for the third person singular, gender (masculine, feminine). That makes about 16 different word forms in each binyan, excluding verbs with direct and indirect object enclitics (e.g., kitib=ha ‘he wrote it’; kitib=li ‘he wrote to me’; kitib=hie=li ‘he wrote it for me’, etc.), which add up to a few hundred word forms. Verb inflection is concatenative, with prefixation and suffixation to a stem-base. The main tense-aspect distinction is between the perfective, formed by suffixes, and the im- perfective, formed by prefixes to mark person and gender and suffixes to mark plural

29 Other studies on Semitic languages, such as Horvath (1981) on Hebrew, also conclude that binyanim are derivational.

38 A Tale of Two Morphologies

(cf. Borg 1981, 1988; Fabri 1995; Ebert 2000; Spagnol 2007, 2009 for discussion on the tense-aspect system in Maltese). Table 2.8 provides the paradigms for the imperfective and perfective forms of two strong regular binyan I verbs, kiteb and xorob. The latter, having a liquid/nasal-medial radical, displays a slightly different imperfective paradigm: in the plural, an epenthetic vowel is inserted immediately before the liquid or , giving rise to the CV structure vCvCC (-ixorb-) rather than vCCC (-iktb-).30 The perfective paradigm of liq- uid/nasal-medial verbs, however, is identical to that of strong regular verbs. Also, note the syncretism of the second-person and third-person singular feminine in the imper- fective, and of the first- and second-person singular in the perfective.

Table 2.8 Paradigms of kiteb ‘write’ and liquid/nasal-medial xorob ‘drink’

Imperfective Perfective Imperfective Perfective

1SG n-ikteb 31 ktib-t n-ixrob xrob-t

2SG t-ikteb 32 ktib-t t-ixrob xrob-t

3SG.M j-ikteb kiteb j-ixrob xorob

3SG.F t-ikteb kitb-et 33 t-ixrob xorb-ot

1PL n-iktb-u ktib-na n-ixorb-u xrob-na

2PL t-iktb-u ktib-tu t-ixorb-u xrob-tu

3PL j-iktb-u kitb-u j-ixorb-u xorb-u

Inflectional affixes are sensitive to root types (cf. Table 2.4). While strong regular, weak-initial, and weak-medial verbs typically have the paradigms displayed in Table 2.8, weak-final verbs such as mexa ‘walk’ and qara ‘read’, behave slightly different. Un- like other verb types, they end in a vowel: one class ends in -i/-ejt, another in -a/-ajt, as

30 An interesting counter-example is lemaħ ‘recognize’, which has the vCCC stem in the plural, e.g., n-ilmħ-u rather than the expected *n-ilimħ-u. Fabri (2009: 13-14) argues that stem allomorphy in this case is condi- tioned not only phonologically. Syllable structure also seems to play a role: compare, for instance, nilmħu, which is syllabified as nilm.ħu with /lm/ as the coda of the first syllable, and n-imird-u ‘IPFV.1-fall ill-PL’ rather than *nimr.du, as the sequence /mr/ is not a permissible coda cluster. 31 Prefixal n- may assimilate completely with weak-medial and reduplicative stems that begin with a liquid or nasal, e.g., ir-rid ‘IPFV.1SG-want’ (for inrid), im-miss-u ‘IPFV.1-touch-PL’ (for inmissu). 32 Complete assimilation of the prefix t- (2SG, 2PL and 3SG.F imperfective) takes place when the stem starts with a coronal, as in id-dur ‘IPFV.2SG-turn’ for *t-dur, and ix-xomm-u ‘IPFV.2-smell-PL’ for *t-xomm-u. Only some binyan I (as dar, xamm), binyan II (iġ-ġedded ‘IPFV.3SG.F-renew’ for *t-ġedded), binyan III (is-siefer ‘‘IPFV.2SG-travel’ for *t-siefer) and QI verbs (iż-żarżar ‘IPFV.3SG.F-hum’ for *t-żarżar) undergo assimilation. 33 The suffix -et has an allomorph -ot, which is generally conditioned by the presence of o in the stem, e.g., xorb-ot ‘drink-PFV.3SG.F’, ħolm-ot ‘dream-PFV.3SG.F’.

Setting the stage 39

illustrated in Table 2.9.34 This final vowel distinction is important in the redistribution of Romance loan verbs within the class of templatic verbs (cf. Sect. 2.3.2).

Table 2.9 Paradigms of weak-final mexa ‘walk’ and qara ‘read’

Imperfective Perfective Imperfective Perfective

1SG n-imxi imx-ejt n-aqra qr-ajt

2SG t-imxi imx-ejt t-aqra qr-ajt

3SG.M j-imxi mexa j-aqra qara

3SG.F t-imxi imx-iet t-aqra qr-at

1PL n-imx-u imx-ejna n-aqr-aw qr-ajna

2PL t-imx-u imx-ejtu t-aqr-aw qr-ajtu

3PL j-imx-u imx-ew j-aqr-aw qr-aw

The perfective forms of weak-final verbs, especially the first and second persons, may be analyzed in the same way as strong regular verbs (cf. √ktb 1SG ktib-t and √mxj

1SG imxej-t). However, in Table 2.9 they are analyzed as perfective stems plus a different set of suffixes (imx-ejt, imx-iet, etc.). This morphemic analysis, whereby the final weak radical loses its radical status and comes to form part of the inflectional suffix, is moti- vated by the manner the weak-final inflection has been extended, either completely or partially, to other verb classes, including reduplicative (e.g., ħass ‘feel’) and silent-final (e.g., sema’ ‘hear’) verbs, as well as to all concatenative verbs (cf. Sect. 2.3). As is evident from Table 2.10, reduplicative and silent-final verbs take the weak-final perfective suf- fixes in the first and second persons. The weak-final inflection has also been extended to other anomalous forms, such as verbs derived from roots with some missing consonant (ġie ‘come’, ra ‘see’) and to some verbs that end in a super-heavy syllable, i.e. long vowel plus consonant (għam ‘swim’, biegħ ‘sell’, strieħ ‘rest’), which have defective perfective paradigms (cf. Mifsud 1995a: 316-317; Spagnol 2009: 66). It has also gained ground with a few verbs whose roots have a stop and a for second and third radicals, such as √ftx, fittex ‘seek’. In an effort to stabilize an anomalous stem by reducing allomorphy (-fittex, -fittx-, fittix-, fittx-), the weak-final suf-

34 A handful of verbs such as nesa ‘forget’ and beda ‘begin’ end in -a in the imperfective (e.g., n-insa ‘IPFV.1SG- forget’) but in -ejt (not -ajt) in the perfective (e.g., ins-ejt ‘forget- PFV.1SG’).

40 A Tale of Two Morphologies

fixes attach to an invariable stem (fittx-), yielding fittx-ejt ‘seek-pfv.1sg’ (for fittix-t), fittx-ejtu ‘seek-pfv.2pl’ (for fittix-tu), and so on (cf. Mifsud 1995a: 315-317). And, as mentioned above, the weak-final inflection is used for all concatenatively formed verbs. In sum, templatic verbs have two sets of inflectional affixes which stand in complemen- tary distribution, and which for ease of reference we may term strong affixes (-t, -na, -tu, etc., cf. Table 2.8) and weak affixes (-ejt, -ejna, -ejtu, etc., cf. Table 2.9).35

Table 2.10 Paradigms of reduplicative ħass ‘feel’ and silent-final sema’ ‘hear’

Imperfective Perfective Imperfective Perfective

1SG in-ħoss ħass-ejt n-isma’ sm-ajt

2SG t-ħoss ħass-ejt t-isma’ sm-ajt

3SG.M j-ħoss ħass j-isma’ sema’

3SG.F t-ħoss ħass-et t-isma’ semgħ-et

1PL in-ħoss-u ħass-ejna n-isimgħ-u sm-ajna

2PL t-ħoss-u ħass-ejtu t-isimgħ-u sm-ajtu

3PL j-ħoss-u ħass-ew/-u 36 j-isimgħ-u semgħ-u

Mifsud (1994; 1995a: 296-318) attributes the ever-increasing importance of the weak affixes to three formal reasons. First, the shape of some of the verbs that took on the weak affixes coincides with that of weak-final verbs. For instance, binyan I redupli- cative verbs (e.g., ħall ‘untie’) coincide with binyan II weak-final verbs (e.g., ħall-a ‘leave’), so much so that their paradigms are in part similar (e.g., ħall-ejt translates both

as ‘untie.PFV.1SG’ and ‘leave.PFV.1SG’). Second, in contrast to strong affixes, most of which begin with a consonant, all weak affixes are vowel-released (e.g., -ejt, -ejna), and as such can be more easily applied to verbal stems ending with a , like those of reduplicative verbs and many concatenative verbs (e.g., ipprogramm- ‘program’, irre- golarizz- ‘regularize’). The third characteristic of weak affixes that makes them a more likely model for the integration of other verb classes is stem regularity. While allomor- phic stem alternation is a typical feature of strong affixes (cf. Table 2.8), the stem weak

35 This distinction applies to quadri-consonantal verbs too. Compare, for instance, strong ħarbat ‘confuse’ (e.g., ħarbat-t ‘confuse-PFV.1SG’) with weak-final kanta ‘sing’ (e.g., kant-ajt ‘sing-PFV.1SG’). 36 There are two variants for the perfective suffixes of reduplicative verbs in the third-person plural: -u as in strong regular verbs, and -ew as in weak-final verbs. Mifsud (1994: 251-252) observes that the latter suffix is gradually replacing the -u suffix. In general, when some verbs have both forms, the -u suffix is used only in some fossilized expressions.

Setting the stage 41

affixes attach to is not subjected to allomorphic alternations, except for stress shift and regular morphophonemic alternations thereby conditioned (cf. Mifsud 1995a: 206- 210). For further discussion on verb inflection in Maltese, see Cremona (1961), Mifsud (1995a), Borg & Azzopardi-Alexander (1997), Fabri (2009), inter alia. The inflection of concatenative verbs is briefly discussed in the next section.

2.3 Concatenative morphology

We have, so far, looked at the Semitic side of Maltese verbal morphology. However, Mal- tese is very different from Arabic and other Semitic languages. Centuries of intense lan- guage contact with Sicilian, Italian and English have brought about two major changes. First, the role of root-and-pattern morphology, as was described in the first part of this chapter, is much reduced, though still very present and productive, in a sense that will be discussed in Sect. 2.4. Second, the pressure exerted on Maltese by non-Semitic lan- guages gave rise to new morphological structures that are not found in Arabic. These morphological structures, in particular the verbal ones, are the focus of this section, which owes a great deal to Mifsud (1994, 1995a, 1996) and Hoberman & Aronoff’s (2003) in-depth treatments of loan verbs in Maltese. The organization of this section is as follows. In Sect. 2.3.1, I discuss the way loans are assimilated in the verbal mechanisms of root-and-pattern morphology. Next, in Sect. 2.3.2, I examine the concatenative innovations of loan verbs that are not incorporated in the templatic system, and briefly describe the inflectional morphology of this class of concatenative verbs.

2.3.1 Templatic loan verbs

The class of templatic verbs includes not only verbs of Semitic origin, but also a large number of verbs from Romance and a couple from English. The Arabicization of loan verbs in Maltese is a complex process involving two main stages. To begin with, 3 or 4 consonants, which will serve as the consonantal root, are extracted from a loan stem. This novel root is then combined with one or more of the verbal patterns, yielding Mal- tese templatic verbs. Tri-consonantal pitter ‘paint’, vara ‘launch’, and quadri- consonantal ċapċap ‘clap’, pinġa ‘draw’ are a few examples of verbs of non-Semitic ori-

42 A Tale of Two Morphologies

gin that have been completely integrated into Maltese root-and-pattern morphology, and may hence be decomposed into a consonantal root and a binyan, e.g., a strong root

√ptr and binyan II (pitter), a weak-final root √vrj and binyan III (vara). Only their ety- mology differentiates these verbs from indigenous verbs. Sometimes templatic loan verbs are also distinguishable from their root consonants, as they may include one or more of the seven phonemes not found in Arabic /p, v, g, tʃ, ts, dz, ʒ/, which Maltese ac- quired through contact with Romance and English (cf. Sect. 1.2.1), and from their vow- els, as they have a different range of vowel sequences than the native binyanim (cf. Sect. 2.2.2). Templatic loan verbs may be separated into two major classes, those derived from weak-final roots (e.g., vara) and those derived from the other root types, i.e. strong regular and reduplicative, and weak-initial and weak-medial (e.g., pitter). In what fol- lows, I outline two key differences between the two groups, which in Mifsud’s (1995a) classification are labeled Type A (R, Red., W-I, W-M) and Type B (W-F) loan verbs. For ease of reference, I use Mifsud’s labels for the two classes. The first difference concerns the source they come from. In the case of Type A verbs, the radicals are generally extracted from previously borrowed nouns and adjec- tives, such as pitter ‘paint’ derived from pittur ‘painter’ (cf. Sicilian pitturi), and werreċ ‘make cross-eyed‘ derived from werċ ‘cross-eyed’ (cf. Old Italian vercio, Sicilian guer- ciu).37 Mifsud (1995a) calls these denominal and deadjectival verbs “second generation loans”, as they are derived from already naturalized nominal or adjectival forms, many of which had probably been subjected to the process of root extraction for the forma-

tion of broken (e.g., werċ – wereċ ‘cross-eyed.PL’).38 By contrast, fully naturalized Type B verbs are usually loans the immediate etymon of which is nonnative, like vara ‘launch’ (cf. Italian varare). The second difference between the two groups of templatic loan verbs lies in the shape of their stem. In their citation form, i.e. third person singular masculine perfec- tive, Type A verbs end in a consonant, as pitter and ċapċap. Type B verbs, on the other hand, always end in a vowel, as vara and pinġa.

37 Of particular interest are cases in which the loan stem has only two consonants and a third consonant, a latent weak radical, is assumed where the loan has a long vowel bounded by two consonants. For instance, from pipa ‘smoking pipe’ the tri-consonantal root √pjp is extracted, as is evident when the root is cast, say, in binyan II pejjep ‘smoke’. 38 Loan verbs completely integrated into Arabic, both standard and dialectal, are to a great extent also de- rived from Arabicized nouns and adjectives (cf. Mifsud 1995a: 49-55).

Setting the stage 43

In addition, Type B verbs exhibit far less phonological variation in inflection than Type A verbs do, as is evident from their paradigms in Table 2.11. Except for stress shift, verbs such as vara and pinġa normally require no change in the stem. Inflectional affixes of Type A verbs, by contrast, subject the stem to several allomorphic variations. The inflectional forms of Type B verbs guarantee the formal integrity of the loan stem, so much so that they can be essentially described in terms of a stem plus conjugational suffixes (cf. Sect. 2.2.3). In view of this, weak-final loan verbs have been interpreted as the link between templatic and concatenative verbs (cf. Mifsud 1994, 1995a, 1996).

Table 2.11 Paradigms of Type A pitter ‘paint’ and Type B vara ‘launch’

Perfective Imperfective Perfective Imperfective

1SG pittir-t in-pitter var-ajt in-vara

2SG pittir-t t-pitter var-ajt t-vara

3SG.M pitter j-pitter vara j-vara

3SG.F pittr-et t-pitter var-at t-vara

1PL pittir-na in-pittr-u var-ajna in-var-aw

2PL pittir-tu t-pittr-u var-ajtu t-var-aw

3PL pittr-u j-pittr-u var-aw j-var-aw

Before turning to a description of the concatenative verb formation strategy in Mal- tese, one final note is in order on loan verbs that, in several respects, lie in between templatic and concatenative verbs. Some examples include kompla ‘complete, continue’ (cf. kampa ‘QI, get by’), irbatta ‘clinch’ (cf. batta ‘II, abate’), skanta ‘amaze‘ and irkanta ‘auction’ (cf. kanta ‘QI, sing’), skorra ‘go beyond’ (cf. korra ‘II, hurt’), mantna ‘maintain’ (cf. majna ‘QI, wane’). While conforming to the syllabic structure of the binyanim, their consonantal content does not match the tri- or quadri-consonantal structure. They may be analyzed as syllabic variants of templatic verbs, where one or more of the root con- sonants was, so to say, substituted by a consonant cluster, underlined above. Evidence for the interpretation of these forms as syllabic variants of binyanim comes from two points. First, some of them take part in the derivational processes available to templatic verbs, e.g., tkompla ‘QII, be continued’, tmantna ‘QII, be main- tained’. Second, they do not take initial gemination (where phonologically permissible), which is a characteristic of concatenative verbs, such as ikkompila ‘compile’ (cf. Sect.

44 A Tale of Two Morphologies

2.3.2). However, other aspects of these verbs point to the opposite conclusion, such as the fact that they exhibit consonant clusters which disrupt the verbal template, and that some of their morphologically related forms, e.g., past (kompl-ut ‘completed’, skorr-ut ‘gone beyond’) and verbal nouns (irkant-ar ‘auctioning’, skant-ament ‘amaze- ment’), are formed concatenatively. For discussion on these in-between cases, see Mif- sud (1995a: 89-110; 1996: 122-126).

2.3.2 Concatenative verbs

Templatic verbs, indigenous and borrowed, constitute a closed list in Maltese. In the comprehensive database of roots and patterns I put together as part of this work, there are 1902 different roots from which 3962 templatic verbs are derived (cf. Ch. 4). Ac- cording to Mifsud (2008: 156), 373 of these verbs (from 247 different roots) are Ro- mance and English loans, a large number of which are recorded in old dictionaries, but are not current in modern standard Maltese. Verb formation through root-and-pattern morphology is no longer productive (available rather than profitable, cf. Sect. 2.4). Other than one-offs, such as qanpen ‘ring out obstinately’ (cf. qanpiena ‘bell’), found in some literary works, virtually no new templatic verb is formed in modern Maltese (cf. Mifsud 1995a: 77). Particularly signifi- cant is the fact that, in spite of the long contact with English stretching over two centu- ries, there are only two verbs of English origin formed by the association of a root and a binyan, fajjar ‘hurl’ and tawwat ‘honk’ (cf. fire, toot). A curious formation is the quadri- consonantal verb tmandar ‘get wasted’ (cf. mandra ‘mess’, sar mandra ‘get wasted, lit. become a mess’ < Italian mandra, mandria) typically used in the sociolect of the youth. The question arises regarding the formation of verbs in Maltese: why was the templatic verb formation strategy interrupted and a new, concatenative system contrived? Many borrowed nouns that enter Maltese without difficulty cannot create templatic verbs, the main reason for this being their inability to fit into the disyllabic or monosyl-

labic binyanim. As Mifsud (1996: 118) notes:

Full naturalization in a S[emitic] sense is an exacting exercise which entails “dehydrating” the borrowed stem-base of its vocalic content, leaving only the consonantal structure which is inter- preted as the root-base on the Semitic pattern, i.e. as a discontinuous morpheme which forms the skeleton of every verbal and nominal derivative.

Setting the stage 45

Other than the difficulty to reduce polyconsonantal stems to 3 or 4 consonants to em- ploy as a root, another serious handicap is the syllabic restructuring that is required to integrate loans into root-and-pattern morphology. Many Romance stems have been considerably changed both phonologically and morphologically to conform to the Se- mitic verbal patterns. Consider the rather extreme cases bata ‘suffer’ from Sicilian patiri, and gaża ‘report’ from Sicilian accusari. Because of such constraints inherent in the binyan system, a more flexible system was required, which would not only accom- modate the new verbal formations with the indigenous verbs, but also exhibit less se- lective resistance to their syllabic configuration. A solution in this sense was offered by the class of weak-final verbs, which was al- ready used as a way out for a number of irregular or anomalous templatic verbs. Among the templatic verbs, which, either fully or partially, adopted the weak-final inflection, are reduplicative and silent-final verbs, and some verbs that end in a super-heavy sylla- ble. It was also extended to all concatenative verbs for several reasons, the main two being the fact that (i) it imposes no restriction on the syllable structure or vowel pat- terns of verbs, other than the phonological restrictions on syllable structures and vow- els that apply throughout the language; and (ii) that, unlike the inflection used with strong verbs, it does not normally bring about any allomorphic change in the stem, ex- cept for regular stress shift (cf. also Sect. 2.2.3). This implies that there are two important differences in the stem structure and in- flectional morphology of templatic and concatenative verbs. While the former comprise no more than two syllables and have a limited number of vowel patterns, there appears to be no restrictions on the number of syllables and the vowel quality in a concatenative verb stem. Secondly, concatenative verbs have a more simplified inflectional morphol- ogy. As discussed in Sect. 2.2.3, templatic verbs have two sets of inflectional affixes, strong (-t, -et, -na, etc.) and weak (-ejt, -iet, -ejna, etc.), with some verb classes having cross-bred paradigms, e.g., silent-final verbs like sema’ ‘hear’ take the strong perfective suffixes in the third person (e.g., semgħ-et ‘hear-PFV.3SG.F’) and the weak ones in the first and second persons (e.g., sm-ajt ‘hear-PFV.1SG’). By contrast, all concatenative verbs have an a/ajt-final paradigm, with a smaller number of verbs taking the i/ejt-final para- digm, as shown in Table 2.12.39

39 Concatenative verbs from Romance are systematically redistributed within the class of weak-final tem- platic verbs (cf. Table 2.9 in Sect. 2.2.3), being assigned the a/ajt-final paradigm when the loan stem be- longs to the first conjugation class in -are (cf. Italian emigr-are, Maltese emigr-ajt, n-emigra ‘emigrate.PFV, IPFV.1SG’) and the i/ejt-final paradigm when it belongs to either the second or third conjugation in -ere or

46 A Tale of Two Morphologies

Table 2.12 Paradigms of a/ajt-final ipparkja ‘park’ and i/ejt-final sploda ‘explode’

Perfective Imperfective Perfective Imperfective

1SG ipparkj-ajt n-ipparkja splod-ejt n-isplodi

2SG ipparkj-ajt t-ipparkja splod-ejt t-isplodi

3SG.M ipparkj-ajt j-ipparkja sploda j-isplodi

3SG.F ipparkj-ajt t-ipparkja splod-iet t-isplodi

1PL ipparkj-ajt n-ipparkj-aw splod-ejna n-isplod-u

2PL ipparkj-ajt t-ipparkj-aw splod-ejtu t-isplod-u

3PL ipparkj-ajt j-ipparkj-aw splod-ew j-isplod-u

As is evident from the table above, for a loan stem such as park to be integrated into the concatenative verbal system of Maltese, it has to undergo two processes, initial gemination (ippark-) and suffixation (-ja). The most distinctive characteristic of con- catenative verbs is perhaps initial gemination, which is found in loan stems with a sin- gle initial consonant (ippark-, cf. English park) and with an initial consonant-sonorant cluster (ikkmand-, cf. Italian comandare ‘command’). Unless the preceding word in the same phonological phrase ends in a vowel, in Maltese there is always a prosthetic vowel before initial geminates. By contrast, loan stems with sibiliant-initial (splod-, cf. Italian esplodere ‘explode’, żbilanċ-, cf. Italian sbilanciare ‘unbalance’) and liquid-initial clusters (irċiev-, cf. Italian ricevere ‘receive’, ilment-, cf. Italian lamentare ‘complain’) do not have geminate-initial stems. Needless to say, vowel-initial stems (ordn-, cf. Italian ordinare ‘order’, urt-, cf. Italian urtare ‘offend’) are also excluded from the gemination process. For an analysis of the origin and distribution of initial gemination, see Mifsud (1995a: 140ff.). Concatenative verbs, therefore, may be formally separated into three main con- stituent elements:

(i) roots with possible gemination of the first consonant and a prosthetic vowel,

e.g., √ORGAN, √PRETEND, √VER;

-ire respectively (cf. Italian discut-ere, Maltese iddiskut-ejt, n-iddiskuti ‘discuss.PFV, IPFV.1SG’; and Italian di- vert-ire, Maltese iddivert-ejt, n-iddiverti ‘enjoy.PFV, IPFV.1SG’). For discussion of this phenomenon also in the case of templatic verbs of Romance origin, see Alex. Borg (1978: 325-327), Mifsud (1995a: 120-126; 1996: 121-122), inter alia.

Setting the stage 47

(ii) verbal suffixes, i.e. -ja (ipparkja ‘park’), -ifika (ivverifika ‘verify’), -izza (or- ganizza ‘organize’), and -ø (ippretenda ‘expect’);40 (iii) inflectional affixes, i.e. those used for weak-final templatic verbs, including the a/ajt- or i/ejt-final distinction in the imperfective (cf. Table 2.9, Table

2.12).41

Finally, it is important to note that the concatenative verb formation strategy is not restricted to Romance and English borrowings, but is a fully productive process which operates in the formation of neologisms. A good case in point are verbs of Arabic origin which are formed concatenatively, such as iżżejtna ‘become oily (hair)’ (cf. żejt ‘oil’, żejtni ‘oily’) and iżżiftja ‘coat with tar’ (cf. żift ‘tar’). The productivity of concatenative (as well as templatic) verb formation awaits further investigation, using (i) psycholin- guistic techniques, e.g., along the lines of Twist (2006), and Anshen & Aronoff (1988) and Baayen’s (1994) production tasks, (ii) dictionary comparison (cf. Bolozky 1999); as well as (iii) corpus-based measures for gauging different aspects of productivity (cf. Baayen 1992, 1993; Baayen & Renouf 1996; Gaeta & Ricca 2006; inter alia).

2.4 Summary

Having examined the major features of templatic and concatenative verbs in Maltese, let us summarize the main conclusions. The shaping of Maltese is marked by intimate inter-language contact between Ara- bic, Romance varieties and, to a lesser extent, English, for about one millennium. Vari- ous layers of borrowings from Old Sicilian, Italian, and later English, were deposited on top of the Arabic foundations laid down between 870, when the Arabs conquered Malta, and 1249, when the Muslims were expelled from the islands. Influence from European languages goes back as early as the 12th century, and grew steadily as Maltese lost con- tact with its Arabic matrix and interacted far more with the European languages. Caught at the crossroads of two cultures and two typologically diverse languages, Mal-

40 A subclass of the verbs that select the zero morph (-ø) take the -ixx- augment, e.g., issuġġerixxa ‘suggest’, which is not always present in the paradigms. See Mifsud (1995a: 169ff.) for an analysis of the distribution of -ixx- in loan verbs. 41 Strong affixes are therefore restricted to the inflection of templatic verbs. From a marginal system in root-and-pattern morphology, the inflection of weak-final templatic verbs has become the main system for the formation of the open-ended class of concatenative verbs.

48 A Tale of Two Morphologies

tese, and in particular its verbal morphology, developed into what Mifsud (1996) calls a “Romance-Arabic Crossbreed”. Perhaps the most striking fact about Maltese is that, as a result of its long, multifari- ous history of language contact, it has two verb formation strategies, root-and-pattern association, on the one hand, and concatenation, on the other. In this chapter, I offered a description of both morphological processes, outlining the main differences between two verb classes. To sum up, templatic verbs differ from concatenative verbs in five im- portant ways. 1. Stem structure. All templatic verbs must be in the form of a binyan. For this rea- son, both their syllabic configuration and vowel sequences are fixed: most of them are disyllabic, a few are monosyllabic, and each binyan allows for a small set of vowel se- quences (cf. Sect. 2.2). By contrast, there appear to be no restrictions on the syllabic structure and vowel quality of concatenative verbs, other than the phonological restric- tions that apply throughout Maltese. 2. Inflection. Templatic verbs may inflect for tense-aspect, mood, person, number, and gender in two main ways, taking either the strong (-t, -et, -na, etc.) or weak (-ejt, -iet, -ejna, etc.) affixes, with a number of them taking both affixes in different paradigm cells. Concatenative verbs, on the other hand, all take the weak inflection. As shown in Sect. 2.2.3 and 2.3.2, the inflectional system of concatenative verbs is more simplified than that of templatic verbs. 3. Derivational potential. While inflection is active in both verb classes, derivational morphology, such as causative, passive and reflexive formation by prosodic change, is limited to templatic verbs. This contrast between consonantal (templatic) roots that may be embedded in various verbal patterns, on the one hand, and syllabic (concatena- tive) roots that are typically assigned one verbal interpretation, on the other, is viewed in this study as a critical point which explains: (i) the irregularity inherent in the binyan system (Ch. 3 & 4), and (ii) the distinction in the formal encoding of the causative- inchoative alternation (Ch. 5). Another difference in this respect concerns forms morphologically related to the two verb classes. For instance, the past participles and verbal nouns associated with templatic verbs may be decomposed into roots and patterns with prefixes and possibly

inherent vowels, such as mvC1C2uuC3 (e.g., miksur ‘broken’), mC1vC2C2vC3 (e.g., imqas-

sam ‘divided out’) for past participles, and tC1vC2C2iiC3 (e.g., tkissir ‘breaking’),

tvC1C2iiC3a (e.g., taqsima ‘division’) for verbal nouns. Those associated with concatena- tive verbs are, however, formed by suffixation, with past participles taking -at or its

Setting the stage 49

variants -ut and -it (e.g., ittajpjat ‘typed’) and verbal nouns taking the suffixes -ar/-ir (e.g., ittajpjar ‘typing’). Other nominalizing suffixes include -ment (e.g., inkoraġġiment ‘encouragement’), -zzjoni (e.g., fissazzjoni ‘fixation’), -nza (insistenza ‘insistence’), -aġġ (e.g., spjunaġġ ‘espionage, spying’), and -ing (e.g., welding ‘soldering alloy used in weld- ing’).42 4. Etymology. Templatic verbs are predominantly of Semitic origin, with a few hun- dred verbs from Romance and a couple from English, which have been subjected to the process of consonantal root extraction and embedding in one or more verbal patterns. Concatenative verbs, on the other hand, are for the most part derived from Italian, Sicil- ian and English, with very few from Semitic. 5. Productivity. One crucial difference between the two verb classes is the fact that one is closed, practically unproductive, and the other is fully productive. A refinement of the notion of morphological productivity is in order at this point. Bauer (2001, 2004) suggests that the term is ambiguous between availability and profitability. A morpho- logical process is said to be available (disponible, Corbin 1987: 177) if it can be used in the production of new words, and profitable (rentable, Corbin 1987: 177) to the extent that it is actively used. In this sense, verbal templatic morphology is unavailable because, under normal conditions, no new Maltese verb is formed through root-and-pattern morphology. Con- sonantal roots could in theory be extracted from several loan stems, such as √prk from park-, √flm from film-, and then combined with binyan morphology (e.g., *parak, *fellem), but in practice this does not take place. Instead, such loans, like any new verb in modern Maltese, are integrated into the verbal system via the concatenative forma- tion, involving initial gemination and suffixation (ipparkja, iffilmja).43

42 There are a number of curious cases, such as: (i) verbs with both a templatic and a concatenative past participle, e.g., pinġa ‘QI, draw’ having both impinġi and pinġut ‘drawn’; (ii) templatic verbs like vara ‘III, launch’ forming past participles and verbal nouns by suffixation (var- at ‘launched’, var-ar ‘launching’) rather than root-and-pattern association; (iii) verbal nouns with two markers, i.e. the templatic t- prefix and the concatenative suffix -ar, e.g., t-ranġ-ar ‘setting right’. For a detailed discussion on the derivational potential of templatic and concatenative verbs and on such hybrid forms, see Mifsud (1995a: 68-72; 126-139; 247-251). 43 Mifsud (1995a, 1995b) argues that Maltese is undergoing a typological shift: from a basically non- concatenative type, its morphology is evolving into a concatenative type, displaying a preference for invari- able stems and marginalizing templatic forms which typically exhibit allomorphic variations. A case in point is the weak inflection, which, because it guarantees the formal integrity of the stem, has gained ground with reduplicative, silent-final, and concatenative verbs, as well as a number of anomalous forms (cf. Sect. 2.2.3).

50 A Tale of Two Morphologies

In view of this difference in the productivity of the two verb classes, Hoberman & Aronoff (2003: 76) have gone as far as to claim that “Maltese is a concatenative lan- guage masquerading as a root-and-pattern language.” The implication is that “the lack of adherence to templatic constraints in borrowed verbs indicates that non- concatenative morphology is not actively functioning in Maltese, suggesting that the profitability of root and pattern morphology is low or non-existent” (Twist 2006: 72). However, for a class of verbs to be diachronically closed does not necessarily mean that it is no longer productive, and that its members can no longer be derived (in syn- tax) at a synchronic level. As Borg & Mifsud (1999: 12) point out, “[e]ven if no new lex- emes enter the language in the root-and-pattern method, this type of morphology is still synchronically at work in the thousands of Semitic verbs and nouns which form an in- tegral part of Maltese”. Although the class of concatenative verbs is undeniably the largest and most pro- ductive, it has been demonstrated in an elicitation experiment designed to measure the relative productivity of the two verb formation strategies in Maltese, that root-and- pattern morphology is productive (Twist 2006). Speakers were able to extract conso- nantal roots from nonce items and merge them with verbal patterns. Quite strikingly, in some instances, speakers chose the templatic strategy even in response to real word stimuli with established verbs that are formed concatenatively. Responses to the nonce stimuli indicate that, in an experimental context centered on word structure that does not take such factors as semantics and language use into account, root-and-pattern morphology is a profitable strategy in Maltese, pace Hoberman & Aronoff (2003). Further research in this respect is required (i) to support or refute either the posi- tion that root-and-pattern morphology is unavailable or that it is unprofitable; (ii) to answer related questions, such as how do speakers decide which strategy to use to form new words; and (iii) to provide statistical calculations of productivity based on measuring profitability in corpora, along the lines of Baayen (1992, 1993), Gaeta & Ricca (2006), and so on. Of particular interest are Bolozky (1999), who adopts different methods of measuring and evaluating productivity of word formation (productivity tests, dictionary comparison, and corpus analysis) in Israeli Hebrew, and Verheij (2000: 43ff.), who applies Baayen’s models to assess the productivity of the binyanim in Bibli- cal Hebrew. Such experimental studies need to be replicated in Maltese in order to measure, among other things, the relative frequency and productivity of the various binyanim. I leave it to future research to investigate these and related issues.

Setting the stage 51

In the discussion that follows, I seek to provide a unified approach to the two verb formation strategies by studying argument structure alternations, in particular the causative-inchoative alternation. However, having in mind the formal and semantic complexities associated with root-and-pattern morphology, I am led first to examine the class of templatic verbs in depth, with the aim of describing the morphology and lexical semantics of the binyan system in Maltese. To begin with, in Ch. 3, I refine some of the intuitions found in previous studies and discuss the assumptions that serve as the basis for the rest of the work. In Ch. 4, I then move on to characterize the binyan system both quantitatively and qualitatively by means of an exhaustive database of verb- creating roots and the patterns they interleave with, before tackling the causative- inchoative alternation in both templatic and concatenative verbs in the next phase (Ch. 5).

Chapter 3 The protagonists: roots and patterns

3.1 The ingredients of roots...... 54 3.1.1 Problems with the traditional approach ...... 56 3.1.2 Roots are underspecified ...... 59 3.2 The functions of the pattern...... 63 3.2.1 When a root appears in various patterns...... 63 3.2.2 What functions do patterns have? ...... 66 3.2.3 Do patterns have functions at all?...... 70 3.3 Root derivation and word derivation...... 71 3.3.1 Cross-linguistic evidence ...... 74 3.3.2 Word formation in Maltese ...... 78

In the previous chapter, I gave an overview of the dual nature of Maltese verbal mor- phology. Maltese verbs were analyzed as belonging to either of two main classes, tem- platic and concatenative. The latter class, consisting mostly of loan verbs from Romance and English, is the only productively available means for building new verbs in modern Maltese. Templatic verbs, by contrast, constitute a closed class of verbs. The formation of new verbs by the interdigitation of a tri- or quadri-consonantal root in a binyan is extremely rare in the current language. Root-and-pattern morphology, however, is fully functional synchronically in the numerous verbs and nouns that make up a large por- tion of the basic lexicon of Maltese. Moreover, the availability of root-and-pattern mor- phology is supported by psycholinguistic evidence from a visual masked priming ex- periment (Twist 2006), though more research is needed in this area to measure the productivity of the two verb formation strategies. The rest of this work aims at providing an analysis of verbs in Maltese as a single system combining the two types of morphology. Root-and-pattern morphology is taken as the starting point for the discussion. It seems reasonable to begin with an analysis of templatic verbs because (i) they historically precede concatenative verbs (cf. Mifsud 1995a: 253), and (ii) inflectional morphology is unmistakably of the Semitic kind for both verb classes (cf. Ch. 2). In addition, since templatic verbs constitute a relatively closed list, it is possible to examine a comprehensive corpus of verbs that belong to this The protagonists: roots and patterns 53

class in order to provide a detailed description of the binyan system in Maltese (Ch. 3 and Ch. 4). The analysis is then extended to concatenative verbs, presenting a unified approach of the entire verbal system of Maltese by focusing on one central feature, ar- gument structure alternations (Ch. 5). This chapter is therefore concerned with templatic verbs, trying to refine some in- tuitions of traditional grammars regarding the contents of roots and the roles of pat- terns. Regular and irregular aspects of templatic verbs are discussed in the light of two general assumptions: one concerning the semantics of roots, another concerning word formation. Instrumental to this discussion is Arad’s (2005) account of Hebrew root- and-pattern morphology. The so-called Root Hypothesis (i.e. the notion that all verbs, nouns and adjectives, whether morphologically simple or complex, are composed of abstract roots) and other principles of Distributed Morphology (cf. Halle & Marantz 1993; Marantz 1997, 2000, 2007; inter alia) related to it will serve as the backbone of the argumentation in this chapter. The aim of this and the following chapter is to offer a comprehensive account of Maltese templatic verbal morphology. My concern is with two interrelated phenomena:

(a) While in many cases the appearance of a root in two or more binyanim cor- responds to semantically transparent alternations (e.g., active-passive kiteb ‘write’ – inkiteb ‘be written’; causative-inchoative qasam ‘split, v.t’ – inqasam ‘split, v.i’), in other cases a single root is assigned different interpretations which may not be related to each other in any systematic way (e.g., √swj, sewa ‘cost’ and sewwa ‘mend’; √nbħ, nebbaħ ‘inspire’ and stenbaħ ‘wake up’. (b) Different roots occurring in the same binyan tend to acquire the same se- mantic interpretation, such as binyan II verbs ċekken ‘make smaller’, mexxa ‘cause to walk’, daħħaq ‘make laugh’, all of which are causative, and binyan IX verbs ċkien ‘become small’, sfar ‘turn yellow/pale’, ħxien ‘put on weight’, all of which are inchoative. However, the correlation between binyanim and specific verb meanings is not one-to-one. The same kind of information can often be encoded through multiple binyanim, e.g. reflexives are found in binyan VI (tbiegħed ‘distance oneself’), VII (intiżen ‘weigh onself’), X (staħba ‘hide oneself’), inchoatives are found in I (nixef ‘dry’), V (tfarrak ‘crumble’), VII (inqasam ‘split’), VIII (ftaqar ‘become poor’), etc.

54 A Tale of Two Morphologies

To anticipate, the main conclusions I draw in this chapter are the following:

(i) The binyan system is neither perfectly transparent nor completely opaque. It has a degree of regularity (argument alternations) and irregularity (gaps and multiple, unpredictable interpretations) in it. (ii) The semantic relatedness of words containing the same radicals and their meanings being too distant to relate derivationally can be explained if we assume that the semantic content of consonantal roots is not fully definable. Roots are underspecified phonologically, in that 3 or 4 radicals are not pro- nounceable on their own, and semantically, in that they acquire different in- terpretations when they combine with different patterns. (iii) The dual nature of the binyan system is also predictable if, along the lines of many theories of word building, I postulate two processes of word forma- tion: one for the regular, productive processes (word derivation), another for the irregular, non-productive processes (root derivation). Verbs derived from roots may be assigned a variety of interpretations in different morpho- logical environments (binyanim), while verbs formed from existing words retain the semantics (and phonology) of the base word.

In what follows I shall first discuss the contents of Maltese roots, arguing that roots are semantically not fully specified (Sect. 3.1). Following that, Sect. 3.2 is concerned with the form-function relation of the binyanim, demonstrating that the characteriza- tion of the binyanim in terms of particular semantic roles (e.g., causative, reflexive) is untenable. Sect. 3.3 develops in detail the idea of regularity and irregularity in the bin- yan system in terms of two processes of word formation, paving the way for the quantitative and qualitative analysis of the database of templatic verbs in Ch. 4.

3.1 The ingredients of roots

“In Arabic as in all Semitic languages the consonants of the word carry the semantic load.” Versteegh’s (1997: 19) claim is more or less the general idea a student of Maltese gets when perusing most grammatical studies on Maltese. Mifsud (1995a: 34) and Borg & Mifsud (1999: 2) define the root as a “discontinuous morpheme of 3 or 4 consonants in fixed sequence, identifying the basic semantic area”. Looking at it from another angle, others hold that “words containing the same root consonants stem from one concept”

The protagonists: roots and patterns 55

(Akkademja tal-Malti 1992: 27, my ). Under this approach, it is the root that carries the core meaning of the word. I will refer to this view as the traditional ap- proach. As discussed above, a single root can build more than one word. Let us consider a few examples.

(1) √ktb C1vC2vC3 (v) kiteb ‘to write’

nC1vC2vC3 (v) inkiteb ‘to be written’

C1vC2C3a (n) kitba ‘writing’

C1vC2C2vvC3 (n) kittieb ‘writer’

C1C2vjjvC3 (n) ktejjeb ‘booklet’

(2) √lbs C1vC2vC3 (v) libes ‘to wear, dress’

C1vC2C2vC3 (v) libbes ‘to dress, v.t’

C1vC2C3a (n) libsa ‘dress’

In (1) above, all the words containing the root consonants [k], [t], [b] share a lexical core, having to do with some notion of writing. Similarly, for libes, libbes, and libsa in (2), the traditional claim would be that the root √lbs means something such as dressing or wearing. However, looking at a larger sample of Maltese data, it becomes difficult to relate certain words with a common root in a systematic fashion. Consider the root √xrb in (3) below. While one may posit that most of the words containing [ʃ], [r], [b] share a common semantic core related to some sort of contact with liquids, they differ consid- erably in meaning. The two verbs xorob and xarrab are derived from the same root, but denote two very different events, drinking and wetting, and cannot be (synchronically) derived from one another in any straightforward way. The same holds true for silet and issielet in (4), which refer to distinct events, extracting and struggling. They are not eas- ily reducible to a derivational relation, at least synchronically.

(3) √xrb C1vC2vC3 (v) xorob ‘to drink’

C1vC2C2vC3 (v) xarrab ‘to wet’

nC1(t)vC2vC3 (v) inxtorob ‘to be drunk/shrink’

C1vC2C3a (n) xarba ‘drink’

tvC1C2iiC3 (n) tixrib ‘wetting’

56 A Tale of Two Morphologies

(4) √slt C1vC2vC3 (v) silet ‘to extract’

tC1vvC2vC3 (v) issielet ‘to struggle’

C1vC2C3a (n) silta ‘extract’

mvC1C2vvC3a (n) misluta ‘earring’

In examining these examples, we are confronted with contrasting evidence. Some word families with the same three or four consonants, such as those in (1) and (2), share a common semantic core and are semantically related. This semantic relation, however, is hard to pin down in other groups of words containing the same consonantal root, such as those in (3) and (4). Moreover, in the latter families, some members ap- pear to be semantically related (e.g., xorob ‘drink’ and xarba ‘a drink’; silet ‘extract’ and silta ‘an extract’), but the meaning of some members is far apart from the meaning of other words with the same root (e.g., xorob ‘drink’, xarrab ‘wet’; silet ‘extract’, issielet ‘struggle’, and misluta ‘an earring’). One particular verb, inxtorob, seems to be a hybrid case of semantic closeness in the passive sense ‘be drunk’ and semantic distance in the sense ‘shrink’. This curious case is discussed in Sect. 3.3.2.1

3.1.1 Problems with the traditional approach

As the data in (5) and (6) below confirm, the main problem regarding the contents of the root is the following. It is very often the case that the meanings of words sharing the same root consonants are closely related on the one hand (e.g., daħal ‘enter’, daħla ‘en- trance’, daħħal ‘insert’; ħarreġ ‘train’, taħriġ ‘training’), and too distant or unrelated on the other (e.g., indaħal ‘interfere’, dħuli ‘friendly’; ħareġ ‘take/go out’, stħarreġ ‘investi- gate’).

(5) √dħl C1vC2vC3 (v) daħal ‘to enter’

C1vC2C2vC3 (v) daħħal ‘to enter, insert’

nC1vC2vC3 (v) indaħal ‘to interfere’

C1vC2C3a (n) daħla ‘entrance’

C1C2vvC3i (a) dħuli ‘friendly’

1 Further research is needed in this area to determine, in a systematic way, the semantic relatedness of words sharing the same root consonants, using Latent Semantic Analysis (cf. Landauer & Dumais 1997; Landauer et al. 1998; Moscoso et al. 2005; inter alia), and such psycholinguistic techniques as the cross- modal priming task (cf., e.g., Marslen-Wilson et al. 1994).

The protagonists: roots and patterns 57

(6) √ħrġ C1vC2vC3 (v) ħareġ ‘to take/go out’

C1vC2C2vC3 (v) ħarreġ ‘to train’

stC1vC2C2vC3 (v) stħarreġ ‘to investigate’

C1vC2C3a (n) ħarġa ‘outing’

tvC1C2iiC3 (n) taħriġ ‘training’

The semantic relations between members of one family are too loose for a deriva- tional account where one word is derived from the other, and at the same time too tight to admit an account where words are simply unrelated.2 Now, in most grammars of Maltese, the root generally comes with a meaning asso- ciated to it. Roots are like dictionary entries with fully specified meanings. The first verbal pattern is presented as the mamma ‘roughly, matrix’, the source from which the other verbal (and sometimes nominal) patterns are derived. In some sense, it is consid- ered as phonologically and semantically more elementary than the rest (cf. Sect. 3.2). Accordingly, it is common practice to take the (primary) meaning of binyan I as the meaning associated with the root. All other forms are then based on the meaning of the root in that binyan. In this way, we often think of the meaning of √ktb as ‘write’, √xrb as ‘drink’, and √dħl as ‘enter’. However, as the data in (3) to (6) show, this approach runs into a number of problems, as roots acquire numerous nominal and verbal interpreta- tions, which may or may not be related to the supposed lexical core. In this traditional view, if words that contain the same root consonants have dif- ferent meanings, this is because of polysemy or homonymy. Under the polysemy per- spective, different meanings are derived from the root in some rather metaphorical or figurative way. Departing from the view that the root √dħl means something such as ‘enter’ (cf. binyan I, daħal), to account for the ‘interfere’ meaning it acquires in the envi- ronment of binyan VII, indaħal, some linguists would argue that interfering means to enter in the middle of something that is not one’s concern. Likewise, because √ħrġ re- fers to a core lexical meaning of taking/going out (cf. binyan I, ħareġ), then to train someone, the interpretation it acquires in binyan II, ħarreġ, would mean causing some skill to come out of the trained entity. And again, the meaning ‘investigate’ in binyan X, stħarreġ, would be understood as the “s of facts or information.

2 It is important to note that here I am not taking into consideration cases that involve homophonous roots, such as √sfr1 sfar ‘turn yellow’, √sfr2 saffar ‘whistle’, √sfr3 siefer ‘travel’. Such roots are distinguished, fol- lowing Aquilina’s (1987-1990) dictionary, which points out different etymological sources for homo- phonous roots. They are generally the result of historical merges of different phonemes (cf. Aquilina 1970; Mifsud 2008). See also the brief discussion in Sect. 4.1.

58 A Tale of Two Morphologies

The homonymy approach is adopted by Borg (1981, 1988). Positing that root meaning can be homonymous, he explains the semantic distance between ħareġ ‘take/go out’, ħarreġ ‘train’ and stħarreġ ‘investigate’ as forms derived from two homo- nymic roots √ħrġ1 (ħareġ ‘take/go out’, inħareġ ‘be taken out’) and √ħrġ2 (ħarreġ ‘train’, tħarreġ ‘train oneself’, stħarreġ ‘investigate’). He takes the meaning ‘to investigate’ to be in some sense derived from the ‘train’ meaning, that is “to exercise (oneself) in relation to a matter” (Borg 1988: 241-242, my translation). Taking a similar approach, Camilleri (1990: 11) argues that ‘to investigate’ is an independent lexical meaning of √ħrġ3, as “it bears no semantic relation to any other verb derived from the [same] root” (my transla- tion). Presumably, we would need a fourth instance of the root, √ħrġ4, to account for the

interpretation it takes on in the nominal pattern C1vC2C2vvC3a, ħarrieġa ‘a stone pro- jecting from the wall’. It follows from this analysis that to explain the various interpretations of words containing the root consonants [d], [ħ], [l], we would need at least three homonymic roots: roughly, one meaning ‘enter’, a second one meaning ‘interfere’, and another one meaning ‘being friendly’. The homonymy approach leads inevitably to circularity, as a root is considered homonymous by virtue of the different meanings the words it ap- pears in acquire. I believe both treatments are influenced by a deterministic approach of the binyan system, where roots, specified with a full dictionary meaning (questioned in Sect. 3.1), interleave with patterns, which serve a number of fixed functions, such as causative, reflexive, passive (questioned in Sect. 3.2), in a semantically transparent way. While dictionaries and grammars present all word meanings as derived from the meaning of more basic elements leading down to the root, a large body of studies on other Semitic languages agrees that root-and-pattern morphology contains a significant element of opacity (cf., e.g., Ravid 1990). In what he terms the “root fallacy”, Barr (1961: 100) ar- gues against the idea of transparency in the verbal system of Hebrew; that is, the view that “there is a “root meaning” which is effective throughout all the variations given to the root by affixes and formative elements, and that therefore the “root meaning” can confidently be taken to be part of the actual semantic value of any word or form which

can be assigned to an identifiable root.” This theory is untenable and the claim that there is a degree of opacity is by now no longer controversial, at least in the treatment of such Semitic languages as Hebrew (cf. Arad 2005: 61). As a consequence of this, some linguists opt for a diametrically opposed view, where roots are devoid of any semantic content. It is only words that have meaning.

The protagonists: roots and patterns 59

The root is no more than a phonological tool for the formation of words, with no mean- ing whatsoever (cf., e.g., Ravid 1990 for Hebrew). In this proposal, which we may term as the opacity or lexicalization view, semantic relatedness of different words sharing the same root is understood as a mere tendency, which can possibly be accounted for on historical grounds. There is no systematic relation between form and meaning, and any lexical item has a meaning which is idiosyncratic or lexicalized and which cannot be related to its form. It goes without saying that the theory of semantically empty roots is also at odds with the evidence presented above. There is, to a certain extent, a correla- tion between the semantic relatedness of words and shared phonological material. As is evident from the data in (1)-(2), but also, in part, from the data in (3)-(6), there is meaning similarity across forms sharing the same root consonants. To sum up, it is equally undesirable to have roots with no semantic content what- soever and to have roots with clear semantic values, which are then inherited composi- tionally by the words they form. Both transparency and opacity are theoretical ex- tremes. In the discussion that follows, I argue that roots are neither semantically naked nor well-dressed. In order to explain the contrast between the semantic relatedness of words formed from a single root and their being semantically too far apart from other members of the same word family, I take an in-between path, which, to my mind, is the most direct analysis of the phenomenon.

3.1.2 Roots are underspecified

Underpinning this analysis are two assumptions which help unravel the puzzle of words containing one root being, at the same time, semantically close and semantically far apart. First, I consider, with Arad (2003a, 2003b, 2005), roots to be underspecified lexical and phonological cores. Second, in step with recent models of word formation, I assume that some words are derived from roots while others are derived from previ- ously formed words. Let us now look more closely at the semantic content of roots. In Sect. 3.3, I then deal with the second assumption. Taken on its own, the root is incomplete from a phonological point of view: [k], [t], [b] form a consonantal skeleton that is not pronounceable on its own. It is only when the root consonants are embedded in one or more patterns that they become a con- tinuous string. Similarly, the root is semantically not fully specified, its meaning is not well-defined. As its phonological shape is not yet determined and as it does not yet have a fixed semantic interpretation attached to it, the root may be incarnated in many dif-

60 A Tale of Two Morphologies

ferent ways, i.e. it may be cast into numerous nominal and verbal patterns, creating words with a specific phonological and semantic content. This is not to say that roots are semantically empty, but rather that, paraphrasing Arad, they are “potentialities”, which may be realized in a number of ways.3 And, in- deed, many Maltese tri- and quadri-consonantal roots acquire several interpretations when they fuse with different nominal and verbal patterns. One of the conclusions that come out of my examination of the Maltese data is that this ability of consonantal roots to create multiple verbs when appearing in several binyanim is a chief characteristic of Maltese consonantal roots (cf. Ch. 4).4 In this respect, some linguists (e.g., Arad 2005: 101; Harley 2009, 2011) draw a parallelism between Semitic roots and Latinate bound roots in English, such as √ceive, √pose or √fer, which create different verbs (conceive, perceive, receive or compose, pro- pose, suppose or confer, infer, refer) that are semantically far apart from one another, and are by no means predictable from the combination of the root and word-forming morphology. See, among others, the discussion in Aronoff (1976) and Baeskow (2006) on the status of bound roots of Latin origin in English. Consonantal roots are underspecified enough to take up meanings that differ con- siderably from one another, such as ‘drink’ (xorob) and ‘wet’ (xarrab), or as ‘take/go out’ (ħareġ), ‘train’ (ħarreġ) and ‘investigate’ (stħarreġ). However, there may be de- grees of underspecification, which are arbitrary, just like the linguistic sign itself. Al- though all roots are semantically underspecified, some roots can be more underspeci- fied than others. There is thus a difference between the degree to which roots such as √ktb and √xrb are underspecified. While the words containing the former root are all related to the activity of writing (kiteb ‘write’, inkiteb ‘be written’, kitba ‘writing’), the agent who performs it (kittieb ‘writer’), and its end result (ktieb ‘book’, ktejjeb ‘book- let’), the words that share the latter root differ greatly in meaning (cf. the examples in (3) above). The more underspecified the lexical core of the root is, the more likely it is to be assigned different interpretations in the environment of different patterns (cf. Arad 2005: Ch. 3). But what does it exactly mean to be semantically underspecified?

3 “Because roots are underspecified potentialities, the interpretations assigned to them in the environment of patterns are to a certain extent arbitrary. They are not predictable from the semantics of the root, nor are they compositionally computed by combining the denotation of the root with that of the pattern” (Arad 2005: 17). 4 While multiple verbs formed from a single consonantal root are typical of Hebrew and similar languages (cf. Arad 2003, 2005), they are uncommon in a language like English. In Maltese, this is typical of templatic verbs but not of concatenative verbs, where roots generally combine with only one verbalizing suffix (Cf. Ch. 5). The issue of language-specific properties of roots is briefly referred to in Sect. 3.2 & Sect. 3.3.

The protagonists: roots and patterns 61

Semantic underspecification The concept of underspecification, that is, the notion that certain features are not ex- pressed in a representation, turns up in different linguistic domains. It emerged in pho- nology but was adopted in semantics in the 1980’s (cf., e.g., Bierwisch 1982, 1988; Bierwisch & Lang 1989). In lexical phonology, underspecified representations have been proposed to account for, among others, variation of phonetic realizations in con- nected speech (cf. Kiparsky 1982; Steriade 1995; inter alia). For instance, the phoneme /n/ is realized in many different ways, depending on the onset of the following word: as a coronal in on duty, a dorsal in on call, or a labial in on board. According to the idea of underspecification, in English the nasal is underspecified in terms of place of articula- tion and voicing. That is, while /n/ can be realized in many different ways depending on the context it is in, the underlying representation of the phoneme is thought to be the same in all cases. At the lexical-semantic level, the notion of underspecification has been particularly fruitful in the discussion on ambiguity and the representation of lexical items with dif- ferent senses (polysemy) and different meanings (homonymy), cf. Bierwisch 1996, 1997), Blutner (1998a, b), inter alia. It is also used in the lexical semantic analysis of affixes. For instance, Lieber (2004) considers derivational suffixes like -er to be seman- tically underspecified (she uses the term “underdetermined”) in English in order to ac- count for the different interpretations, such as agent (writer), patient (loaner), instru- ment (opener), etc., associated with it. Similarly, von Heusinger & Schwarze (2002, 2006) avail themselves of the notion of semantic underspecification to explain the am- bivalent nature of the prefix s- in Italian denominal verbs of removal. The prefix may in fact refer both to the removal of the base noun in verbs like s-bucciare ‘peel’ (cf. buccia ‘peel’) as well as to the removal of another entity from the nominal base, as in s-tanare ‘cause to come out of the burrow’ (cf. tana ‘burrow’). In line with the model of two-level semantics, which posits a distinction between a level of meaning defined by grammar and a level of interpretation based upon conceptual knowledge, von Heusinger & Schwarze argue that the two subtypes of denominal verbs of removal have one under- specified semantic representation, which is then specified by conceptual categorization. Much work in this area has been carried out in computational linguistics and psy- cholinguistics, and a number of proposals have been made of what an underspecified lexical representation might look like (cf. Rayner & Duffy 1986; Copestake & Briscoe 1995; Pustejovsky 1995; Poesio 1996; Frisson & Pickering 2001; Bos 2004; Sereno et al.

62 A Tale of Two Morphologies

2006; Frisson 2009; Egg 2010a; inter alia). In the realm of psycholinguistics, the idea of underspecification was put forward to account for experimental data, especially eye- tracking research in homonymous and polysemous words (cf., e.g., Frisson 2009 for an overview). In Natural Language Processing, semantic underspecification is seen as an important technique to keep semantic representations tractable in spite of the prob- lems posed by ambiguity (cf., e.g., Egg 2010b). However, crucially, it is still not clear what an underspecified meaning actually is. A number of theoretical views have been proposed. Some consider the lexical rep- resentation to be semantically rich, as in Pustejovsky’s (1995) Generative Lexicon, where the lexical representation of a word like book includes a qualia structure that lists a number of essential attributes, such as the formal attribute that it is an object, the telic property that it is intended to be read, etc. Others assume a semantically impover- ished representation, as in Carston (2002) and Wiese (2004), where the semantic com- ponent of the lexical representation of a word is not much more than a “gateway” to conceptual space.5 Without going into details, some theories argue that, in essence, the meaning of a word is the meaning of that word in a context, implying that individual senses are not actually stored and looked up but are rather computed on the basis of a broader context (cf., e.g., Wilson & Carston 2007). What appears to be common to all these contrasting views, whether the lexical representation is taken to be rich or not, and whether the discrete senses are stored or constructed ad hoc, is the idea that what language users activate first is not a full-fledged interpretation of a word. Rather, some- thing more abstract is initially accessed and this serves, if need be, as the starting point for a more detailed specific interpretation (cf. Frisson 2009). In view of all these different approaches, the issue of lexical-semantic under- specification is still open to question. While it is not the intent of this study to resolve semantic underspecification, which must be the topic of future work, it seems promis- ing to resort to this notion to account for the wide semantic variation of words contain- ing the same radicals. To summarize the first assumption in this study, which draws heavily on Arad’s (2003a, 2003b, 2005) theory of the morphosyntactic makeup of the Semitic verb, we can make two related points. First, roots are phonologically and semantically under-

5 Often invoking Wittgenstein’s (1953) discussion of the word game and the inability to define it in such a way that it can refer to a game of poker, chess, etc., some linguists question the existence of abstract repre- sentations, as in many cases it is impossible to find a definition that covers all the possible senses of a word. Others reject the notion of a core meaning, as it would be impossible to account for idiosyncratic and un- predictable senses of a word (cf. Lehrer 1990; Rice 1992).

The protagonists: roots and patterns 63

specified. They are potentialities which are fully specified with respect to form, mean- ing and lexical category only when they are embodied in the various nominal and verbal patterns. Second, there is a degree of arbitrariness in the assignment of interpretation. Not only is the root an arbitrary linguistic Saussurean sign, but the interpretations as- sociated with that sign in different environments are also to a degree arbitrary (cf. Arad 2003a). With this picture of the semantic contents of roots in hand, we can now take a closer look at the roles of patterns. The next sections are concerned with the form- function relation of the verbal patterns in Maltese.

3.2 The functions of the pattern

The second protagonist of this tale is the pattern, which is the environment where the interpretation of the consonantal root is determined. As discussed above, roots are re- garded as phonological and lexical underspecified cores, which potentially may be in- carnated in many different ways. Patterns are the actual environments where these cores are incarnated, turning a root into a verb, a noun, or an adjective. It is in these verbal or nominal environments that roots take a particular phonological form and ac- quire a specific semantic interpretation. In Sect. 3.2.1, I discuss the appearance of a root in more than one binyan and the two distinct phenomena this results in, namely argument alternations and multiple in- terpretations. Following that, in Sect. 3.2.2, I give an overview of how binyanim are as- signed specific semantic roles in a number of Maltese grammars. In Sect. 3.2.3, I then move on to demonstrate that the relation between the semantic and syntactic proper- ties of verbs and their morphological form is not one-to-one.

3.2.1 When a root appears in various patterns

We have, so far, established two important conclusions regarding Maltese tri- and quadri-consonantal roots. First, these roots may be inserted in different binyanim. This property of Semitic roots to create multiple verbs is different from the way verb- creating roots behave in a language like English, where a root (√SIMPL, √WID) usually takes one and only one verbal morpheme (-ify, -en). In other words, if a root in English takes the verbal morpheme -ify, then it does not take -en as well, and vice-versa. The

64 A Tale of Two Morphologies

existence of simpl-ify and wid-en rules out forms like *simpl-en and *wid-ify.6 Maltese tri- or quadri-consonantal roots, by contrast, select for multiple verb-creating (and noun-creating) morphemes, i.e. binyanim. From this property of roots to appear in more than a single binyan follows our sec- ond conclusion. When a tri- and quadri-consonantal root combines with different bin- yanim, this combination may either result in two different interpretations of the root (recall, for instance, √ħrġ in ħareġ ‘take/go out’, ħarreġ ‘train’ and stħarreġ ‘investi- gate’), or – as I show below – it correlates with particular argument structure configura- tions.7 In the latter case, a root occurring in two different binyanim corresponds to a regular change in the structure of the arguments. The main alternations marked by Mal- tese templatic verbs include the active-passive, causative-inchoative, transitive- reflexive, and causative-noncausative, as illustrated in Table 3.1.8

Table 3.1 The main argument alternations in the Maltese binyan system

Active Passive

√ktb kiteb I, write inkiteb VII, be written √qsm qassam II, distribute tqassam V, be distributed √wġb wieġeb III, answer twieġeb VI, be answered

Causative Inchoative

√ftħ fetaħ I, open infetaħ VII, open √sħn saħħan II, warm up saħan I, warm up √wsgħ wessa’ II, widen twessa’ V, widen

Transitive Reflexive

√ħsl ħasel I, wash inħasel VII, wash oneself √skr sakkar II, lock up issakkar V, lock oneself in √bgħd biegħed III, distance tbiegħed VI, distance oneself

6 Two counterexamples to this generalization are the pairs (to) gas and gasify, and (to) winter and winter- ize, in which one item takes zero morphology, just like write, and the other takes -ify and -ize respectively (Arad 2005: 59). 7 As discussed in detail in Ch. 4, a root appearing in two or more binyanim may also yield synonymous verbs, e.g., √slf in binyan I silef and binyan II sellef both mean ‘to lend’. 8 In view of a number of differences between verbs alternating with causatives, it is useful to separate in- choative and noncausative verbs, even though the alternation itself is roughly the same (cf. Ch. 4 & Ch. 5).

The protagonists: roots and patterns 65

Causative Noncausative

√dħq daħħaq II, make laugh daħaq I, laugh √fwq fewwaq II, cause to burp tfewwaq V, burp √ħsd ħasad I, frighten inħasad VII, get a fright

Unlike roots which create verbs that are semantically distant from one another (cf. Sect. 3.1), the roots listed in the table above create verbs that constitute alternations related to the event structure of the verb. In these examples, two binyanim sharing a common root are alternants that express a difference in the event type, such as a change of state (saħan ‘warm up, v.i’) vs. an induced change of state (saħħan ‘warm up, v.t’), or an externally caused event (ħasel ‘wash’) vs. a self-induced event (inħasel ‘wash oneself’). Also, a difference in event structure is paralleled by a difference in argument structure. For instance, in causatives, an argument is added, that is the causer of the event. By contrast, the external argument is not projected in passives and is identified with the internal argument in reflexives.9 Note that there is a degree of syncretism in the binyan system. The same two bin- yanim may participate in more than one alternation.10 For instance, the pair wessa’– twessa’ takes part both in the causative-inchoative alternation, as in (7), and in the ac- tive-passive alternation, as in (8). This point is taken up again in Ch. 4.

(7a) Il=ħajjat-a wessgħ-et il=libsa.

DEF=tailor-F widen-PFV.3SG.F DEF=dress ‘The tailor let out the dress.’ (7b) Il=libsa t-wessgħ-et.

DEF=dress INCH-widen-PFV.3SG.F ‘The dress became loose.’

(8a) Se j-wessgħ-u t=triq.

FUT IPFV.3-widen-PL DEF=road ‘They are going to widen the road.’

9 See, among others, Hale & Keyser (1986), Levin (1993), Levin & Rappaport Hovav (1995), for a discussion of this kind of argument structure alternations. 10 By contrast, in some cases, there is, so to say, a division of labor among the binyanim. For instance, the passive of kisser ‘break, v.t’ is in binyan V, tkisser ‘be broken’, while its inchoative is in binyan VII, inkiser ‘break, v.i’. Likewise, ħeba ‘hide, v.t’ has a passive in binyan VII, inħeba ‘be hidden’, and a reflexive in binyan X, staħba ‘hide (oneself)’.

66 A Tale of Two Morphologies

(8b) It=triq t-wessgħ-et u reġgħ-et

DEF=road PASS-widen-PFV.3SG.F and do again-PFV.3SG.F in-għat-at it=tarmak.

PASS-give-PFV.3SG.F DEF=tarmac ‘The road has been widened and resurfaced.’

To sum up this part, then, we have established that roots appearing in more than one binyan may give rise to two distinct phenomena, which we can refer to as argument alternations (AA) and multiple interpretations (MI). One criterion, suggested by Arad (2005: Ch. 3), that is useful in teasing apart cases of AA and cases of MI is semantic en- tailment. In the case of ħasel ‘wash’ and inħasel ‘wash oneself’ or kiteb ‘write’ and inkiteb ‘be written’, the verb pairs share the same semantic entailments: washing oneself en- tails washing, and being written entails writing. On the other hand, in the case of MI, no semantic entailments hold between, say, xorob ‘drink’ and xarrab ‘wet’, or silet ‘extract’ and issielet ‘struggle’, cf. (3)-(4). Likewise, daħal ‘go in’ and indaħal ‘interfere’, cf. (5), although somehow ‘closer’ in meaning than these last two pairs, do not include the meaning of each other in the same way that inħasel entails the meaning of ħasel. The two phenomena, AA and MI, do not exclude one other. Consider the examples discussed in (3)-(6) above. In each group of words containing the same root conso- nants, there are instances of MI, where a single root creates verbs that are difficult to relate systematically. However, while ħareġ ‘take out’ and ħarreġ ‘train’ are a case of MI, ħareġ and its passive counterpart inħareġ ‘be taken out’ do semantically entail each other. The same holds for ħarreġ and its reflexive alternant tħarreġ ‘train (oneself)’. Analogously, √xrb involves instances of MI when cast in binyan I xorob ‘drink’ and II xarrab ‘wet’, and an instance of an AA, namely the alternation between the causative xarrab and the inchoative ixxarrab ‘get wet’.

3.2.2 What functions do patterns have?

When dealing with the binyanim, grammars devote many pages to discussions of the various primary and secondary functions or meanings that the patterns can have (e.g., Sutcliffe 1936: 81-101; Borg 1981: 5-12, 1988: 197-247; Cachia 1984). These functions are generally presented as specific interpretations the root acquires in the environment of a particular binyan. The causative or intensive for binyan II, the reflexive and/or pas- sive for binyan V, the inchoative for binyan IX, are some of the functions one commonly

The protagonists: roots and patterns 67

comes across in the literature. It is often the case that the functions of a given binyan are taken in relation to those of other binyanim. For instance, binyan V is thought of as the passive or reflexive of binyan II, which in turn is said to be the causative or inten- sive of binyan I. Table 3.2 summarizes the general properties of the nine tri-consonantal binyanim and their functional relations, as discussed in some Maltese grammars, namely Sutcliffe (1936), Borg (1981, 1988), Cachia (1984, 1994), Mifsud (1995a). Quadri-consonantal binyanim are usually glossed over. In Mifsud (1995a: 36) and Borg & Mifsud (1999: 5), QII is presented as the passive and/or reflexive of QI, which is said to have a “basic ac- tive meaning (transitive or intransitive).”

Table 3.2 Canonical functional relations between binyanim

Binyan Sutcliffe Borg Cachia Mifsud

I simple form inchoative primary basic active causative reflexive causative (perception verbs)

II intensive of I causative intensive of I intensive of I causative (intensive) initiative transitive of I transitive of I

III conative causative active transitive of I intensive causative

V reflexive of II passive causative reflexive-passive passive of II passive reflexive causative reciprocal reflexive of II effective reciprocal

VI reflexive of III passive causative reflexive-passive passive of III passive reflexive causative reciprocal reflexive of III effective reciprocal

VII reflexive of I passive causative passive-reflexive passive of I passive reflexive causative reflexive of I effective

68 A Tale of Two Morphologies

VIII reflexive of I causative active passive of I effective reflexive causative passive-reflexive reflexive of I passive reciprocal reciprocal

IX acquisition of inchoative acquisition of inchoative color or quality color or state

X in origin reflexive causative active transitive originally reflexive of IV or intransitive inchoative

To better understand the functions listed in Table 3.2, three terminological points are in order. First, Sutcliffe uses effective to refer to what I here term inchoative or anti- causative (cf. Ch. 5). The grammars that followed Sutcliffe generally ignore this observa- tion that anticausatives may be marked by binyan V, VI and VII verbs. Cremona (1962: 41ff.) calls them impersonal passives, verbs in which an external cause or agent is not relevant. Second, in his studies developed within the localist framework, Borg uses the labels passive causative and reflexive causative. The former basically refers to passives (of causatives). By reflexive causative he means both reflexives proper, such as staħba ‘hide oneself’, as well as verbs like mexa ‘walk’ that have a single argument with both the role of an initiator and an undergoer (cf. Ramchand 2008: 82). The third point concerns Cachia’s use of initiative, by which he means causative, “to induce someone to perform an action” (Cachia, 1994: 197, my translation). He uses other terms such as debilitative for marrad ‘infect’ and the like, and restorative for such verbs as fejjaq ‘heal’, which in essence are causative (Cachia 1984: 7-10, 1994: 203- 207). At least three principal observations come out of this characterization of binyanim as summarized in the above table. The traditional point of departure is that binyanim make up a system of transparent relations, which may be obscured by some exceptions to the rules. For instance, Cachia (1984: 20) argues that tkellem ‘speak’, being (in some contexts) transitive, is an exception, as binyan V verbs should have a passive, reflexive or reciprocal meaning. I demonstrate, in Sect. 3.2.3, that these meanings attributed to each binyanim are at best tendencies. They generally exist only within the most fre- quent forms, from which they are often unduly generalized.

The protagonists: roots and patterns 69

Another observation that emerges is that there is “one basic tri-consonantal form with nine [rather eight] derived forms” (Borg & Mifsud 1999: 3). The way binyanim are presented in grammars suggests that binyan I is the lexical prime, the phonologically and semantically default form. The other binyanim are derived from it through such operations as vowel lengthening, gemination of radicals, and the addition of the so- called “servile consonants”. However, a large number of roots do not appear in binyan I (cf. Appendix I), and some roots appear in only one pattern which is not binyan I, such as tbissem ‘smile’ (√bsm, V), issara ‘wrestle’ (√srj, VI), infexx ‘burst into, erupt in’ (√fxx, VII), smar ‘get tanned’ (√smr, IX), etc. The derivational view is so weighty and pervasive that, even in these cases, grammars and dictionaries would nevertheless take binyan I as the source, either by positing hypothetical forms, “[w]ords coined by grammatical analogy of word-pattern” (Aquilina 1987-1990: xv), or by presuming that these base forms were actually present at some earlier stage in the development of the language (e.g., Cremona 1973: 97; Cachia 1994: 203). The third observation is that the binyanim may be characterized not only semanti- cally (causative, reflexive, etc.) but also syntactically in terms of transitivity. Having ex- amined a comprehensive database of templatic verbs in Maltese, we can say that binyan VI, VII, and IX are inherently intransitive, II and III are predominantly transitive, and V is primarily intransitive. By contrast, binyan I, VIII and X may host either transitive or intransitive verbs. Borg (1981, 1988) and Borg & Mifsud (1999) show that the transitiv- ity of binyanim correlates with their canonical functions, yielding a series of interde- pendencies of the sort: the interpretation of binyan II is causative (e.g., saħħan ‘warm up, v.t’) when the corresponding binyan I is intransitive (saħan ‘warm up, v.i’), and in- tensive (e.g., ġebbed ‘pull again and again’) when binyan I is transitive (ġibed ‘pull’). This issue, which was dealt with in detail in Borg’s (1981, 1988) “semantic classification of the forms [i.e. the binyanim]”, is taken up again in Ch. 4. The key points discussed in this section can be summed up as follows. The binyan system in Maltese is presented in various grammars as a transparent system, where each binyan is associated with semantic roles such as causative, reflexive, etc. The bin- yanim can also be distinguished in terms of transitivity, and this syntactic characteriza- tion determines which function a given binyan may acquire. There is a widespread ten- dency to take binyan I as the most basic form and to treat it as the source from which the other binyanim are derived. In the following section, I bring into question the neat association of canonical functions with particular binyanim.

70 A Tale of Two Morphologies

3.2.3 Do patterns have functions at all?

So far, I have shown that the binyan system in Maltese accommodates both regularly formed argument alternations and irregularly created multiple interpretations. If we had to focus on the regular aspect of the system, AA, can we establish a direct correla- tion between semantic and syntactic features of verbs and their morphological form? In other words, can we characterize the binyanim in semantic terms, as is conventionally done in grammars? This is the question that I seek to answer in this section. It is tempting, if not necessary, for linguists to seek a compositional analysis for a given grammatical contrast with semantic import. For verbal patterns, such an analysis would assign to each binyan-marking operation a distinct function. In line with this, for the Hebrew verbal system, Arad (2005: 107-108) notes:

Given that there are several different verbal patterns (binyanim), it would obviously be desirable to find certain regularities between the morphological form of a verb—binyan—and its syntactic and semantic properties. Otherwise, had there been a complete irregularity in the Hebrew verbal system, speakers would have had to learn separately each combination of a root with a binyan, and this, while not impossible, is also theoretically undesirable, at least from the acquisition standpoint.

However, looking at the data in Table 3.1, it becomes obvious that the contribution of each verbalizing morpheme or binyan is not semantically transparent. That is, for each combination of a root and a pattern, we cannot predict what kind of event is denoted by the verb. The binyan system, therefore, is far from being transparent or deterministic from two different angles. First, there is no exclusive mapping between the semantic and syntactic property of verbs and their morphological realization. As a matter of fact, in Table 3.1, reflexives appear in binyan V, VI, and VII, causatives appear in I and II, pas- sives appear in V, VI, and VII, and inchoatives appear in I, V, and VII. This means that the same verb alternation may be marked by different verbal patterns. Second, several verbs of certain semantic and syntactic types may select for differ- ent binyanim. In other words, the same binyan is used to mark both causatives and noncausatives or inchoatives, as is the case, for instance, with binyan I (cf., e.g., causa- tive fetaħ ‘open’ with inchoative saħan ‘warm up’). Moreover, one argument alternation is realized in more than one pair of patterns. For instance, the active-passive alternation may be marked, among others, by:

The protagonists: roots and patterns 71

binyan II and V ħassar ‘cancel’ tħassar ‘be cancelled’ binyan I and VII bagħat ‘send’ intbagħat ‘be sent’ binyan III and VI nieda ‘launch’ tnieda ‘be launched’.

This means that in the binyan system there is a degree of syncretism, with some pat- terns displaying more syncretism than others (cf. Ch. 4). In a nutshell, there is no direct mapping from verbal semantic properties to binyan forms. On the one hand, the same binyan may host verbs of different semantic types (with the exception of binyan IX, which is associated almost exclusively with in- choatives). On the other, verbs of the same semantic type (e.g., causative) may appear in different binyanim (e.g., I, II, III). While rejecting the correlation, so neatly presented in grammars, between certain binyanim and specific interpretations, one cannot fail to acknowledge that there are pockets of regularity in the binyan system, as has already been observed in Borg (1981, 1988). In Ch. 4, I show that argument alternations are to a large extent morphologically regular and quantify the regular and irregular share in the system. Before, however, I introduce the second assumption underpinning this work, which helps predict and explain the regularity and irregularity found in Maltese tem- platic verbs.

3.3 Root derivation and word derivation

One central argument put forward in this chapter is that the binyan system in Maltese is neither perfectly transparent nor completely opaque. When looking at a large body of data, it turns out that, while the semantic relations between certain binyanim created from the same root can be predicted, there is a great deal of variation and idiosyncrasy in the system. Interleaving a root with different binyanim may result either in regular changes in the semantic and syntactic configuration of the verbs they form, or in verbs that are semantically unpredictable and too distant from each other to allow for a deri- vational account, even if they share some phonological material, e.g., the consonantal root. To account for these regular and irregular aspects within the binyan system, I posit, following Marantz (1997, 2001, 2007) and Arad (2003a, 2003b, 2005), a distinction in the syntactic composition of words, which in several respects echoes the distinction found in traditional lexicalist theories of morphosyntax between lexical and syntactic word formation (cf. Siegel 1974; Wasow 1977; Kiparsky 1982; Spencer 1991; Giegerich 1999; inter alia). Words formed in the lexicon are more prone to phonological and se-

72 A Tale of Two Morphologies

mantic irregularities; they often receive an idiosyncratic interpretation and lexical pho- nology may change their form. Syntactic word formation, by contrast, is regular both morphonologically and with respect to the semantic interpretation of words. In Marantz and Arad’s model, however, it is not assumed that there are two places where words can be constructed. The basic idea is that all morphological composition takes place in syntax. The basic building blocks in syntax are the roots, atomic, non- decomposable elements that are not yet categorized as noun (n), verb (v) or adjective (a). In this syntactic approach to word formation, known as Distributed Morphology,11 words can be built by combining a category-neutral root with a syntactic head, thus turning the root into a n, v or a, but also by combining an already constructed word with a new syntactic head. Figures 3.1 and 3.2 capture this contrast between word formation from roots and word formation from previously formed words.

x

x x n, v, a

√ n, v, a √ n, v, a

Figure 3.1 Root-derived word Figure 3.2 Word-derived word

As these representations illustrate, a category head x may merge either with a root or an existing noun, verb or adjective. This introduces a locality constraint on the interpre- tation of roots, which Arad (2005: 247) explains as follows, “[r]oots are assigned an in- terpretation in the environment of the first category-assigning head with which they

11 The central claim of this theoretical framework is that there is no unified Lexicon as understood in ear- lier generative treatments of word formation, i.e. as a place for the creation of words and the storage of special meanings assigned to words. Rather, the functions of what is traditionally taken as the Lexicon are “distributed” over three components that come into play at different times in the process of syntactic computation. The first component is an inventory of atomic building blocks, containing roots and features, from which all other complex elements are built. The vocabulary, the second component, provides the phonological forms, which are inserted post-syntactically, in a process called Vocabulary Insertion. At the interface with our knowledge of the world is the third component, the Encyclopedia, which lists the variety of contexts each root appears in, enabling the output of the syntax to be interpreted semantically (cf. Halle & Marantz 1993; Marantz 1997, 2000, 2007; Embick 2000; Arad 2003a, 2003b; Embick & Marantz 2008; inter alia).

The protagonists: roots and patterns 73

are merged. Once this interpretation is assigned, it is carried along throughout the deri- vation”. In other words, while root-derived verbs may pick numerous interpretations in different environments, word-derived verbs are tied to the meaning of the verb, noun or adjective from which they are derived. They have no access to the root, but only to the words they are derived from (cf. Marantz 1997; Embick 2008; Alexiadou 2009; inter alia). In order to motivate the generalization that root-derived words are more likely to display morphonological irregularities and semantic idiosyncrasies than word-derived words, Arad (2005: Ch. 7) draws a connection with the Chomskyan condition on phase impenetrability (Chomsky 2001, 2004, 2005), even though she does not fully discuss the implications of linking the generalization to phases. The main idea is that the com- bination of a root with a category-assigning head defines a phase, and phases are gen- erally considered to be impenetrable once they have been spelled out. In short, the form and meaning of a root-derived word are set immediately after it is formed by the spell- ing-out of the phase. Any further (word) derivations are compositional and do not have access to the root, but only to the morphonological and semantic interpretation of the base form they are derived from (cf. Marantz 2007). This connection to phases has be- come something of an Achilles’ heel for the Marantz-Arad approach because, as dis- cussed in Sect. 3.3.2, it makes predictions that turn out to be somewhat restrictive. In addition, word formation from roots is quite often non-productive, as the combi- nation of a root with a particular category-defining head but not with others is to a large extent arbitrary. Recall the examples from English in Sect. 2.2.2: the root √RELIG arbitrarily combines with the adjectivalizing head -ious (relig-ious), but not with -y

(*relig-y), and √STICK arbitrarily combines with the head -y (stick-y), but not with -ious (*stick-ious). It is therefore expected that word derivation, in contrast to root deriva- tion, displays no or, at any rate, very few gaps. The main differences between root-derived and word-derived words are listed in Table 3.3 (cf. Marantz 2001, 2007). From this follows the second general assumption in this study. The prediction is that root-derived words are more prone to phonological and semantic irregularities (special phonological processes, idiosyncratic and idiomatic meanings, etc.), while word-derived words are morphonologically regular and have semantically predictable meanings.

74 A Tale of Two Morphologies

Table 3.3 Key differences between root derivation and word derivation

Root-derived words Word-derived words

merger with a category-assigning head merger above a category-bearing head idiosyncratic, idiomatic meanings predictable meanings apparent semi-productivity apparent complete productivity independent of argument structure possible operations on argument structure

The Marantz-Arad model makes the correct predictions with respect to word for- mation processes in a number of languages, such as zero derivation in English and Dutch, denominal verbs (as well as other verbs) in Hebrew, and plural formation in . I now discuss each one of these cases (Sect. 3.3.1), and then turn to an exami- nation of the phenomenon in the binyan system in Maltese, motivating the use of the framework in this study (Sect. 3.3.2). Finally, in Sect. 3.3.3, I tackle one weak point of Arad’s analysis, concerning the impenetrability of the root and categorizing head.

3.3.1 Cross-linguistic evidence

Arad (2005) argues that the locality constraint introduced above is a universal prop- erty. In other words, (i) the ability to be assigned multiple interpretations is strictly re- served to roots, and (ii) the assumption that once the root merges with a category head and forms a word, its interpretation is fixed and is carried along throughout the deriva- tion, constitute a constraint that holds across all languages. On the other hand, the abil- ity of roots to be assigned a variety of interpretations in different morphological envi- ronments is language specific. For instance, while the roots in a language like Hebrew typically appear in various verbal (and nominal) environments, the roots in a language like English generally appear in only one verbal environment. In view of the claim that the locality constraint on the interpretation of root-derived and word-derived words is a universal property of languages, in this section, I review some studies which put to the test the syntactic distinction between root derivation and word derivation in four languages from two typologically distinct families. I first con- sider studies on zero derivation in two , English and Dutch. Follow- ing that, I turn to studies on Semitic languages, namely plurals in Amharic and denomi- nal verbs in Hebrew.

The protagonists: roots and patterns 75

Zero derivation Zero derivation or conversion, that is items which can function as different grammatical categories, is a good testing ground in which to start a discussion on category-less roots that receive their grammatical category when inserted in syntax. A typical example is verb-noun pairs like searchV and searchN or paintV and paintN. In such cases, it is as- sumed that the roots √SEARCH and √PAINT are combined with a verbal or nominal functional node, depending on the syntactic environments in which a verb or a noun is required. The relation between such pairs thus consists in their common underlying root, with their separate categories taking shape in syntax. However, in a number of studies it is argued that not all verb-noun conversion pairs constitute root level derivations. Rather, one form is more basic than the other; that is, they are either nouns derived from verbs or verbs derived from nouns. Building on ob- servations by Kiparsky (1982), Myers (1984), and others, Arad (2003a, 2005) distin- guishes between root-derived and noun-derived verbs on the basis of stress assignment and semantic relations in English zero-related pairs. The generalization is that when the verb and the noun share phonological properties, such as the same stress pattern, as in cómmentN,V, dísciplineN,V and cóntactN,V, there tends to be a tight semantic relation be- tween the two members of the pair. By contrast, in pairs like óbjectN and objéctV, récordN and recórdV, cóntestN and contéstV, not only is stress assignment different in the verbs and the nouns, but they also stand in a looser semantic relation to each other. In light of the assumptions made above, the conclusion is that the latter cases are root-derived while the former are word-derived. In short, being derived from a common root, pairs like óbjectN and objéctV have the typical nominal and verbal stress in English and distant semantics. Pairs like cómmentN,V, on the other hand, are noun-derived because they have the same stress properties (the typical nominal stress in English) and the semantic relation between the two members is clear. A similar directional approach to verb-noun conversion pairs is taken by Don (2004, 2005) and Kraaikamp (2008) for Dutch (cf. also Acquaviva 2009). Two proto- typical properties of nouns and verbs in Dutch are gender and conjugational class.

Nouns are either neuter, as in het huis ‘the.NEUT house’, or non-neuter, as in de weg

‘the.NON-NEUT road’. Verbs are either regular, if they are stem-invariant, or irregular, if they display stem vowel alternation. Compare, for example, the regular present–past– participle triplet of ‘to work’, werk – werk-te – ge-werk-t, with the irregular triplet of ‘to walk’, loop – liep – gelopen.

76 A Tale of Two Morphologies

Now, it has been shown that, in the case of zero derivation, the verb is regular if it is derived from a noun, and the gender of the noun is non-neuter when it is derived from a verb. Considering this, the assumption, then, is that irregular verbs and neuter nouns, being root-derived, have more or less deviant semantic interpretations as opposed to word derivations. This prediction is in fact borne out. To take an example, consider the noun-verb pair prijs ‘price’ and prijzen, which has two meanings: a denominal interpretation ‘put a price on’, and the meaning ‘to praise’, which cannot be considered to be transparently derived from the noun. As expected, when used in the denominal meaning, the verb is regular, but when used in the ‘to praise’ meaning, it is an irregular verb (cf. Kraaikamp 2008: 12-13). This is taken as evidence that prijzen ‘to praise’ is a root derivation, whereas prijzen ‘put a price on’ is derived above the root level, i.e. it is derived from the noun prijs ‘price’, as it inherits both the semantics and phonology of the base noun. Thus, as in the case of English zero- related pairs, morphonological regularity seems to go hand-in-hand with a semantics that suggests a derivational relation with the word, rather than with the root.12

Plural nouns and denominal verbs Let us now turn to data from two Semitic languages, starting with Amharic plurals. Am- haric has a regular and an irregular way to form the plural of nouns. The regular forma- tion involves attaching the suffix -očč (or one of its allomorphs, generally conditioned

by nouns ending in a vowel) to a noun like ‘house’, giving bet-očč ‘house-PL’. Many nouns, however, take the so-called irregular plural markers, which may be external, namely the suffixes -an, -at or -t, or internal, i.e. ablaut and a change in the prosodic template (cf. Leslau 1995). Once again, irregular morphology runs parallel to unpre- dictable semantics, as the examples in (9) demonstrate.

(9) wär ‘month’ wär-at ‘season’ näfs ‘soul’ näfs-at ‘small insects’

12 However, this does not mean that any morphonological irregularity corresponds to semantic idiosyn- crasy. Take, for instance, verb inflection in English: the interpretation of an irregular past tense form like broke is just as predictable as that of a regular verb like braked. Such examples seem to rest uncomfortably with the assumption in Marantz (2007) that the domain for morphonological irregularity is the same as the domain for semantic idiosyncrasy. Distributed Morphology generally accounts for this by invoking ‘Read- justment rules’ to change break to broke, etc. Embick (2010) deals with this mismatch between morpho- nological irregularity and predictable interpretations by delaying the phonological spell-out of morpho- syntactic features until the next cycle head is merged. See also the discussion in Svenonius (2011).

The protagonists: roots and patterns 77

There seems to be a lot of variation in the semantic relations between the singular and the (root-derived) irregular plural forms. The relation may be reasonably straightfor- ward, as in a set of months making up a season, or more tenuous, as in souls and small insects. In addition, according to Kramer (2009: Ch. 4), the (word-derived) regular forms are very productive; that is, there are no paradigmatic gaps in regular plural for- mation. “All nominals in Amharic may be regularly pluralized, even if they also have an irregular plural” (Kramer 2009: 145). By contrast, irregular plurals are non-productive and have paradigmatic gaps, which suggests that they are formed by the combination of a root with n[+PL]. Finally, let us consider Hebrew denominal verbs, which take us back to binyan morphology. Arad (2003a: 745-747, 2003b: 89-90, 2005: 245-247) shows how a root like √sgr interleaves with various verbal and nominal patterns, some of which are listed in (10).

(10) √sgr C1aC2aC3 (v) sagar ‘to close’

hiC1C2iC3 (v) hisgir ‘to extradite’

C1oC2C3ayim (n) sograyim ‘parentheses’

miC1C2eC3et (n) misgeret ‘frame’

CiC1C2eC3 (v) misger ‘to frame’

Particularly interesting for our discussion is the verb at the end, misger, which bears not only morphonological similarity to the noun misgeret (i.e. the typically nominal pre- fix m-, apart from the same root consonants), but also shares an interpretation with the noun from which it is derived. Arad (2003b: 90) notes that “[i]t may seem natural or even trivial that the verb made from the noun frame means to frame. But this, I argue, is a crucial property of noun-derived verbs.” While the root-derived verbs sagar and his- gir are assigned different, hardly related interpretations, misger, which is word-derived, as suggested by the presence of the prefix m- that is carried over from the nominal pat- tern miC1C2eC3et into the verbal form, is semantically tied to the noun it is derived from. On the basis of this data, Arad (2003b: 90) concludes that:

Although the verb misger contains the consonants of the root √sgr, it cannot have access to the underspecified core meaning of the root, or to all the interpretations assigned to that root in dif- ferent environments: something seems to interfere between the verb misger and the root √sgr. This interfering element, I argue, is the noun misgeret.

With this cross-linguistic evidence in hand, we can now examine whether the Aradian predictions also hold for the Maltese verbal system. This is the topic of the next section.

78 A Tale of Two Morphologies

3.3.2 Word formation in Maltese

In this section, I demonstrate that the transparent (regular meaning associations, phonological similarity, and full productivity) and opaque aspects (idiosyncratic, un- predictable meanings, phonological irregularity, and gaps) of the binyan system in Mal- tese can be explained by drawing a distinction between word formation from roots and non-roots, along the lines of Marantz (2001, 2007) and Arad (2003a, 2003b, 2005). More specifically, on the basis of morphonological and semantic evidence, I argue that argument alternations (namely the active-passive and transitive-reflexive alternations) are word-derived, whereas all other instances found in the binyan system, including the phenomenon of multiple interpretations (cf. Sect. 3.2.1), roots which derive singleton verbs, and roots which derive verbs with identical meanings (cf. Ch. 4), are root deriva- tions. The status of the causative-inchoative/noncausative alternation with respect to root derivation vs. word derivation is discussed in Ch. 4 and Ch. 5. Consider, to begin with, the data in (3) above, reproduced here as (11), which pro- vide an example of a root creating different verbs and nouns, some of which are semantically close, such as xorob ‘drink’, inxtorob ‘be drunk’, xarba ‘a drink’, and others which are not, such as xorob ‘drink’ and xarrab ‘wet’. (I will, for the time being, pass over the secondary meaning of inxtorob ‘shrink’. The problems posed by this and similar verbs are discussed in some detail below).

(11) √xrb C1vC2vC3 (v) xorob ‘to drink’

C1vC2C2vC3 (v) xarrab ‘to wet’

nC1(t)vC2vC3 (v) inxtorob ‘to be drunk/shrink’

C1vC2C3a (n) xarba ‘drink’

tvC1C2iiC3 (n) tixrib ‘wetting’

Interestingly, xorob ‘drink’ and xarrab ‘wet’ are not only semantically distant, but also have a different vowel pattern (cf. Ch. 4 for discussion on vowels). By contrast, the vow- els of xorob ‘drink’ and inxtorob ‘be drunk’ are the same and the semantic relation be- tween the two is straightforward: they mark a regular change in the structure of the arguments, namely the active-passive alternation, as the sentences in (12) illustrate.

(12a) It=tifel xorob l=ilma.

DEF=boy drink.PFV.3SG.M DEF=water ‘The boy drank the water.’

The protagonists: roots and patterns 79

(12b) Matul ir=Ramadan ix=xarba ta-l=ħarrub

during DEF=Ramadan DEF=drink of-DEF= t-in-xtorob minn ħafna Musulman-i 13

IPFV.3SG.F-PASS-drink from many Muslism-PL ‘The carob drink is drunk by many Muslims during Ramadan.’

In light of the semantic relations that hold between verbs containing the radicals [ʃ], [r], [b] and their vowel patterns, I claim that xorob and xarrab are root derivations, whereas inxtorob is derived from xorob. I assume that the formation of inxtorob is as follows: first, the radicals of √xrb are combined with the verbalizing head, morphologi- cally spelled out as C1vC2vC3 (Figure 3.3). The verb xorob is then embedded under an- other verbalizing head, nC1vC2vC3 (Figure 3.4). As discussed in Sect. 2.2.2, the prefix n- has a phonologically conditioned allomorph nt- and, since the first root consonant is a sibilant, [ʃ] undergoes metathesis with prefixal t-, resulting in in-x-torob (for *int-xorob).

Vnxtorob

Vxorob Vxorob v, nC1vC2vC3

√xrb v, C1vC2vC3 √xrb v, C1vC2vC3

Figure 3.3 Root-derived xorob Figure 3.4 Word-derived inxtorob

While word-derived verbs retain the vowel patterns of the verbs they are derived from, root-derived verbs may or may not have the same sequence of vowels. Recall, for in- stance, the root √ħrġ and the examples in (6) above: even though ħareġ ‘take out’, ħarreġ ‘train’ and stħarreġ ‘investigate’ share the same vowel sequence a-e, they are clearly cases of root derivations, as their semantic relations are tenuous to say the least. On the other hand, the binyan VII verb inħareġ ‘be taken out’ and the binyan V tħarreġ ‘train oneself’, I argue, are derived from ħareġ and ħarreġ respectively. Having taken the phonology and semantics of ħareġ as its input, inħareġ bears phonological similarity to the binyan I verb and shares an interpretation with it. Basically, it is the passive coun-

13 The example is taken from the newspaper ‘Il-ġabra tal-miżwed’, Illum, 13 September 2009.

80 A Tale of Two Morphologies

terpart of ħareġ ‘take out’. Likewise, tħarreġ is derived above the root level and receives a phonological form (note the gemination of the second radical) and a semantic inter- pretation based upon the binyan II verb, i.e. it is the reflexive of ħarreġ ‘train’. Particularly significant in this respect is the fact that root-derived verbs like ħareġ are typically more polysemous than verb-derived verbs like inħareġ. While ħareġ can mean ‘to go out, take out, publish/issue, turn out, defray, break out, etc.’, its passive counterpart can mean ‘to be taken out, be published/issued, be defrayed’. Note that an active reading of each of these three meanings is (already) included in the base it is de- rived from, supporting the assumption that word-derived words inherit interpretations that have been assigned above the root level. To take another example, the binyan II verb niġġeż, which is directly derived from the root √ngż, means ‘to prick, hurt/upset, administer an injection, hook a fish which snaps off, etc.’ By contrast, binyan V tniggeż, being verb-derived, is tied to the interpre- tations of niggeż, and receives either a passive ‘to be pricked’ or a reflexive ‘to prick oneself’ interpretation. These and many other cases where root derivations, unlike word derivations, are highly polysemous suggest that verb-derived verbs take as their input items whose meaning has already been fixed at the first merging of a root with a category-assigning head.14 As a side note, we can put the syntactic distinction in word formation to test also on some concatenative verbs and nouns in Maltese. Consider the triplets in (13).

(13a) iwweldja ‘to weld’ iwweldjar ‘the act of welding’ welding ‘soldering alloy, the act of welding’

(13b) iċċita ‘to quote, cite’ iċċitar ‘the act of quoting, citing’ ċitazzjoni ‘citation, summons, fine’

(13c) aġixxa ‘to act, behave’ aġir ‘way of acting, behaving’ azzjoni ‘action, behavior, share, (religious) action’

14 In a discussion on affixal polysemy (i.e. that an affix like -ize can sometimes mean ‘cause to become x’, e.g., privatize, sometimes ‘cause to go into x’, e.g., containerize, sometimes ‘perform x’, e.g., anthropologize, etc.), Lieber (2004: 43) relates polysemy to semantic underspecification, claiming: “One reason that affixes tend to be highly polysemous is that their actual semantic content is vastly abstract and underdetermined.” This fits neatly with our analysis that consonantal roots are underspecified and words directly derived from roots tend to be more polysemous than words derived from (semantically and phonologically speci- fied) words.

The protagonists: roots and patterns 81

It appears that the difference between the -ar/-ir suffixed nouns and the -ing/-zzjoni suffixed nouns can be reduced to a distinction between word derivation (the -ar/-ir cases) and direct root derivation (the -ing and -zzjoni cases). Thus, iwweldjar, iċċitar, and aġir are taken to be verbal nouns because they share interpretations (roughly, ‘an act of verb-ing’) and morphonological properties (such as initial gemination, the verbal suffix -ja, the /dʒ/ segment in the stem of aġ-ixxa) with the verbs (they are derived from). By contrast, welding, ċitazzjoni and azzjoni must be root-derived nouns for two reasons: (i) they do not inherit formal elements, such as initial gemination and verbal suffixes, from the verb (cf. welding vs. iwweldjar), and (ii) they take on interpretations such as ‘soldering alloy’, ‘fine’, ‘share’ respectively, which are not strictly speaking in- cluded in the relative verbs. This difference in word formation is illustrated below.

Nwweldjar

Nwelding Vwweldja n, -ar

√WEL D n, -ing √WELD v, -ja

Figure 3.5 Root-derived welding Figure 3.6 Word-derived iwweldjar

For a short discussion on the interpretations of concatenative noun pairs suffixed with -ar and -ing, see Mifsud (1995a: 249-250) and Borg & Azzopardi-Alexander (1997: 252). Further research on in Maltese is required to resolve, among other things, issues related to their semantics and morphonology, in particular initial gemination, which seems to be optional, at least in some registers, for some -ar suffixed nouns (e.g., the heading Applikazzjoni retroattiva ta’ politika ġdida dwar iċ-ċitar ‘Retro- active application of a new citation policy’).15

15 Retreived from http://www.ombudsman.europa.eu/en/cases/summary.faces/mt/4989/html.book mark;jsessionid=3413 BFD8AB0C5505FEBF08BC44B0A7B6

82 A Tale of Two Morphologies

Counterexamples? One potential counterexample to this generalization is inxtorob. If, as I have argued above, it is a case of word derivation, then, in line with the Aradian predictions, there should be no interpretative leeway, and the semantic relation between the two words, xorob and inxtorob, must be straightforward. Since the grammar has already assigned a meaning to the verb xorob ‘drink, absorb’, that meaning must be the base of the mean- ing for the corresponding passive verb. However, inxtorob not only has the predicted compositional meaning ‘be drunk, be absorbed’ but can also mean ‘shrink, become emaciated’. It therefore has two idiosyncratic interpretations that do not build on the semantics of xorob, which can mean ‘to drink, absorb, take in, etc.’, but crucially none of its meanings are transparently associated with the notions of shrinking and emaciating. This poses a problem to Arad’s proposal that phases, once spelled out, become im- penetrable. Put differently, if a verb is created from a previously formed verb, it has no access to the closed domain defined by the verbal head. Word-derived verbs, therefore, cannot, so to say, look inside the underspecified core meaning of the root (and take on one of its numerous interpretations), but can only see the semantic interpretation of the unit that has been fixed in that particular verbal environment. How come, then, inxtorob takes on interpretations (‘shrink, become emaciated’) that are not available in the verb it is supposedly derived from? Along the same lines, how is it possible that tkabbar ‘be grown, enlarged’, which is derived from the verb kabbar ‘grow, enlarge’ (as the phonology and semantics suggest), can also mean ‘be- come proud’? This interpretation clearly forms part of the underspecified meaning of the root √kbr, as it becomes apparent, for instance, in the environment of the adjectival

pattern C1C2vvC3i, kburi ‘proud’. However, it is not one of the interpretations that the root is assigned in the environment of binyan II, kabbar ‘grow, enlarge, increase, con- sider one older than one really is, etc.’16 There are (at least) two possible explanations to this apparent incongruence. We can, on the one hand, argue that similar surface forms may have different representa- tions; that is, that inxtorob ‘be drunk, be absorbed’ and tkabbar ‘be grown, be enlarged’ are word-derived because they mark a regular argument alternation (active-passive) whereas inxtorob ‘shrink, become emaciated’ and tkabbar ‘become proud’ are root- derived, as their meanings are idiosyncratic. Otherwise, we can try to explain non-

16 Aquilina (1987-1990: 647) lists the past participle imkabbar ‘proud’ under kabbar. However, according to the semantics of the verbs, it must be the past participle of tkabbar. Note that in many cases the past participles of binyan II and binyan V have the same form (cf., e.g., Mifsud 1995a: 70).

The protagonists: roots and patterns 83

predictable meanings in word derivations by questioning the impenetrability of the root plus categorizing head. Let us look at the two explanations in some more detail. The idea that similar surface forms can have distinct structures was introduced in- directly in Sect. 3.3.1, in the discussion on zero-related pairs. Recall the Dutch verb pri- jzen, which is root-derived in the meaning ‘praise’ and noun-derived in the meaning ‘put a price on’. The arguments Kraaikamp (2008) provides in support of this conclu- sion are based on the phonological and semantic regularity of the noun-derived prijzen ‘put a price on’, which is a regular, stem-invariant verb with a transparent denominal interpretation, in stark contrast to the morphonologically irregular and semantically idiosyncratic root-derived prijzen ‘praise’. Embick & Marantz (2008) reach the same conclusion, that an adjective like formal has two distinct structures, one derived from a noun, denoting the property of ‘having form’ (Figure 3.6), the other derived from a root, used in a mathematical sense (Figure 3.5), cf. Borer (2009a).17

Aformal

Aformal Nform a, -al

√FORM a, -al √FORM n, -ø

Figure 3.5 Root adjectivalization Figure 3.6 Denominal formal

In step with these analyses, it is possible to interpret Maltese verbs like inxtorob and tkabbar as units with different underlying structures in spite of their identical sur- face forms. While there is no morphonological evidence in the Maltese verbs to support this claim (unlike in the Dutch verbs cited above), this variability in the representation of similar surface forms can perhaps be motivated from a diachronic perspective. An examination of the entries in a rather comprehensive dictionary of Maltese (Aquilina 1987-1990) reveals that only the idiosyncratic meanings of inxtorob are listed; it is de- fined as ‘shrink, become emaciated’ but not as ‘be drunk, be absorbed’. This seems to

17 It is worth noting that this intuition is in line with the argument in Plank (2010) that in the case of polysemous words, individual senses rather than lexical items as a whole are involved in derivation.

84 A Tale of Two Morphologies

suggest that the binyan VII verb started off as a root derivation, with unpredictable se- mantics, but that in the last few decades native speakers felt the need to form the pas- sive of binyan I.18 In connection with this diachronic remark, Kraaikamp (2008: 9) points out that:

It has to be underlined that the kind of derivation that is discussed here, and within Distributed Morphology in general, is the way words are created/represented in the grammar of the speaker. If a verb is said to be ‘derived’ from an underlying verb, this is not a claim about a diachronic de- velopment. What is meant is that the verb is created in syntax from an underlying verbal category, which is a synchronic situation. This means that derivation structures are not unchangeable. At one point in time, a verb may be derived from an underlying verb, while at another their deriva- tional relationship may have disappeared and both verbs have become direct root derivations.

On the basis of diachronic evidence, Kraaikamp (2008) demonstrates that, over time, a verb that was originally represented as a word derivation can become a direct root derivation (and, arguably, a root derivation can become a word derivation). She argues that the relative frequency of the words in question and the similarity of their seman- tics are among the factors influencing the historic change of derivation structures. Let us now consider the second explanation. The presence of unpredictable mean- ings in a number of word-derived words can be accounted for by challenging the pro- posal that category-defining heads are phase-defining; i.e. that they play the central role of fixing both morphonological and semantic information before further word forma- tion takes place. In different parts of this chapter, it was argued that one of the crucial differences between root derivation and word derivation is that semantic idiosyncra- sies and idiomatic interpretations are strictly reserved to the former. This is because, according to Arad (2003a, 2003b, 2005), once a root merges with a functional head, the interpretation is fixed, and any subsequent derivation must operate with this interpre- tation. However, in a recent strand of research, the notion of phasehood and phase im- penetrability has been called into question. Anagnostopoulou & Samioti (2009, 2010), Borer (2009a, 2009b), Gribanova (2009), Harley (2009, 2011), Panagiotidis (2011), Svenonius (2011), inter alia, have brought up cases of word-derived words which do

not retain the original interpretation of the root, such as these examples from English:

18 Such forms tend to be used in particular registers, such as journalistic Maltese. And, in fact, the example in (12) above was taken from a newspaper, as it is not easy to find naturally occurring examples of inxtorob used in the passive reading. In this case, pseudo-passives (it-tifel xorb-u l-ilma ‘the boy drank-it the water’) are preferred to regular passives (l-ilma nxtorob mit-tifel ‘the water was drunk by the boy’). See Borg (1988), among others, on pseudo-passives in Maltese.

The protagonists: roots and patterns 85

(14a) nature natural naturalize ‘make natural’ ‘grant full citizenship’

(14b) edit editor editorial ‘of or relating to an editor or editing’

‘opinion piece in a newspaper, magazine, etc.’

Consider also two analogous cases from Maltese in (15).

(15a) għallaq ‘hang’ ‘hold a stranglehold on’ tgħallaq ‘be hanged, hang oneself’ ‘(jocular) get married’

(15b) laqqam ‘graft’ ‘vaccinate’ tlaqqam ‘be grafted’ ‘be vaccinated’ ‘imitate or follow the (bad) example of someone’

Under Arad, the interpretations of nature, edit, għallaq and laqqam are spelled out once the roots √NATURE, √EDIT, √għlq and √lqm merge with the first functional head: a nominal head in (14a) and verbal heads in (14b) and (15). These interpretations cannot be altered and subsequent derivation can only take these interpretations as their input. However, naturalize, editorial, tgħallaq and tlaqqam are not only interpreted compositionally (i.e. ‘make (more) nature-like’; ‘of or relating to an editor or editing’; the passive and reflexive of ‘hang’; and the passive of ‘graft, vaccinate’), but they also acquire idiosyncratic, non-predictable interpretations that do not build on the interpretation of the root-derived base. Such cases where verb-derived verbs are more polysemous than the root-derived verbs are quite rare and difficult to find, at least in the binyan system of Maltese. For the languages under discussion, therefore, the impenetrability of the root and categorizer appears to be too strict. What really seems to be the case is that, as Kramer (2009: 147) puts it, “the unit formed by the root and its first categorizing head is more

86 A Tale of Two Morphologies

likely to be phonologically and semantically idiosyncratic” (boldface in original). Harley (2009) argues along the same lines that the first combination of a root with a functional head tends to exhibit more morphonological and semantic idiosyncrasies than further derivation from that initial merger.19 To summarize so far, then, the Marantz-Arad model makes clear predictions about how words derived from roots and words derived from existing words behave with re- spect to morphonology and semantics. The distinction between root derivation and word derivation as originally conceived in Arad and Marantz’s work, however, seems to be at times too strict. Further research is required to study the conditions under which idiomatic, non-predictable meaning arises in word formation, and to resolve the issue of idiosyncratic semantics being assigned outside the first cycle, even if rarely so. In spite of this, the model is nonetheless a useful tool that helps predict and explain se- mantic and phonological differences between root-derived and verb-derived verbs (and nouns) in Maltese. This principled distinction is used in this study to account for the inherent regularity and irregularity in the Maltese verbal system, in particular templatic verbs. The full implications of the regular and irregular sides within the binyan system are discussed in Ch. 4, where a quantitative and a qualitative analysis of the verbal root- and-pattern morphology of Maltese are provided.

19 Harley (2009: 7) also argues that subsequent derivation can be idiosyncratic until the “first semantic cyclic node”. However, the issue of what a semantic cycle node really is remains open. See discussion in Alexiadou (2009), Borer (2009b), Harley (2009), and others, on different proposals to which node is re- sponsible for fixing the interpretation.

Chapter 4 When patterns put down roots

4.1 Database of roots and patterns ...... 88 4.1.1 Data collection ...... 89 4.1.2 Productivity of roots and patterns ...... 92 4.1.3 Co-occurrence of patterns ...... 98 4.2 The routes patterns take ...... 103 4.2.1 Morphological marking of categories...... 104 4.2.1.1 Singletons...... 105 4.2.1.2 Argument structure alternations...... 107 4.2.1.3 Multiple interpretations...... 115 4.2.1.4 Synonyms ...... 118 4.2.2 Section summary...... 122 4.3 A(n ir)regular binyan system ...... 125 4.3.1 Main findings and assumptions...... 125 4.3.2 Verb formation and the four categories ...... 129

We have, so far, established that the binyan system in Maltese displays both a degree of systematicity and a complexity of interactions between morphological components, i.e. roots and verbal patterns. Let us briefly recapitulate three main points discussed in the previous chapters resulting in the regular and irregular aspects of templatic verbs. First, while all templatic verbs must be in the form of a binyan, there are many unfilled binyan slots. As will become apparent in this chapter, a large number of roots are in- serted in only one or two patterns. Second, the occurrence of a root in more than one binyan correlates with systematic meaning associations (argument structure alterna- tions) on the one hand, and with multiple interpretations and synonymous verbs on the other. Third, while many verbs derived from the same consonantal root have one vowel sequence throughout, other morphologically related verbs have different vowels. Recall the vocalic variation in xorob ‘drink’ and xarrab ‘wet’, which was then linked to the ob- servation that a change in the vowel patterns may correspond to unpredictable seman- tics. This point is further developed in the rest of this chapter. Drawing on some insights from Distributed Morphology, in particular the struc- tural distinction between word formation from roots and word formation from existing words, I tried to track down the cause of the regularity and irregularity in the system. 88 A Tale of Two Morphologies

The conclusion, which will be fleshed out in the following sections, is that verbs derived from roots tend to exhibit gaps and non-predictable morphonology (vocalic variation and a fair degree of freedom in binyan selection) and semantics (idiosyncratic mean- ing). Verbs derived from previously formed verbs, by contrast, have relatively predict- able semantics, do not involve vowel alternations, and, as we shall see below, are fully productive, with no gaps in the system. The aim of this chapter is twofold. First, it quantifies the regularity and irregularity in Maltese templatic verbs; now that we have shed light on why the binyan system is not completely ir/regular, the next step is to seek to answer the question of how regular and irregular the system actually is. Second, it spells out in detail the range of semantic categories and verb alternations that are marked by binyan morphology, giving a com- plete picture of the relations between morphological forms and the different functions they express. I present the findings in three stages. Sect. 4.1 gives a quantitative analysis of roots and verbal patterns, examining the binyanim separately and paying particular attention to their productivity and preference for specific root types. A qualitative analysis is then provided in Sect. 4.2, which looks at the complexity of relations among binyanim, de- scribing the way morphosyntactic features are mapped into the verbal patterns. Follow- ing these detailed analyses of templatic verbs, Sect. 4.3 brings to completion the argu- ment made in Ch. 3 by setting out the principal sources of regularity and irregularity in the binyan system. In so doing, this section also summarizes the main conclusions drawn in Ch. 3 and Ch. 4.

4.1 Database of roots and patterns

Having established that the relations between templatic verbs in Maltese are partly sys- tematic and partly unpredictable, it becomes very important to quantify the regular and the irregular share in the binyan system. This section, in fact, takes a quantitative ap- proach, addressing such questions as the following. How many verb-creating tri- and quadri-consonantal roots are there in Maltese? How many verbs does each root create? Which verbal patterns are the most or least productive? In order to resolve these and other related issues, it was crucial to compile an exhaustive list of consonantal roots in Maltese and the binyanim they appear in. In what follows, I describe the process of data collection and provide detailed analyses of the distribution and productivity of roots and patterns.

When patterns put down roots 89

4.1.1 Data collection

In the previous chapters, Maltese templatic verbs were described as a combination of two bound morphemes, a consonantal root and a binyan interwoven within each other in a non-concatenative fashion. In Maltese there are around 2000 verb-creating roots and eleven binyanim, nine for tri-consonantal roots and two for quadri-consonantal roots (cf. Table 2.5 for the allomorphic variations of each binyanim, which are dictated by different root types, such as reduplicative, weak-final, etc.). I have also shown that, subject to certain combinatorial constraints (cf. Sect. 2.2.2), a tri-consonantal root can in principle combine with any of the nine binyanim to create different verbal lexemes. In practice, however, no tri-consonantal root appears in all nine patterns. Rather, around two thirds of all tri-consonantal roots appear in only one or two binyanim. Quadri-consonantal roots, by contrast, do combine with both QI and QII binyanim, but about half of them combine with one pattern only (see data below). The discussion that follows revolves primarily around tri-consonantal roots and patterns. I occasionally make reference to quadri-consonantal verbs, but a fuller treatment of these verbs must await future work (cf. Ellul 2010 for a brief analysis). Although roots seem to combine freely with patterns, the distribution of templatic verbs within the lexicon is not entirely random. As we shall see in the rest of this chap- ter, there is a degree of organization in the binyan system, particularly in the greater than chance co-occurrence of patterns for a given root when one verb is derived from another. In order to achieve a full understanding of the way consonantal roots are mapped into the various binyanim, it was necessary to set up an exhaustive database of tri- and quadri-consonantal roots and record the number of verbs in each pattern. Using Serracino-Inglott’s (1975-1989) monolingual dictionary, Aquilina’s (1987- 1990) bilingual dictionary, and Mifsud’s (1995a: 272-295) list of loan verbs that have been fully integrated into the root-and-pattern system of Maltese, a database of all con- sonantal roots which give rise to one or more patterns was compiled in the form of a spreadsheet. A small extract from the database for tri-consonantal roots is reproduced in Table 4.1. A full listing of the data tabulated by root and pattern is given in Appendix I (tri-consonantal roots) & Appendix II (quadri-consonantal roots). The database contains all the templatic verbs recorded in the mentioned works, without discriminating between dialectal, slang, bookish, current or dated forms. For instance, under the root √ktb one finds not only the verbs kiteb ‘I, write, enroll (v.t)’ and inkiteb ‘VII, be written, enroll (v.i)’, which have a high frequency of use, but also the verb

90 A Tale of Two Morphologies

tkieteb ‘VI, engage in mutual correspondence’, which has fallen out of use. I also added some templatic verbs, unrecorded in dictionaries of Maltese, which I came across in lit- erary works (e.g., lajjem ‘II, slow down’, in Rużar Briffa’s poem Quo Vadis?) or heard other native speakers use (e.g., irħas ‘IX, become cheap’ for more common raħas ‘I, be- come cheap’) or which I know as a native speaker (e.g., tmandar ‘QII, get wasted’).

Table 4.1 Extract from the database of roots and patterns

Root I II III V VI VII VIII IX X

krh kerrah tkerrah krieh stkerrah

krm korom *karrem *tkarrem nkaram

krr *kerr stkerr

krt karrat

ksb kiseb kisseb tkisseb nkiseb

ksħ kesaħ kessaħ tkessaħ

ksr kiser kisser tkisser nkiser

kss kess

ktb kiteb *kitteb tkieteb nkiteb

ktf kittef tkittef

In principle, it is also possible to compile an inventory of only the verbal lexemes that an educated speaker of modern Maltese has. The list would be drastically reduced, with some roots appearing in fewer binyanim (e.g., under √sgħl one would list sagħal ‘I, cough’ but not the obsolete siegħel ‘III, cause one to cough’), and other roots being omit- ted altogether (e.g., √kbx and √wrx deriving, among others, kebbex ‘be unfaithful’ and werrex ‘slap someone’s face’ respectively). Since at the time of compiling the database there was no reliable dictionary or language resource that provides categorization of lexemes on the basis of their usage, it was not always easy to decide, for instance, which forms are current and which are dated.1 With such resources in hand, it is possible to replicate the exercise, providing additional results in terms of the usage and the fre- quency of use of the lexemes. A few verbs, which dictionaries mark as unused, obsolete

1 Carrying corpus studies would certainly help make such usage-based decisions. However, because of the normative nature of writing, a number of verbs, being dialectal, tend to be used more often in the spoken rather than in the written language. The present lack of corpora of spoken Maltese and dialectal Maltese clearly does not facilitate the undertaking of such a task.

When patterns put down roots 91

or hypothetical are, however, entered in the database, marked by an asterisk, but are excluded from the calculations given here. The discussion that follows is therefore based on an all-embracing database of templatic verbs, which add up to almost 4000 verbs derived from over 1900 different consonantal roots. A breakdown of the results is given in Table 4.2, which shows three things: the total number of tri- and quadri-consonantal roots, the number of verbs de- rived from both kinds of roots, and finally the mean number of patterns represented by root. One striking observation that comes out of this table is that there are numerous roots that are not used to form actual templatic verbs. Only 25% of all possible combi- nations of tri-consonantal roots with binyanim are attested verbs in Maltese, leaving 9572 binyan slots empty. By contrast, quadri-consonantal roots exhibit much fewer gaps, creating 718 verbs, which make up 75% of the total number of possible combina- tions of quadri-consonantal roots with binyan QI and binyan QII.

Table 4.2 Total consonantal roots and verbs they create

No. of Roots No. of verbs Mean patterns/root

Tri-consonantal 1424 3244 2.28 Quadri-consonantal 478 718 1.5 Total 1902 3962

Before presenting more results, a word is in order on the classification of homo- phonous roots in the database. In all, there are around 400 homophonous roots, such as √bjd1 bajjad ‘whitewash’, √bjd2 bied ‘lay eggs’, and √sfr1 sfar ‘turn yellow’, √sfr2 saffar ‘whistle’, √sfr3 siefer ‘travel’. They are usually the result of historical merges of different phonemes, in particular emphatic consonants with their non-emphatic counterparts (cf. Aquilina 1970; Mifsud 2008; inter alia). In general, it seems reasonable to have two (or more) separate entries for such roots both on etymological and semantic grounds. However, because a number of con- sonantal roots are assigned multiple interpretations that are difficult to relate deriva- tionally, at least synchronically (recall verbs like ħareġ ‘take out’, ħarreġ ‘train’ and stħarreġ ‘investigate’, which are all derived from the same root, √ħrġ), it is not always easy to tell apart cases of homophony from cases of multiple interpretations. In some studies, such as Borg (1981, 1988), verbs derived from two etymologically distinct

92 A Tale of Two Morphologies

roots, like għalaq ‘close’ and għallaq ‘hang, strangle’, are treated on a par with verbs such as fetaħ ‘open’ and fettaħ ‘loosen’, which etymologically belong to the same root and which in this study are analyzed as multiple interpretations of one root. In relation to this, Fabri (2009: 4) writes:

It is not clear to what extent these forms (= binyanim) are productive in modern Maltese, and very little is known about native speaker intuitions about the relations between the members of a ‘family’. For example, to what extent does a native speaker see the verb għallaq ‘strangle’ as a sec- ond form derivation of għalaq ‘close’, even though historically the two are unrelated (see Aquilina 1990: 954, 957)? In other words, it is not at all clear how psychologically real these families are to native speakers.

From this observation arises an interesting question concerning the semantic related- ness of words sharing the same root consonants which may be of identical or different historical origin. However, this is an issue which requires psycholinguistic and corpus- based research that makes use of such methods as the Latent Semantic Analysis (cf. Landauer & Dumais 1997; Landauer et al. 1998; inter alia). For the present purposes, it is important to avoid confusion between semantic un- relatedness that is due to different etymologies, on the one hand, and tenuous semantic relations among verbs derived from the same root by virtue of the characteristic of con- sonantal roots to be assigned multiple interpretations, on the other. For this reason, in line with Aquilina’s (1987-1990) dictionary, which points out the etymological sources of roots, homophonous roots are distinguished in the database by means of superscript numbers, such as √bjd1 and √bjd2.

4.1.2 Productivity of roots and patterns

There are a number of interesting ways to analyze the raw data in order to gain insights into how roots interact with patterns. In this section, I take a closer look at the database from three different yet related angles. I first examine the distribution of roots across different binyanim. Following that, I give a quantitative analysis of the morphological productivity of binyanim derived from tri-consonantal roots. Finally, I reinterpret these results in light of the binyanim’s preference for different root types, weak and strong, with the aim of determining the extent to which binyanim are conditioned by the mor- phonological shape of the root.

When patterns put down roots 93

Distribution of roots Let us begin by looking at the frequencies for patterns per root. Table 4.3a shows the number of tri-consonantal roots that appear in different patterns. Very few roots are cast in more than two patterns: only a third of tri-consonantal roots appear in three or more patterns. It is striking that most of the roots appear in only one (27%) or two pat- terns (40%), leaving a significantly large number of binyan slots empty. This means that a very large number of patterns are not used for creating actual templatic verbs .

Table 4.3a Actual number of patterns per tri-consonantal roots

Patterns/root 1 2 3 4 5 6 7 No. of roots 380 567 249 161 64 2 1 Percentage 26.69 39.82 17.49 11.31 4.49 0.14 0.07

The exercise was repeated for quadri-consonantal roots. The results, as summarized in Table 4.3b, show that a little more than half of the total of quadri-consonantal roots cre- ate verbs in both patterns. The rest typically appear only in binyan QI, such as √gdwd gedwed ‘mutter’, √ħbrk ħabrek ‘strive’, √slpj salpa ‘sail’, etc. A much smaller number is cast only in QII, such as √gxtr tgexter ‘become unsociable’, √rnġj tranġa ‘be set right’. The entire list of quadri-consonantal roots and the verbs derived from them is given in Appendix II.

Table 4.3b Actual number of patterns per quadri-consonantal roots

Patterns/root 1 2 No. of roots 238 240 Percentage 49.79 50.21

The rest of this chapter deals with the interrelations of tri-consonantal roots with patterns, leaving an in-depth analysis of quadri-consonantal roots and patterns to fu- ture research. Let us now turn to the morphological productivity of the nine tri- consonantal binyanim.

94 A Tale of Two Morphologies

Morphological productivity of binyanim A more interesting question concerns the morphological productivity of patterns. Say- ing that a given binyan is productive does not mean that it is more often used for the production of neologisms or for the integration of loan verbs. Rather, in this subsection, I measure the profitability of the various binyanim in the lexicon; that is, the extent to which they are actively used. An analysis of patterns in terms of their availability re- quires future psycholinguistic and corpus-based research.2 Research on Semitic lan- guages along these lines has been carried out, among others, by Bolozky (1999) for Is- raeli Hebrew and Verheij (2000) for . In Table 4.4, I give the productivity rates of the nine binyanim, indicated by Roman numerals. Over three fourths of the total number of tri-consonantal roots create verbs in binyan I, II and V. Another relatively productive pattern is VII, accommodating around 10% of tri-consonantal roots. The other patterns are virtually unproductive, with only 13% of the roots appearing in the five of them put together.

Table 4.4 Productivity of binyanim derived from tri-consonantal roots

Pattern I II III V VI VII VIII IX X Frequency 696 994 102 790 137 337 91 73 24 % of roots 21.45 30.64 3.14 24.35 4.22 10.39 2.80 2.25 0.74

Here, binyanim were treated individually. In the next section, I tackle the interrelations among patterns on the basis of their co-occurrence on the same roots, thereby assess- ing the degree to which the binyanim can be said to form a system. Before going into this analysis, however, let us study the relationship, if any, between the formal aspects of the root and the productivity of the patterns.

Binyanim and root types Having looked at how roots are distributed across patterns and at the tendency of roots to combine with particular patterns, I now address the issue of possible preferences a given binyan might have for different root types. In Sect. 2.2.1, roots were categorized into different types, drawing a main distinc-

2 In Sect. 2.4, I briefly discuss the distinction between the availability and profitability of templatic verbs in Maltese.

When patterns put down roots 95

tion between strong and weak roots. While the consonants of strong roots are always audible and normally not affected by morphonological processes, weak roots have the glides /j/ or /w/ as one of their radicals, which are silent in some words, giving rise to

‘irregular’ forms. Strong roots were further subdivided into regular C1C2C3 and redupli- cative C1C2C2. There are three different subtypes of weak roots, depending on the posi- tion of the glide, in initial, medial or final position. As discussed in detail in Ch. 2, root types determine the syllabic structure of verbs and have an effect on their inflectional morphology. These five root types are summarized in Table 2.4, reproduced here with some minor changes as Table 4.5.

Table 4.5 The main types of tri-consonantal roots in Maltese

Root type Example(s) Verb(s) Meaning(s)

Strong regular √ktb kiteb write reduplicative √xmm xamm smell

Weak initial √wħl, √jsr weħel, jassar stick, enslave medial √dwr, √tjr dar, tar turn, fly final √dgħw, √qrj dagħa, qara swear, read

Let us now look into the issue of whether the morphonological shape of the root, i.e. if it is strong or weak, puts a constraint on the selection of a binyan. If this were the case, then this would suggest that the distribution of binyanim is to some degree condi- tioned by formal properties of the root. To begin with, the total number of roots and the verbs they form are laid out in Ta- ble 4.6 according to the root type.3 Two main observations can be drawn from this ta- ble. First, over two thirds of the tri-consonantal roots in Maltese are strong, the major- ity of which are regular rather than reduplicative. Second, around 27% of templatic verbs (excluding those formed from quadri-consonantal roots) are derived from weak tri-consonantal roots.

3 Note that one particular root, √wċċ, is grouped with reduplicative roots, but can instead be entered with weak-initial roots. Also note that the 12 irregular roots, i.e. the roots that for historical reasons have the first radical missing (e.g., √’ħd creating ħa ‘I, take’ and ittieħed ‘VI, be taken’), are classified with the strong regular roots, unless their last radical is a glide (e.g., √’dj forming idda ‘VIII, shine’), in which case they are listed with weak-final roots.

96 A Tale of Two Morphologies

Table 4.6 Tri-consonantal roots and verbs they create grouped by root type

Root type R. Red. W-I W-M W-F No. of roots 826 158 38 210 192 Percentage 58.01 11.10 2.67 14.75 13.45 No. of verbs 1986 370 91 445 352 Percentage 61.22 11.41 2.81 13.72 10.85 Mean verbs/root 2.40 2.34 2.39 2.12 1.83

Table 4.6 also shows that the verbs, taken together without reference to their bin- yan, are distributed rather regularly over the root types. For each root type, in fact, the mean number of verbs per root is about 2. However, the proportions are not exactly the same. This emerges from an examination of the possible correlations between root type and binyan selection by calculating the distribution of the binyanim over the five root types. When the total number of roots by type is plotted against the nine patterns (cf. Table 4.7), a bar chart like the one in Figure 4.1 is obtained.

Table 4.7 Root types across binyanim

R. Red. W-I W-M W-F

no. % no. % no. % no. % no. % I 424 60.83 74 10.62 11 1.58 87 12.48 100 14.35 II 587 59.05 127 12.78 32 3.22 172 17.30 77 7.75 III 64 62.75 2 1.96 5 4.90 1 0.98 30 29.41 V 475 60.20 98 12.42 31 3.93 127 16.10 58 7.35 VI 100 72.99 1 0.73 5 3.65 7 5.11 24 17.52 VII 216 64.09 45 13.35 4 1.19 26 7.72 46 13.65 VIII 65 71.43 11 12.09 0 0 5 5.50 10 10.99 IX 48 65.75 6 8.22 2 2.74 16 21.92 1 1.37 X 7 29.17 6 25 1 4.17 4 16.67 6 25

When patterns put down roots 97

The chart reveals that: • All binyanim have a high tendency to combine with strong regular roots, especially VI and VIII. The only exception is X, which shows an almost equal preference for regular, reduplicative and weak-final roots. • Relatively speaking, reduplicative roots appear most frequently in X. They are em- bedded frequently in I, II, V, VII and VIII, but almost never in III and VI. • Weak-initial roots never appear in VIII and hardly ever in I and VII. They are rela- tively more commonly cast in II, III, V, VI and X. • Binyan III, VI and VIII do not generally accommodate weak-medial roots, which tend to interleave with II, V, IX and X. • In proportion to all types, weak-final roots combine most frequently with III and X. It is relatively common for this root type to be inserted in I, VI and VII, but not in II, V and IX.4 • Binyan III and VI tend to combine with regular and weak-final roots, and are underrepresented on the other three root types. • It is more often that binyan II, V and IX combine with regular and weak-medial roots than with other root types. However, II and V also show a relative preference for reduplicative roots.

4 There is only one weak-final root which combines with IX, √ħlw, creating ħliel ‘become sweet’, where the root is reanalyzed as a reduplicative one, with the final glide turning into a liquid.

98 A Tale of Two Morphologies

These findings suggest that there is some morphological conditioning of binyanim by root type, which can be couched in terms of (dis)preferences. While regular roots generally combine with all binyanim, reduplicative roots have a strong dispreference for III and VI. The combination of weak-initial roots with I, VII and VIII is dispreferred or even not attested. There is a relatively low number of verbs formed by the combina- tion of weak-medial roots and binyan III, VI and VIII, on the one hand, and of weak-final roots and binyan II, V and IX, on the other. These observations are summarized in Table 4.8.

Table 4.8 (Dis)preferences of combinations of root types and binyanim

Patterns Preferred type Example Dispreferred type

I, VII, VIII R. √xħt W-I Red. √sdd II, V, IX R. √ċkn W-I W-M √twl W-F III, VI R. √ħrs Red. W-F √vrj W-M X R. √nbħ W-I Red. √knn W-F √ħbj

4.1.3 Co-occurrence of patterns

We have, so far, examined the binyanim separately, observing that, among other things, I, II and V are the predominant patterns in the system, and that VII, although much smaller, is still considerably larger than the other five binyanim, which together ac- commodate only 13% of all tri-consonantal roots. In this section, I analyze the patterns as one system by looking at the relations between binyanim, especially in terms of their co-occurrence on the same roots. Table 4.9 gives the frequencies of roots appearing in only two binyanim. When a root combines with two patterns, there is a very high chance it selects II and V, e.g., √wtq, wettaq ‘implement’, twettaq ‘be implemented’. It is also relatively frequent for

When patterns put down roots 99

roots creating two binyanim to appear in I and VII, e.g., √bdj, beda ‘begin’, inbeda ‘be begun’. Other less frequent combinations include I and II, e.g., √għlj, għola ‘go up (prices)’, għolla ‘put up (prices)’, and III and VI, as √ħrs, ħares ‘protect’, tħares ‘be pro- tected’.

Table 4.9 Roots creating two verbs

Patterns No. of roots Percentage Patterns No. of roots Percentage

II – V 386 68 I – VIII 3 0.5 I – VII 80 14 I – X 2 0.4 I – II 37 6.5 II – VI 2 0.4 III – VI 30 5.3 II – III 1 0.2 I – III 6 1.1 III – IX 1 0.2 II – IX 6 1.1 V – VI 1 0.2 I – V 5 0.9 V – IX 1 0.2 I – VI 5 0.9 VII – VIII 1 0.2

As expected from the data in Table 4.4, when three different templatic verbs share the same root, the greatest probability is that their patterns are I, II and V. It is often the case that in these triplets, the binyan II verb is both the causative of the inchoative verb in I as well as the active of the passive in V, as in √ksħ, kesaħ ‘cool, v.i’, kessaħ ‘cool, v.t’ and tkessaħ ‘be cooled’. The same holds for triplets involving binyan II, V and IX. The verb in binyan II tends to be syncretic, typically with an inchoative counterpart in IX and a passive counterpart in V, e.g., √qsr, qassar ‘shorten, v.t’, tqassar ‘be shortened’, qsar ‘shorten, v.i’ (cf. the discussion in Borg 1981, 1988, and Sect. 4.2 concerning the morphosyntactic associations between binyanim). Other frequent three-place co-occurrences include I, II and VII, e.g., √ftq, fetaq ‘un- stitch’, fettaq ‘pick holes’, infetaq ‘be unstitched’, and I, VII and VIII, where the verbs in the latter two binyanim are very often synonymous, e.g., √sdd, sadd ‘clog (v.t.), instadd and stadd ‘clog up, v.i’ (cf. the discussion in 4.2.1.4 on synonyms). The results are sum- marized in Table 4.10.

100 A Tale of Two Morphologies

Table 4.10 Roots creating three verbs

Patterns No. of roots Percentage Patterns No. of roots Percentage

I – II – V 113 45.4 II – III – V 2 0.8 I – II – VII 31 12.4 II – V – VI 2 0.8 II – V – IX 25 10 III – VI – IX 2 0.8 I – VII – VIII 18 7.2 I – II – VIII 1 0.4 II – V – VII 14 5.6 I – III – IX 1 0.4 I – III – VI 8 3.2 I – V – VI 1 0.4 I – VI – VII 6 2.4 I – V – X 1 0.4 I – V – VII 5 2 I – VI – VIII 1 0.4 II – V – VII 4 1.6 II – III – VI 1 0.4 II – V – X 4 1.6 III – V – VI 1 0.4 I – II – VI 2 0.8 III – VI – VII 1 0.4 I – II – IX 2 0.8 III – VI – VIII 1 0.4 I – VII – X 2 0.8

Let us now take a closer look at roots that create four different verbs. As the num- bers and percentages in Table 4.11 illustrate, nearly two thirds of these roots appear in I, II, V and VII, e.g., √ftħ, fetaħ ‘open, v.t’, fettaħ ‘loosen, v.t’, tfettaħ ‘be loosened’, infetaħ ‘open, v.i’. The remaining roots in this group are most frequently cast in I, II, V and IX, e.g., √sbħ, sebaħ ‘dawn’, sebbaħ ‘beautify’, issebbaħ ‘be beautified’, sbieħ ‘become beauti- ful’. As I show in some detail in Sect. 4.2, roots embedded in four or more patterns gen- erally involve multiple interpretations and/or verbs with identical meanings.

Table 4.11 Roots creating four verbs

Patterns No. of roots Percentage Patterns No. of roots Percentage

I – II – V – VII 94 58.39 II – III – V – VI 2 1.2 I – II – V – IX 20 12.42 1 – II – III – VIII 1 0.6 I – II – V – VI 10 6.2 I – II – VII – IX 1 0.6 I – III – VI - VII 8 5 I – II – VII – X 1 0.6 I – II – VI – VII 5 3.1 I – VI – VII – VIII 1 0.6 I – II – V – VIII 3 1.9 II – V – VI – VIII 1 0.6 I – II – V – X 3 1.9 II – V – VII – VIII 1 0.6

When patterns put down roots 101

I – II – VII – VIII 3 1.9 II – V – VII – IX 1 0.6 I – III – VII – VIII 2 1.2 II – V – VIII – IX 1 0.6 I – V – VII – VIII 2 1.2 II – V – IX – X 1 0.6

Finally, consider the 64 roots (around 5% of all tri-consonantal roots) that com- bine with five different patterns in Table 4.12. Almost half of them involve the co- occurrence of I, II, V, VII and VIII, with these last two verbs usually being synonymous, e.g., √nsj, nesa ‘forget’, nessa ‘cause to forget’, tnessa ‘be caused to forget’, and both int- nesa and intesa meaning ‘be forgotten’. Another one fifth of the roots in this group ap- pear in I, II, V, VI and VII. It is usually the case that the verb in VI is (i) synonymous with V, (ii) is assigned a different interpretation from the other verbs sharing the same root, or (iii) has fallen out of use, e.g., √qtgħ, qata’ ‘cut’, qatta’ ‘tear, v.t’, tqatta’ ‘tear, v.i’, inqata’ ‘be cut’, and VI tqata’ ‘struggle, fight’, which is an instance of multiple interpreta- tions.

Table 4.12 Roots creating five verbs

Patterns No. of roots Percentage

I – II – V – VII – VIII 30 46.9 I – II – V – VI – VII 14 21.9 I – II – V – VI – VIII 4 6.3 I – II – V – VII – IX 4 6.3 I – III – VI – VII – VIII 3 4.7 I – II – III – V – VI 2 3.1 I – II – III – V – VII 1 1.6 I – II – III – VI – IX 1 1.6 I – II – III – VI – X 1 1.6 I – II – V – VI – IX 1 1.6 I – II – V – VII – X 1 1.6 I – II – VI – VII – VIII 1 1.6 II – V – VII – VIII – X 1 1.6

102 A Tale of Two Morphologies

With this data in hand, we are now in a position to look at the distribution of all tri- consonantal roots across patterns. The co-occurrence frequencies for all pairs of pat- terns are given in Table 4.13.5 The data suggest that the distribution of patterns in the lexicon is not random. Rather, some of them tend to occur together, while others are almost mutually exclusive. As expected from the results in the previous tables, the most recurring combinations of patterns include binyan I, II, V and VII. The most frequent pair is by far II and V (making up 25% of the total number of roots appearing in more than one pattern), followed by I and II (13%), I and V (11%), I and VII (10%). On the other hand, very few are the roots co-occurring in III and VII, or VI and X, to mention two examples.

Table 4.13 Co-occurrence frequencies for tri-consonantal roots

I II III V VI VII VIII IX X

I II 390 III 36 14 V 319 746 10 VI 97 52 57 42 VII 298 216 11 176 41 VIII 75 49 8 45 17 62 IX 29 63 5 54 4 6 1 X 11 12 1 11 1 5 1 1

In a nutshell, in this section, I have shown that Maltese templatic verbs lend them- selves to a quantitative analysis, which reveals that the distribution of verbal patterns by root within the lexicon is not completely random, but is systematic in several re- spects. It was observed that, even if the phonological constraints on the combination of roots and patterns are minimal (cf. Sect. 2.2.2 and Sect. 4.3.1), the majority of tri- consonantal roots occur in only one or two patterns. It was also noted that binyan I, II and V are the most productive; binyan VII is also relatively productive. Another conclu- sion concerned the extent to which the selection of a given binyan is conditioned by the

5 The statistics in Table 4.13 include the figures in Tables 4.9-4.12 plus two roots, √bdl and √slt, which ap- pear in 6 different binyanim, and another root, √frq, which creates 7 different verbs, even if some of them are dated or have fallen out of use.

When patterns put down roots 103

root type. The data suggests that there indeed is a correlation between the formal as- pects of the root and the patterns it selects, with certain patterns (e.g., III, VI) showing preferences for certain root types (regular and weak-final) and dispreference for other types (reduplicative, weak-medial). Finally, the quantitative analysis points to a degree of regularity in the co-occurrence of various patterns for the same roots: some combi- nations of patterns often occur together while others tend to exclude each other. With the above discussion on the interactions of roots with patterns in mind, we can now move on to a qualitative analysis of the database in two main stages. After de- scribing the morphosemantic features of templatic verbs, a mapping of these features into their morphological realizations is carried out, thereby giving a complete picture of the relations between verb alternations and morphological form.

4.2 The routes patterns take

In this section, I offer a qualitative analysis of the database of tri-consonantal roots and the patterns they combine with. After compiling the database, listing each root accord- ing to the patterns in which it occurs, the roots were sorted out according to the rela- tions that hold between their different occurrences. One important result that comes out of this categorization is that, when roots are inserted in patterns, they fall into four main categories. The first category contains a considerable number of roots which de- rive singleton verbs. The other categories involve roots that interleave with more than one binyan. The relations between verbs that are derived from a common root may be an instance of argument alternations, multiple interpretations, or synonyms. Similar ob- servations on the morphosyntactic categories of consonantal roots across different bin- yanim have been made for Hebrew by Arad (2005). In the following paragraphs, I discuss the morphological marking of the four cate- gories mentioned above, giving a complete view of the way different functions relate to binyan morphology. Two important observations arise from this examination. The sys- tem is regular in terms of directionality, i.e. the tendency of features to be marked by certain combination of patterns but not others. However, the system is irregular due to syncretism, i.e. that several features are realized by the same combination of patterns.

These two aspects are tackled in some detail in Sect. 4.2.2 and Sect. 4.3.

104 A Tale of Two Morphologies

4.2.1 Morphological marking of categories

We have already seen that the binyanim do not have a fixed semantic contribution, such as causative, reflexive, inchoative, etc. On the contrary, there are cases where one and the same pattern, e.g., binyan I, marks both causative (e.g., √ftħ, fetaħ ‘open, v.t’, cf. infetaħ ‘VII, open, v.i’) as well as inchoative (e.g., √ksħ, kesaħ ‘cool, v.i’, cf. kessaħ ‘II, cool, v.t’) verbs, depending on the root it combines with. The question to ask next, then, is, if we had to shift our attention from individual binyanim to the binyan system as a whole, can we find generalizations and regularities in the system of relations that hold be- tween patterns derived from the same root? In order to provide an answer to this question, in what follows, I analyze the data- base of roots and patterns qualitatively, throwing light on the relations among each of the four categories (singletons, argument alternations, multiple interpretations, syno- nyms) and their morphological form. Among the issues such an analysis allows us to address are the following. Are some categories marked by certain groups of patterns but not others, or do the same morphological patterns mark more than one syntactic structure? Which argument alternations occur most frequently in the database? How many roots are assigned multiple interpretations? Which particular combinations of patterns give rise to verbs with identical meaning? Let us now turn to these and other questions, taking a closer look at each of the four categories separately. A methodological note is, however, in order before presenting the results of an ex- amination of the four categories. In the analyses that follow, it is not roots (types) but categories (tokens) that are counted, taking into account the individual senses of polysemous verbs. Consider this example taken from the discussion in Sect. 3.3.2:

għallaq tgħallaq alternation 1a. hang 1a. be hanged active-passive 1b. hang 1b. hang oneself transitive-reflexive 2. oppress 2. (jocular) get married

The root √għlq creates two verbs, one in binyan II and another in binyan V, which add up to two tokens, as it participates both in the active-passive (sense 1a) and the transi- tive-reflexive alternation (sense 1b). Other idiomatic meanings (sense 2) that do not involve alternations in the event structure of the verb or synonyms are not included in the calculations. This means that there are 1424 tri-consonantal roots, but the total

When patterns put down roots 105

number of categories is 3145. The number of categories is so high because argument alternations in particular are very syncretic (cf. Sect. 4.3). From a quantitative point of view, the largest category is argument alternations, whereas the smallest is the cate- gory of multiple interpretations. The figures are listed in Table 4.14.

Table 4.14 The four categories

Category No. of tokens Percentage

Argument alternations 1922 61.11 Synonyms 557 17.71 Singletons 380 12.08 Multiple interpretations 286 9.09

4.2.1.1 Singletons

As observed in Sect. 4.1, 380 tri-consonantal roots in Maltese (making up 27% of all tri- consonantal roots, and 12% of the total number of categories) are embedded in one pattern only, each root leaving eight binyan slots empty. Some examples include:

Root Verb Binyan Meaning Root Verb Binyan Meaning

√bsm tbissem V smile √ġrj ġara I happen √dnb dineb I sin √jqr stejqer X revive √fċj tfaċċa V appear √kżj tkaża VI find fault √fġġ feġġ I emerge √lqq leqq I shine √flj falla II fail √nħr naħar I snore √fxx infexx VII erupt in √sjd stad VIII fish √ġj’ ġie I come √smr smar IX get tanned

The data above show that roots of this kind appear in several patterns. An examination of all the roots in the database reveals that singletons may be in the form of all nine pat- terns. Their distribution across patterns is given in Table 4.15, together with the per- centage of the total number of roots that create only one verb. The majority of single- tons (almost 80%) appear in binyan I or II. Put differently, if a root is inserted in one of the other seven binyanim (in particular binyan VII, VIII, IX and X), there is a very high chance that it is also inserted in at least one other binyan.

106 A Tale of Two Morphologies

Table 4.15 Distribution of roots appearing in one binyan only

Pattern I II III V VI VII VIII IX X No. of roots 143 153 22 28 14 3 3 7 7 % of singletons 37.63 40.26 5.79 7.37 3.68 0.79 0.79 1.84 1.84

Note that verbs are listed under this category on purely morphological grounds; that is, they form a group by virtue of the fact that there is no other corresponding verb derived from the same root. Still, in some cases their being a singleton seems to be se- mantically and syntactically motivated. Nearly all verbs in this category are intransitive; very few of them are transitive, e.g., gaża ‘report’, or ambitransitive, e.g., falla ‘II, fail’, ċeda ‘III, concede’. Since most of them do not normally take a direct object, they cannot participate in alternations in- volving a direct object, such passivization and reflexivization. This, however, does not explain why dineb ‘I, sin’, għer ‘I, be jealous’, smar ‘IX, get tanned’ and the like cannot be causativized. The semantics of the verbs must be part of the explanation in this case. Semantically, singletons do not constitute a homogenous class: they include stative verbs, as seta’ ‘I, can’ and √’mb amba ‘I, need’,6 and verbs of light and sound emission, like leqq ‘I, shine’, idda ‘VIII, sparkle’, naħaq ‘I, bray’, qaqa ‘III, cackle’, idden ‘II, crow’, which cannot be construed as having an external causer, and hence do not causativize. Another large number of singletons are denominal. Examples of such roots include the following, some of which are dated:

√bħħ tbaħħaħ ‘V, become void’ cf. baħħ ‘void’ √ċrk ċerrek ‘II, move in a circle’ cf. ċirku ‘circle’ √mħħ maħħaħ ‘II, think deeply’ cf. moħħ ‘brain, mind’ √psj passa ‘II, pace, walk back and forth’ cf. pass ‘pace’ √rġl traġġel ‘V, become a man’ cf. raġel ‘man’ √tbb tabbab ‘II, see a doctor frequently’ cf. tabib ‘doctor’ √tjġ tejjeġ ‘II, wed’ cf. tieġ ‘wedding’

6 The share of singletons in the database would have been higher had I excluded verbal forms that are dated or not found in my idiolect, such as saffar ‘cause to depart’, għattas ‘cause to sneeze’, and considered the current forms siefer ‘go abroad’, għatas ‘sneeze’ only. By removing unusued forms from the database, the list of singleton stative verbs would also grow to include such items as af ‘know’, bagħad ‘hate’, ried ‘want’, which in my idiolect do not have other corresponding verbs derived from the same root, like intaf ‘be known’, inbagħad ‘be hated’, intried ‘be wanted’, which are, however, recorded in dictionaries.

When patterns put down roots 107

√tkn takkan ‘II, walk on (high) heels’ cf. takkuna ‘heel(s)’ √trq terraq ‘II, walk along roads’ cf. triq ‘road’ √zpn zappan ‘II, dig with a mattock’ cf. zappun ‘mattock’

Clearly, there are other denominal verbs which have other morphologically related verbs to them, such as:

√pjp pejjep tpejjep cf. pipa ‘II, smoke’ ‘V, be smoked’ ‘smoking pipe’ √trġ tarraġ ittarraġ cf. taraġ ‘II, terrace’ ‘V, be terraced’ ‘stairs’ √xmx xemmex ixxemmex cf. xemx ‘expose to the sun’ ‘sun oneself’ ‘sun’

However, while the latter roots create at least one transitive verb, pejjep, tarraġ, xemmex (and their passive or reflexive counterparts), the former roots give rise to in- transitive verbs, which cannot be passivized or reflexivized. Since most of them are cast in binyan II, they cannot create a causative verb either, because of the unidirectional relations that hold among binyanim. As we shall see below, the system does not allow an in binyan II to have a causative counterpart in any of the other bin- yanim. Therefore, it seems that the appearance of tri-consonantal roots in only one verbal pattern can be accounted for on semantic and syntactic grounds, at least for some verbs. It is nevertheless not at all clear why these roots are not assigned multiple inter- pretations. Nor can we explain why these roots do not create synonymous verbs. An in- depth investigation of the verbs in this category must await future research. For the present purposes it suffices to identify a morphological category of singleton verbs.

4.2.1.2 Argument structure alternations

Having briefly surveyed singletons, let us now turn our attention to roots that combine with more than one binyan. The discussion in this subsection revolves around four ma- jor alternations, the active-passive, transitive-reflexive, causative-inchoative, and causative-noncausative. They all involve a difference in the event structure of the verb, syntactically expressed by a change in the transitivity of the verb. From a quantitative point of view, the most common alternation marked by the binyan system in Maltese is

108 A Tale of Two Morphologies

the active-passive alternation, followed by the causative-inchoative alternation. Table 4.16 lists the number of tokens and their percentages from the total number of in- stances of an argument structure alternation in the binyan system.

Table 4.16 Number of tokens of argument alternations

Argument alternation No. of tokens Percentage

Active-passive 1008 52.45 Causative-inchoative 578 30.07 Causative-noncausative 212 11.03 Transitive-reflexive 124 6.45

Active-passive alternation Let us now take a closer look at the morphological realization of each of the four verb alternations, starting with the active-passive. In this morphosyntactic operation, the external argument fails to be projected in the syntax. It may still be expressed as an oblique or adjunct, such as the optional by-phrase, as in (1). The (suppressed) external argument of the passive can also license agent-oriented adverbials (2), and control into a purpose clause (3).

(1a) Dun Karm kiteb l=Innu Malt-i.

Dun Karm write.PFV.3SG.M DEF=anthem Maltese-SG.M ‘Dun Karm wrote the Maltese national anthem.’ (1b) L=Innu Malt-i n-kiteb

DEF=anthem Maltese-SG.M PASS-write.PFV.3SG.M f=l=1922 (minn Dun Karm).

in=DEF=1922 from Dun Karm ‘The Maltese national anthem was written in 1922 (by Dun Karm).’

(2) L=ittra n-kitb-et apposta/bl-addoċċ.

DEF=letter PASS-write-PFV.3SG.F on purpose/carlessly ‘The letter was written on purpose/carlessly.’

(3) L=ittra n-kitb-et biex il=membr-i

DEF=letter PASS-write-PFV.3SG.F in order to DEF=member-PL

When patterns put down roots 109

kollha j-kun-u avż-at-i.

all IPFV.3-be-PL notify-PST.PTCP-PL ‘The letter was written so that all members are notified.’

The distribution of templatic verbs taking part in the active-passive alternation is remarkably regular. From Table 4.17, which summarizes all the tokens of active and passive verbs in the binyan system, we can observe that the form of the passive is, al- most without exception, predictable from the form of the active, and vice versa. The system is unidirectional. That is, when speakers or language learners come across a new active verb in binyan II, they can infer that its passive counterpart will be in binyan V. The same holds true for the combinations I – VII and III – VI. These and similar mor- phosyntactic correspondences among binyanim have already been noted by Borg (1981, 1988), cf. Sect. 4.3.

Table 4.17 Morphological marking of active-passive pairs

Patterns No. of roots Percentage Active Passive Gloss

II – V 552 54.76 ħassar tħassar cancel I – VII 337 33.43 qatel inqatel kill I – VIII 56 5.55 niseġ intiseġ weave III – VI 33 3.27 nieda tnieda launch Other 30 2.98 xtara inxtara buy

Note, however, that around 6% of the binyan I active verbs have a passive in binyan VIII not in VII, as expected. The number is relatively high because binyan VII and VIII tend to be synonymous (cf. Sect. 4.2.1.4). For instance, √nfħ creates an active verb in binyan I, nefaħ ‘inflate (v.t)’, with a corresponding passive in VII (intnefaħ) and VIII (intefaħ), which are absolute synonyms, both translating as ‘be inflated, inflate, v.i’. Only four roots, √nfx, √nsġ, √rhn, √rqm, form an active verb in I and their passive in VIII, without appearing in binyan VII too. The category ‘other’ includes a small number of exceptional cases, such as the ex- ample given in the table, where the active is in VIII and its passive counterpart in VII. In another isolated combination, the passive of II is not in V, but in VII, such as √għlm, għallem ‘teach’, intgħallem ‘be taught’, and √mxj, mexxa ‘lead’, intmexxa ‘be led’. A few pairs have an active in I and a passive in VI rather than VII, as expected. These generally

110 A Tale of Two Morphologies

include (defective) roots which give rise to anomalous forms, e.g., ħa ‘I, take’ and ittieħed ‘VI, be taken’; kiel ‘I, eat’ and ittiekel ‘VI, be eaten’. Setting these few exceptions aside, the morphological realization of the active- passive alternation is extremely regular. Complementing this morphosyntactic regular- ity is the semantic regularity of the verbs entering this alternation. Recall the observa- tion made in Ch. 3, that the meanings assigned to passives are typically included in the semantics of their active counterparts, which tend to be more polysemous. This follows from the analysis that passives are verb-derived verbs, merging with a category- bearing head above the root level. Not only is the interpretation of passives generally tied to the semantics of the active verbs, but they also inherit phonological traits, such as gemination and vowel sequences, from the active verbs they are derived from. This issue is taken up in more detail in Sect. 4.3.

Transitive-reflexive alternation This alternation, which is the alternation least represented in the database (6.45%), involves a change in the number of verb arguments. One alternant, the reflexive, is in- transitive, while the other is transitive. In line with many studies on reflexivization (cf. Siloni (2001); Doron & Rappaport Hovav (2009); inter alia), I do not separate reflexive and reciprocal verbs. Arguably, they involve the same syntactic process, and differ only in terms of how events are prototypically performed. Compare, for instance, the event of washing and shaving with the event of marrying or fighting. While one can wash or shave oneself, it is not prototypical for one to marry or fight oneself. Syntactically, reflexivization or reciprocalization is a valency-reducing operation, which generates a reflexive or reciprocal verb out of a transitive verb form. It is basi- cally an operation which identifies two of the variables in a verb’s argument structure, indicating that they have the same referent (cf. Grimshaw 1981). Thus, roughly, reflex- ives are verbs that denote an event that the Agent argument performs on or applies to her/himself, as in (4a). The interpretation of reflexive verbs can also be paraphrased by means of a reflexive argument, as in (4b). In this morphosyntactic analysis, however, I treat as reflexives only those verbs that denote reflexive meaning without realizing a reflexive argumental object, as in (4a).

(4a) It=tifel in-ħasel bi-l=misħun.

DEF=boy REFL-wash.PFV.3SG.M with-DEF=hot water ‘The boy washed with hot water.’

When patterns put down roots 111

(4b) It=tifel fisser ruħ=u ħażin.

DEF=boy explain.PFV.3SG.M soul=his bad ‘The boy explained himself badly.’

An examination of the binyanim that mark reflexive and non- pairs reveals that this alternation displays a significant degree of morphosyntactic and se- mantic regularity. Over three fourths of all tokens are unidirectional: if a reflexive oc- curs in binyan V, VI and VII, then, almost without exception, its transitive counterpart will be in binyan II, III and I respectively. As in the active-passive alternation, there is a small percentage of I – VIII combinations, which result from the presence of a fairly large number of absolute synonyms in VII and VIII. Take, for instance, the root √xħt: the non-reflexive counterpart of VII, inxteħet, and VIII, xteħet, ‘throw oneself’, is in I for both verbs. A handful of binyan VI reflexive or reciprocal verbs (e.g., tħamel ‘tolerate one an- other’, ixxiebah ‘resemble one another’) have their transitive counterpart in I (ħamel ‘tolerate’, xebah ‘resemble’) rather than III, as one would expect. These are taken as ex- ceptions, together with some other pairs, such as I – X (√ħbj, ħeba ‘hide’, staħba ‘hide oneself’), II – VIII (√lqgħ, laqqa’ ‘cause to meet’, iltaqa’ ‘meet’; √sbt, sabbat ‘plump’, sta- bat ‘plump oneself’), and II – VI (√ġld, ġelled ‘provoke a fight’, iġġieled ‘fight’). These ob- servations are summarized in Table 4.18.

Table 4.18 Morphological marking of transitive-reflexive pairs

Patterns No. of roots Percentage Transitive Reflexive Gloss

II – V 61 49.19 sakkar issakkar lock in I – VII 25 20.16 wiżen intiżen weigh III – VI 9 7.26 miera tmiera contradict I – VIII 9 7.26 xeħet xteħet throw I – VI 5 5.65 ħamel tħamel like, tolerate Other 13 10.48 kenn stkenn shelter

As we shall see in Sect. 4.3, with the exception of some verb pairs, the transitive- reflexive alternation is regular in two aspects. First, it is generally possible to predict the form of the transitive from the reflexive, and vice versa. Second, reflexives, being an instance of word (as opposed to root) derivation, inherit the semantics and phonology

112 A Tale of Two Morphologies

of their non-reflexive counterparts. The vowel sequence and interpretation of reflexive verbs are dependent on those of the transitive verbs they alternate with.

Causative-inchoative/noncausative alternation I now turn to verbs that take part in an argument structure alternation, which, like the other alternations discussed so far, generally involves a change in transitivity, reflected semantically by a different event structure. Consider the following pairs:

Root Binyan Verb Meaning Binyan Verb Meaning √fqgħ VII infaqa’ burst (v.i) I faqa’ burst (v.t) √ġrj I ġera run II ġerra cause to run √ħrq VII inħaraq burn (v.i) I ħaraq burn (v.t) √kmx V tkemmex crease (v.i) II kemmex crease (v.t) √ksħ I kesaħ cool (v.i) II kessaħ cool (v.t) √qbl I qabel rhyme (v.i) II qabbel rhyme (v.t) √rbħ I rebaħ win II rebbaħ cause to win √xbh I xebah resemble II xebbah liken

The members on the right are the causative of the verbs to the left, which for the pur- poses of this analysis have been broken into two subgroups, inchoative and ‘other’ non- causative verbs. While the alternation itself is basically the same, there are a number of semantic and syntactic differences between inchoative and noncausative verbs. In- choatives are intransitive and typically denote a change of state, such as burst, cool, burn, crease or become wrinkled.7 Their causative counterparts, then, express a bring- ing about of this change of state. Noncausative verbs, by contrast, may be both transi- tive (win, resemble) and intransitive (run, walk, etc.), and semantically they do not form a homogenous class. As a first observation, the examples above show that noncausa- tives include both eventive (run, win) and stative (resemble, rhyme) verbs (cf. Spagnol 2007, 2009 on criteria for distinguishing states from events in Maltese). Let us now consider the different ways templatic verbs effect the two alternations, starting with the causative-inchoative alternation. Unlike the active-passive and transi- tive-reflexive alternations, the causative-inchoative alternation is morphologically ir- regular. As made evident by Table 4.19, which lists the causative-inchoative pairings in the database, there is a considerable variety in the marking of this alternation.

7 One exception to this generalization is tgħallem ‘learn’, which is transitive.

When patterns put down roots 113

Table 4.19 Morphological marking of causative-inchoative pairs

Patterns No. of roots Percentage Causative Inchoative Gloss

II – V 258 44.64 tappan ittappan blur II – I 97 16.78 saħħan saħan warm up I – VII 74 12.80 qasam inqasam split II – IX 56 9.69 qassar qsar shorten I – VIII 21 3.63 żied żdied increase III – VI 14 2.42 sieħeb issieħeb associate Other 58 10.03 kisser inkiser break

For instance, the inchoative of a causative verb in binyan II may be in I, V and IX. There is almost an equal chance that a binyan I verb participating in this alternation is a causative to binyan VII or an inchoative to binyan II. In addition, there are some roots, listed under ‘other’, that mark the causative-inchoative contrast by one and the same binyan; that is, they are labile, ambitransitive verbs, with one verb form being used both in a causative and an inchoative sense (cf. Ch. 5), as the following examples illus- trate:

Root Binyan Verb Meaning Binyan Verb Meaning √bdj I beda begin (v.t) I beda begin (v.i) √għmj I għama blind I għama grow blind √għġl II għaġġel hurry (v.t) II għaġġel hurry (v.i) √sdd II saddad rust (v.t) II saddad rust (v.i)

Among the less frequent combinations of patterns that mark this alternation are II – VII as well as II – VIII. A few examples include:

Root Binyan Verb Meaning Binyan Verb Meaning √fqr II faqqar impoverish VIII ftaqar become poor √għqd II għaqqad entangle VII ingħaqad get entangled √ksr II kisser break (v.t) VII inkiser break (v.i) √xrd II xerred disseminate VIII xtered spread (v.i)

114 A Tale of Two Morphologies

To anticipate the discussion in Sect. 4.3, the irregularity in the morphological marking of the alternation is paralleled by:

(i) gaps in the system – while passives and reflexives always presuppose an active or transitive counterpart, there are a number of inchoatives that lack a corresponding causative verb (e.g., smar ‘get tanned’), and vice versa (e.g., sann ‘sharpen, v.t’); (ii) vocalic variation – unlike passives and reflexives, which have the same vowels as the transitive verbs they are derived from, causative and inchoative verbs do not always share the same vowel patterns, as in √kbr, kiber ‘grow, v.i’ and kabbar ‘grow, v.t’; (iii) idiosyncratic interpretations – contrary to passives and reflexives, which are se- mantically tied to their transitive counterparts and hardly ever acquire an inter- pretation that is not included in their corresponding active and non-reflexive verbs, causatives and inchoatives often have unpredictable, idiosyncratic interpre- tations.

While causative-inchoative pairings are irregular in several respects, the alternation is not entirely unsystematic. It is regular in terms of directionality. That is, although the inchoative of a binyan II verb can be in different patterns, such as I, V, VIII, IX, it is never the other way round, i.e. that the causative is in I or V and the inchoative in II. This contrast between the irregularity in the morphosyntax and semantics of alter- nating verbs, on the one hand, and the unidirectionality of the binyanim in marking ar- gument alternations, on the other, is also found in causative-noncausative verb pairs. Consider the data in Table 4.20.

Table 4.20 Morphological marking of causative-noncausative pairs

Patterns No. of roots Percentage Causative Noncausative Gloss

II – I 115 54.25 mexxa mexa walk II – V 44 20.75 fewwaq tfewwaq burp III – I 11 5.19 qiegħed qagħad place, stand I – VII 10 4.72 ħasad inħasad startle III – VI 9 4.25 miegħek tmiegħek wallow Other 23 10.85 sabbar stabar console

When patterns put down roots 115

The morphological realization of this alternation is quite varied, like the marking of causative-inchoative verbs. When a causative verb appears in binyan II, its noncausa- tive counterpart can be in the form of either binyan I or binyan V. Similarly, we cannot predict if the noncausative alternant of a causative in III will be in I or VI. The probabil- ity is almost the same. This alternation provides two pieces of evidence against the notion that the bin- yanim serve a specific semantic role. First, a verb in binyan I can be the causative of binyan VII as well as the noncausative of II and III. Second, there are a handful of roots that create a causative and a nonocausative verb within the same verbal pattern. These are four examples of such ambitransitive verbs:

Root Binyan Verb Meaning Binyan Verb Meaning √ħrġ I ħareġ take out I ħareġ go out √pġj II poġġa make one sit II poġġa sit √sfr III siefer take abroad III siefer go abroad √tmj VI ittama make one hope VI ittama hope

This variation in the morphological expression of the alternation goes hand in hand with gaps (e.g., while, qabeż ‘I, jump’ alternates with qabbeż ‘II, make one jump’, there is no causative form corresponding to qomos ‘I, skip, bounce’) and vowel changes (e.g., żifen ‘I, dance’ and żeffen ‘II, make one dance’). Still, the causative-noncausative alterna- tion is to a small extent regular. As with the other three argument alternations dis- cussed above, the relations between most of the pairs of binyanim are constantly direc- tional. In other words, while for a given causative-noncausative alternation involving binyan III, we cannot tell if the corresponding verb will be in I or VI (or another pat- tern), we can nonetheless predict that binyan III will accommodate the causative alter- nant and not the noncausative one (cf. Sect 4.3).

4.2.1.3 Multiple interpretations

The third category comprises roots that acquire very different interpretations in differ- ent verbal patterns. As the following examples taken from Ch. 3 illustrate, the semantic relations that hold between these verbs are tenuous to say the least.

116 A Tale of Two Morphologies

Root Binyan Verb Meaning Binyan Verb Meaning √dħl I daħal go in VII indaħal interfere √slt I silet extract VI issielet struggle √xrb I xorob drink II xarrab wet, drench

In spite of the semantic distance between such verbs, some pairs are semantically closer (‘go in’ and ‘interfere’) than others (‘extract’ and ‘struggle’). Some meaning pairs may be historically related. Compare, for instance, the verb forms derived from √xrb with the English drink and drench, which hark back to the same Germanic etymon. Still, the semantic association has faded away synchronically in Maltese as in English. What is crucial for our categorization is that there is no systematic alternation known in the literature that can relate such verbs. Unlike argument alternations, such as ħasel ‘wash’ and inħasel ‘wash oneself’, where the reflexive verb semantically entails its transitive counterpart, there is no such semantic entailment in the case of multiple interpreta- tions. This criterion, as introduced in Ch. 3, was used to distinguish semantically related verbs (argument alternations and synonyms) from multiple interpretations. There is considerable chaos in the morphological marking of this category. No clear pattern comes out as regards which combinations of binyanim tend to realize mul- tiple interpretations of roots. Consider the following examples:

Root Binyan Verb Meaning Binyan Verb Meaning √fdl I fadal remain II faddal save V tfaddal be saved √frd I fired separate (v.t) II farrad make unpaired VII infired separate (v.i) V tfarrad become unpaired √ftħ I fetaħ open (v.t) II fettaħ loosen, expand VI infetaħ open (v.i) V tfettaħ be loosened √ħqq I ħaqq deserve II ħaqqaq assert, contend X stħaqq deserve V tħaqqaq be contended √ħsb I ħaseb think II ħasseb worry (v.t) VI inħaseb be thought V tħasseb worry (v.i) VI tħaseb worry (v.i) √krh II kerrah make ugly X stkerrah abhor V tkerrah be made ugly IX krieh become ugly

When patterns put down roots 117

√lqt I laqat hit II laqqat collect VII intlaqat be hit V tlaqqat be collected VIII iltaqat be hit √mwt I miet die II mewwet discourage V tmewwet lose heart √nfs II niffes breathe into V tniffes utter a sound √swj I sewa cost II sewwa repair V issewwa be repaired √tbgħ I tebagħ print II tebba’ stain (v.t) VII intebagħ be printed V ittebba’ stain (v.i) √tnj I tena fold II tenna repeat VI intena be folded V ittenna be repeated √xtj I xita rain II xitta winter

The occurrence of one root in three or more different patterns often marks a combina- tion of argument alternations, multiple interpretations, and synonyms. For instance, √krh forms an active-passive alternation in II and V, a causative-inchoative alternation in II and IX, while the interpretation of the verb in X is not semantically related to the other three verbs in any systematic or predictable way. Two recurring combinations involving multiple interpretations are I – II/V, as in √fdl, √mwt, √swj, and I/VII – II/V, as in √frd, √ftħ, √tbgħ, √tnj. This last combination of patterns was also noted by Borg (1981, 1988), who analyzes the data as cases of root homonymy, with one root creating verbs in I and VII, and another (homophonous) root appearing in II and V. Finally, there are a small number of roots that give rise to two instances of multiple interpretations. Let us take two examples. Depending on the binyanim it merges with, √ħrġ can express an event of taking out, an event of training or an event of investigat- ing. Likewise, the appearance of √slt in different verbal patterns can correspond to an event of extracting/unsheathing, an event of fraying, or an event of struggling. Accord- ing to Aquilina’s (1987-1990) dictionary, the dated verb in binyan VIII, stilet, may be both the passive of silet ‘unsheathe’ and the inchoative of II sellet ‘fray’, forming a syno- nym with the verb in VII and another synonym with the verb in V respectively.

Root Binyan Verb Binyan Verb Meaning Category

√ħrġ I ħareġ VII inħareġ take ACT-PASS

II ħarreġ V tħarreġ train TRANS-REFL

X stħarreġ investigate MULT. INTERP.

118 A Tale of Two Morphologies

√slt I silet VII insilet extract, unsheathe ACT-PASS

II sellet V issellet fray CAUS-INCH

VIII stilet unsheathe, fray PASS, INCH

VI issielet struggle MULT. INTERP.

4.2.1.4 Synonyms

One important finding of the qualitative analysis of the database is that the binyan sys- tem in Maltese contains a significant number of roots (18%) which create synonymous verbs in different patterns, as the following examples show:

Root Binyan Verb Binyan Verb Meaning √bdl I bidel II biddel change (v.t) √ġgħl I ġagħal III ġiegħel force √ksr I kiser II kisser break (v.t) √nfħ VII intnefaħ VIII intefaħ inflate (v.i) √qtr I qatar II qattar drip (v.i) √sdd II saddad V issaddad rust (v.i) √slf I silef II sellef lend √slħ VII instelaħ VIII stelaħ get bruised √sqj I seqa II saqqa give water/liquid to √tql I teqel IX tqal become heavy √xbh VI ixxiebah VIII xtiebah resemble each other √żqq V iżżaqqaq IX żqaq become potbellied

Some of the above pairs, such as those derived from √ġgħl, √nfħ, √slf, √slħ and √xbh, are absolute synonyms. The rest are near or contextual synonyms. For instance, the dif- ference between kiser and kisser lies in the nominal arguments they take: while the former typically collocates with body parts, laws, silence, and the like (5a), the latter collocates with other breakable entities (glass, furniture, appliances, etc.), traditions, families, and others (5b), including a few body parts kiser does not co-occur with (5c).

(5a) Kiser sieq=u.

break.PFV.3SG.M leg=his ‘He broke his leg.’ (5b) Kisser il=vażun.

break.PFV.3SG.M DEF=vase ‘He broke the vase.’

When patterns put down roots 119

(5c) Kisser ras=u.

break.PFV.3SG.M head=his ‘He broke his head; he racked his brain.’

Turning to another root, √sqj, the difference between the verb in binyan I and the verb in binyan II is also in terms of collocation. While seqa is generally used with humans and animals, e.g., sqejt it-tarbija ‘I gave the baby water, milk, etc.’, saqqa tends to co- occur with crops, plants, and the like, e.g., saqqejt il-fjuri ‘I watered the flowers’. In other synonymous couples, by contrast, there is a lot of overlap. A case in point is the couple sharing the root √bdl. Both bidel and biddel collocate freely with a noun like kors ‘course’, e.g., bidel il-kors and biddel il-kors ‘he changed his course (of study)’. However, clothing items and topics, for example, typically co-occur with biddel, e.g., biddel il-qmis ‘he changed his shirt’, biddel id-diskors ‘he changed the subject’, but not with bidel. Note that bidel il-qmis is actually attested, but with a different meaning, ‘he swapped or exchanged the shirt’. One member of certain synonymous pairs is likely to be more polysemous than the other. Consider the root √sdd. In determinable contexts, the verbs it creates have iden- tical meanings, as in (6a, 6b). However, binyan II saddad can also mean ‘cause to rust’ (7a), ‘last for a long time’ (7b), and so on.

(6a) Il=ħadid qed i-saddad.

DEF=iron PROG IPFV.3SG.M-rust ‘The iron is rusting.’ (6b) Il=ħadid qed ji-s-saddad.

DEF=iron PROG IPFV.3SG.M-INCH-rust ‘The iron is rusting.’

(7a) L=ilma baħar i-saddad il=ħadid.

DEF=water sea IPFV.3SG.M-rust DEF=iron ‘Seawater rusts iron.’ (7b) Dawk it=tnejn sadd-u flimkien.

those DEF=two rust-PFV.3PL together ‘Those two have been together for ever, lit. those two rusted together.’

120 A Tale of Two Morphologies

For some other synonymous pairs, the difference is that one verb is dated (e.g., teqel and żqaq from the data above) and the other is current (e.g., tqal ‘become heavy’ and iżżaqqaq ‘become potbellied’). Besides absolute and contextual synonyms, this category comprises a group of verbs with subtle differences in their aspectual interpretation. Consider the examples below: the verbs on the right express iterative aspect, which in the literature (e.g., Sut- cliffe 1936; Mifsud 1995a), has been referred to as “intensive meaning”. As we shall see below, the distinction between roots that create absolute and contextual synonyms, on the one hand, and roots that form iterative and non-iterative verbs, on the other, is re- flected in how they are mapped into the various binyanim.

Root Binyan Verb Meaning Binyan Verb Meaning √bkj I beka cry V tbekka cry often √bws I bies kiss II bewwes kiss repeatedly √ġbd I ġibed pull II ġebbed pull and pull √gdm I gidem bite II giddem bite frequently √xmm I xamm smell II xammem smell repeatedly

Having surveyed the main types of synonymous verbs in the binyan system, let us now examine their distribution in the database. As Table 4.21 shows, around four fifths of the 557 roots in this category are absolute or contextual synonyms. The rest are cases of verbs with identical meaning which mark an aspectual distinction, namely it- erative phasal aspect (cf. Spagnol 2009 for a distinction between grammatical, lexical and phasal aspect).

Table 4.21 Roots creating synonymous verbs

Subgroups of synonyms No. of tokens Percentage

Synonyms 440 79 Iterative aspect 117 21

Taking the two subgroups separately, I will now analyze which patterns the roots in this category combine with. Table 4.22 lists the major combinations of patterns for - solute and contextual synonyms, with the number of roots appearing in each combina-

When patterns put down roots 121

tion. The table displays the most frequent pairs and triplets only. There are numerous other combinations, which are, however, infrequent. I group them all together under ‘other’.

Table 4.22 Most frequent combinations of absolute and contextual synonyms

Patterns No. of roots Percentage Examples Meanings

I – II 93 21 basar bassar foresee VII – VIII 75 17 instadd stadd clog up V – VII 51 12 iddellek indilek be smeared II – V 37 8.4 sewwes issewwes become worm-eaten I – V 29 6.6 korob tkarrab complain I – VI 18 4.1 nagħas tniegħes doze off V – IX 17 3.9 twebbes ibbies become hard I – IX 13 3 boloq blieq lose one’s prime V – VII – VIII 13 3 ixxerred inxtered scatter xtered Other 94 21.4 twieġeb intwieġeb be answered

Table 4.23 summarizes the occurrences for verbs expressing iterative aspect across dif- ferent patterns. Combinations of patterns constituting less than 2% of the total number of roots creating iterative and non-iterative pairs are grouped under ‘other’.

Table 4.23 Most frequent combinations of iterative aspect

Patterns No. roots %age Example Meaning Example Meaning

I – II 86 74 rass press rassas press repeatedly VII – V 10 8.5 inġibed be pulled iġġebbed be pulled repeatedly I – V 7 6 talab ask for ittallab beg, ask many a time I – VI 5 4.4 daħak laugh iddieħak laugh frequently I – III 3 2.6 leheġ pant lieheġ pant frequently Other 6 5.1 liegħaq lick tliegħaq lick here and there

From the data above we can note that the most common combination of patterns in which synonyms, whether absolute, contextual or aspectual, occur is I – II. However,

122 A Tale of Two Morphologies

there appears to be a significant difference between the two subgroups of synonyms. While nearly three fourths of the aspectual distinctions are marked by binyan I and II, absolute and contextual synonyms tend to be distributed over three different combina- tions, namely I – II, VII – VIII, and V – VII. It is interesting that while these last two combinations are reserved for synonyms, the combination I – II is also used to mark non-causative/inchoative and causative verbs. This suggests that overall there is a tendency for complementary distribution. That is, while a combination such as I – VII is most likely to mark an argument alterna- tion (cf. Sect. 4.2.1.2), the combinations VII – VIII and V – VII are almost exclusively as- sociated with synonyms. Finally, the presence of a fairly large number of synonymous verbs derived from the same root argues against the deterministic approach to the binyan system, which holds that the binyanim serve fixed functions, such as II being the causative of I. In the examples discussed above, the binyan II verbs are not the causative of their binyan I counterparts. Rather they are cases of verbs with the same meaning in multiple forms.

4.2.2 Section summary

A clear picture of how roots interact with verbal patterns emerges from the qualitative analysis of the binyanim system, together with the quantitative results in Sect. 4.1. The main findings of this morphosyntactic investigation are the following. We have established that 88% of all tri-consonantal roots appear in more than one verbal pattern. Around two thirds of these mark argument alternations. The rest either create synonymous verbs (18%) or are assigned multiple interpretations in the context of different binyanim (9%). Note that the percentages are from the total number of categories tri-consonantal roots fall into. The four categories were then examined in terms of the patterns in which they are realized. The analysis revealed that the traditional characterization of binyanim in terms of lexical or semantic roles, such as causative and reflexive, are no more than

tendencies. For instance, the so-called “causative” binyan II, C1vC2C2vC3, has non- causative verbs appearing in it. Recall intransitive denominal verbs, such as passa ‘walk back and forth’ and tejjeġ ‘wed’ (Sect. 4.2.1.1), and labile, ambitransitive verbs, such as għaġġel ‘hurry up’ and saddad ‘rust’, which create a causative and inchoative verb in binyan II (Sect. 4.2.1.2). Conversely, there are many causative verbs that do not occur in this binyan, but in I, III and VIII. Therefore, we have to give up the one-to-one corre-

When patterns put down roots 123

spondences between verbal patterns and canonical functions. This, however, need not force a rejection of the notion of systematicity in the binyan system. While the contribution of individual binyanim is not transparent, the contribution of combinations of binyanim is, to a certain degree, predictable. There is a strong ten- dency for complementary distribution. For instance, with the exception of four isolated cases,8 the combination VII – VIII never marks argument alternations. It typically corre- sponds to absolute or contextual synonyms. By contrast, the combination I – VII almost always marks an argument alternation. There are one or two exceptions to this. For in- stance, when cast in I and VII, both √qlgħ and √qtgħ express argument alternations, such as active-passive, qala’ ‘I, earn’, inqala’ ‘VII, be earned’, and causative-inchoative, qata’ ‘cut, v.t’, inqata’ ‘cut, v.i’, as expected. But in two particular senses, they are syn- onymous: bit-talb qala’/inqalgħetlu grazzja ‘through prayer he earned a grace’, and qata’/inqatagħlu nifsu ‘he was out of breath’. And, within the category of argument al- ternations, the combination of I – II marks causative and inchoative/noncausative verb pairs but never marks active-passive or transitive-reflexive pairs. Still, not all combinations exclude each another. There are almost equal chances that the appearance of a root in binyan I – II or in II – V expresses an argument alterna- tion or creates synonymous verbs. This is a consequence of the fact that binyan mor- phology, as morphology in general, does not have a one-to-one character. Rather, the relation between morphological marking and syntactic structure is syncretic in two ways. First, the same morphological patterns mark more than one syntactic feature, e.g., I – II may, among others, represent a causative-inchoative alternation (√qbd, qabad ‘catch fire’, qabbad ‘set on fire’) as well as an aspectual distinction (√gdm, gidem ‘bite’, giddem ‘bite repeatedly’). Second, the same syntactic feature is marked by more than one pair of patterns, e.g., the active-passive alternation may be realized in I – VII (kiteb ‘write’, inkiteb ‘be written’), in II – V (kisser ‘break’, tkisser ‘be broken’), III – VI (wieġeb ‘answer’, twieġeb ‘be answered’), and so on. To counterbalance the syncretic nature of the verbal patterns is the regularity that arises from the directionality of verb alternation. As was noted in Borg (1981, 1988) and Borg & Mifsud (1999), binyanim interact markedly with syntax. Because patterns like V, VI, VII and IX are inherently or predominantly intransitive, then, in an argument

8 Namely, √ħjġ, ħtieġ ‘VIII, need’, inħtieġ ‘VII, be needed’; √xrj, xtara ‘VIII, buy’, inxtara ‘VII, be bought’; √xrr, xtarr ‘VIII, ponder’, inxtarr ‘VII, be pondered’; √xwq, xtaq ‘VIII, desire’, inxtaq ‘VII, be desired’.

124 A Tale of Two Morphologies

alternation they are bound to realize the passive, reflexive, and/or inchoative alternant. By contrast, II and III, being predominantly transitive, typically mark the active, non- reflexive and causative members of an alternating pair. In other words, given two verbs derived from the same root in patterns II and V, by their morphology we cannot tell whether they express an argument alternation, form synonyms, or have multiple interpretations. However, if we do know that it is an in- stance of a verb alternation, then we can predict that the transitive alternant will be hosted by II and the intransitive one will appear in V. Other pairs are also unidirec- tional. For instance, it is never the case that a passive is in III and its corresponding ac- tive in VI. Likewise, if two verbs in I and VII take part in a transitive-reflexive alterna- tion, then we can tell in advance that the reflexive verb will be in the shape of binyan VII and never the other way round. Slightly more complicated is the situation involving patterns I, VIII and X, as they may host either transitive or intransitive verbs. Therefore, in an argument alternation, a verb in binyan I can, for example, be both the causative (e.g., √ftħ, fetaħ ‘I, open, v.t’, infetaħ ‘VII, open, v.i’) as well as the inchoative (e.g., √nxf, nixef ‘I, dry, v.i’, nixxef ‘II, dry, v.t.’) alternant. Still, for these verb pairs, knowing that the combination of patterns in- volves a typically transitive binyan (e.g., II) and an inherently intransitive binyan (e.g., VIII), we can nonetheless tell in advance that the verb in binyan I will be intransitive if its alternant is in II, but it will be transitive if its alternant is in VII. Taking another example, from the above analysis we know that a root inserted in VII and VIII is most likely to yield two synonymous verbs or, possibly, a pair with differ- ent interpretations. However, in four isolated cases (see footnote 8), the combination VII – VIII marks argument alternations. Given that a verb in VII is always intransitive whereas a verb in VIII may be both transitive and intransitive, then, in this case too, it follows from the directionality of verb alternations that the active will, with no excep- tion, be realized in VIII and the passive in VII. In a nutshell, then, an examination of the morphological marking of the four cate- gories roots fall into brings to the surface two conflicting elements. On the one hand, the system is morphologically transparent, imposing transitivity constraints on the bin- yanim, which then resurface as consistent directionality in the morphological realiza- tion of argument alternations. On the other hand, templatic verbs exhibit a large degree of syncretism. While it is true that some combinations of patterns are restricted to cer- tain categories (e.g., I – VII and III – VI mark almost exclusively argument alternations, and VII – VIII and V – VII are reserved for synonyms), other combinations may corre-

When patterns put down roots 125

spond to more than one category. This is so with verbs appearing in II and V or I and II, which may be synonyms, alternating verbs, or an instance of multiple interpretations. Moreover, even if we take a group of patterns that are most likely to express an argu- ment alternation, such as I – VII, we cannot tell whether they will take part in the active- passive, transitive-reflexive, or in causative-inchoative alternation. These two extremes of transparency and opacity, I argue, are two major character- istics of templatic verbs in Maltese. Now that we have delineated the main sources of regularity (directionality) and irregularity (syncretism) and the extent to which the sys- tem is regular and irregular, the next step is to give an explanation for these two diver- gent properties of templatic verbs by linking the above findings with the structural dis- tinction between root-derived and verb-derived verbs introduced in Ch. 3.

4.3 A(n ir)regular binyan system

The aim of this section is to summarize the most important conclusions drawn in the previous and the current chapter, which tried to (i) identify the regular and irregular aspects of the binyan system; (ii) describe the extent to which the system is transparent and opaque; and (iii) account for the dual nature of templatic verbs by means of a prin- cipled distinction in the process of word formation. Let us give a brief outline of the as- sumptions and results (Sect. 4.3.1), in order to bring together the key contributions of this analysis of templatic verbs in Maltese (Sect. 4.3.2).

4.3.1 Main findings and assumptions

Semantic (un)relatedness The point of departure for the preceding discussion was the puzzle that the semantics of words containing the same root consonants may be close and relatively predictable (e.g., xorob ‘I, drink’, xorb ‘drinking’), on the one hand, and distant and idiosyncratic (e.g., xarrab ‘II, wet’, inxtorob ‘VII, shrink’), on the other.

Underspecified roots In order to resolve the above puzzle, I resorted to (i) one tenet of Distributed Morphol- ogy, the so-called Root Hypothesis; that is, the idea that nouns, verbs and adjectives, regardless of their morphological complexity, may be decomposed into roots, atomic

126 A Tale of Two Morphologies

lexical elements; and to (ii) the assumption that roots are underspecified on three lev- els. First, roots are category-neutral. It is only when they merge with a word-creating head that an actual verb, noun or adjective is formed. Second, roots are incomplete from a phonological point of view: √ktb, for instance, is unpronounceable on its own. Third, the semantic content of roots is not fully definable. Just as √ktb becomes a continuous, pronounceable string when inserted in a verbal or nominal pattern, roots acquire spe- cific interpretations in the environment of different patterns.

Contribution of patterns Contrary to the deterministic approach taken by traditional grammars of Maltese, there is no exclusive mapping between the semantic and syntactic property of verbs and their morphological realization. Rather, (i) the same binyan may host verbs of different se- mantic types, e.g., a verb in binyan VII may be passive, reflexive, inchoative;9 and (ii) verbs of the same semantic type (e.g., reflexive) may appear in different binyanim (e.g., V, VI, VII, VIII, X). A significant contribution of the binyanim to morphological transpar- ency is instead in terms of transitivity. Some of them (VI, VII, IX) are inherently intransi- tive, which means that verbs that are typically transitive, i.e. causative, active and non- reflexive verbs, will not appear in these non-transitive binyanim. The same holds for binyan V, which is predominantly intransitive. Other verbal patterns are almost always transitive (II, III) and are therefore incongruent with inchoatives, passives and reflex- ives. The valency of the remaining three patterns, I, VIII, X, is, however, inconsistent.

Distribution of roots The mean number of patterns per root is around 2, which means that the binyan system is full of gaps. In fact, on examining the distribution of roots across patterns, it turns out that over two thirds of tri-consonantal roots combine with only one or two patterns, leaving several binyan slots empty. The rest typically occur in three (17%) or four (11%) different patterns. While no root appears in all nine binyanim, there are very few roots (less than 5%) that are inserted in five, six, or seven patterns.

Combinatorial constraints The combination of roots and patterns is subject to a few phonological constraints. Bin- yan VIII, which involves a -t- infixed between the first C- and V-slot, does not combine

9 The sole exception is binyan IX, which is associated almost exclusively with inchoatives.

When patterns put down roots 127

with roots whose first consonant is a stop /b, p, t, d, k, g, ʔ/, an affricate /tʃ, dʒ, ts/, a glide /j, w/, or silent għ and h. By contrast, it hosts roots with liquids, nasals, fricatives, and sibilants as first radical. For phonotactic reasons, certain roots are not inserted in binyan X. Being characterized by a prefix st-, it does not combine with roots starting with /t, tʃ, d, dʒ, ts/ etc., as they would create onset clusters that are not permissible in Maltese. Silent-medial roots tend not appear in binyan II and V, as the virtual phonemes għ and h cannot be geminated. In addition, there are some dispreferences for root types. For instance, reduplicative roots tend not to be embedded in binyan III and VI; weak-initial roots do not appear in VIII (see restriction above on glides as first radical), and they rarely appear in I and VII; weak-medial roots hardly ever occur in III, VI and VIII; and weak-final roots do not usually combine with II, V and IX.

Correspondences among patterns Subject to such combinatorial constraints, tri-consonantal roots can in principle com- bine with any of the nine binyanim to create different verbal lexemes. However, an ex- amination of the distribution of roots across patterns revealed that only four binyanim are really productive, I, II, V and VII. The other four are underrepresented, especially X. Several grammars of Maltese observe four similarities between the productive and the non-productive patterns. First, on the equivalence of binyan II and III, Sutcliffe (1936: 84) remarks, “[t]he third form [… ] is practically an extension of the second, and has the same meanings”. Similar observations have been made by Borg (1981, 1988) and Cachia (1994: 207), among others. Second, a parallel equivalence obtains between binyan V and VI, which, as we have seen above, regularly mark the passives, reflexives and inchoatives of the correspond- ing transitive verbs in II and III respectively.10 Third, the qualitative analysis in Sect. 4.2 has also brought to light the fact that 17% of the absolute and contextual synonyms are marked by VII and VIII. That several verbs in VII and VIII have identical meanings was already observed by Sutcliffe (1936: 97), “[s]ome verbs are found both in the seventh and in the eighth form without change of meaning.” Fourth, Borg (1981: 90) identifies a parallelism between intransitive verbs in bin- yan I and verbs in IX: “Semantically, first form verbs such as ‘kiber’ “he grew” and

10 Borg (1981: Ch.3) notes a semantic equivalence of V, VI and VII, which is confirmed by the results in Sect. 4.2.

128 A Tale of Two Morphologies

‘għolob’ “he became lean” are indistinguishable from ninth form verbs such as ‘ċkien’ “he grew small” and ‘ħxien’ “he grew fat”.” The semantic and syntactic equivalence between the productive I, II, V, VII and the respective non-productive IX, III, VI, VIII is mirrored by morphological similarity:

• Both binyan V (tC1vC2C2vC3) and VI (tC1vvC2vC3) involve a prefix t-, which under- goes complete assimilation when the first radical of the root is a coronal.

• Binyan I (C1vC2vC3) and IX (C1C2vvC3), by contrast, are prefixless. In addition, some inflectional forms of I and IX are identical, namely the first and second persons of

the perfective. Compare, for instance, kiteb ‘I, write’, 1SG ktib-t and 1PL ktib-na, with

swied ‘IX, get dark/tanned’, 1SG swid-t and 1PL swid-na. • Some grammars, such as Cachia (1994: 207), claim that the only difference between

II (C1vC2C2vC3) and III (C1vvC2vC3) is that the latter usually combines with roots whose second radical is an ungeminatable silent għ or h, though they also note that some binyan III verbs are derived from roots with a /r/ or /f/ as their middle radi- cal, which can be geminated.

• Finally, consider the formal equivalence of binyan VII (ntC1vC2vC3) and VIII

(C1tvC2vC3), especially when they combine with nasal- and liquid-initial roots, such as √nfħ, creating intnefaħ and intefaħ, which, in addition, happen to be synonymous, ‘inflate, v.i’. Besides, recall that metathesis takes place with the prefixal t- of binyan VII and roots that have a sibilant for first radical, yielding forms such as inxtorob (for *int-xorob) ‘shrink’. At first glance, these verb forms appear to be hybrids of binyan VII (prefix n-) and VIII (infix -t-), and, in fact, in some dictionaries they are listed as VII+VIII verbs (cf. Aquilina 1987-1990).

These correspondences suggests that the four pairs of binyanim have a quasi- allomorphic status. They are virtually in complementary distribution. This claim is sup- ported by the analysis of the co-occurrence of patterns on the same root in Sect. 4.1.3, which revealed that less than one percent of all tri-consonantal roots appear in both I – IX (0.98%) and II – III (0.47%).11 An examination of Table 4.13 shows that 1.4% of tri- consonantal roots appear in V and VI, and that 2.09% co-occur in VII and VIII. Note that when two verbs in V and VI share the same root, the latter verb form is very often dated

11 A root appearing in I and IX may correspond to an instance of multiple interpretation (e.g., √sbħ, sebaħ ‘dawn’, sbieħ ‘become beautiful’) or, more likely, to two synonymous verbs, e.g., √twl, tal, twal ‘become long’, √smn, simen, smien ‘get fat’. One tends to be more frequent or current (twal, simen) than the other. When a root appears in II and III, the verbs created are synonymous, e.g., √wld, welled, wieled ‘give birth’, √ġld, ġelled, ġieled ‘provoke a fight’. The verb in III is usually dated.

When patterns put down roots 129

and synonymous with the binyan V verb, e.g., iġġieneb ‘move aside’, tbiedel ‘be ex- changed’, tħaseb ‘worry’, tħalat ‘be mixed’. The number of roots co-occurring in VII and VIII is relatively ‘higher’ because of the propensity roots have for creating synonyms in these two patterns. In terms of productivity, morphosyntactic similarity, and distribution in the lexi- con, we can therefore conclude that Maltese templatic verbs make up a four-way sys- tem consisting of binyan I, II, V and VII, which in a few cases alternate with their ‘shadow’ binyanim IX, III, VI and VIII respectively. Binyan selection is often phonologi- cally conditioned. For instance, silent-medial roots almost always appear in III rather than II, because they cannot be geminated, and weak-medial roots are more likely to be inserted in V than VI (cf. the (dis)preferences of root types by binyanim Sect. 4.1.2). Fi- nally, binyan X is highly underrepresented, hosting only 24 roots (0.74%), seven of which have practically fallen out of use (cf. Sect 2.2.2).

Vowel sequences of binyanim In passing, it was noted that the vowel sequence of some verbs derived from one root is the same throughout, e.g., ħareġ ‘I, take out’, ħarreġ ‘II, train’, tħarreġ ‘V, train, v.i’, inħareġ ‘VII, be taken out’, stħarreġ ‘X, investigate’, while the vowels of other verbs change from one pattern to the other, kiber ‘grow, v.i’, kabbar ‘grow, v.t’, tkabbar ‘be grown’.

Four categories A qualitative analysis of the interactions between roots and patterns has shown that tri- consonantal roots fall into four major categories. Argument alternations, such as active- passive and causative-inchoative, make up the largest group. The other three categories consist of roots that derive singleton verbs, roots that acquire multiple interpretations, and roots that form synonymous verbs.

4.3.2 Verb formation and the four categories

To conclude this section on the binyan system in Maltese, I integrate the discussion on the morphological realization of the four categories roots fall into with the structural distinction between root derivation and word derivation, which was introduced in Ch. 3

130 A Tale of Two Morphologies

to explain why the system has both transparent and opaque elements. Let us go over the most important observations concerning the distinction in word formation. In line with Marantz and Arad’s model of word formation, couched within the framework of Distributed Morphology, I argued in Ch. 3 that words can be formed in syntax in two ways: (i) by merging a category-free root with a syntactic head, turning the root in a noun, verb or adjective, or (ii) by combining a previously formed word with a new syntactic head. From this distinction follows a locality constraint on the in- terpretations and forms of roots: root-formed words are more likely to display mor- phonological irregularities and semantic idiosyncrasies than word-formed words. Refer to Sect. 3.3 for empirical evidence from different languages, including Maltese, in favor of the distinction between root and word derivation, and for potential counterexamples to the predictions made by the model. Based on these assumptions and generalizations, we can reinterpret the morpho- syntactic realization and semantic behavior of the four categories (singletons, argument alternations, multiple interpretations, and synonyms) in light of the distinction in word formation. To preempt, the result is that singleton verbs, synonyms, multiple interpre- tations, causative, inchoative and noncausative verbs are root-derived whereas pas- sives and reflexives are word-derived.12 Let us now survey the two groups in some de- tail, beginning with cases of root derivation.

Root-derived verbs Unsurprisingly, root derivation gives rise to verbs belonging to very different catego- ries. Verbs directly created from a root range from verbs that stand on their own (sin- gletons) to verbs that are morphologically related to other verbs (causatives, synonyms, etc.); from verbs whose meanings are systematically related (synonyms, verbs entering the causative-inchoative alternation) to verbs that are semantically quite far apart (multiple interpretations). Singletons are root-derived for two main reasons. First, they display many gaps in the system. The actual presence of singletons means there are eight other binyanim, which, at the intersection with that same root, yield a phonologically ill-formed verb (e.g., a stop-initial root in binyan VIII) or a verb form that cannot be interpreted. Sec- ond, the morphological marking of singletons is relatively variable, in the sense that

12 Strikingly, these findings echo to a great extent the conclusions Arad (2005: Ch. 6) draws for templatic verbs in Hebrew, suggesting strong similarities in the verbal systems of Semitic languages.

When patterns put down roots 131

they may appear in any of the nine patterns, as the following examples show, even though most of them occur in I and II (cf. Sect 4.2.1.1).

Root Binyan Verb Meaning

√xrq I xeraq suit √slj II salla swear √lgħb III liegħeb drool √ħsr V tħassar take pity on √bgħl VI tbagħal work hard √fxx VII infexx erupt in √sjd VIII stad fish √żrq IX żrieq become azure √’dn X stieden invite

It goes without saying that verbs with multiple interpretations are also formed at the first merging of a root with a functional head. As Arad (2005: 203) rightly observes, the only difference between these verbs and singletons lies in “the number of verbal environments in which they are assigned an interpretation.” In fact, unlike roots that take on an interpretation in one pattern only, the roots belonging to the category of multiple interpretations acquire a different interpretation in more than one pattern. And these interpretations are irreconcilable. Consider a couple of examples mentioned in Sect. 4.2.1.3: the interpretations that √lqt or √tbgħ are assigned in binyan I (laqat ‘hit’, tebagħ ‘print’) and in binyan II (laqqat ‘collect’, tebba’ ‘stain’) are not mutually de- rived. In addition, the analysis of the morphological realization of the roots in this cate- gory demonstrated that there is no clear pattern as regards the forms that mark verbs with multiple interpretations. Although synonyms tend to be realized in particular groups of patterns such as I – II (e.g., bidel, biddel ‘change’) and VII – VIII (e.g., instadd, stadd ‘clog up’), as shown in Sect 4.2.1.4, they nonetheless constitute a case of root derivation because they exhibit a lot of variation in the morphological marking. In fact, synonymous verbs may appear in a variety of 26 different pairs, 15 different triplets, and 7 different quadruplets. In addition, both multiple interpretations and synonyms may include vocalic varia- tion, which is typical of root derivation. Some examples of synonyms derived from the same root that have different vowel sequences include the following:

132 A Tale of Two Morphologies

Root Binyan Verb Meaning

√ġbd V ġebbed be pulled repeatedly VII inġibed be pulled √qlj I qela fry II qalla fry √slf I silef lend V sellef lend √sqj I seqa give water to II saqqa water, irrigate

Roots that are assigned multiple interpretations in the context of different binyanim may also have different vowel patterns. Consider the following verb forms:

Root Binyan Verb Meaning

√tkk II tikkek put dots on letters V ittekkek have blemishes √xrb I xorob drink II xarrab wet

Of particular interest are the already mentioned roots √frd and √tbgħ, each creating four different verbs in I, II, V and VII. The verbs in I – VII and II – V mark different argu- ment alternations, such as active-passive tebba’ ‘stain’, ittebba’ ‘be stained’, and transi- tive-reflexive fired ‘separate, v.t’, infired ‘separate, v.i’.

Root Binyan Verb Meaning Binyan Verb Meaning √frd I fired separate (v.t) II farrad make unpaired VII infired separate (v.i) V tfarrad become unpaired √tbgħ I tebagħ print II tebba’ stain (v.t) VII intebagħ be printed V ittebba’ stain (v.i)

The two alternating pairs are, however, difficult to relate derivationally. Because all verb forms are derived from the same semantically underspecified root, it is perhaps conceivable to metaphorically associate the state of being unpaired with that of being separated. Similarly, it is not impossible to find figurative links between the event of staining and the event of printing. However, they do not semantically entail each other in the same way that, for instance, an event of separating reciprocally entails the event

When patterns put down roots 133

of separating. Evidence in support of this claim that these two roots are assigned multiple inter- pretations in binyan I and II comes from the morphonology of the verb pairs. The verb forms derived from √frd show vocalic variation, i-e in binyan I, a-a in binyna II. The verbs created from √tbgħ retain the same vowel sequence, however the virtual pho- neme għ gets realized as /ħ/ in binyan I, /ˈtɛbɐħ/, and has no phonetic realization in binyan II, /ˈtɛbbɐ/. I argue that the verbs in binyan VII and V are verb-derived from the verbs in I and II respectively. For this reason, they inherit not only the form of the base verbs (same vowel sequence, same phonetic realization of għ), but also the semantics of the corresponding verbs (they, in fact, constitute cases of argument alternations). These observations are in line with the predictions about the morphonological and semantic behavior of root-derived and word-derived words. Let us now turn to more cases of verb derivation, leaving causative and non-causative verbs for last.

Verb-derived verbs The above cases of root derivation, where verbs are built at the first merging of a root with a functional head, involve gaps in the system, special meanings assigned in differ- ent verbal environments, vowel change, and a great deal of variation in the morphologi- cal realization of the verbs falling in a given category. Passives and reflexives, on the contrary, are almost always morphologically and semantically regular. An examination of their behavior in the binyan system suggests that they must be instances of verb derivation, as they are dependent on the form and meaning of their transitive counter- parts. Evidence for this claim comes first of all from the asymmetry between actives and passives, on the one hand, and non-reflexives and reflexives, on the other. While all pas- sives have a corresponding active verb, there are some active verbs, such as ċeda ‘con- cede’, għama ‘blind’, stieden ‘invite’, which stand on their own, without a passive coun- terpart. In such cases it is possible to resort to the periphrastic passive (7a) or the pseudo-passive (7b).

(7a) Il=kelliem ġie mistieden

DEF=speaker come.PFV.3SG.M inivite.PST.PTCP mil-l=kumitat organizzattiv.

from-DEF=committee organizing ‘The speaker was invited by the organizing committee.’

134 A Tale of Two Morphologies

(7b) Lil ħu=h sawt-u=h

OBJ brother=his beat-PFV.3PL=him u għam-ew=h minn waħd-a.

and blind-PFV.3PL=him from one-F ‘His brother was beaten and blinded in one eye.’

Similarly, reflexives are derivationally related to their corresponding transitive verbs. There are many transitive verbs which lack a reflexive counterpart, but there is not one reflexive templatic verb that exists independent of a transitive verb form. Such verbs as xorob ‘drink’, wera ‘show’, wieġeb ‘answer’, can only be reflexivized periphrastically, as in (8), if they can be conceived in a reflexive sense at all.

(8a) It=tifel wera ruħ=u sodisfatt.

DEF=boy show.PFV.3SG.M soul=his pleased ‘The boy seemed pleased, lit. the boy showed himself pleased.’ (8b) It=tifel wieġeb lil=u nnifs=u.

DEF=boy answer.PFV.3SG.M to=him self=him ‘The boy answered his own question, lit. the boy answered himself.’

Inchoative verbs, by contrast, do not exhibit this dependence on their transitive alter- nants. There are several inchoatives, as we shall see below, that lack a corresponding causative form. The second piece of evidence is morphonological. While synonyms, verbs with multiple interpretations, and verbs taking part in the causative-inchoative and causa- tive-noncausative alternation may have different vowel sequences, the vowels of pas- sives and reflexives are the same as those of the transitive verbs they alternate with. Moreover, in Sect. 4.2.1.2, it was observed that the active-passive and transitive- reflexive alternations display a great deal of morphological regularity. With the excep- tion of a few isolated cases, they are unidirectional. That is, the form of the passive or reflexive verb is predictable from the form of the transitive verbs they are derived from. More specifically, all passive verbs appear in binyan V, VI or VII.13 Over 91% of these verbs have their corresponding active verbs in II, III and I respectively.

13 A few of them appear in VIII, and this is generally a result of the fact that many binyan VIII verbs have a synonymous verb in VII.

When patterns put down roots 135

The remaining exceptional cases are nonetheless regular. Consider, for example, pairs where the passive of verbs in II is in VII, rather than V, as expected:

Root Binyan II Binyan VII √għlm għallem ‘teach’ intgħallem ‘be taught’ √mxj mexxa ‘lead’ intmexxa ‘be led’.

Clearly, in these cases the passives are also verb derived. The gemination of the medial radical, a characteristic of binyan II and V but not of binyan VII, is carried over from the binyan II verbs into the passive forms. Analogously, 76% of all instances of a transitive-reflexive alternation in the binyan system involve a reflexive in V, VI or VII and a transitive counterpart in II, III and I, in this order. In this case too, there are a few unexpected combinations, such as II – VII, which, however, show regular morphonology. For instance, the reflexive form intgħatta ‘cover oneself’: being verb-derived, inherits the shape (same vowel sequence and gemi- nation of the middle radical) of its transitive counterpart għatta ‘cover’. The semantics of passives and reflexives are the third source of evidence for their being derived from transitive verbs, not directly from roots. The morphonological simi- larities observed above are paralleled by the fact that passives and reflexives do not generally acquire idiosyncratic interpretations that are unavailable for their corre- sponding transitive verbs.14 The opposite is widely attested. In these alternations, it is the active and non-reflexive verbs that tend to be polysemous: they very often have specialized or idiomatic meanings that are lost under passivization or reflexivization, as in the following examples:

Root Binyan Verb Meaning Binyan Verb Meaning √ġjb I ġab bring VII inġab be brought esteem esteem each other work out behave imagine, etc. √ħrġ I ħareġ take out VII inħareġ be taken out publish be published defray be defrayed

14 Refer to Sect. 3.3.2 for a number of counterexamples to this generalization.

136 A Tale of Two Morphologies

go out break out turn out, etc. √ngż II niggeż prick V tniggeż be pricked hurt, offend prick oneself inject hook a fish

What about causatives and non-causatives? Finally, let us discuss the derivation of causative and inchoative/noncausative verbs. Many theories of argument structure argue in favor of a derivational relation among verbs entering the causative-inchoative and the causative-noncausative alternation. Some theories derive the causative from the non-causative one (Lyons 1968; Lakoff 1970; Pinker 1989; Jackendoff 1990; inter alia), while others claim that the order of derivation is reversed (Levin & Rappaport Hovav 1995; Chierchia 2004; Reinhart 2000; and others). However, a closer examination of the templatic verbs that undergo causa- tive alternations in Maltese suggests that causatives, inchoatives and noncausatives are derived from the root not from one another.15 Interestingly, Arad (2005: Ch. 6) reaches the same conclusion for Hebrew verbs, and some of the arguments she brings in favor of this claim are analogous to the evidence from Maltese verbs I will review next. To start with, there are several gaps in the formation of causative and non- causative verbs. While passives and reflexives require the presence of a corresponding transitive verb, causatives may exist on their own without an inchoative counterpart, e.g., bexx ‘spray’, għallaq ‘hang’, sann ‘sharpen, whet’. Likewise, it is not uncommon to have inchoatives, such as batta ‘abate’, stenbaħ ‘wake up’, tbrekken ‘become shrewd’, and noncausatives, like baħħar ‘sail’, qomos ‘skip, bounce v.i’, tfaċċa ‘appear’, that do not alternate with a morphologically related causative verb. Additional support for the claim that causatives and non-causatives are root- derived comes from the vowel sequences of the verbs. We have seen that passives and reflexives, being derived from other verbs, inherit the vowels of their active and non- reflexive counterparts. By contrast, several causative-inchoative/noncausative pairs, especially in binyan I and II, exhibit vocalic variation, as these examples illustrate:

15 Similar proposals, claiming that alternating causatives and inchoatives do not stand in a derivational relationship (but are rather derived from a common base), have been advanced by Doron (2003), Embick (2004a, b), Alexiadou, Anagnostopoulou & Schäfer (2006), and Pylkkänen (2008), among others.

When patterns put down roots 137

Root Binyan Verb Meaning Binyan Verb Meaning √brd I bired cool off II berred cool (v.t) √għds I għodos submerge (v.i) II għaddas submerge (v.t) √għrq I għereq sink, drown (v.i) II għarraq sink, drown (v.t) √kbr I kiber grow (v.i) II kabbar grow (v.t) √kbs I kibes catch fire II kebbes set on fire √ktr I kotor increase (v.i) II kattar increase (v.t) √qrb I qorob get near II qarrab bring near √skr I siker get drunk II sakkar make one drunk √tlf I tilef lose II tellef cause one to lose √wħl I weħel stick (v.i) II waħħal stick (v.t) √żfn I żifen dance II żeffen make one dance

Moreover, in terms of morphological complexity or markedness of alternating verbs, there is a stark contrast between root-derived and verb-derived verbs. In the ac- tive-passive alternation, passives are always more morphologically complex or marked than their active counterparts (cf. Table 4.17). The same pattern is observed in the tran- sitive-reflexive alternation: the non-reflexive alternant carries less morphology than the reflexive verb (cf. Table 4.18). On the contrary, in the causative-inchoative and causative-noncausative alternations both alternants may be morphologically marked or unmarked. Consider, for example:

Inchoative Causative għola ‘rise’ għolla ‘raise’ ingħalaq ‘close’ għalaq ‘close’

In the first pair, the causative is morphologically more marked than the inchoative, dis- playing gemination of the middle radical. The reverse is observed in the second pair: the inchoative contains an additional prefix n-, which makes it morphologically more complex than its causative counterpart. In sum, then, while cases of verb derivation (i.e. passives and reflexives) are more morphologically complex than the transitive verbs from which they are derived, the direction of morphological markedness for causatives and non-causatives can go both

138 A Tale of Two Morphologies

ways. Besides, there are cases where causatives and non-causative are equally marked, as with ambitransitive verbs, where the same form is used both causatively and non- causatively (cf. Ch. 5). It never happens that a passive and an active or a reflexive and a non-reflexive are derived from the same root and have the same form. Finally, causatives and non-causatives differ from word-derived verbs in the range of interpretations they may have. Unlike passives and reflexives, inchoatives and non- causatives do acquire specialized meanings that do not exist with their corresponding causatives. As is evident from the examples below, this is also true for causatives, which may be assigned idiomatic or idiosyncratic interpretations that are not present in the non-causative verbs they alternate with.

Root Binyan Inchoative Meaning Binyan Causative Meaning √għrq I għereq sink II għarraq sink drown drown spoil, ruin get married aggravate squander √ksħ I kesaħ cool II kessaħ cool calm down irritate be shocked √nxf I nixef dry II nixxef dry grow thin shrivel wither numb be shocked shock shrink (wood) kill game lack words

Even though these conclusions pose a challenge to derivational accounts of the causative-inchoative alternation, they are, however, not incompatible with a directional analysis in terms of morphological markedness, similar to studies carried out by Has- pelmath (1993) amd Comrie (2006), among others. This question and other issues re- lated to the causative-inchoative alternation in Maltese are addressed in greater detail in the next chapter. With a clear picture of the interaction between roots and patterns in mind, we can now turn to the question of describing the verb morphology of Maltese as one dichoto-

When patterns put down roots 139

mous system. In the last part of this study, I examine the causative-inchoative alterna- tion in Maltese, with the aim of providing a unified approach to templatic and concate- native verbs.

Chapter 5 Towards a unified account

5.1 Getting to the root of the alternation...... 142 5.1.1 The alternation ...... 142 5.1.2 On the typology of causatives and inchoatives ...... 145 5.2 Maltese: between (anti)causative and labile verbs ...... 152 5.2.1 Morphological instability...... 152 5.2.2 Labile verbs...... 160 5.2.2.1 Internal and external causation ...... 160 5.2.2.2 Sentence creation task ...... 164 5.2.2.3 Corpus study ...... 168 5.3 Summary...... 171

The tale takes a slight twist in this chapter. In the previous chapters, we have identified two verb formation strategies in Maltese, root-and-pattern association and concatena- tion, which result in two main verb classes, templatic and concatenative verbs. They differ with respect to stem structure, derivational and inflectional morphology, etymol- ogy, and productivity, among others. Templatic verbs, in fact, are profitable, in the sense that most of them have a high frequency of use. However, they are unavailable, because, under normal conditions, no new Maltese verb is created by inserting a tri- or quadri- consonantal root in a binyan. Instead, concatenative verbs constitute an open, fully pro- ductive class of verbs in modern Maltese. In view of the double-sided nature of the verbal morphology of Maltese, the main question asked in this work is the following. Is the difference between templatic and concatenative verbs a pure morphological phenomenon or are there deeper semantic or syntactic differences between the two? Argument structure alternations were chosen as the starting point for such an investigation for two main reasons. First, from a quan- titative point of view, the analysis of templatic verbs carried out in Ch. 4 revealed that argument alternations make up two thirds of all the associations of a tri-consonantal root with a verbal pattern. Second, argument alternations present an important testing

Earlier versions of this chapter appeared in Spagnol (2011b) and Gatt & Spagnol (2011). Towards a unified account 141

ground where we can consider the interrelations between morphological, syntactic and lexical semantic aspects of verbs. In the literature on argument realization, it has been argued that morphosyntactic structure is directly related to the lexical semantic struc- ture of verbs. More specifically, the lexical semantic structure of a verb represents that portion of the verb’s meaning that (i) determines the morphosyntactic structures in which the verb can appear; and (ii) sets out the verb’s arguments and the relations among them. On this assumption, the idea is that it should be possible to surmise lexical structure from the alternation patterns a verb is associated with. Various studies have, for this reason, focused their attention on the ability of verbs to show argument alterna- tions, that is, alternate expressions of their arguments (cf. Hale & Keyser 1986; Pinker 1989; Jackendoff 1990; Levin 1993; Levin & Rappaport Hovav 1995, 2005; Ramchand 2008; inter alia). Argument alternations are a useful gauge to determine morphosyntac- tically relevant semantic aspects of verbs that pattern together, for it is a verb’s ability to participate in a particular set of alternations that reveals its semantic affiliations. This chapter is concerned with one such alternation, the causative-inchoative al- ternation, also known as the (Haspelmath 1993; Levin & Rappa- port Hovav 1995; Schäfer 2009), causative-anticausative alternation (Alexiadou, Anag- nostopolou & Schäfer 2006), unaccusativity alternation (Kiparsky 1997), ergative alter- nation (Matsuzaki 2001). I focus on the causative-inchoative alternation because it is one of the basic types of valency changing alternations, as the set of verbs that take part in it constitutes the prototypical class of transitive verbs (Tsunoda 1985; Levin & Rap- paport Hovav 1995: Ch. 3). In addition, according to the quantitative and qualitative investigation of the binyan system in Ch. 4, nearly one third of all argument alternations marked by templatic verbs are instances of the causative-inchoative alternation, con- firming it as a salient verb alternation in Maltese. The discussion unfolds in four steps. In Sect. 5.1, I describe the formal marking of the causative-inchoative alternation from a typological point of view. Following that, I survey some of the main theoretical accounts proposed for the alternation. In Sect. 5.2, I then turn to Maltese. First, I discuss the implications of two verb formation strategies on argument alternations, which are mediated through a system of roots and binyanim, on the one hand, and are not encoded in derivational morphology, on the other. Finally, I zoom in on labile verbs, i.e. verb pairs that use the same form as causative and in- choative, and tackle the much debated issue of whether they are basically transitive or intransitive (or neither). Concluding remarks and points for future directions are laid out in Sect. 5.3.

142 A Tale of Two Morphologies

5.1 Getting to the root of the alternation

This section is concerned with derivational approaches to the causative-inchoative al- ternation. It is structured as follows. In Sect. 5.1.1, I introduce the main properties of the alternation. In Sect. 5.1.2, I review the formal encoding of the alternation from a cross-linguistic perspective and consider the role morphological markedness has played in determining the directionality of derivation. Next, I look into a number of theoretical accounts of the alternation and discuss some problems they run into. Fi- nally, on the basis of the findings in Ch. 4, I argue that causatives and inchoatives are derived from a common underspecified root, paving the way for the discussion on the alternation in Maltese in Sect. 5.2.

5.1.1 The alternation

The causative-inchoative alternation characterizes pairs of verbs standing in a semantic relation to each other, where the inchoative verb is intransitive and typically expresses a change of state, and the causative verb is transitive and typically denotes a bringing about of this change of state. The following examples from English illustrate the para- digm that has been at the center of much typological and theoretical discussion:

(1a) The boy broke the glass. causative (1b) The glass broke. inchoative

One prominent feature of the alternation pattern [NP2 V NP1 // NP1 V] is that the sub- ject of the intransitive alternant bears the same semantic relation to the verb as the ob- ject of the transitive. The transitive member of the pair expresses a causative event, while its intransitive counterpart expresses an event in which an external cause need not be assumed or is irrelevant. The transitive subject causes some sort of change in the direct object, and has the semantic role of agent or actor. The intransitive subject is un- derstood as self-causing or self-affecting. Formally, researchers (e.g., Dowty 1979; Pinker 1989; Jackendoff 1990; Levin & Rappaport Hovav 1995; Rappaport Hovav & Levin 1998; and others) make use of lexi- cal semantic representations to spell out the internal architecture of a verb’s meaning.

Following one core assumption in decompositional lexical semantics, I draw on CAUSE

Towards a unified account 143

and BECOME as operators that factor out the relations between events as well as states. The representation or template assigned to causative verbs makes explicit two subevents, a causing and a change of state subevent, which are connected by the rela- tion CAUSE:

((∝) CAUSE (BECOME (x )))

Hence, sentence (1a) above is represented as ((The boy ACT) CAUSE (BECOME (glass ))). The events described by inchoative verbs are conceptualized as having no ex- ternal cause. Accordingly, the lexical semantic template of an event such as (1b) above has a simple structure of only one subevent (Rappaport Hovav & Levin 1998; McKoon & Macfarland 2000):

(BECOME (x )))

Other than break, examples of verbal concepts that frequently display the causa- tive-inchoative contrast in the world’s languages include boil, close, freeze, improve, melt, open, sink (cf. Haspelmath 1993; Levin 1993; Nichols, Peterson & Barnes 2004; Comrie 2006). In general, verbs undergoing the alternation have a stable semantic core. They primarily denote change of states, as the above examples. Some are verbs of movement, such as bounce, roll, spin. Assuming these verbs of motion express a change in location, we can subsume them under the notion of change of state. Since not all change of state verbs alternate (recall the discussion in Ch. 4 on in- choatives that lack a transitive counterpart, and causatives that stand on their own, without a corresponding inchoative), one issue that has been the subject of much dis- cussion in linguistic theory concerns the meaning components of change of state verbs that determine whether they will participate in the alternation or not (cf. Smith 1970; Hale & Keyser 1986; Levin & Rappaport Hovav 1995; Reinhart 2000, 2002; inter alia). With the underlying assumption that verbs undergoing the same alternation sets constitute a semantically coherent class, Levin (1993: 27-33) distinguishes three cate- gories of what she calls “causativity alternations”: (i) causative-inchoative verbs, (ii) induced action verbs (She walked the horse – The horse walked), and (iii) a general cate- gory of other types of verbs that enter the transitivity alternation (The nurse burped the baby – The baby burped). On my account, the second and third categories are regarded as instances of the causative-noncausative alternation. In spite of the similarities be- tween the causative-inchoative and the causative-noncausative alternation, the latter

144 A Tale of Two Morphologies

will not be further discussed in this chapter, as the two structures involve slightly dif- ferent phenomena, as argued in Levin & Rappaport Hovav (1995: 113ff.) and Reinhart (2002), among others (see also the discussion in Sect. 4.2.1.2 on differences between inchoatives and noncausatives). From a typological point of view, Schäfer (2009) argues that the causative-inchoative alternation is “a widespread, probably universal phe- nomenon”, while the causative-noncausative alternation is “clearly restricted to a sub- set of languages.” It should, however, be kept in mind that the relation between the two argument alternations remains an unsettled topic (cf. Doron 2003; Schäfer 2009; inter alia). The alternation is sensitive to the kind of theme argument the verb takes. That is, it is not always the case that verbs which occur transitively with certain arguments have an intransitive counterpart, and vice versa (cf. Levin & Rappaport-Hovav 1995: 85, 105; Alexiadou, Anagnostopoulou & Schäfer 2006; Koontz-Garboden 2009: 90-1; Schäfer 2009). Consider the following examples:

(2a) The girl broke the vase/toaster.

(2b) The vase/toaster broke.

(3a) The boy broke his promise/the world record. (3b) *His promise/The world record broke.

The same verb takes part in the alternation in (2) but not in (3). While entities like vases, toasters, and windows can undergo a breaking event both as the object of a causative and the subject of an inchoative, entities like promises, world records, and contracts cannot undergo such a change of state event as the subject of an inchoative, at least not in English. Corpus-based and cognitive approaches (e.g., Montemagni 1994; Montemagni & Pirelli 1995; Montemagni, Pirelli & Ruimy 1995; McKoon & Macfarland 2000; Thepkanjana 2003) show that selectional restrictions for the subject of intransi- tives and the object of transitives are not always identical, in particular with idiomatic senses of verbs and figurative meaning extensions, namely metaphor and metonymy. Semantically, verbs entering the causative-inchoative alternation tend to consti- tute one class, i.e. (non-agentive) verbs of change of state. Formally, however, they are very diverse. Typological investigation has shown that there is considerable variation in the morphological marking of the alternation across languages. Consider the event of burning and opening in these four different languages, in (4)-(7), taken from Haspel- math (1993).

Towards a unified account 145

Marking on the causative:

(4) (Khalka) Mongolian ongoj-lg-ox ‘open, v.t’ ongoj-x ‘open, v.i’

(5) - jal-aa-naa ‘burn, v.t’ jal-naa ‘burn, v.i’

Marking on the inchoative:

(6) Arabic fataħa ‘open, v.t’ in-fataħa ‘open, v.i’

(7) (Modern) Greek kéo ‘burn, v.t’ kéome ‘burn, v.i’

While in the verb pairs from Arabic and Greek the inchoative member bears more mor- phology than the causative one, in the examples from Mongolian and Hindi-Urdu it is the causative verb that is morphologically more complex. Morphological variation in the formal encoding of the alternation is found also within the same language. For in- stance, for the verbal concept of spreading, in Mongolian it is the inchoative alternant that is marked by special morphology (delge-r-ex ‘spread, v.i’), while the causative is unmarked (delg-ex ‘spread, v.t’), in contrast to the coding of the notion of opening, cf. (4) above.

5.1.2 On the typology of causatives and inchoatives

With the above as background, I now briefly review the literature on the typological variation in the morphological marking of causative-inchoative pairings. Following that, I take a look at theoretical accounts that, despite the morphological differences across and within languages, claim that all alternating verbs start off either from an inherently causative verb or from a basically inchoative verb. After considering the problems these approaches are faced with, I offer a solution along the lines of Embick (2004a, b), Pylk- känen (2008) and other syntactic approaches, arguing that verbs undergoing the causa- tive-inchoative alternation are derived from a common root. Finally, I attempt to bridge the gap between the mismatch of the root-based approach taken in this study and the iconic association of morphological markedness with directionality of derivation.

146 A Tale of Two Morphologies

Haspelmath (1993) As mentioned above, there are many ways cross-linguistically of effecting the alterna-

tion morphologically. Haspelmath (1993) provides a comprehensive survey of such formal types. Comrie (2006), then, builds on this analysis by establishing a profile for each language based on the marking of verbal notions, and by tracing the historical sta- bility of patterns of causative-inchoative pairings in the lexicon. In this section, I give an overview of the first cross-linguistic investigation. Comrie’s study is discussed in some detail in Sect. 5.2. Haspelmath’s typological approach has twofold significance. In dealing with the formal encoding of causative and inchoative pairs in terms of morphological marked- ness, the study draws conclusions about lexical semantic restrictions on the verbs that enter the alternation. I now briefly review both aspects of the work. Cross-linguistically, the relationship between causative and inchoative verbs with a common lexical meaning is marked with affixes, stem modification, suppletion, and other means. Haspelmath approaches the transitivity alternation from the perspective of morphological complexity: the alternant carrying less morphology is simple or basic while the alternant taking extra morphology is considered derived. Basing his analysis on morphological direction of derivation, he proposes a classification of formal rela- tions in causative-inchoative verb pairs. A distinction is first made between directed pairs, in which one member is derived from the other, and non-directed pairs, where there is no greater morphological mark- ing on either member. Directed pairs are further subdivided into causative and anti- causative alternations. In causative alternations, the causative member is morphologi- cally more marked than the inchoative, as in (4) and (5) above. Conversely, in anti- causative alternations, the inchoative member is morphologically more marked than the causative, as in (6) and (7) above. There are three non-directed types, illustrated by examples from Maltese: labile, where the same form is used both as causative and in- choative (e.g., għama ‘blind’ – għama ‘go blind’); equipollent, where neither member is more complex than the other, but the forms differ (e.g., faqqar ‘make poor’ – ftaqar ‘be- come poor’); suppletive, where different lexical items with different argument struc- tures are used (e.g., qatel ‘kill’ – miet ‘die’). From an investigation of a set of 31 verbal notions that can be construed both causatively and inchoatively across 21 languages, Haspelmath draws two interrelated generalizations, which are concerned with the idea that morphological patterns are sensitive to the kind of event named by the verbs in question. According to the first

Towards a unified account 147

conclusion, the semantic properties of verbs impose a constraint on whether verbs can participate in the causative-inchoative alternation or not. The claim is that verbs can be placed along a cline, going from those most likely to be conceptualized as occurring spontaneously (e.g., laugh, bloom) to those that are more easily conceived as occurring through the initiation of an external force (e.g., wash, decapitate). The most likely can- didates to take part in the causative-inchoative alternation are those lying closer to the middle than to the ends of this gradient, that is, verbs that are intrinsically neither agent-oriented nor self-causing. The second generalization concerns the relationship between morphological com- plexity and this cline or “scale of increasing likelihood of spontaneous occurrence”, as Haspelmath (1993: 105) labels it. Events such as breaking, closing, splitting, and gather- ing, which are typically instigated by an external agent, are predicted to pattern with anticausative formations, that is, their more marked form is the intransitive. Events un- likely to be directly causable from an external agent are expected to show a preference for the causative alternation, i.e. for pairs whose morphologically marked form is the transitive one. Among the events that belong to this group are melting, freezing, and drying. In Maltese, Haspelmath’s prediction is borne out to a great extent, as shown in Sect. 5.2. The notion of melting presents an interesting case in Maltese, where it may be expressed by either of two (contextually) synonymous verb pairs: dab (v.i.) – dewweb (v.t.) and inħall (v.i.) – ħall (v.t.). While the first pair, as predicted, patterns with the causative alternation, the second one shows the opposite direction of derivation. Note that Haspelmath offers general typological tendencies in the lexicalization of causative- inchoative pairs and does not predict the behavior of any given verb in any given lan- guage. Even though the notion of spontaneous occurrence is quite vague and in need of elaboration, there seems to be some degree of parallelism between the scale he pro- poses and transitivity gradients put forward by Hopper & Thompson (1980) and other researchers (cf. McMillion 2006: 20). The distinction between self-causing events and eventualities with external agents receives more attention in Sect. 5.2. Note that this analysis differs from my account of the causative-inchoative alterna- tion in two ways. First, in Haspelmath’s study, verbs, not roots, are taken as the basic elements in the lexicon. Second, it assumes that for a given pair of morphologically re- lated verbs, the member that bears more morphological material is derived from the less complex verb. Since under my approach all verbs are composite entities, there is always the possibility that the morphologically simple and the morphologically complex

148 A Tale of Two Morphologies

verb are derived not from each other, but from an underlying root. Indeed this is the conclusion drawn Ch. 4. Below, I identify points of convergence between the two appar- ently conflicting analyses.

Theoretical approaches In the studies delineated above, it is generally assumed that verbs entering the causa- tive-inchoative alternation are derivationally related. Not surprisingly, one issue that has received much attention in the literature on the alternation concerns the deriva- tional relationship among causative and inchoative verb pairs: which alternant is basic and which one is derived? Two main views have been put forward: causativization and detransitivization. Let us review both proposals briefly. For various researchers (e.g., Lyons 1968; Lakoff 1970; Pinker 1989; Jackendoff 1990; Ramchand 2008), alternating verbs have a monadic lexical semantic representa- tion, i.e. they lack an implicit external argument, and causatives are derived from in- choatives via causativization or transitivization. The lexical semantic structure of in- choatives is simple, as they are conceptualized as having no external cause (cf. Sect. 5.1.1). Because they have fewer participants and contain less information, inchoatives are taken to be less complex than causatives, whose structure consists of both a causing and a change of state subevent. Through an operation of causativization, in lexicalist theories (e.g., Hale & Keyser 1986), a causative predicate is added to the lexical repre- sentation of the anticausative base. According to the second view, the order of derivation is reversed. One such influ- ential analysis, which has it that inchoatives are derived from basically dyadic causa-

tives via deletion of the CAUSE operator, is Levin & Rappaport Hovav (1995: Ch. 3). They argue that the lexical semantic representation is dyadic for both causatives and in- choatives. Inchoatives are derived from their causative counterparts by means of a rule of detransitivization, which consists in the “lexical binding” of the causer argument of the causative verb. The causer argument is present in the lexical conceptual structure but not at the level of argument structure of the inchoative verb. It is bound in the map- ping from lexical semantic representation to argument structure and is therefore pre- vented from being projected into the syntax. In other words, although causatives and inchoatives have the same lexical semantic representation, they differ in their argument structure: causatives are dyadic, inchoatives are monadic due to the lexical binding of

the causer argument. Other studies that advocate a detransitivization analysis include Grimshaw (1981)

Towards a unified account 149

and Reinhart (2000, 2002). Starting with the observation that there is widespread syn- cretism between reflexive and anticausative morphology (e.g., Romance reflexive ), Chierchia (2004) and Koontz-Garboden (2009) propose a reflexivization analy- sis of anticausativization. Under this analysis, inchoatives are underlyingly transitive and are derived by a process of reflexivization. While this analysis caters for the fact that reflexive morphology is often found with anticausatives, it has been the subject of criticism. From a morphological point of view, it runs into the problem of accounting for passive/anticausative syncretism that is found in language after language, as no analy- sis successfully achieves to reduce passives to reflexivization. On the semantic side, it is not always conceivable to interpret anticausatives (e.g., the boat sank) reflexively (as ‘(some property) of the boat sank the boat’), especially when dealing with such exam- ples as the anticausative for ‘to be born’, which would be represented as ‘give birth to oneself’ (cf. Doron 2003; Piñón 2001b; inter alia). Derivational approaches have been questioned for two main reasons (cf. Alex- iadou, Anagnostopolou & Schäfer (2006) and Alexiadou (2006) for detailed criticism). First, not all causative verbs have an inchoative counterpart, and vice versa, as it was shown in Ch. 4. In such cases, verbs would have to be derived from a corresponding hy- pothetical base or from some abstract verb which is “frozen”, as in Chierchia 2004. Sec- ond, even if one had to take only alternating verbs into account, there would still be a mismatch between the assumed derivational and overt morphological complexity, given the cross-linguistic variation found in the formal encoding of the alternation (cf. Haspelmath 1993; Piñón 2001a, 2001b; Alexiadou, Anagnostopolou & Schäfer 2006; Sect. 5.1.2). Crucially, both views, causativization and detransitivization, do not do jus- tice to the crosslinguistic variation in the morphological marking of the alternation. Moreover, these approaches leave unexplained the fact that in Maltese causative and inchoative templatic verbs display morphonological and semantic behavior that are not typical of verb derivation, as discussed next.

Where do causatives and inchoatives come from? In spite of the morphological variation found cross-linguistically, derivational studies generally assume there is one direction of (syntactic) derivation for all alternating verbs. However, in Ch. 4, I discussed evidence that templatic verbs undergoing the causative-inchoative alternation are both derived from category-neutral roots by dif- ferent verbalizing heads. It was shown that causatives and inchoatives may both in- volve:

150 A Tale of Two Morphologies

• gaps, i.e. causatives sometimes lack an inchoative counterpart, and vice versa; • vocalic variation, i.e. alternating verbs may have different vowel sequences, e.g., għereq ‘drown, v.i’ – għarraq ‘drown, v.t’; • morphological markedness, i.e. causatives may be more complex than in- choatives, and vice versa (it is also possible that both are equally complex or that they have the same form); • relative freedom in binyan selection, e.g., the corresponding inchoative of a causative in binyan II can be in I, V, IX, etc.; • specialized meanings that are not available in the other alternant.

Irregularities of this sort are not found in passives and reflexives, which, I argue, are cases of verb derivation. In fact, passives and reflexives never stand alone, without a transitive counterpart. They always have the same vowel sequences as the active and non-reflexive verbs they alternate with, and there are strict correspondences among binyanim, to the extent that, for instance, an active in binyan II will almost always have its passive in V. In addition, passives and reflexives are always morphologically more marked than their transitive counterparts. Lastly, on the semantic side, it is very un- common that passives and reflexives acquire idiosyncratic interpretations that are un- available in their corresponding actives and non-reflexives. This analysis that causatives and inchoatives are derived from a common base is incidentally the same solution proposed in other works. This common base in a lexical- ist approach, such as Davis & Demirdache (2000), is an underlying causative event from which all change of state verbs are derived by a process of “event foregounding”, which builds on Pustejovsky’s (1995) event structure representation, where all change of state verbs are decomposed into a process (P) that leads to the change of state (T/S). Since only foregrounded events are syntactically realized, their proposal is that in causatives both P and T/S are foregrounded, whereas in inchoatives only T/S is fore- grounded. Foregrounding can then result in morphological marking, thus accounting for the cross-linguistic variation. In syntactic approaches developed in Distributed Morphology, such as Embick (2004a, b), Alexiadou, Anagnostopolou & Schäfer (2006), and Pylkkänen (2008), the common base from which causatives and inchoatives are derived is the root. The ap- pearance of a root in particular syntactic frames is then restricted by the encyclopedic or conceptual knowledge roots are associated with. In some studies, roots fall into a number of classes according to their encyclopedic semantics. In Harley & Noyer (2000)

Towards a unified account 151

and Alexiadou, Anagnostopolou & Schäfer (2006), for instance, roots are categorized along Levin & Rappaport Hovav’s (1995) distinction between internal and external cau- sation (cf. Sect. 5.2): if a root is externally caused, it must occur in a causative syntax (e.g., kill, destroy), if it is internally caused, it must occur in an anticausative syntax (e.g., blossom, wilt). If, on the other hand, the cause is unspecified, it may freely alternate be- tween a causative and an anticausative syntax (e.g., break, cool). This issue is taken up again in Sect. 5.2. In principle, proposals of this kind, where both causatives and inchoatives are de- rived from a common root, provide a way to account not only for the morphonological and semantic irregularities associated with these verbs in a language like Maltese, but also for cross-linguistic differences in the formal encoding of the alternation.

Morphological markedness vs. directionality of derivation Although, on my account, both causatives and inchoatives are derived from an underly- ing root, it is still possible to make observations in terms of morphological markedness, as in Haspelmath (1993). Derivational approaches generally assume an iconic reason- ing. That is, the morphologically marked form is considered to be derived from the un- marked form. However, an analysis like Haspelmath’s, which is based on the morpho- logical markedness of verb forms, is compatible not only with a derivational account, where the morphologically complex verb is derived from the less complex one. From a general, methodological point of view, it is also compatible with a root-based account, such as the one presented in this work, where both verbs are derived from a common root, and not from one another. Morphological variation may then be accounted for in terms of markedness, with the main assumption being that there is a cross-linguistic tendency for semantically marked forms to be more morphologically marked too. In other words, Haspelmath’s approach can be interpreted as a cognitive rather than a derivational account. Indeed, in more recent studies, Haspelmath (2005, 2008a, 2008b) explains the variation in the coding of causative and inchoative verb pairs in terms of frequency and economy, rather than in terms of derivationality. Whichever member of a pair occurs more fre- quently tends to be unmarked, while the rare, less expected, member tends to be overtly marked. In the rest of this chapter, therefore, morphological marking is not assumed to play a direct role in determining the directionality of the derivation, as both causatives and inchoatives are taken to be derived from a category-free root which merges with two

152 A Tale of Two Morphologies

different verbalizing heads. The discussion is not going to be couched in terms of direc- tionality of derivation among causative and inchoative verb pairs, but rather in terms of directionality of markedness. That is, while in an anticausative pairing, as described above, the inchoative is morphologically more marked than its causative counterpart, this is not taken to mean that the causative is the source from which the inchoative is derived. The same holds for the other formal types.

5.2 Maltese: between (anti)causative and labile verbs

The aim of this section is twofold. First, in Sect. 5.2.1, I consider the applicability of cross-linguistic analyses to Maltese templatic and concatenative verbs, arguing that the alternation is undergoing considerable change, as it displays a gradual increase of labile verbs. In Sect. 5.2.2, then, I deal with the correlation between the labile verbs and the lexical semantic nature of verbs entering the causative-inchoative alternation. Focusing mainly on the link between the dichotomy of internal and external causation and mor- phological markedness, I present data from a sentence creation task and a corpus study on the transitivity biases of a set of labile verbs in Maltese.

5.2.1 Morphological instability

The main question I raise in this section concerns the historical directionality of mor- phological markedness. The point of departure is Comrie’s (2006) study, which is an elaboration of Haspelmath (1993), in that it extends the analysis of verbal concepts as expressed morphologically to Maltese, Swedish and Tsez. By quantifying the verbal con- cepts for which every language makes use of each of the five formal types discussed above (i.e. causative, anticausative, labile, equipollent, suppletive), he also proposes language profiles that would have Finnish as overwhelmingly causative, French and Russian as predominantly anticausative, English as overwhelmingly labile, showing an areally atypical preference for this non-directed type, and so on. It is argued that, in terms of the contrast between predominance of anticausative and causative pairings, the transitivity profile of a language like Maltese, is diachroni- cally stable (Comrie 2006: 314). In what follows, I analyze a corpus of verbal concepts that enter the causative-inchoative alternation in Maltese and argue, contra Comrie, that the alternation is not time stable in Maltese, displaying a shift in morphological

Towards a unified account 153

markedness, from anticausative to labile pairs. Let us take a closer look at the analysis. Comrie (2006), following Haspelmath (1993), proposes a method for investigating the morphological relations that hold between 31 notions in 24 languages, including Maltese. From an analysis of the five formal types that are used to express such verbal concepts as boil, burn, gather, open, split, etc. it turns out that Maltese has an overall preference for anticausative formations, as Table 5.1 shows. This profile is shared with the majority of the European languages under investigation, which also have a propen- sity for anticausative pairings. It is, however, at odds with the global cross-linguistic tendency to prefer causative formations (cf. Nichols 1993; Comrie 2006).

1 Table 5.1 Comrie’s (2006) sample of Maltese causative-inchoative pairs

Gloss Causative Binyan Inchoative Binyan Type

boil għalla II għala I Causative freeze reżżaħ II reżaħ I Causative dry nixxef II nixef I Causative wake up qajjem II qam I Causative go/put out tefa I intefa VII Anticausative sink għerreq II għereq I Causative learn/teach għallem II tgħallem V Anticausative melt dewweb II dab I Causative stop waqqaf II waqaf I Causative turn dawwar II dar I Causative dissolve terraħ II teraħ I Causative burn ħaraq I inħaraq VII Anticausative destroy qered I inqered VII Anticausative fill mela I imtela VIII Anticausative finish temm I intemm VII Anticausative begin beda I beda I Labile spread firex I infirex VII Anticausative roll gerbeb QI tgerbeb QII Anticausative develop żviluppa żviluppa Labile

1 I suggest some minor changes to Comrie’s analysis of Maltese. I consider break to constitute an equipol- lent pair (kisser–inkiser), cf. the discussion on synonyms in Sect. 4.2.1.4. Alternative lexical choices may be given for connect (equipollent għaqqad–ingħaqad), freeze (labile iffriża), dissolve (anticausative ħall–inħall), split (anticausative qasam–inqasam), and rise/raise (causative għola–għolla).

154 A Tale of Two Morphologies

get lost/lose tilef I intilef VII Anticausative rise/raise qajjem II qam I Causative improve tejjeb II tjieb IX Causative rock bandal QI tbandal QII Anticausative connect rabat I irtabat VIII Anticausative change bidel I inbidel VII Anticausative gather ġabar I inġabar VII Anticausative open fetaħ I infetaħ VII Anticausative break kiser I inkiser VII Anticausative close għalaq I ingħalaq VII Anticausative split feraq I inferaq VII Anticausative die/kill qatel I miet I Suppletive

Note that only one of the 31 items in Comrie’s sample is a concatenative verb pair, namely labile żviluppa ‘develop’. This gives rise to the issue of whether concatenative verbs also show a preference for anticausative pairings or whether they effect the causative-inchoative alternation in a different way from templatic verbs. The question that I try to address in this section can be couched in the following way. Has the heavy influence of Romance and, to a certain extent, English on Maltese – including the devel- opment of non-templatic processes of verb formation and a reanalysis or simplification of verb inflection – resulted in a change in the formal encoding of the causative- inchoative alternation in the language?

Templatic verbs To begin with, let us consider the statistical validity of Comrie’s (2006) sample for Mal- tese by looking at the distribution of the causative-inchoative alternation in the data- base of tri-consonantal roots and the patterns they create. That templatic verbs (at least those derived from tri-consonantal roots) have a strong preference for anticausative alternations emerges from Table 4.19, reproduced here with some changes as Table 5.2. Considering that binyan II is morphologically more marked than I and IX, 26.20% of templatic verbs mark the alternation causatively. If for the other pairs of binyanim, i.e. II – V, I – VII, I – VIII, and III – VI, we take I, II and III to be morphologically un- marked, then 63.72% of all instances of a causative-inchoative contrast express anti- causative alternations. Most of the remaining 10% are cases of labile (e.g., √għmj, għama ‘blind, go blind’) and equipollent (e.g., √ksr, kisser ‘break, v.t’ – inkiser ‘break, v.i’)

Towards a unified account 155

pairs. In sum, then, the qualitative analysis carried out in Ch. 4, goes to confirm that Maltese templatic verbs indeed have a propensity for anticausative pairings.

Table 5.2 Morphological marking of templatic causative-inchoative verbs

Patterns Type No. roots Percentage Causative Inchoative Gloss

II – V Anticausative 258 44.79 tappan ittappan blur II – I Causative 97 16.84 saħħan saħan warm up I – VII Anticausative 74 12.85 qasam inqasam split II – IX Causative 56 9.72 qassar qsar shorten I – VIII Anticausative 21 3.65 żied żdied increase III – VI Anticausative 14 2.43 sieħeb issieħeb associate Other Varied 56 9.72 kisser inkiser break

The next step is to examine the morphological realization of concatenative verbs that undergo the causative-inchoative alternation. Before going into this issue, a word is in order on the differences between anticausative and non-active morphology in Mal- tese.

Distinguishing anticausatives from passives In general, grammars of Maltese do not account for anticausatives, i.e. inchoatives that are morphologically more marked than their causative counterparts. Inchoatives are only discussed when they are morphologically ‘simple’, that is, when they occur in bin- yan I or IX. When inchoatives contain more morphology than their corresponding causatives, i.e. when they occur in binyan V, VI or VII (e.g., issaħħab ‘become cloudy’, tbaċċaċ ‘become chubby’, ittappan ‘blur, v.i’, infetaħ ‘open, v.i’, inqasam ‘split, v.i’), they are treated as “passives and/or reflexives” (cf. Borg 1981; Mifsud 1995a; inter alia). To be fair, Sutcliffe (1936) does have a class of anticausatives, called “effectives”. So does Cremona (1962), who labels them “impersonal passives”, by which he means verbs whose external cause or agent is not relevant (cf. Ch. 3). One reason why several gram- mars set anticausatives aside is the fact that they are very often syncretic with passives. Even though both passivization and anticausativization entail the advancement of a direct object, passives and anticausatives differ in certain well-known ways, as has been shown by Marantz (1984), Comrie (1985), Levin & Rappaport Hovav (1995), Alex-

156 A Tale of Two Morphologies

iadou, Anagnostopolou & Schäfer (2006), among others. Some of the tests mentioned in the literature apply to Maltese too. Passive verbs like issawwat ‘be beaten’, inxtara ‘be bought’, inqatel ‘be killed’, can be modified by by-phrases (8a), and agent-oriented ad- verbs (8b). They can license instruments (8c), and allow control into purpose clauses (8d).

(8a) Ir=raġel is-sawwat mil-l=pulizija.

DEF=man PASS-beat.PFV.3SG.M from-DEF=police ‘The man was beaten by the police.’ (8b) Ir=raġel is-sawwat apposta.

DEF=man PASS-beat.PFV.3SG.M purposely ‘The man was beaten on purpose.’ (8c) Ir=raġel is-sawwat bi frosta.

DEF=man PASS-beat.PFV.3SG.M with whip ‘The man was beaten with a whip.’ (8d) Ir=raġel is-sawwat biex

DEF=man PASS-beat.PFV.3SG.M in order ma j-ivvuta-x.

NEG IPFV.3SG.M-vote-NEG ‘The man was beaten to keep him from voting.’

By contrast, anticausatives like issaħħab ‘become cloudy’ and ittappan ‘blur, v.i’ do not generally license external arguments, such as agents and instruments. They do not li- cense causers or causing events introduced by the preposition minn ‘by’, as in (9a). Some anticausatives, however, do license causers or causing events if these are intro- duced by the preposition bi ‘with’, as in (9b).

(9a) ??Il=ħġieġ it-tappan mi-t=tifel.

DEF=glass INCH-blur.PFV.3SG.M from-DEF=boy ‘The glass fogged up by the boy.’ (9b) Il=ħġieġ it-tappan bi-l=fwar.

DEF=man INCH-blur.PFV.3SG.M from-DEF=steam ‘The glass fogged up from the steam.’

The situation is, however, more complicated with verbs such as infetaħ ‘open, v.i’ and inqasam ‘split, v.i’, which are syncretic between passives and anticausatives. In such

Towards a unified account 157

cases, one useful diagnostic to distinguish between the two is the waħ(e)d- ‘by itself, lit. alone’ modifier. Inchoatives, whether morphologically unmarked (10a) or marked (10b), generally take the waħ(e)d- modification.2 Passives, on the other hand, do not typically license this modifier, as (10c) and (10d) illustrate.

(10a) Id=dgħajsa għerq-et waħed=ha.

DEF=boat sank.PFV.3SG.F alone=her ‘The boat sank by itself.’ (10b) Il=bieb in-fetaħ waħd=u.

DEF=door INCH-open.PFV.3SG.M alone=him ‘The door opened by itself.’ (10c) ??Ir=raġel in-qatel waħd=u.

DEF=man PASS-kill.PFV.3SG.M alone=him ‘The man killed by himself.’ (10d) ??Il=biljett in-xtara waħd=u.

DEF=ticket PASS-buy.PFV.3SG.M alone=him ‘The ticket bought by itself.’

After this brief excursus, let us return to the issue of the formal encoding of con- catenative verbs taking part in the causative-inchoative alternation.

Concatenative verbs In root-and-pattern morphology, the (consonantal) root is often inserted in more than one binyanim, creating different morphologically related verbs. In concatenative mor- phology, on the contrary, the (syllabic) root takes a single verbal morpheme, thus as- signing only one verbal interpretation to each root. Therefore, while in non- concatenative morphology valency changing alternations are typically mediated through the verbal patterns, by the combination of one root with different binyanim, in concatenative morphology they are not marked morphologically. Rather, argument al- ternations are either expressed periphrastically (e.g., aċċerta ‘(re)assure’ – aċċerta ruħu ‘assure oneself’) or by a single verb, which is used both transitively and intransitively (e.g., skura for both ‘make dark’ and ‘become dark’). Indeed, concatenative verbs express the causative-inchoative contrast with a change in syntactic patterns, but with no formal change in the verb. They are character-

2 Note that there are some inchoatives which do not typically take the waħ(e)d- modifier (see footnote 6).

158 A Tale of Two Morphologies

ized by labile pairings, where single verb forms syntactically occur in both transitive and intransitive clausal patterns, as sentences (11) to (14) show.

(11a) Oħt=i ċċarġja-t il=batterija.

sister=my charge-PFV.3SG.F DEF=battery ‘My sister recharged the battery.’ (11b) Il=batterija ċċarġja-t.

DEF=battery charge-PFV.3.SG.F ‘The battery recharged.’

(12a) San Franġisk immansa l=lupu.

Saint Francis tame.PFV.3SG.M DEF=wolf ‘Saint Francis tamed the wolf.’ (12b) Il=lupu mmansa.3

DEF=wolf tame.PFV.3SG.M ‘The wolf became tame/tamed.’

(13a) Omm=ha ttowja-t il=laħam.

mother=her thaw-PFV.3SG.F DEF=meat ‘Her mother thawed the meat.’ (13b) Il=laħam ittowja.

DEF=meat thaw.PFV.3SG.M ‘The meat thawed.’

(14a) Il=gvern issoda l=ekonomija.

DEF=government strengthen.PFV.3.SG.M DEF=economy ‘The government has strenghtened the economy.’

(14b) L=ekonomija ssoda-t. 4

DEF=econonmy strenghten-PFV.3.SG.F ‘The economy has strengthened.’

3 Some loans appear both as templatic and concatenative verbs, such as nerveż ‘make nervous’–tnerveż ‘become nervous’ and innervja with both uses, and mannas ‘tame’–tmannas ‘become tame’ and immansa as a labile alternation. 4 Reflexive verbs, especially non-templatic ones, take the marker ruħ- ‘soul’ plus direct object , such as concatenative iddefenda ruħu ‘defend oneself’ and templatic ħejja ruħu ‘prepare oneself’. This marker is sometimes used to express the intransitive member of an otherwise labile pair e.g., in-negozju ssoda ruħu ‘the market has strengthened (itself)’ (cf. Spagnol 2011a on the assimilation of reflexive verbs of Romance origin in Maltese). This is in line with the cross-linguistic tendency for anticausativization and reflexiviza- tion to be marked in morphologically identical ways (cf. Koontz-Garboden 2009). Worth noting in this re- spect is Haspelmath’s (1990) finding that anticausative markers often develop diachronically from reflex- ive markers via bleaching of agent entailments.

Towards a unified account 159

A number of these labile verbs (e.g., issoda, żviluppa) hark back to Romance anticausa- tive formations (assodare–assodarsi, sviluppare–svilupparsi). In such cases, Maltese does not borrow or calque the Romance pattern, neutralizing the anticausative formation and marking the causative-inchoative contrast as a labile alternation. Summing up this section, then, evidence has been provided that the causative- inchoative alternation in Maltese is undergoing change, which goes parallel to the di- chotomy in verb formation, templatic and concatenative. The alternation starts off with a strong preference for anticausatives in templatic verbs. Maltese is to a certain extent losing its morphology in expressing the alternation, and is developing an overwhelm- ingly non-directional profile, with a strong increase of labile verbs. Maltese is extending the set of labile verbs to express the causative-inchoative contrast quite generally. The class of labile verbs is, in fact, constantly increasing, just like in English, Greek and some Daghestan languages, and unlike Sanskrit where it is decreasing (cf. Lavidas forthcom- ing). Considering that few other languages appear to have developed this means of marking the contrast (cf. Comrie 2006), one may well wonder about the factors in- volved in this development. Language contact and other external factors must be part of the diachronic explanation of this labilization process.5 The large number of loan verbs from English (e.g., iffriża ‘freeze’, impruvja ‘improve’, ittowja ‘thaw’), a language with a predominantly labile profile, must have had a part in the emergence and expansion of labile verbs in Maltese. Even though verbs in Italian, like templatic verbs in Maltese, show a preference for anticausative formations (e.g., sciogliere (v.t) – scioglier-si (v.i) ‘melt’), the causative-inchoative alternation is also commonly expressed by labile verbs, e.g., affondare ‘sink’, diminuire ‘reduce’, guarire ‘heal’ (for in-depth analyses of the phe- nomenon in Italian, see Burzio 1986; Montemagni 1994; Centineo 1995; Montemagni, Pirelli & Riumy 1995; Folli 2001). The rise of a concatenative verb formation strategy along the templatic one has therefore had a significant impact on the encoding of the alternation in Maltese. Other than language contact, internal mechanisms such as analogy and reanalysis could have speeded up the revision of the alternation in Maltese. For templatic verbs, the alternation exhibits a large degree of freedom in morphological marking (cf. Ch. 4). Inchoative verbs may, for instance, belong to binyan I (nixef ‘dry’), IX (krieh ‘become

5 Following McMillion (2006), the term labilization is here used to refer to language change that involves an increase in labile verb types to the extent that it becomes a conventional means for effecting argument al- ternations.

160 A Tale of Two Morphologies

ugly’), V (tgħawweġ ‘bend’), VII (inkiser ‘break’), etc. This could have been a catalyst for the change to different means for expressing the causative-inchoative contrast. Still, further diachronic research needs to be undertaken to determine which external and internal factors are crucial for the process of labilization.

5.2.2 Labile verbs

So far in this chapter, we have seen that there is a great deal of variation in the formal encoding of the causative-inchoative alternation both across and within languages. Haspelmath (1993) argues that verbal concepts may be conceptualized as occurring spontaneously (melt, dry) or as being instigated by an external agent (break, gather). This conceptual difference, which is represented as a spontaneity scale, is then assumed to be reflected in the morphological marking of verbs. The prediction is that spontane- ous events are more likely to have the inchoative alternant morphologically unmarked. The reverse, i.e. that the inchoative is the more marked form, is expected when the event in question typically requires an external agent that is responsible for bringing about the change of state. In the rest of this section, I give an overview of what may be regarded as a re- statement of Hasplemath’s spontaneity scale, namely Levin & Rappaport Hovav’s (1995) distinction between internally and externally caused verbs. Following that, I discuss the role verbs that lack overt morphological marking, i.e. labile verbs, play in the relationship between the formal encoding of verbs and their lexical semantics.

5.2.2.1 Internal and external causation

Under Levin & Rappaport Hovav’s (1995) formulation of the Unaccusativity Hypothesis, it comes to light that there is not one, but two semantic classes of inchoative verbs. The basic lexical difference is the source of change: an inchoative verb may denote either an

internally caused or an externally caused event. In the context of change of state verbs, the idea is that verbs like bloom, deteriorate, and rust are internally caused because the means of bringing about the change of state event is conceptualized as an inherent property of the entity undergoing the change. Verbs such as break, crumble, and ex- plode, on the other hand, are conceptualized as coming about due to a force external to the entity undergoing the change of state. Stated another way, an externally caused in- transitive construction such as the glass broke is a derivation of the underlying transi- tive structure, implying that someone or something broke the glass. By contrast, flow-

Towards a unified account 161

ers bloom and pipes rust because of something internal to them; the change of state event is brought about by the entity undergoing the change itself. Among the diagnostics to distinguish externally from internally caused verbs are (a) the by itself test, (b) the scope of negation, and (c) adverbial modifiers. Relying on data mostly from Spanish, Koontz-Garboden (2009: 106ff.) shows how por sí solo ‘by itself’ is acceptable with externally caused (e.g., el barco se hundió por sí solo ‘the boat

REFL sank by itself’) but not with internally caused events (e.g., ??Juan empeoró por sí solo ‘Juan worsened by himself’), because the latter are argued by Levin & Rappaport

Hovav to lack a CAUSE operator as part of their denotation. Since the by itself modifier must be the sole agent or cause of the event named by the verb heading the clause in which it appears, it is judged unacceptable in clauses headed by internally caused verbs that lack agentive or causer subjects.6 According to the second diagnostic, there are two different interpretations under negation for external causation verbs, as shown in examples (15) and (16), taken from Koontz-Garboden (2009: 112-113).

(15) Father: ¿Se rompió el vaso? ‘Did the glass break?’ Son: No, no se rompió el vaso. ‘No, the glass did not break.’

(16) Father: ¿Que pasó, hijo? ‘What happened, child?’ Son: El vaso se rompió. ‘The glass broke.’ Father: No se rompió sino que tú lo rompiste! ‘The glass didn’t break–you broke it!’

6 In Maltese, the waħ(e)d- modifier in general fits more naturally with externally caused (a, b) than inter- nally caused verbs (c, d):

a. Il-bieb ingħalaq waħdu. c. ??Il-kafè kesaħ waħdu. ‘The door closed by itself.’ ‘The coffee cooled down by itself.’ b. Il-fjamma ntfiet waħedha. d. ??Xagħarha twal waħdu. ‘The flame went out by itself.’ ‘Her hair got longer by itself.’

However, there are internally caused verbs that do license waħ(e)d- (e), and externally caused verbs that are judged unacceptable with waħ(e)d- (f):

e. Il-ħwejjeġ nixfu waħedhom. f. ??Il-ġelat inħall waħdu. ‘The clothes got dry by themselves.’ ‘The ice-cream melted by itself.’

In Schäfer (2007), which is an analysis of the by itself phrase and its counterparts in German, Greek and Italian, it is observed that the phrase can be found cross-linguistically modifying predicates which arguably do not involve an external causer argument.

162 A Tale of Two Morphologies

In negating the former case, the son denies that the glass has undergone the change of state. It did not break. In the latter example, the glass does break, even though the verb naming the change of state is negated. What the father denies is not that the glass broke, but that the vase was the cause of its own breaking. In other words, negation has

scope over a BECOME operator in (15) and over a CAUSE operator in (16), loosely speak- ing. According to Koontz-Garboden (2009: 112-119), such ambiguity is not found with internally caused verbs, which lack a causative lexical semantic representation. The ability to take agent-oriented adverbial modifiers is another piece of evidence for the internal–external causation dichotomy. Centineo (1995) observes that external but not internal causation verbs in Italian may be modified by the violentemente ‘violently’. Compare externally caused la porta si è chiusa violentemente ‘the door closed violently’ with internally caused *la nave è affondata violentemente ‘the boat sunk vio- lently’. Similar data has been observed for Greek by Alexiadou & Anagnostopoulou (2004: 131ff.). Additional support for Levin & Rappaport Hovav’s dichotomy comes from corpus analysis and psycholinguistic experimentation. McKoon & Macfarland (2000) found out that the range of subjects of transitive patterns for internally caused verbs is restricted to lexically delimited entities: bloom, for instance, is restricted to flowers, plants, trees, and the like. For external causation verbs, by contrast, there is no such restriction. The range of transitive subjects of externally caused events like break is open to animate entities, natural forces, conditions, and instruments.

Similar findings are reported by Wright (2001, 2002). Corpus data indicate that in- ternally caused verbs occur significantly less often in transitive constructions. When they are used transitively, they are more likely (i) to involve a nature-related causer as opposed to externally caused verbs, which tend to involve a human causer, and (ii) to have a metaphorical interpretation. Psycholinguistic evidence supports the view that change of state verbs can gener- ally be divided into two classes. The experiments in McKoon & Macfarland (2000: 847ff.) show processing differences between sentences with internally caused and sen- tences with externally caused verbs. External causation verbs, both transitive and in- transitive, require longer comprehension times than verbs of internal causation. This conclusion follows from the assumption that verbs describing externally caused events are more complex than verbs denoting internally caused events, because the former have two subevents (a causing and a change of state subevent) in their lexical semantic representation while the latter have only one (a change of state subevent).

Towards a unified account 163

Wright (2001, 2002) observes another difference between the two main types of change of state verbs with regard to transitivity. Results from survey data show that internally caused verbs are rated less acceptable than externally caused verbs in transi- tive constructions. Internally and externally causation verbs differ in terms of the fre- quency and acceptability with which they are used transitively. She suggests that causer type (human-driven vs. non-human driven events), controllability (internal vs. external locus of control), and selectional restrictions in relation to subject-modification, are three factors that play a role in determining the transitive behavior of change of state verbs. The difference in lexical semantic representation between internally and exter- nally caused verbs results in different predictions about the morphological marking of verbs. Rephrasing Haspelmath’s (1993) generalization on the association between the formal and lexical semantic aspects of verbs, we can say that externally caused verbs, which lexicalize a CAUSE operator, favor anticausativization, while internally caused events, which lack a causative lexical semantic representation, favor causativization. In other words, externally caused verbs are morphologically unmarked as causatives, while internally caused verbs are morphologically unmarked as inchoatives. What about labile alternations, where neither verb is overtly coded as marked or unmarked? In order to establish whether labile verbs are basically causative or in- choative, researchers (e.g., Visser 1970; Kitazume 1996; McMillion 2006) carry out studies under a diachronic perspective, tracing the earliest recorded appearances of the items to determine whether they started out as transitive or intransitive verbs. The idea is that the more salient type would be the basic type in an argument structure alterna- tion and that it will be the first to appear in the language. Such an exercise is difficult to conduct for languages without a long written tradition like Maltese. Besides, there is no good reason why a diachronic sequence should necessarily translate into synchronic preeminence of one of the senses of the current labile verbs. It may be therefore prema- ture to assume that a verb is structurally of one rather than another pattern because it is used historically in the first pattern before it appears in the second. The internal-external causation dichotomy can be viewed not only diachronically, but also in terms of frequency. Preliminary corpus evidence from English demonstrates that verbs such as break and open occur more frequently as transitives, while the occur- rence of freeze, dry, and the like in intransitive clauses is higher than that for transitive ones (Wright 2001: 127-8; Haspelmath 2008a: 13, 2008b: 14). This finding incidentally correlates with Haspelmath’s (1993) spontaneity scale: the percentage of transitive oc-

164 A Tale of Two Morphologies

currences of items that are typically instigated by an external agent is higher than that for items that are typically conceptualized as occurring spontaneously. In what follows, I report on two experiments, a sentence creation task and a cor- pus-based analysis, that have been carried out on a set of labile verbs in Maltese with the aim of determining whether these verbs evince a transitivity bias. The prediction is that verbs naming internally caused events occur more often in intransitive frames, whereas verbs of external causation appear more frequently in transitive frames.

† 5.2.2.2 Sentence creation task

Evidence that labile verbs are associated with two different basic syntactic frames, comes from native speakers’ responses to a sentence creation task. For this online study, 40 labile verbs were selected (cf. Table 5.3a), which were then divided into five groups of eight verbs each. A total of ten subjects per group, all native speakers of Mal- tese, were asked to write sentences using the labile verbs. Each participant wrote a sen- tence for each of the eight labile verbs and for 24 items, which consisted of nouns (e.g., xemx ‘sun’), transitive only verbs (e.g., sab ‘find’), and intransitive only verbs (e.g., żelaq ‘slip’). Each labile verb was therefore seen by ten participants. To make sure sub- jects write the first sentence they could think of, for each item there was a 60 second countdown. Subjects with timeouts were not included in the analysis. The sentences created by the subjects were coded as transitive, intransitive or other. This last category includes metalinguistic uses, adjectival uses, and/or nulls. Three verbs in particular, beda ‘begin’, kompla ‘continue’, and spiċċa ‘finish’, had a pre- dominant ‘other’ categorization, which is primarily due to their being often used as as- pectuals, as in beda jikteb ‘he started writing’, rather than as alternating verbs, such as il-film beda fis-6 ‘the movie started at 6’ and beda l-ħidma tiegħu ‘he started his work’. The corpus data discussed in the following section suggests that this is also true in cor- pora. Since data from these verbs is unlikely to be informative, they were dropped from the present analysis. Table 5.3 and Figure 5.1, which list the results for the sentence creation task, re- veal that verbs such as evapora ‘evaporate’, ikkalma ‘calm down’, sparixxa ‘vanish’, sploda ‘explode’ show a clear intransitive bias, whereas other verbs, including iċċarġja ‘charge’, illixxa ‘smooth’, ippurifika ‘purify’, issikka ‘tighten’, display a preference for

† The results in this section and Sect. 5.2.2.3 stem from joint work with Albert Gatt (cf. Gatt & Spagnol 2011)

Towards a unified account 165

transitive frames. Note that a few verbs show an equal or almost equal likelihood of be- ing used in a transitive or an intransitive frame. Among these verbs are irdoppja ‘dou- ble’, irrombla ‘roll’, varja ‘vary’, żvojta ‘empty’. These are summarized in Table 5.3b. These results seem to suggest an analysis along the lines of Harley & Noyer (2000) and Alexiadou, Anagnostopolou & Schäfer (2006), who posit that roots fall into a num- ber of categories depending on their encyclopedic semantics, which then has an impact on their appearance in given syntactic frames. More specifically, a verb like sparixxa ‘vanish’ must be derived from a root associated with internal causation. The fact that it is almost exclusively used in intransitive constructions follows from this encyclopedic meaning associated with the root. Analogously, verbs like illixxa ‘smooth’, being derived from roots associated with external causation, are conceptualized as change of state events that are typically brought about by an external argument, and hence hardly oc- cur in intransitive frames. Verbs that do not show a clear preference for transitive or intransitive constructions, such as irdoppja ‘double’, must be derived from roots with an unspecified cause, as proposed by Harley & Noyer (2000), Alexiadou, Anagnostopolou & Schäfer (2006), and may thus alternate between causative and inchoative syntax.

Table 5.3a Labile verbs: results of the sentence creation task

Labile verb Gloss Intransitive Transitive Other

1 ibbawnsja bounce 8 1 1 2 ibbilanċja balance 3 7 0 3 iċċarġja charge 2 8 0 4 iċċattja flatten 0 7 3 5 iddisintegra disintegrate 9 1 0 6 iddifrostja defrost 6 4 0 7 iddritta straighten 2 8 0 8 espanda expand 9 1 0 9 evapora evaporate 10 0 0 10 iffriża freeze 7 3 0 11 għaġġel hurry up 10 0 0 12 għama blind/go blind 8 2 0 13 impruvja improve 3 7 0 14 intensifika intensify 4 6 0 15 ikkalma calm down 10 0 0

166 A Tale of Two Morphologies

16 ikklirja clear 2 7 1 17 ikkuntenta please/be pleased 7 3 0 18 illaxka loosen 7 2 1 19 illixxa smooth 0 10 0 20 immansa tame 1 7 2 21 innokkla curl 7 2 1 22 ippurifika purify 1 8 1 23 irdoppja double 5 5 0 24 irrombla roll 5 5 0 25 slowja slow down 9 1 0 26 sparixxa vanish 10 0 0 27 sploda explode 10 0 0 28 issikka tighten 1 8 1 29 issoda strengthen 7 3 0 30 ittowja thaw 3 6 1 31 varja vary 5 5 0 32 xegħel light, turn/go on 10 0 0 deflower/lose 33 żverġna 2 7 1 one’s virginity 34 żviluppa develop 3 7 0 35 żvina bleed to death 10 0 0 36 żvojta empty 6 4 0 37 żvolġa develop 10 0 0

Table 5.3b Labile verbs in sentence creation task grouped according to transitive bias

Intransitive bias Transitive bias No significant bias

evapora espanda illixxa immansa irdoppja għaġġel slowja iċċarġja żverġna irrombla ikkalma ibbawnsja iddritta żviluppa varja sparixxa għama ippurifika iddifrostja sploda iffriża issikka intensifika xegħel ikkuntenta ibbilanċja ittowja żvina illaxka iċċattja żvojta żvolġa innokkla impruvja iddisintegra issoda ikklirja

Towards a unified account 167

168 A Tale of Two Morphologies

5.2.2.3 Corpus study

Having established that labile verbs do not constitute a homogeneous class with respect to transitivity, the next step is to confirm the results obtained from the sentence crea- tion task by carrying out a corpus analysis of the same set of labile verbs. In order to test whether the frequency of use reflects the biases observed above, we ran searches on labile verbs and categorized them as transitive or intransitive. In the case of polyse- mous verbs, sentences that involved senses that do not take part in the causative- inchoative alternation (e.g., iċċarġja in the sense ‘to require payment’ rather than in the intended, alternating sense ‘to supply with or receive electrical energy’) were not taken into account. The corpus data was obtained from web crawling and from the MLRS corpus, a text collection of around 100 million tokens, consisting of press articles, academic and legal documents, blogs, speeches, literary works, and so on.7 The analysis is based on all in- flected forms of the verbs, including forms with pronominal clitics. Since the stems of templatic verbs are subjected to several allomorphic variations (cf. Ch. 2), the corpus investigation has so far concentrated on concatenative verbs only, leaving the three templatic verbs, għaġġel ‘hurry up’, għama ‘blind/go blind’, and xegħel ‘light’, for future research. Another nine verbs were left out of the analysis because very few tokens were found. These include iċċattja ‘flatten’, innokkla ‘curl’, ittowja ‘thaw’. In what follows, I therefore present the results of corpus searches conducted on 25 labile verbs. As Table 5.4a and Figure 5.2 illustrate, the corpus study confirms that verbs such as iċċarġja ‘charge’, illixxa ‘smooth’ and issikka ‘tighten’ have a strong transitive bias. As expected, verbs like evapora ‘evaporate’, sparixxa ‘vanish’ and sploda ‘explode’ occur most frequently in intransitive frames. It also confirms that irdoppja ‘double’ has almost equal chances to be used in a transitive as well as an intransitive construction. The re- sults are summarized in Table 5.4b. However, a comparison of the two studies reveals that the correlation is far from perfect for a few of the verbs. For instance, varja ‘vary’ shows a 50-50 split in the sen- tence creation task, but has a strong preference for intransitive constructions in the corpus study. This seems to suggest that more data from more subjects is required to yield more clear-cut results, especially for the verbs that show no particular bias. Tak- ing another verb, żvina ‘bleed to death’ is an internally caused verb according to the

7 The corpus can be accessed on http://mlrs.research.um.edu.mt.

Towards a unified account 169

sentence creation task, but it displays a propensity for transitive frames in the corpus investigation. In this case, the mismatch appears to be due to the fact that most of the texts in which żvina occurred dealt with the exsanguination of animals. Extending the corpus analysis to a larger number of texts from a wider range of genres will certainly provide less equivocal results.

Table 5.4a Labile verbs: results of the corpus study

Labile verb Gloss % Intransitive % Transitive No. of items

2 ibbilanċja balance 10.6 89.4 104 3 iċċarġja charge 13.3 86.7 15 5 iddisintegra disintegrate 66.7 33.3 15 7 iddritta straighten 22.8 77.2 57 8 espanda expand 52 48 98 9 evapora evaporate 79.8 20.2 89 10 iffriża freeze 13.2 86.8 121 13 impruvja improve 84.6 15.4 39 14 intensifika intensify 25 75 56 15 ikkalma calm down 56.9 43.1 167 17 ikkuntenta please/be pleased 56.2 43.8 153 18 illaxka loosen 62.3 37.7 61 19 illixxa smooth 12.5 87.5 32 20 immansa tame 16 84 25 22 ippurifika purify 20 80 35 23 irdoppja double 60.4 39.6 96 24 irrombla roll 75 25 56 26 sparixxa vanish 83.8 16.2 111 27 sploda explode 76 24 179 28 issikka tighten 9.1 90.9 44 29 issoda strengthen 73.3 26.7 15 31 varja vary 97.5 2.5 199 34 żviluppa develop 43.4 56.6 408 35 żvina bleed to death 25 75 16 37 żvolġa develop 71.2 28.8 125

170 A Tale of Two Morphologies

Towards a unified account 171

Table 5.4b Labile verbs in corpus study grouped according to transitive bias

Intransitive bias Transitive bias No significant bias

varja żvolġa issikka iddritta espanda impruvja ibbilanċja intensifika ikkuntenta sparixxa illixxa żvina żviluppa evapora iffriża ikkalma sploda iċċarġja irdoppja irrombla immansa illaxka issoda ippurifika iddisintegra

5.3 Summary

In this chapter, I have tried to give a unified account of the verbal system in Maltese by examining the formal properties of verbs undergoing the causative-inchoative alterna- tion. The analysis revealed that there is considerable variation in the morphological marking of the alternation in Maltese, which runs parallel to the distinction in word formation between templatic and concatenative verbs. Let us consider templatic verbs first. From a qualitative analysis of tri-consonantal roots and the patterns they combine with, it was observed that templatic verbs gener- ally express the alternation by casting one consonantal root in two different binyanim. It is often the case that the inchoative alternant is embedded in a pattern that is mor- phologically more marked than the binyan that hosts the causative alternant (e.g., fetaħ ‘open, v.t’ – infetaħ ‘open, v.i’). There are also a small number of cases where the direc- tion of markedness is the opposite, i.e. the inchoative verb carries less morphology than the causative one (e.g., nixxef ‘dry, v.t’ – nixef ‘dry, v.i’). The following list includes the different ways causative-inchoative pairs may be marked in root-and-pattern morphol- ogy (cf. Ch. 4). Only the combination of patterns that make up over 2% of the total num- ber of causative-inchoative alternations in the binyan system are given below.8

8 Among the infrequent combinations, which add up to 8% of the total, is a small group of roots that form both causative and inchoative verbs in binyan I or II (cf. Sect. 4.2.1.2).

172 A Tale of Two Morphologies

Binyan (causative) Binyan (inchoative)

II C1vC2C2vC3 V tC1vC2C2vC3

II C1vC2C2vC3 I C1vC2vC3

I C1vC2vC3 VII nC1vC2vC3

II C1vC2C2vC3 IX C1C2vvC3

I C1vC2vC3 VIII C1tvC2vC3

III C1vvC2vC3 VI tC1vvC2vC3

II C1vC2C2vC3 VII nC1vC2vC3

As discussed earlier and as is evident from the above list, there is considerable variation in the morphological realization of the alternation. The choice of the binyan for causative and inchoative verbs is, to a certain extent, up to the root. This would be unexpected if we had to consider causatives to be derived from inchoatives, or vice versa. If, on the other hand, we take the causative-inchoative alternation to involve two root-derived verbs, then this morphological behavior, along with other factors such as vocalic variation, specialized meanings, and gaps in the paradigm, is expected. The root may be freely inserted in the environment of a causative and an inchoative verb, as il-

lustrated below:

Vcausative Vinchoative

√root v, binyan √root v, binyan

Morphological spell-out: Morphological spell-out:

C1vC2vC3 (I) C1vC2vC3 (I)

C1vC2C2vC3 (II) tC1vC2C2vC3 (V)

C1vvC2vC3 (III) tC1vvC2vC3 (VI)

nC1vC2vC3 (VII)

C1tvC2vC3 (VIII)

C1C2vvC3 (IX)

Concatenative verbs, by contrast, do not express the alternation overtly. Since in concatenative morphology roots are typically associated with a single verbal mor- pheme, the causative-inchoative contrast is generally marked by labile verbs, where the same form is embedded in both transitive and intransitive constructions (e.g., iċċara ‘brighten, v.i, v.t’). In a few cases, the reflexive construction, involving the marker ruħ-

Towards a unified account 173

‘soul’ plus a direct object clitic, is used to mark the inchoative alternant (e.g., irdoppja ‘double, v.t’ – irdoppja ruħu ‘double, v.i’). Like in root-and-pattern morphology, concate- native causatives and inchoatives display irregularities, including gaps, e.g., iddeterjora ‘deteriorate’ and ixxrinkja ‘shrink’ have no corresponding causative, while ibbliċja ‘bleach’ and immodifika ‘modify’ stand on their own, without an inchoative counterpart; and idiosyncratic interpretations unavailable in the other alternant, e.g., there is no causative for illaxka in the idiomatic sense of ‘become licentious’. These irregularities can be accounted for if we assume the root to be in a local relation with the causative and inchoative verbal heads that merge with it:

Vcausative Vinchoative

√root v, suffix √root v, suffix

The morphological variation found in Maltese, as in other languages (cf. Sect. 5.1) raises two important questions. The first one is of a theoretical nature. What is the deri- vational relation, if any, between verbs undergoing the alternation? After surveying two main views that relate causatives and inchoatives derivationally, I observed that both accounts, causativization and detransitivization, leave unexplained half of the morpho- logical paradigm found cross-linguistically. On the basis of evidence from the morpho- nological and semantic behavior of templatic verbs in Maltese, I argued that the alterna- tion is better captured by a non-derivational account that assumes causatives and in- choatives are both derived from an underlying root. This analysis has the further bene- fit of accounting for the cross-linguistic differences in the morphological realization of the alternation. The second question concerns the lexical semantics of the verbs. On the assump- tion that the morphosyntactic structure is directly related to the lexical semantic struc- ture of verbs, I argued that the morphological variation found in Maltese is sensitive to the kind of event named by the verbs in question, whether internally or externally caused. Since externally caused events are conceptualized as coming about because of something external to the entity undergoing the change of state, the prediction is that in templatic verbs, the causative alternant will be unmarked and the inchoative one will

174 A Tale of Two Morphologies

be marked. On the contrary, when the inchoative is morphologically unmarked and the causative marked, the verbs must be denote an internally caused event, one in which the means for bringing about the change of state is conceptualized as residing in the entity undergoing the change itself. Along the same lines, for concatenative, labile verbs it is expected that externally caused verbs will be used mostly in their transitive form, whereas internally caused verbs will occur more often in intransitive frames. A sentence creation task and a cor- pus study on a set of around 40 labile verbs revealed that these change of state verbs generally show a strong bias for either transitive or intransitive constructions, even if concatenative (and a few templatic) verbs fail to show any morphological mark that would distinguish causative or inchoative as formally more marked than the other. The presence of a small number of labile verbs that show an almost equal likelihood of ap- pearing in transitive or intransitive frames suggests that some roots may be unspecified for the type of causation involved, allowing them to surface both with and without an external argument.

Chapter 6 Bringing down the curtain

This final chapter serves two purposes. It marks out the route we followed in this work, and indicates other paths that future research on Maltese verbs may take. I shall wind up the argument made in this study in two steps. First, I give an overview of the key dif- ferences between templatic and concatenative verbs. Then, I show how, over and above these differences, the two verb classes converge in one system.

6.1 Contrasting the two morphologies

In view of the formal differences between root-and-pattern morphology and concatena- tive morphology, a comprehensive grammar of Maltese is perhaps best presented using the split page method. I now outline some of the ways in which templatic verbs differ from concatenative verbs, as discussed throughout this work, in particular in Ch. 2 and Ch. 5.

Templatic verbs Concatenative verbs

Structural may be decomposed into a con- formed by the combination of a makeup sonantal root and a verbal pat- syllabic root and a verbal suffix; tern or binyan, often marked by initial gemina-

e.g., √ktb + C1vC2vC3 tion, e.g., √TAJP + -ja = kiteb ‘write’ = ittajpja ‘type’

Stem shape fixed prosodic structure; must no restrictions on the syllabic be in the form of a monosyllabic configuration and vowel quality, or disyllabic binyan, which al- other than the phonological re- lows for a limited set of vowel strictions that apply throughout sequences the language 176 A Tale of Two Morphologies

Inflection take either of two sets of inflec- take weak affixes only (-ajt, -at, (tense-aspect, tional affixes, strong (-t, -et, -na, -ajna, etc.) person, num- etc.) or weak (-ajt, -at, -ajna, ber, gender) etc.), or a combination of the two, as in the case of silent-final verbs

Distribution consonantal roots generally syllabic roots typically take one of roots combine with more than one verbalizing suffix only verbal pattern

Argument mediated through the verbal marked periphrastically or by alternations patterns, as every verb must be means of labile or ambitransi- in the form of a binyan tive verbs

Past formed by the combination of a built by the suffixation of -at, -it participles root and a word pattern con- or -ut to a verbal stem, taining the prefix m-, e.g., ittajpja + -at

e.g., √ktb + mvC1C2uuC3 = ittajpjat ‘typed’ = miktub ‘written’

Verbal created by inserting a root into formed by suffixing -ar, -ment, nouns a nominal pattern, etc. to a verbal stem,

e.g., √ktb + C1vC2C3a e.g., ittajpja + -ar = kitba ‘writing’ = ittajpjar ‘typing’

Productivity closed, virtually unproductive open, highly productive (avail- (unavailable) class, as no new able) class verb is formed through root- and-pattern morphology

Etymology mostly of Semitic origin, with a predominantly of Romance and few hundred verbs derived English origin, with very few from Romance, and a couple of verbs derived from Semitic, re- verbs derived from English analyzed in a concatenative fashion

Bringing down the curtain 177

6.2 Comparing the two morphologies

In order to be able to compare the two systems and arrive at a unified account of the two verb formation strategies in Maltese, it was important to study root-and-pattern morphology first. Since templatic verbs constitute a closed class of verbs, it was possi- ble to compile an exhaustive database of all consonantal roots and the patterns they occur in, and come up with a classification based on the lexical semantics and morpho- logical marking of the verbs. In what follows I illustrate the main findings of this exami- nation, as discussed in Ch. 3 and Ch. 4, and propose a way of extending this analysis to concatenative verbs, as shown in Ch. 5.

6.2.1 Analyzing templatic verbs in Maltese

In this study, templatic verbs are understood as the end result of the interdigitation of a category-neutral, consonantal root with a binyan, a verbal pattern which has a specific syllabic structure, and slots for vowels and root consonants to fit in. The binyan not only makes the root consonants pronounceable, but it also determines the lexical cate- gory: it is only when roots are inserted in a particular pattern that an actual verb, noun or adjective is formed. This implies that all templatic verbs must be in the form of a bin- yan. I argued that the binyanim belong to the domain of derivation, not to the marking of inflectional categories such as person, number, tense-aspect, and gender. They display two main characteristics of word formation. First, they are not completely productive: the binyan system is full of gaps (cf. Sect. 4.1). Second, word formation creates words with lexicalized meanings, and when cast in different verbal patterns, roots may take on specialized, non-transparent interpretations. See the discussion in Sect. 2.2.2 for further arguments that point to the conclusion that binyan morphology is essentially deriva- tional. A root can in principle combine with several binyanim to form different verbs. However, no tri-consonantal root combines with all nine patterns. Over two thirds of all tri-consonantal roots appear in only one or two patterns. The mean number of patterns per root is around 2, which means that many binyan slots are left empty. The combination of roots and binyanim is subject to a number of phonological con- straints. For instance, roots starting with /k, t, tʃ, dʒ/ etc. cannot appear in binyan X,

178 A Tale of Two Morphologies

which is characterized by the prefix st-, as they would create onset consonant clusters that are not permissible in Maltese, *kst, *tst etc. In addition, an examination of the pos- sible relationships between patterns and root types, strong (regular, reduplicative) and weak (initial, medial, final), revealed that binyan III and VI disfavor reduplicative roots, and that weak-final roots are unlikely to be inserted in binyan II, V and IX. See Sect. 4.1 for further (dis)preferences. An analysis of the distribution of roots across patterns showed that only four pat- terns are really productive in Maltese, namely binyan I, II, V and VII. Moreover, I argued in Sect. 4.3 that, in terms of morphological productivity, morphosyntactic similarity, and complementarity in distribution, templatic verbs in Maltese make up a four-way system: III, VI, VIII, IX are, as it were, the shadow binyanim of II, V, VII and I respec- tively. In general, this quasi-allomorphy of patterns is phonologically conditioned. The remaining pattern, binyan X, is highly underrepresented, as it combines with only 24 roots, making up less than one percent of the total number of tri-consonantal roots in Maltese. In Ch. 3, I showed that the binyan system is inherently regular and irregular from three different angles. First, when a root is embedded in more than one pattern, this very often correlates with a regular argument structure alternation (e.g., √ktb, kiteb ‘write’ – inkiteb ‘be written’). However, two or more verbs containing the same root consonants may have two semantically distant interpretations (e.g., √xrb, xorob ‘drink’ – xarrab ‘wet’) or may be synonymous (e.g., √slf, silef ‘lend’ – sellef ‘lend’). The qualita- tive analysis carried out in Ch. 4 brought to light the fact that 61% of all combinations of tri-consonantal roots with patterns mark an argument alternation, 18% are synony- mous verbs, 9% correspond to multiple, semantically distant interpretations, and 12% are cases of roots that derive singleton verbs. Second, from a formal point of view, tem- platic verbs sharing a single consonantal root may either have the same vowel se- quence, as in kiteb and inkiteb, or display vocalic variation, as in xorob and xarrab. Fi- nally, an examination of the distribution of verbs in the lexicon reveals that verbs de- rived from the same root may be fully productive (e.g., passives in binyan V, tqassam ‘be distributed’, always have an active counterpart, typically in binyan II, qassam ‘distrib- ute’) or exhibit gaps in the system (e.g., inchoatives in IX, smar ‘get tanned’, do not al- ways have a corresponding causative). These conflicting aspects of the binyan system, I argue, can be explained if we:

Bringing down the curtain 179

(a) assume that roots are category-neutral and semantically underspecified (it is only when they are inserted in a word pattern that their lexical category is de- termined and that they acquire a specific interpretation); (b) recognize that there is no one-to-one relation between the semantic and syntac- tic properties of verbs and their morphological form (the same binyan may host verbs of different semantic types (e.g., a verb in binyan I may be active, causa- tive, inchoative), and verbs of the same semantic type (e.g., active) may appear in different patterns (e.g., I, II, III, VIII)); (c) assume, in line with Marantz (1997, 2001, 2007), Arad (2003a, 2003b, 2005), and others, that: (i) words can be formed either by merging a root with a category-bearing head or by combining an existing word with a new syntactic head; (ii) word formation from roots gives rise to irregular, non-productive proc- esses, whereas word formation from previously formed words results in regular, productive processes.

From the qualitative analysis in Ch. 4, we concluded that one main source of irregu- larity in the binyan system is syncretism. There is no one-to-one mapping between the morphological form of verbs and their semantic-syntactic structure. This can be ob- served in two ways. First, the same combination of binyanim may mark more than one semantic-syntactic category. A root in, say, binyan I and II may represent a causative- inchoative alternation (e.g., √għlj, għola ‘rise’, għolla ‘raise’), a pair of synonyms (e.g., √bdl, bidel, biddel ‘change’), multiple interpretations (e.g., √tbgħ tebagħ ‘print’, tebba’ ‘stain’), and others. Second, the same semantic-syntactic feature may be marked by more than one combination of patterns. For instance, the transitive-reflexive alterna- tion may be realized in II – V, I – VII, III – VI, etc. (cf. Sect. 4.2). We also noted that regularity in the system stems from two main sources. First, there is a strong tendency for complementary distribution among the four categories roots fall into (argument alternations, multiple interpretations, synonyms, singletons). For instance, when a root is inserted in VII and VIII, this almost always correlates with two synonyms and hardly ever with an argument alternation. Second, argument alter- nations are strikingly regular with respect to directionality. While rejecting the deter- ministic approach of traditional grammars of Maltese that find neat associations be- tween patterns and semantic roles such as causative and passive, we can observe the significant contribution of the binyanim in terms of transitivity. The system is, to a cer-

180 A Tale of Two Morphologies

tain extent, morphologically transparent as it imposes transitivity constraints on the binyanim. From the fact that some of them are inherently or predominantly intransitive (V, VII, IX) and others are predominantly transitive (II, III), it follows that argument al- ternations are very often unidirectional, with passives, reflexives and non-causatives appearing in the patterns that typically host intransitive verbs and their active, non- reflexive and causative counterparts appearing in the predominantly transitive pat- terns. Although argument alternations display a great degree of regularity, we observed that two alternations, the causative-inchoative and causative-noncausative, are irregu- lar in several respects. On the basis of the structural distinction between root deriva- tion and word derivation, I argue that passives and reflexives are verb derived, while causatives and non-causatives are root derived. In Sect. 4.3, I presented morphonologi- cal and semantic evidence for this claim. Let us go over the key differences between root-derived and verb-derived verbs. To begin with, passives and reflexives have the same formal properties (e.g., se- quence of vowels, medial gemination) as the transitive verbs they are derived from. Causatives and inchoatives, by contrast, may display vocalic variation. In an overall analysis, it turns out that vocalic variation among verbs containing the same root is al- ways associated with root derivations, i.e. multiple interpretations (e.g., fired ‘separate’ – farrad ‘make unpaired’), synonyms (e.g., silef – sellef ‘lend’), causative-inchoative (e.g., kiber ‘grow v.i’ – kabbar ‘grow v.t’), causative-noncausative (e.g., żifen ‘dance’ – żeffen ‘make one dance’). Related to this is the observation that passives and reflexives are morphologically more complex than the transitive verbs they alternate with. Causatives and inchoatives, on the other hand, may be both morphologically marked. Besides, there are cases where one labile verb marks both the causative and inchoative verb. The third piece of evidence is in terms of productivity. While passives and reflexives always have a transitive counterpart, it is not difficult to find causatives that stand alone, without a corresponding inchoative (e.g., sann ‘sharpen’), and inchoatives that lack a causative counterpart (e.g., batta ‘abate’). Finally, from a semantic point of view, it is very rare for passives and reflexives to take on an interpretation that is not available in their active and non-reflexive counter- parts, as discussed in Ch. 3. Causatives, by contrast, acquire specialized and idiomatic meanings that do not exist with their corresponding inchoatives. The same holds for

Bringing down the curtain 181

inchoatives, which may be assigned idiosyncratic interpretations that are not present in the causative verbs they alternate with.

6.2.2 Extending the analysis to concatenative verbs

The causative-inchoative alternation was chosen as the point of departure for a com- parative treatment of templatic and concatenative verbs in Maltese because it is a basic argument alternation which makes up around one third of all verb alternations in tem- platic verbs, and because it brings into play various morphological, syntactic and lexical semantic aspects of the verbs. An examination of the formal encoding of the alternation in Maltese revealed that there is variation in the marking of the alternation, which goes hand in hand with the distinction between the two verb formation strategies. Templatic verbs typically mark the alternation by means of two different morphologically related verbs, with the in- choative verb being generally morphologically more complex than the causative one. By contrast, concatenative verbs do not express the alternation overtly: labile verbs, i.e. ambitransitive verbal forms, are used to mark both the causative and the inchoative alternant, as in it-tifel iċċarġja l-mobile ‘the boy recharged the mobile’ and il-mobile iċċarġja ‘the mobile recharged’. In a few cases, the reflexive marker ruħ- plus a direct object clitic is used to mark the inchoative alternant, as in il-problema intensifikat ruħha ‘the problem has intensified’. In Ch. 4 and Ch. 5, I argued that the morphological irregularities of the causative- inchoative alternation are best captured by a root-based analysis. In spite of the differ- ences in the formal encoding of the alternation by templatic and concatenative verbs, a unified analysis of the causative-inchoative alternation in Maltese can be provided if we assume that all verbs taking part in the alternation are cases of root derivation. From a sentence creation task and a corpus study on a set of around 40 labile verbs in Maltese, we concluded that labile verbs do not all behave in the same way. Even though they fail to show any morphological mark that would distinguish the causative or inchoative alternant as formally more marked than the other, some of them are more likely to be used in transitive constructions, while others tend to occur in intransitive frames. This difference among alternating labile verbs was explained in lexical semantic terms. Assuming that the morphosyntactic structure of verbs is directly related to their lexical semantic structure, I argue that the variation found in labile verbs in Maltese is

182 A Tale of Two Morphologies

sensitive to the kind of event named by the verbs in question, whether it is internally or externally caused. Externally caused events, like iċċarġja ‘recharge’, are conceptualized as coming about due to a force external to the entity undergoing the change of state. By contrast, in internally caused events, like sparixxa ‘vanish’, the means for bringing about the change of state is conceptualized as residing in the entity undergoing the change itself. It follows from this that externally caused labile verbs are used more frequently in transitive constructions, whereas internally caused labile verbs occur more often in intransitive frames.

6.3 Tales in search of an author

This study opens up new avenues for research on the two morphologies in Maltese. As a continuation of this study, a quantitative and qualitative analysis of templatic verbs de- rived from quadri-consonantal roots (cf. Appendix II) is needed to have a complete pic- ture of the morphology and lexical semantics of all templatic verbs in Maltese. Second, the comparative approach taken in Ch. 5 of this study needs to be extended to the other argument alternations, namely the active-passive, transitive-reflexive and causative- noncausative. Further psycholinguistic and corpus-based studies on labile verbs are required to confirm the bias in transitivity observed in this work, and to shed light on the class of verbs that show an almost equal likelihood to appear in either transitive or intransitive patterns. In addition, comparative studies can be undertaken to determine possible dif- ferences and similarities between labile verbs of English origin, such as ibbawnsja, iċċarġja, iffriża, impruvja, ikklirja, slowja and their English counterparts bounce, charge, freeze, improve, clear, slow (down). From a more general perspective, in order to identify more points of convergence and divergence between templatic and concatenative verbs, further research is needed on the morphological productivity of the two verb classes to establish the availability and profitability of root-and-pattern and concatenative morphology, and to address such questions as how speakers decide which strategy to use to form new words. Ex- perimental studies aimed to provide statistical calculations of productivity based on measuring profitability in corpora, along the lines of Baayen (1992, 1993), Gaeta & Ricca (2006), inter alia, need to be replicated for Maltese. Also, the relative frequency and productivity of the various binyanim can be measured following studies such as Bolozky (1999), who adopts different methods of measuring and evaluating productiv-

Bringing down the curtain 183

ity of word formation (productivity tests, dictionary comparison, and corpus analysis) in Israeli Hebrew, and Verheij (2000), who applies Baayen’s models to assess the pro- ductivity of binyanim in Biblical Hebrew.

References

Acquaviva, Paolo. 2009. Roots and Lexicality in Distributed Morphology. In Alexandra Galani, Daniel Red- inger & Norman Yeo (eds.), York Papers in Linguistics, 1-21. York. Agius, Dionysius. 1996. Siculo-Arabic. London and New York: Kegan Paul International. Akkademja tal-Malti. 1992. Aġġornament tat-Tagħrif fuq il-Kitba Maltija. Malta. Alexiadou, Artemis. 2006. On (anti-)causative alternations. Handout, EALing 2006, http://ealing.cognition.ens.fr/ealing2006/handouts/alexiadou2.pdf. Alexiadou, Artemis & Elena Anagnostopoulou. 2004. Voice morphology in the causative-inchoative alterna- tion: Evidence for a non-unified structural analysis of unaccusatives. In Artemis Alexiadou, Elena Anagnostopoulou & Martin Everaert (eds.), The unaccusativity puzzle: Explorations of the syntax- lexicon interface, 114-136. Oxford: Oxford University Press. Alexiadou, Artemis, Elena Anagnostopoulou & Florian Schäfer. 2006. The properties of anticausatives crosslinguistically. In Mara Frascarelli (ed.), Phases of Interpretation, 187-211. Berlin: Mouton. Anagnostopoulou, Elena & Yota Samioti. 2009. Domains for idioms. Paper presented at Roots: Word forma- tion from the perspective of ‘core lexical elements’, Stuttgart, 10-13 June. Anagnostopoulou, Elena & Yota Samioti. 2010. Locality Constraints on the Interpretation of Words: Greek Participles”. Paper presented at the Workshop on Linguistic Interfaces, University of Ulster, 2-4 De- cember. Anshen, Frank & Mark Aronoff. 1988. Producing morphologically complex words. Linguistics 26. 641-655. Aquilina, . 1958. Maltese as a mixed language. Journal of Semitic Studies 3(1). 58-79. Aquilina, Joseph. 1959. The Structure of Maltese. A Study in Mixed Grammar and Vocabulary. Malta: The Royal . Aquilina, Joseph. 1961a. Maltese ax a mixed language. Papers in Maltese Linguistics, 42-62. Malta: Royal University of Malta. Aquilina, Joseph. 1961b. Some historical phonetic changes of Maltese. Papers in Maltese Linguistics, 117- 165. Malta: Royal University of Malta. Aquilina, Joseph. 1979. Maltese Arabic Comparative Grammar. Malta: University of Malta. Aquilina, Joseph. 1987-1990. Maltese-English Dictionary, 2 volumes. Malta: Midsea. Arad, Maya. 2003a. Locality constraints on the interpretation of roots: The case of Hebrew denominal verbs. Natural Language & Linguistic Theory 21. 737-778. Arad, Maya. 2003b. Why syntax matters. Recherches Linguistiques de Vincennes 32. 83-108. Arad, Maya. 2005. Roots and patterns: Hebrew morpho-syntax. Dordrecht: Springer. Aronoff, Mark. 1976. Word formation in generative grammar. Cambridge, MA: MIT Press. Aronoff, Mark. 1994. Morphology by Itself. Cambridge, MA: MIT Press. Baayen, R. Harald. 1992. Quantitative aspects of morphological productivity. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 1991, 109-149. Dordrecht: Kluwer. Baayen, R. Harald. 1993. On frequency, transparency and productivity. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 1992, 181-208. Dordrecht: Kluwer. Baayen, R. Harald. 1994. Productivity in language production. Language and Cognitive Processes 9(3). 447- 469. Baayen, R. Harald. 2001. Word-Frequency Distributions. Dordrecht: Kluwer. Baayen, R. Harald & Antoinette Renouf. 1996. Chronicling the Times: Productive lexical innovations in an English newspaper. Language 72. 69-96. Bachra, Bernard. 2001. The Phonological Structure of the Verbal Roots in Arabic and Hebrew. Leiden: Brill. Baeskow, Heike. 2006. A revival of Romance roots. Morphology 16. 3-36. Bakker, Peter & Maarten Mous. 1994. Mixed languages. 15 Case Studies in Language Intertwining. Amst- erdam: Ifott. Barr, James. 1961. The Semantics of Biblical Language. Oxford: Oxford University Press. Bat-El, Outi. 1994. Stem modification and cluster transfer in Modern Hebrew. Natural Language and Natu- ral Linguistic Theory 12. 571-596. References 185

Bat-El Outi. 2001. In search for the roots of the C-root: The essence of Semitic morphology. Handout from The Workshop on Roots and Template Morphology. Los Angeles: USC. Bat-El, Outi. 2003a. Semitic verb structure within a universal perspective. In Joseph Shimron (ed.), Lan- guages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 29-59. Amsterdam: John Benjamins. Bat-El, Outi. 2003b. The fate of the consonantal root and the binyan in Optimality Theory. Recherches lin- guistiques de Vincennes 32. 31-69. Benmamoun, Elabbas. 2003. The role of the imperfective template in Arabic morphology. In Joseph Shim- ron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 99- 114. Amsterdam: John Benjamins. Berent, Iris & Joseph Shimron. 2003. What is a root? In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 201-222. Amsterdam: John Benjamins. Berent, Iris, Vered Vaknin & Gary Marcus. 2007. Roots, stems, and the universality of lexical representa- tions: Evidence from Hebrew. Cognition 104. 254-286. Berman, Ruth A. 2003. Children’s lexical innovations. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 243-291. Amsterdam: John Benjamins. Bickel, Balthasar, Bernard Comrie & Martin Haspelmath. 2008. Leipzig glossing rules. http://www.eva.mpg.de/lingua/pdf/ LGR09_02_23.pdf (5 July 2011). Bierwisch, Manfred. 1982. Formal and lexical semantics. Linguistische Berichte 80/82. 3-17. Bierwisch, Manfred. 1988. On the grammar of local prepositions. In M. Bierwisch, W. Motsch & I. Zimmer- mann (eds.), Syntax, Semantik und Lexikon, 1-65. Berlin: Akademie Verlag. Bierwisch, Manfred. 1996. How much space gets into language? In P. Bloom, M. Peterson, L. Nadel & M. Garrett (eds.), Language and Space, 31-76. Cambridge, Mass.: The MIT Press. Bierwisch, Manfred. 1997. Lexical information from a minimalist point of view. In Ch. Wilder, H.-M. Gärtner & M. Bierwisch (eds.), The Role of Economy Principles in Linguistic Theory, 227–266. Berlin: Akademie Verlag. Bierwisch, Manfred & Ewald Lang. 1989. Somewhat longer – much deeper – further and further. Epilogue to the dimension adjective project. In M. Bierwisch & E. Lang (eds.), Dimensional Adjectives: Grammati- cal Structure and Conceptual Interpretation, 471-514. Berlin: Springer. Blutner, Reinhard. 1998a. Lexical pragmatics. Journal of Semantics 15. 115-162. Blutner, Reinhard. 1998b. Lexical Underspecification and Pragmatics. In Ludewig & Bart Geurts (eds), Lexikalische Semantik aus kognitiver Sicht, 141-171. Tübingen: Gunter Narr Verlag. Bolozky, Shmuel. 1999. Measuring producitivity in word formation. The case of Israeli Hebrew. Leiden: Brill. Borer, Hagit. 2003. Computing argument structure. In Joseph Shimron (ed.), Languages Processing and Ac- quisition in Languages of Semitic, Root-based, Morphology, 321-362. Amsterdam: John Benjamins. Borer, Hagit. 2009a. Roots and categories. Talk presented at the 19th Colloquium on Generative Grammar, University of the Basque Country, April 1-3. Borer, Hagit. 2009b. Very late insertion. Paper presented at Root Bound. Los Angeles: USC, 21 February. Borg, Albert. 1981. A study of aspect in Maltese. Ann Arbor: Karoma Publishers. Borg, Albert. 1988. Ilsienna: Studju grammatikali. Malta. Borg, Albert. Forthcoming. Maltese as a national language. To appear in Stefan Weninger (ed.), Semitic Lan- guages. An International Handbook on their Structure, their History and their Investigation, Berlin: Wal- ter de Gruyter. Borg, Albert & Manwel Mifsud. 1999. 'Il-forom' (Binyanim) in Maltese: inflection or derivation? Paper pre- sented in the Second Mediterranean Meeting on Morphology, Malta, 10-12 September. Borg, Albert & Manwel Mifsud. 2002. Maltese Object Marking in a Mediterranean Context. Paolo Ramat & Thomas Stolz (eds.), Mediterranean Languages 32-46. Bochum: Universitätsverlag Dr. N. Brockmeyer. Borg, Alexander. 1978. A Historical and Comparative Phonology and Morphology of Maltese. , Is- rael: The Senate of the Hebrew University dissertation. Borg, Alexander. 1994. Language. In Henry Frendo & (eds.). Malta: Culture and Identity, 27- 50. Malta: Ministry of Youth and the Arts. Borg, Alexander. 1996. On some Levantine linguistic traits in Maltese. Journal of Studies. Studies in Modern Semitic Languages 16. 133-152. Bos, Johan. 2004. Computational semantics in discourse: Underspecification, resolution, and inference. Journal of Logic, Language and Information 13. 139-157. Brame, Michael. 1972. On the Abstractness of Phonology: Maltese ʕ. In Michael Brame (ed.), Contributions to Generative Phonology, 53-77. Austin: University of Texas Press. Brincat, Joseph M. 1998. A Malta l’italiano lo insegna la televisione. Italiano e Oltre XIII. 52-58. Brincat, Joseph M. 1999. Languages Across Frontiers. The Acquisition of Italian in Malta by Viewers of Ital- ian TV programmes. In Bernard Caron (ed.), Proceedings of the XVIth International Congress of Lin- guists. , 20-25 July 1997. Oxford: Elsevier Sciences. Brincat, Joseph M. 2000. Il-Malti: Elf Sena ta’ Storja. Malta: PIN. Brincat, Joseph M. 2004. Malta. Una Storia Linguistica. Genova: Le Mani.

186 A Tale of Two Morphologies

Brincat, Joseph M. 2008. Malta. In Kees Versteegh (ed.), Encyclopedia of Arabic Language and Linguistics, vol. III, 141-145. Buccellati, Giorgio. 1997. Akkadian. In Hetzron (ed.), The Semitic languages, 69-99. New York: Rout- ledge. Buckley, Eugene. 2003. Emergent vowels in Tigrinya templates. In Jacqueline Lecarme (ed.), Research in Afroasiatic Grammar II, 105-125. Philadelphia: John Benjamins. Burzio, Luigi. 1986. Italian Syntax. A Government and Binding Approach. Dordrecht: Reide. Centineo, Giulia. 1995. The distribution of si in Italian transitive/inchoative pairs. In Mandy Simons & Teresa Galloway (eds.), Proceedings from Semantics and Linguistic Theory V, 54-71. Ithaca: Cornell Uni- versity. Cachia, Lawrenz. 1984. Il-Verbi fil-Malti. Studju dwar il-forom. Malta: CharVin Press & Co. Ltd. Cachia, Lawrenz. 1994. Grammatika ġdida tal-Malti. Malta: Veritas Press. Camilleri, Marthese. 1990. Analiżi sintattika-semantika ta’ xi verbi tal-għaxar forma fil-Malti. Malta: Univer- sity of Malta B.Ed. (Hons.) thesis. Carston, Robyn. 2002. Thoughts and utterances. The pragmatics of explicit communication. Oxford: Black- well. Caruana, Sandro. 2006. Trilingualism in Malta. Maltese, English and italiano televisivo. International Journal of Multilingualism 3(3). 159-172. Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences. In Artemis Alex- iadou, Elena Anagnostopoulou & Martin Everaert (eds.), The unaccusativity puzzle, 22–59. Oxford: Ox- ford University Press. Chomsky, Noam. 2001. Derivation by phase. In M. Kenstowicz (ed.), Ken Hale: A Life in Language, 1-52. Cambridge, MA: MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In A. Belletti (ed.), Structures and Beyond, 104-131. New York: Oxford University Press. Chomsky, Noam. 2005. On phases. Unpublished ms., MIT. Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. Comrie, Bernard. 1986. The Maltese pharyngeal. Zeitschrift für Phonetik, Sprachwissenschaft und Kommuni- kationsforschung 39. 12-18. Comrie, Bernard. 2006. Transitivity pairs, markedness, and diachronic stability. Linguistics 44. 303-318. Comrie, Bernard. 2009. Maltese and the World Atlas of Language Structures. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz, & Martine Vanhove (eds.), Introducing Maltese Linguis- tics: Selected Papers from the 1st International Conference on Maltese Linguistics. Bremen, 18-20 Octo- ber, 2007, 3-11. Amsterdam: John Benjamins. Copestake, Ann & Ted Briscoe. 1995. Semi-productive polysemy and sense extension. Journal of Semantics 12. 15-67. Corbin, Danielle. 1987. Morphologie dérivationelle et structuration du lexique. Tübingen: Niemeyer. Cowan, William. 1966. Loss of emphasis in Maltese. Journal of Maltese Studies 3. 27-32. Cremona, Ninu. 1962. Tagħlim fuq il-kitba Maltija II. Malta: Lux Press. Cremona, Ninu. 1973. Tagħlim fuq il-kitba Maltija II. Malta: Lux Press. Davis, Henry & Hamida Demirdache. 2000. On Meanings: Evidence from Salish. In Carol Tenny & James Pustejovsky (eds.), Events as Grammatical Objects, 97-142. Stanford, CA: CSLI. Dell, François & Mohamed Elmedlaoui. 2002. Syllables in Tashlhiyt Berber and in . Dordrecht: Kluwer. Deutsch, Avital, Ram Frost, & Kenneth Forster. 1998. Verbs and nouns are organized and accessed differ- ently in the mental lexicon: Evidence from Hebrew. Journal of Experimental Psychology: Learning, Memory, and Cognition 24. 1238-1255. Deutsch, Avital & Ram Frost. 2003. Lexical organization and lexical access in a non-concatenated morphol- ogy. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 165-186. Amsterdam: John Benjamins. Doron, Edit. 2003. Agency and Voice: the semantics of the Semitic templates. Natural Language Semantics 11. 1-67. Doron, Edit & Malka Rappaport Hovav. 2009. A Unified Approach to Reflexivization in Semitic and Ro- mance. Brill’s Annual of Afroasiatic Languages and Linguistics 1. 75-105. Dowty, David. 1979. Word Meaning and Montague Grammar. The Semantics of Verbs and Times in Generative Semantincs and in Montague’s PTQ. Dordrecht: Reidel. Drewes, A.J. 1994. Borrowing in Maltese. In Peter Bakker & Maarten Mous (eds.), Mixed Languages. 15 Case Studies in Language Intertwining, 83-111. Amsterdam: Ifott. Ebert, Karen. 2000. Aspect in Maltese. In Ölsen Dahl (ed.), Tense and Aspect in the , 753- 785. Berlin: Mouton de Gruyter. Egg, Markus. 2010a. Semantic underspecification. Language and Linguistics Compass 4(3). 166-181. Egg, Markus. 2010b. Semantic underspecification: concepts and applications. Paper presented at the Se- maine Nancéienne de sémantique formelle, Nancy, 24 March. Ellul, Leanne. 2010. Il-Verbi Kwadrilitteri fil-Malti. Malta: University of Malta B.A (Hons) thesis.

References 187

Embick, David. 2000. Features, syntax, and categories in the Latin perfect, Linguistic Inquiry 31(2). 185- 229. Embick, David. 2004a. Unaccusative Syntax and Verbal Alternations. In Artemis Alexiadou, Elena Anag- nostopoulou & Martin Everaert (eds.), The unaccusativity puzzle: explorations of the syntax-lexicon interface, 137-158. Oxford, UK: Oxford University Press. Embick, David. 2004b. On the structure of resultative participles in English. Linguistic Inquiry 35. 355-392. Embick, David & Alec Marantz. 2008. Architecture and blocking. Linguistic Inquiry 39(1). 1-53. Ephratt, Michal. 1997. The psycholinguistic status of the root in Modern Hebrew. Folia Linguistica XXXI. 77- 103. Erwin, William M. 2004. A Short Reference Grammar of Iraqi Arabic. Washington: Georgetown University Press. Fabri, Ray. 1995. The Tense and Aspect System of Maltese. In Rolf Thieroff (ed.), Tense Systems in European languages, 327-343. Tübingen: Niemeyer. Fabri, Ray. 2009. Stem allomorphy in the Maltese verb. In Thomas Stolz (ed.), Ilsienna, 1-20. Bochum: Uni- versitätsverlag Brockmeyer. Farrugia, Ġorġ. 1998. Il-Malti lsien imħallat. In Karl Borg (ed.), Lingwa u Lingwistika, 227-291. Malta: Klabb Kotba Maltin. Faust, Noam & Ya’ar Hever. 2010. Empirical and theoretical arguments in favor of the discontinuous root in Semitic languages. Brill’s Annual of Afroasiatic Languages and Linguistics 2. 80-118. Fenech, Edward. 1978. Contemporary Journalistic Maltese: an Analytical and Comparative Study. Leiden: Brill. Fenech, Edward. 1980. Lingwistika Ġenerali. Studji Lingwistiċi tas-Seklu XX. Malta: AC Aquilina & Co. Hamp- ton. Folli, Raffaella. 2001. On the Relation of Priority Between Causative and Inchoative structures. In Yves D’Hulst, Johan Rooryk & Jan Schroten (eds.), and Linguistics Theory 1999, 143- 166. Amsterdam/Philadelphia: John Benjamins. Folli, Raffaella & Heidi Harley. 2006. Waltzing Matilda around and around: On the licensing of directed- motion resultatives. Studia Linguistica 60. 1-35. Frisch, Stefan. 2004. Language processing and segmental OCP effects. In Bruce Hayes, Robert Kirchner & Donca Steriade (eds.), Phonetically Based Phonology, 346-371. Cambridge: Cambridge University Press. Frisch, Stefan, Janet Pierrehumbert & Michael Broe. 2004. Similarity avoidance and the OCP. Natural Lan- guage & Linguistic Theory 22. 179-228. Frisson, Steven & Martin J. Pickering. 2001. Obtaining a figurative interpretation of a word: support for underspecification. Metaphor and Symbol 16. 149-171. Frisson, Steven. 2009. Semantic underspecification in language processing. Language and Linguistics Com- pass 3(1). 111-127. Frost, Ram, Kenneth Forster & Avital Deutsch. 1997. What we can learn from the morphology of Hebrew? A masked-priming investigation of morphological representation. Journal of Experimental Psychology, Learning, Memory, and Cognition 23. 829-856. Gafos, Adamantios. 1998. Eliminating Long-distance Consonant Spreading. Natural Language and Linguistic Theory 16. 223-278. Gafos, Adamantios. 2001. The Initial State and Verbal Stems in Arabic. Unpublished ms., Utrecht Institute of Linguistics, OTS. Gafos, Adamantios. 2003. Greenberg’s asymmetry in Arabic: A consequence of stems in paradigms. Lan- guage 79. 317-355. Gatt, Albert & Michael Spagnol. 2011. Labile verbs in Maltese. Talk given at the GĦILM Third international conference on Maltese linguistics, , April 8-10. Giegerich, Heinz. 1999. Lexical strata in English: Morphological causes, Phonological effects. Cambridge: Cambridge University Press. Goldenberg, Gideon. 1994. Principles of Semitic word-structure. In Gideon Goldenberg & Shlomo Raz (eds), Semitic and Cushitic Studies, 29-64. Wiesbaden: Harrassowitz. Goldsmith, John A. 1990. Autosegmental and metrical phonology. Cambridge, MA: Blackwell. Goral, Mira & Loraine K. Obler. 2003. Root-morpheme processing during word recognition in Hebrew speakers across the adult life span. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 223-242. Amsterdam: John Benjamins. Greenberg, Joseph. 1950. The patterning of root morpheme in Semitic. Word 6. 162-181. Gribanova, Vera. 2009. The Phonology and Syntax of Sub-words. Paper presented at Generative Linguistics in the Old World 32, Nantes, 16 April. Gribanova, Vera. 2011. Russian bracketing paradoxes as a window into the nature of morphosyntactic and phonological cyclicity. Paper presented at Phonetics and Phonology Workshop: Crosslinguistic Inves- tigations in Syntax-Phonology, Standford, 28 February. Grimshaw, Jane. 1981. On the lexical representation of Romance reflexivie clitics. In Joan Bresnan (ed.), The Mental Representation of Grammatical Relations, 87-148. Cambridge: MIT Press.

188 A Tale of Two Morphologies

Hale, Kenneth & Samuel J. Keyser. 1986. Some Transitivity Alternations in English. (Lexicon Project Work- ing Papers 7). Cambridge, MA: Center for Cognitive Science, MIT Press. Harley, Heidi. 2009. Roots: Identity, Insertion, Idiosyncracies. Paper presented at Root Bound, Los Angeles: USC, 21 February. Harley, Heidi. 2011. On the Identity of Roots. Paper presented at Approaches to the Lexicon (Roots III), Jerusalem, 13-16 June. Harley, Heidi & Rolf Noyer. 2000. Formal vs. Encyclopedic properties of vocabulary: evidence from nomi- nalization. In Bert Peeters (ed.), The Lexicon-Encyclopedia Interface, 349-374. Amsterdam: Elsevier Press. Haspelmath, Martin. 1990. The Grammaticalization of passive morphology. Studies in Language 14. 25-72. Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitives, 87-120. Amsterdam: Benjamins. Haspelmath, Martin. 2005. Universals of causative verb formation. LSA Institute class handout, August 2. Cambridge, MA. Haspelmath, Martin. 2008a. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive Lin- guistics 19. 1-33. Haspelmath, Martin. 2008b. Causatives and anticausatives. Syntactic Universals and Usage Frequency, Leipzig Spring School on Linguistic Diversity, March 2008. Leipzig. Hayes, John. 2001. The Integration of Romance Vocabulary in Maltese. Romance Philology 54. 393-405. Hoberman, Robert D. 2007. Maltese Morphology. In Alan S. Kaye (ed.), Morphologies of Asia and Africa 1, 257-281. USA: Eisenbrauns. Hoberman, Robert D. & Mark Aronoff. 2003. The verbal morphology of Maltese. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 61-78. Amst- erdam: John Benjamins. Hopper, Paul J. & Sandra A. Thompson. 1980. Transitivity in Grammar and Discourse. Language 56. 251- 299. Horvath, Julia. 1981. On the status of vowel patterns in Modern Hebrew: Morphological rules and lexical representations. In UCLA Occasional Papers 4. 228-261. Hume, Elizabeth. 1991. Consonant/Vowel Interactions in Maltese: Implication for Feature Theory. In B. Birch, K. Hunt & V. Samiian (eds.), WECOL 20, 138-151. California State University at Fresno. Hume, Elizabeth. 1996. , parallels in Maltese. Natural Language & Linguisitc Theory 14(1). 163-203. Hume, Elizabeth, Jennifer Venditti, Alexandra Vella & Samantha Gett. 2009. Vowel Duration and Maltese ‘għ’. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz, & Martine Vanhove (eds.), Introducing Maltese Linguistics: Selected Papers from the 1st International Conference on Maltese Linguistics. Bremen, 18-20 October, 2007, 15-46. Amsterdam: John Benjamins. Idrissi, Ali, Jean-François Prunet & Renée Béland. 2008. On the abstractness of Arabic roots. Linguistic In- quiry 39(2). 221-259. Jackendoff, Ray. 1990. Semantic structures. Cambridge: MIT Press. Jacobsen, Wesley. 1992. The Transitive Structure of Events in Japanese. Tokyo: Kuroshio. Kaye, Alan S. & Judith Rosenhouse. 1997. Arabic dialects and Maltese. In Robert Hetzron (ed.), The Semitic Languages, 263-311. New York: Routledge. Kiparsky, Paul. 1982. From cyclic to lexical phonology. In H van der Hulst & N. Smith (eds.), The structure of phonological representations, 131-175. Dordrecht: Foris. Kiparsky, Paul. 1997. Remarks on Denominal Verbs. In Alex Alsina, Joan Bresnan & Peter Sells (eds.), Com- plex Predicates, 473-499. Stanford: CSLI. Kitazume, Sachiko. 1996. Middles in English. Word 47. 161-183. Koontz-Garboden, . 2009. Anticausativization. Natural Language and Linguistic Theory 27. 77-138. Lakoff, George. 1970. Irregularity in Syntax. New York: Holt, Rinehart and Winston. Lavidas, Nikolaos. forthcoming. The diachrony of labile verbs: evidence from the history of English and Greek. In Seppo Kittilä & Leonid Kulikov (eds), Diachronic typology of voice and valency-changing cate- gories. Amsterdam: Benjamins. Lyons, John. 1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. Kontzi, Reinhart. 1981. L’elemento maltese nel maltese. Journal of Maltese Studies 14. 37-42. Kraaikamp, Margot. 2008. Root derivation or word derivation: A note on Dutch conversion pairs and pre- fixed verbs. Unpublished paper. Kramer, Ruth. 2006. Root and Pattern Morphology in Coptic: Evidence for the Root. In Davis, Amy Rose Deal & Youri Zabbal (eds.), The Proceedings of the 36th Annual Meeting of the North East Lin- guistic Society (NELS 36), 399-412. GLSA 2006. Kramer, Ruth. 2009. Definite markers, Phi-features, and Agreement: A Morphosyntactic Investigation of the Amharic DP. Santa Cruz: University of California dissertation. Laks, Lior. 2009. Valence Changing Operations: Where does Morpho-phonology Interfere?. Proceedings of the 16th Conference of the Student Organization of Linguistics in Europe, ConSOLE XVI. Paris: Université Paris Diderot.

References 189

Landauer, Thomas & Susan Dumais. 1997. A solution to Plato’s problem: the Latent Semantic Analysis the- ory of acqusition, induction and representation of knowledge. Psychological Review 104(2). 211-240. Landauer, Thomas, Peter W. Foltz & Darrell Laham. 1998. Introduction to Latent Semantic Analysis. Dis- course Processes 25. 259-284. Lederman, Shlomo. 1982. Problems in a prosodic analysis of Hebrew morphology. Studies in the Linguistic Sciences 12. 141-163. Lehrer, Adrienne. 1990. Polysemy, conventionality, and the structure of the lexicon. Cognitive Linguistics 1.207-246. Leslau, Wolf. 1995. Reference grammar of Amharic. Wiesbaden: Harrossowitz. Levin, Beth. 1993. English Verb Classes and Alternations. Illinois: The University of Chicago Press. Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity at the Syntax–Lexical Semantics Interface. Cambridge, MA: MIT Press. Levin, Beth & Malka Rappaport Hovav. 2005. Argument realization. Cambridge: Cambridge University Press. Lieber, Rochelle. 2004. Morphology and lexical semantics. Cambridge: Cambridge University Press. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge: MIT Press. Marantz, Alec. 1997. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. U. Penn Working Papers in Linguistics 4. 201-225. Marantz, Alec. 2000. Words. Unpublished ms., Talk presented at WCCFL 20. Marantz, Alec. 2007. Phases and words. In S.-H. Choe (ed.), Phases in the Theory of Grammar, 191-222. Seoul: Dong In. Marshall, David R. 1968. A Comparative Table of the Meaning Patterns of the Derived Forms of the Verb in Arabic and Maltese. Journal of Maltese Studies 5. 14-24. Marslen-Wilson, William, Lorainne K. Tyler, Rachelle Waksler & Lianne Older. 1994. Morphology and meaning in the English mental lexicon. Psychological Review 101. 3-33. Matsuzaki, Toru. 2001. Verb meanings and their effects in syntactic behaviors: A study with special reference to English and Japanese ergative pairs. Florida: University of Florida dissertation. Mayer, Thomas, Christian Rohrdantz, Frans Plank, Peter Bak, Miriam Butt, & Daniel A. Keim. 2010. Conso- nant co-occurrence in stems across languages: Automatic analysis and visualization of a phonotactic constraint. In Proceedings of the ACL 2010 Workshop on NLP and Linguistics: Finding the Common Ground (NLPLING 2010), 67-75. McCarthy, John. 1981. A prosodic theory of nonconcatenative morphology. Linguistic Inquiry 12. 373-418. McCarthy, John. 1986. OCP effects: Gemination and antigemination. Linguistic Inquiry 17. 207-263. McCarthy, John. 1993. Template form in prosodic morphology. Papers from the Third Annual Formal Lin- guisticts Society of Midamerica Conference, 187-218. McMillion, Alan. 2006. Labile Verbs in English. Their Meaning, Behaviour and Structure. Stockholm: Stock- holm University dissertation. McKoon, Gail & Talke Macfarland. 2000. Externally and internally caused change of state verbs. Language 76. 833-858. Montemagni, Simonetta. 1994. Non Alternating Argument Structures: the Causative/Inchoative Alternation in Dictionaries. In Willy Martin, Willem Meijs, Margreet Moerland, Elsemiek ten Pas, Piet van Sterken- burg & Piek Vossen (eds.), Proceedings of the Sixth Euralex International Congress on Lexicography, Amsterdam, 30-8/3-9. 349-359. Montemagni, Simonetta & Vito Pirelli. 1995. Do Lexical Rules Apply Across the Board? A Corpus-Based In- vestigation in the Machinery of the Causative-Inchoative Alternation in Italian. Proceedings of the Ac- quilex Workshop on Lexical Rules. University of Cambridge Computer Laboratory. Cambridge. Montemagni, Simonetta, Vito Pirelli & Nilda Ruimy. 1995. Ringing things which nobody can ring. A corpus- based study of the causative-inchoative alternation in Italian. Textus, English Studies in Italy 8. 371- 390. Mifsud, Manwel & Albert Borg. 1994. Arabic in Malta. Indian Journal of Applied Linguistics 20. 89-102. Mifsud, Manwel. 1994. The Weak-final Conjugation in the Semitic Component of Maltese. In Joseph M. Brin- cat (ed.), Languages of the Mediterranean, 244- 265. Malta: Institute of Linguistics. Mifsud, Manwel. 1995a. Loan Verbs in Maltese. Leiden: Brill. Mifsud, Manwel. 1995b. Productivity of Arabic in Maltese. Proceedings of the 2nd International Conference of AIDA, 151-160. Cambridge: Cambridge University Press. Mifsud, Manwel. 1996. The Loanverb in Maltese: A Romance-Arabic Crossbreed. In Jens Lüdtke (ed.), Ro- mania Arabica. Festschrift für Reinhold Kontzi zum 70 Geburtstag, 117-128. Tübingen: Narr. Mifsud, Manwel. 2008. Maltese. In Kees Versteegh (ed.), Encyclopedia of Arabic Language and Linguistics, vol. III, 146-159. Leiden: Brill. Mori, Laura. 2009. The shaping of Maltese throughout the centuries: Linguistic evidence from a diachronic- typological analysis. In Bernard Comrie, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz, & Martine Vanhove (eds.), Introducing Maltese Linguistics: Selected Papers from the 1st International Con- ference on Maltese Linguistics. Bremen, 18-20 October 2007, 291-307. Amsterdam: John Benjamins.

190 A Tale of Two Morphologies

Moscoso del Prado, Fermín, Avital Deutsch, Ram Frost, Vija H. de Jong, Robert Schreuder & R. Harald Baayen. 2005. Changing places: a cross-language perspective on frequency and family size in Hebrew and Dutch. Journal of Memory and Language 53. 496-512. Myers, Scott. 1984. Zero Derivation and Inflection, MIT Working Papers in Linguistics 7, MIT Workshop in Morphology, Cambridge, MA, 53-69. Nichols, Johanna. 1993. Transitive and causative in the Slavic lexicon: evidence from Russian. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitivity, 69-86. Amsterdam: John Benjamins. Nichols, Johanna, David A. Peterson & Jonathan Barnes. 2004. Transitivizing and detransitivizing languages. Linguistic Typology 8. 149-211. Ornan, Uzzi. 1990. Machinery for Hebrew word formation. In M. Golumbic (ed.), Advances in Artificial Intel- ligence, 75–93. Heidelberg and New York: Springer Verlag. Owens, Jonathan. 1997. The Arabic grammatical tradition. In R. Hetzron (ed.), The Semitic languages, 46-48. New York: Routledge. Padgett, Jean, 1991. Stricture in feature geometry. Amherst: University of Massachusetts dissertation. Panagiotidis, Phoevos. 2011. Root ‘content’ and lexical semantics. Paper presented at Approaches to the Lexicon (Roots III), Jerusalem, 13-16 June. Pinker, Steven. 1989. Learnability and Cognition: The Acquisition of Argument Structure. Cambridge, MA: MIT Press. Piñón, Christopher. 2001a. Modelling the causative-inchoative alternation. Linguistische Arbeitsberichte 76. 273-293. Piñón, Christopher. 2001b. A finer look at the causative-inchoative alternation. In Rachel Hastings, Brendan Jackson & Zsofia Zvolenszky (eds.), Proceedings of Semantics and Linguistic Theory 11, Ithaca, NY: CLC Publications, Cornell University. Plank, Frans. 1983. Transparent versus functional encoding of grammatical relations: A parameter for syn- tactic change and typology. Linguistische Berichte 86. 1-13. Plank, Frans. 2010. Variable direction in zero-derivation and the unity of polysemous lexical items. Word Structure 3(1). 82-97. Plank, Frans, 2011. Direction of derivation is determined by semantic complexity, isn’t it. Handout, Seminar on The Direction of derivation, Universität Konstanz. Poesio, Massimo. 1996. Semantic ambiguity and perceived ambiguity. In K. van Deemter & S. Peters (eds.), Semantic ambiguity and undespecification, 159-201. Stanford: CSLI. Pozdniakov, Konstantin & Guillaume Segerer. 2007. Similar Place Avoidance: A statistical universal. Lin- guistic Typology 11(2). 307-348. Prunet, Jean-François. 2006. External evidence and the . Morphology 16. 41-67. Prunet, Jean-François, Renée Béland & Ali Idrissi. 2000. The Mental Representation of Semitic Words. Lin- guistic Inquiry 31. 609-648. Puech, Gilbert. 1978. A Cross-dialectal Study of in Maltese. In Donka Farkas, Wesley M. Jacobsen & Karol W. Todrys (eds.), Proceedings of CLS 14, Chicago Linguistic Society, University of Chi- cago. Pustejovsky, James. 1995. The Generative Lexicon. Cambridge, MA: MIT Press. Pylkkänen, Liina. 2008. Introducing Arguments. Cambridge, MA: MIT Press. Ramchand, Gillian. 2008. Verb meaning and the lexicon: a first-phase syntax. Cambridge: Cambridge Univer- sity Press. Rappaport Hovav & Beth Levin. 1998. Morphology and Lexical Semantics. In Arnold Zwicky & Andrew Spencer (eds.), Handbook of Morphology, 248-271. Oxford: Blackwell. Ravid, Dorit. 1990. Internal structure constraints on new-word-formation devices in Modern Hebrew. Folia Linguistica 24. 289-347. Ravid, Dorit. 2003. A developmental perspective on root perception in Hebrew and Palestinian Arabic. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Mor- phology, 293-319. Amsterdam: John Benjamins. Rayner, Keith & Susan A. Duffy. 1986. Lexical complexity and fixation times in reading: effects of word fre- quency, verb complexity, and lexical ambiguity. Memory & Cognition 14. 191-201. Reinhart, Tanya. 2000. The Theta System: Syntactic realization of verbal concepts, OTS Working papers. Reinhart, Tanya. 2002. The Theta System – An Overview. Theoretical Linguistics 28. 229-290. Rice, Sally A. 1992. Polysemy and lexical representation: the case of three English prepositions. Proceedings of the fourteenth annual conference of the cognitive science society, 89-94. Hillsdale: Erlbaum. Rose, Sharon. 2003. The formation of Ethiopian Semitic internal reduplication. In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 79-98. Amst- erdam: John Benjamins. Schäfer, Florian. 2007. By itself, ms, Universität Stuttgart. Schäfer, Florian. 2009. The Causative Alternation. Language and Linguistics Compass 3. 641-681. Schwarze, Christoph. 2001. Aspetti semantici della formazione delle parole. Arbeitspapier Nr. 107, Fach- bereich Sprachwissenschaft, Unversität Konstanz.

References 191

Sereno, Sara C., Patrick J. O’Donnell & Keith Rayner. 2006. Eye movements and lexical ambiguity resolution: investigating the subordinate-bias effect. Journal of Experimental Psychology: Learning, Memory, and Cognition 27. 556-573. Serracino-Inglott, Erin. 1966. The Triliterality of Quadriliterals in Semitic Maltese. Journal of Maltese Studies 3. 47-67. Serracino-Inglott, Erin. 1975-1989. Il-Miklem Malti. Malta: Klabb Kotba Maltin. Shimron, Joseph (ed). 2003a. Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology. Amsterdam: John Benjamins. Shimron, Joseph. 2003b. Semitic languages: Are they really root-based? In Joseph Shimron (ed.), Languages Processing and Acquisition in Languages of Semitic, Root-based, Morphology, 1-28. Amsterdam: John Benjamins. Siegel, Dorothy. 1974. Topics in English Morphology. Cambridge, MA: MIT dissertation. Siloni, Tal. 2001. Reciprocal verbs. In Yehuda W. Faik (ed.), Proceedings of Association for Theoretical Linguistics 17, http://atar.mscc.huji.ac.il/-english/IATL/17/Siloni.pdf. Smith, Carlotta. 1970. Jespersen’s ‘Move and Change’ Class and Causative Verbs in English. In Mohammad A. Jazayery, Edgar C. Polomé & Werner Winter (eds.), Linguistic and Literary Studies in Honor of Archibald A. Hill. Vol. 2: Descriptive Linguistics, 101-109. The Hague: Mouton de Gruyter. Spagnol, Michael. 2007. L-Aspett Lessikali fil-Verb Malti. Malta: University of Malta MA thesis. Spagnol, Michael. 2009. Lexical and grammatical aspect in Maltese. In Thomas Stolz (ed.), Ilsienna, 51-86. Bochum: Universitätsverlag Brockmeyer. Spagnol, Michael. 2011a. L’integrazione delle costruzioni italiane con il si in maltese. In Walter Breu (ed.), L’influsso dell’italiano sul sistema del verbo delle lingue minoritarie. Resistenza e mutamento nella mor- fologia e nella sintassi, Resistenza e mutamento nella morfologia e nella sintassi. Atti del 2° Convegno In- ternazionale Costanza, 10-13 dicembre 2008 (Diversitas Linguarum 29), 211-226. Bochum: Brock- meyer. Spagnol, Michael 2011b. The causative-inchoative alternation in Maltese. In Sandro Caruana, Ray Fabri & Thomas Stolz (eds.), Variation and Change: The Dynamics of Maltese in Space, Time and Society, 155- 179. Berlin: Akademie Verlag. Spagnol, Michael & Thomas Mayer. 2011. The Matter of the Root. Poster presentation, GĦILM Third interna- tional conference on Maltese linguistics, Valletta, April 8-10. Spencer, Andrew. 1991. Morphological Theory. Oxford: Blackwell. Steriade, Donca. 1995. Underspecification and markedness. In G. John (ed.), The Handbook of phonological theory, 114-174. Cambridge: Blackwell. Stolz, Thomas. 2003. Not quite the right mixture: Chamorro and Malti as candidates for the status of mixed language. In Yaron Matras & Peter Bakker (eds.), The Mixed Languages Debate. Theoretical and Empiri- cal Advances, 271-315. Berlin: Mouton de Gruyter. Sutcliffe, Edmund F. 1936. A grammar of the Maltese language with chrestomathy and vocabulary. London: Oxford University Press. Svenonius, Peter. 2011. RootsI, Modularity, and Idiosyncracy. Paper presented at Approaches to the Lexicon (Roots III), Jerusalem, 13-16 June. Thepkanjana, Kingkarn. 2003. A cognitive account to the causative/inchoative alternation in Thai. In Gene Casad, & Gary B. Palmer (eds.), Cognitive and Non-Indo-European Languages, 247-274. Berlin: Mouton de Gruyter. Tosco, Mauro. 1993. Morfologia italiana in maltese. In Riccardo Contini, Fabrizio A. Pennacchietti & Mauro Tosco (eds.), Serta Semitica et Philologica Constantino G. Tsereteli Dicata, 319-332. Torino: Zamorani. Trimble, Louis P. 1971. Phonemic Change and the Growth of Homophones in Maltese. Journal of Maltese Studies 7. 92-98. Tsunoda, Tasaku. 1985. Remarks on Transitivity. Journal of Linguistics 21. 385-396. Tucker, Matthew A. 2009a. The high-low attachment analysis and the Arabic verb. Ms., University of Cali- fornia, Santa Cruz. Tucker, Matthew A. 2009b. The Root-and-Prosody Approach to . Ms., University of California, Santa Cruz. Tucker, Matthew A. 2010. The Morphosyntax of the Arabic verb: toward a unified syntax-prosody. Ms., Uni- versity of California, Santa Cruz. Ussishkin, Adam, 1999. The inadequacy of the consonantal root: Modern Hebrew denominal verbs and output-output correspondence. Phonology 16. 401-442. Ussishkin, Adam. 2000. Root-and-pattern morphology without roots and patterns. In M. Hirotani, A. Coet- zee, N. Hall & J.-Y. Kim (eds), Proceedings of NELS, 30: 655-670. Ussishkin, Adam. 2005. A fixed prosodic theory of nonconcatenative templatic morphology. Natural Lan- guage & Linguistic Theory 23. 169-218. Vassalli, Mikiel Anton. 1827. Grammatica della lingua maltese. Malta. Vella, Francis. 1831. Maltese grammar for the use of the English. Leghorn: Glaucus Masi. Verheij, Arian J.C. 2000. Bits, Bytes, and Binyanim: A Quantitative Study of Verbal Lexeme Formations in the . Orientalia Lovaniensia Analecta 93. Leuven: Peeters, Leuven.

192 A Tale of Two Morphologies

Versteegh, Kees. 1997. Landmarks in linguistic thought III: the Arabic linguistic tradition. London, New York: Routledge. Versteegh, Kess. 2004. From Classical Arabic to the modern Arabic . In: G. Booij, C. Lehmann, J. Mugdan, S. Skopeteas & W. Kesselheim (eds.), An International Handbook on Inflection and Word- Formation, 1740–1754. Berlin, New York: Walter de Gruyter. Visser, Fredericus. 1970. A Historical Syntax of the . Part I: Syntactical Units with one Verb. Leiden: E. J. Brill. von Heusinger, Klaus & Christoph Schwarze. 2002. Underspecification in the semantics of word formation – The case of denominal verbs of removal in Italian. Arbeitspapier Nr. 111, Fachbereich Sprachwissen- schaft, Unversität Konstanz. von Heusinger, Klaus & Christoph Schwarze. 2006. Underspecification in the Semantics of Word-Formation. The Case of Denominal Verbs of Removal in Italian. Linguistics 44. 1165-1194. Walter, Mary Ann. 2006. Effects in Maltese Arabic. In Sami Boudelaa (ed.), Perspectives on Arabic Linguistics. Proceedings of the 16th Arabic Linguistics Society Symposium, 161-178. Amst- erdam: John Benjamins. Wasow, Thomas. 1977. Transformations and the Lexicon. In Peter W. Culicover, Thomas Wasow & Adrian Akmajian (eds.), Formal Syntax, 327-360. New York: Academic Press. Wiese, Heike. 2004. Semantics as a gateway to language. In H. Härtl & H. Tappe (eds.), Mediating between concepts and language, 197-222. Berlin: Mouton de Gruyter. Wilson, Deirdre & Robyn Carston. 2007. A unitary approach to lexical pragmatics: relevance, inference, and ad hoc concepts. In N. Burton- (ed.), Pragmatics, 230-259. New York: Palgrave Macmillan. Wittgenstein, Ludwig. 1953. Philosophical investigations. Oxford: Basil Blackwell. Wright, Saundra K. 2001. Internally caused and externally caused change of state verbs. Evanston: North- western University dissertation. Wright, Saundra K. 2002. Transitivity and change of state verbs. In Julie Larsen & Mary Paster (eds.), Pro- ceedings of the 28th Annual Meeting of the Berkeley Linguistics Society, 339-350. Berkeley, CA: Berkeley Linguistics Society. Yamaguchi, Toshiko. 1998. Lexical Semantic Analysis of Causative/Inchoative Alternation in Japanese: A preliminary investigation of subclasses of verbs. In Arnold Doug (ed.), Essex Graduate Students Papers in Language and Linguistics Vol. II. Department of Language and Linguistics, University of Essex http://www.essex.ac.uk/linguistics/publications/egspll/volume_2/pdf/Yamaguchi.pdf

Appendix I Tri-consonantal roots and patterns

strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X b-ċ-ċ¹ baċċaċ tbaċċaċ b-ċ-ċ² biċċeċ tbiċċeċ baċċan, tbaċċan, b-ċ-n beċċen tbeċċen b-ċ-r biċċer tbiċċer b-d-d bidded tbidded biddel, b-d-l bidel tbiddel nbidel beddel biġġel, b-ġ-l tbiġġel nbiġel beġġel tbiegħad, b-għ-d¹ bagħad *bagħgħad bagħad nbagħad tbiegħed b-għ-d² biegħed tbiegħed b-għ-l bagħgħal tbagħgħal tbagħal nbagħat, b-għ-t bagħat ntbagħat b-h-m biehem tbhejjem tbiehem b-ħ-ħ tbaħħaħ b-ħ-n *beħħen b-ħ-r¹ baħħar tbaħħar b-ħ-r² baħħar tbaħħar b-ħ-t baħat bieħet *tbaħħat b-k-m bikem bikkem tbikkem b-k-r bakar bakkar *tbakkar b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X b-k-t *beket bekket tbekket bala', balla', tballa', nbela', b-l-għ bela' bella' tbella' nbala' bellah, tbellah, b-l-h blieh belleh, tbelleh bl -ħ-n bl eħħen tbl eħħen ntbell, b-l-l bell bellel tbell btell b-l-q boloq bellaq *tbellaq blieq b-l-s¹ belles tbelles b-l-s² ballas b-l-t¹ ballat tballat b-l-t² bellet tbellet b-n-n¹ bennen tbennen b-n-n² bennen tbennen b-n-s bannas bannan, b-n-t *tbannat bannat b-q-għ baqa' baqqa' b-q-n baqqan tbaqqan b-q-q¹ baqqaq tbaqqaq b-q-q² baqqa *tbaqqaq b-q-r baqqar b-q-t baqat baqqat tbaqqat b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X b-r-d¹ barad barrad nbarad b-r-d² bired berred tberred b-r-ġ barraġ tbarraġ nbaraġ b-r-ħ beraħ berraħ tberraħ nberaħ b-r-k¹ birek berrek tberrek b-r-k² bierek tbierek br -k-n tbr ekken b-r-m baram barram tbarram nbaram btaram b-r-n berren tberren b-r-q berraq tberraq b-r-r barr b-r-x¹ barax barrax *tbarrax nbarax b-r-x² berrex b-s-l bassal b-s-m *bissem tbissem nbasar, b-s-r basar bassar tbassar ntbasar b-s-s bass bassas tbassas ntbass b-t-ħ bettaħ tbettaħ b-t-l batal battal tbattal b-t-m battam tbattam b-t-n batan battan *nbatan b-t-r batar nbatar b-x-q bexxaq tbexxaq b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X b-x-r baxar baxxar tbaxxar nbexx, b-x-x bexx bexxex tbexxex ntbexx baża', bażża', tbażża', b-ż-għ beża' beżża' tbeżża' b-ż-l¹ biżel biżżel tbiżżel b-ż-l² beżżel tbeżżel nbeżaq, b-ż-q beżaq beżżaq tbeżżaq ntbeżaq b-ż-r bażżar *tbażżar b-z-n bezzen tbezzen

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X b-d-j beda nbeda b-d-w/j bieda b-j-d¹ bajjad tbajjad bjad bad, nbad, b-j-d² bied nbied b-j-għ biegħ nbiegħ b-j-n bejjen tbejjen b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X tbejjet, b-j-t biet bejjet ibbejjet nbeka, b-k-j beka bekka tbekka ntbeka b-n-j bena nbena b-r-j¹ barra tbarra b-r-j² bera b-t-j¹ bata tbata b-t-j² batta tbatta b-w-b bewweb bewweġ, b-w-ġ bawwax, bewwex b-w-għ *bawa' tbawwa' b-w-ħ bewwaħ tbewwagħ b-w-j buwa b-w-l biel bewwel *tbewwel b-w-q bewwaq tbewwaq b-w-r bawwar tbawwar nbies, b-w-s bies bewwes tbewwes ntbies b-w-t bewwet b-x-j baxxa tbaxxa strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ċċaċċar, ċ-ċ-r ċaċċar ċċaċċra ċiegħek, ċ-għ-k ċċiegħek ċagħaq ċ-ħ-d ċaħad ċaħħad ċċaħħad nċaħad ċ-k-n ċekken ċċekken ċkien ċ-k-r ċakkar ċ-l-m ċallam ċ-l-q ċellaq ċċellaq ċ-l-s ċallas ċċallas ċ-p-n ċeppen ċ-p-p *ċappap ċ-p-r ċpar ċ-p-s ċappas ċċappas ċ-q-m ċaqqam ċċaqqam ċ-r-k ċerrek ċ-r-t ċarrat ċċarrat ċ-t-r ċattar weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ċ-ċ-j ċiċċa ċeda, ċ-d-j ċieda ċ-f-j ċefa ċ-j-ċ ċieċ ċ-j-k ċiek ċ-j-m ċejjem ċ-j-q ċejjaq ċ-m-j ċama ċ-r-j ċara ċ-w-ċ ċewweċ ċċewweċ ċ-w-l ċawwal ċ-w-r *ċewwer strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X d-b-b ddebbeb d-b-ħ debaħ *debbaħ d-b-l debbel ddebbel dbiel d-b-n debben ddebben d-b-r¹ dabbar ddabbar d-b-r² deber debber ddebber d-f-n¹ difen *deffen ndifen d-f-n² deffen ddeffen d-f-r dafar *daffar ndafar d-f-s deffes ddeffes d-għ-m dagħam diegħem dgħam d-h-b *dehheb dieheb ddieheb d-h-n¹ dehen *dehhen ndehen d-h-n² diehen ddiehen d-h-r deher dieher ndiehes, d-h-s dehes dehhes ndehes, ndess d-h-x dehex diehex ddiehex ndiehex daħak, daħħak, ddieħak, ndaħak, d-ħ-k / d-ħ-q daħaq daħħaq ddieħek, ndaħaq ddieħaq d-ħ-l daħal daħħal ddaħħal ndaħal d-ħ-n *daħan daħħan ddaħħan b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X d-ħ-s *deħes d-k-k dekkek ddekkek d-k-r dakkar ddakkar d-l-k dilek dellek ddellek ndilek d-l-l dell dellel ddellel ndell d-l-m dalam dallam ddallam *dlam damma', d-m-għ ddemma' demma' demmel, d-m-l ddemmel demmen d-m-m¹ demmem ddemmem dammam, d-m-m² damm ddammem ndamm dammem d-m-s demmes ddemmes d-n-b¹ denneb ddenneb *ddieneb d-n-b² dineb danna, d-n-n dann ddanna d-n-s dennes ddennes d-q-q daqq daqqaq ddaqqaq ndaqq d-q-s daqqas ddaqqas d-r-b darab darrab ddarrab ndarab d-r-k direk dierek d-r-r darr darrar ndarr b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X d-r-s¹ darras ddarras d-r-s² dires *derres ndires d-t-r ddetter d-x-x dexxex ddexxex weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X d-għ-w dagħa ddiegħa d-h-j deha dgħief, d-j-f¹ dgħajjef ddgħajjef dgħaf d-j-f² diefa dief d-j-n dejjen ddejjen d-j-q dejjaq ddejjaq djieq d-j-r dar dajjar d-k-j daka d-n-j denna ddenna d-r-j¹ dara darra ddarra ndara d-r-j² derra dderra d-r-j³ dira dab, d-w-b dewweb ddewweb dieb b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X d-w-d dewwed ddewwed d-w-ħ *dewwaħ d-w-j¹ dewa ndewa d-w-j² dewwa ddewwa d-w-l dawwal ddawwal dwal dam, dawwem, d-w-m diem dewwem, ddewwem dejjem daq, d-w-q dewwaq ddewwaq ndaq dieq dawwar, ddawwar, d-w-r dar ndar dewwer ddewwer d-w-r dewwer ddewwer d-x-j *dixxa ddixxa strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X f-d-d fidded tfidded f-d-l¹ fadal faddal tfaddal f-d-l² feddel tfeddel f-ġ-ġ¹ feġġ f-ġ-ġ² faġġaġ f-ġ-l fiġel faġar, nfaġar, f-ġ-r *faġġar faraġ nfaraġ f-għ-l *fagħal nfehem, ftehem, f-h-m fehem *fehhemfiehem ftiehem nftehem, ftiehem nftiehem f-ħ-l *faħal faħħal tfaħħal f-ħ-m faħħam tfaħħam f-ħ-r faħar faħħar tfaħħar ftaħar f-ħ-x faħħax tfaħħax f-k-k *fekk fekkek tfekkek f-k-r *fakar fakkar tfakkar ftakar f-l-ġ fileġ nfileġ f-l-ħ felaħ nfelaħ f-l-k fellek tfellek f-l-l fellel tfellel f-l-q fellaq f-l-s¹ felles tfelles b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X f-l-s² files felles f-l-s³ felles f-l-z fallaz flaz f-n-d fannad tfannad fnad f-q-d faqad *faqqad ftaqad f-q-għ faqa' faqqa' tfaqqa' nfaqa' f-q-m faqqam tfaqqam ftaqam f-q-r *faqar faqqar tfaqqar ftaqar fqar f-q-s faqas faqqas tfaqqas nfaqas f-r-d fired farrad tfarrad nfired f-r-ġ *faraġ farraġ tfarraġ f-r-għ¹ forogħ ferragħ tferragħ friegħ f-r-għ² ferragħ tferragħ f-r-ħ¹ feraħ ferraħ *tferraħ nferaħ f-r-ħ² ferraħ tferraħ nferaħ f-r-k¹ *farak farrak tfarrak f-r-k² forok f-r-q feraq ferraq fieraq tferraq tfieraq nferaq fteraq f-r-x firex ferrex tferrex nfirex f-s-d¹ fasad *fassad nfasad f-s-d² fised fissed tfissed f-s-l fassal tfassal f-s-q¹ fesaq nfesaq f-s-q² fessaq tfessaq b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X f-s-r fisser tfisser f-t-ħ fetaħ fettaħ tfettaħ nfetaħ f-t-l fitel fettel tfettel nfitel f-t-m fatam fattam nfatam f-t-q fetaq fettaq *tfettaq nfetaq f-t-r fatar fattar tfattar fettet, tfettet, f-t-t fittet tfittet f-t-x fittex tfittex ntfettex f-x-l fixel nfixel f-x-x nfexx f-ż-r fażżar tfażżar weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X f-ċ-j tfaċċa afda, f-d-j¹ nfada fada f-d-j² feda nfeda f-g-j faga nfaga f-ħ-j feħa b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X fajjad, f-j-d fied fjad fawwad f-j-q fieq fejjaq tfejjaq f-j-r fajjar tfajjar f-l-j¹ fela nfela f-l-j² falla f-n-j fena nfena f-r-j¹ fera f-r-j² fiera tfera f-s-w/j fesa nfesa f-t-j fotta tfotta f-w-ġ fewweġ tfewweġ fwieħ, f-w-ħ fieħ fewwaħ tfewwaħ fjieħ f-w-q fewwaq tfewwaq f-w-r far fawwar tfawwar f-w-t fiet *fejjet strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ġ-b-d ġibed ġebbed ġġebbed nġibed ġ-b-l ġebbel ġġebbel ġ-b-r ġabar ġabbar ġġabbar nġabar ġ-b-s ġebbes ġġebbes ġ-d-d ġedded ġġedded ġ-d-m ġiddem ġġiddem ġġaddar, ġ-d-r ġġidder ġ-għ-d ġagħad ġiegħed ġġiegħed nġagħad ġ-għ-l ġagħal ġiegħel ġġiegħel nġagħal ġ-h-r ġahar ġ-h-ż ġieheż ġġieheż ġileb, ġ-l-b ġlejjeb ġ-l-d¹ ġelled ġieled ġġieled ġ-l-d² ġelled ġġelled ġ-m-d ġimed ġemmed ġġemmed ġama', ġamma', ġġamma', nġama', ġ-m-għ ġema' ġemma' ġġemma' nġema' ġ-m-l ġemmel ġġemmel ġ-m-r ġammar ġġammar ġ-n-b ġenneb ġieneb ġġenneb ġġieneb ġ-n-n ġennen ġġennen ġ-n-t ġannat ġġannat b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ġ-r-b *ġarab ġarrab ġġarrab ġ-r-d ġarad *ġarrad nġarad ġ-r-f *ġaraf ġarraf ġġarraf ġ-r-għ *ġeragħ ġerragħ ġġerragħ ġ-r-ħ ġeraħ ġerraħ ġġerraħ nġeraħ iġtar, ġ-r-r ġarr *ġarrar nġarr xtarr ġ-s-m ġissem ġġissem ġ-s-s *ġess ġesses ġ-ż-ż ġeżż ġeżżeż nġeżż

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ġ-j-' / ġ-j-j ġie ġab, nġab, ġ-j-b ġieb nġieb ġejjef, ġġejjef, ġ-j-f / ġ-w-f ġewwef ġġewwef ġela, nġela, ġ-l-w ġila nġila ġ-r-j¹ ġara ġ-r-j² ġera ġerra ġġerra ġġiera nġera b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ġ-v-j ġova ġaħ, ġ-w-ħ ġewwaħ ġġewwaħ ġieħ ġ-w-ż ġewweż ġġewweż ġeża, ġ-ż-j nġeża iżża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X gebbeġ, g-b-ġ / g-b-x gebbex g-d-b gideb giddeb tgiddeb tgiedeb ngideb g-d-l geddel tgeddel g-d-m¹ gidem giddem tgiddem tgiedem ngidem g-d-m² tgeddem geddes, tgeddes, g-d-s / g-z-z gezzez tgezzez g-n-ċ gannaċ g-r-f giref gerref tgieref ngiref g-r-m gerrem tgerrem g-r-r garr ngarr g-r-s gerres tgerres g-r-x¹ gerrex tgerrex g-r-x² girex ngirex g-r-ż gireż gerreż g-s-s gess geżżeż ngeżż, g-ż-ż geżż geżgeż ngess g-z-z gezz weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X g-d-j goda gara, g-r-j tgara garr g-ż-j gaża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għ-b-d għabad għ-b-r¹ għabbar tgħabbar għ-b-r² għabbar tgħabbar għ-b-x għabbex tgħabbex għ-d-b¹ għadab għaddab tgħaddab għ-d-b² għaddeb tgħaddeb ngħadd, għ-d-d għadd għadded tgħadded ntgħadd għ-d-l għaddel tgħaddel għ-d-m għaddam tgħaddam ngħader, għ-d-r¹ għader ntgħader għ-d-r² għaddar tgħaddar stagħdar għ-d-s għodos għaddas tgħaddas għ-f-ġ għafeġ għaffeġ tgħaffeġ għ-f-n¹ għafen għ-f-n² għefen ngħafas, għ-f-s għafas għaffas tgħaffas ntgħafas ngħoġob, għ-ġ-b¹ għoġob ntgħoġob għ-ġ-b² għaġġeb tgħaġġeb stagħġeb għ-ġ-l għaġġel tgħaġġel għ-ġ-m għaġem b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ngħaġen, għ-ġ-n għaġen *għaġġen ntgħaġen għ-ġ-ż għaġeż tgħaġġeż għ-k-k għakkek tgħakkek għ-k-m għakem għ-k-r¹ għakkar tgħakkar għ-k-r² għokor għ-k-s għokos għakkes tgħakkes ngħakes għ-l-b¹ għolob għalleb tgħalleb ngħeleb, għ-l-b² għeleb għalleb ntgħeleb ngħalef, għ-l-f għalef *għallef ntgħalef għ-l-k għallek tgħallek għ-l-l¹ għallel tgħallel stgħall għ-l-l² *għall għallel tgħallel ngħall għallem, tgħallem, għ-l-m¹ ntgħallem għellem tgħellem għ-l-m² għelem ngħalaq għ-l-q¹ għalaq *għallaq ntgħalaq għ-l-q² għallaq tgħallaq għ-l-t għalat għallat tgħallat għ-m-b għamba b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għ-m-d¹ għammad tgħammad għ-m-d² għammed tgħammed ngħamel, għ-m-l għamel *għammel ntgħamel għ-m-m għammem tgħammem għ-m-q għammeq tgħammeq għ-m-r¹ għamar għammar tgħammar ngħamar għ-m-r² għammar tgħammar għ-m-s għames għammes tgħammes ngħames għ-m-x għammex tgħammex għ-m-ż għemeż għammeż tgħammeż tgħiemeż ngħameż għ-n-b għanneb għ-n-q għannaq tgħannaq għ-n-t għanet għ-q-b għaqqab għ-q-d għaqad għaqqad tgħaqqad ngħaqad għ-q-l¹ għaqel għaqqel tgħaqqel għ-q-l² għoqol ngħaqar, għ-q-r għaqar għaqqar tgħaqqar ntgħaqar għ-q-x għaqqex għ-r-b¹ għarrab għ-r-b² għarrab tgħarrab għ-r-b³ *għarreb tgħarreb b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għarraf, tgħarraf, tgħaraf, ngħaraf, għ-r-f għaraf għarref tgħarref tgħieref ntgħaraf għarreġ, għ-r-ġ għaraġ *tgħarreġ għarrex tgħierek, ngħorok, għ-r-k għorok ttgħierek ntgħorok għ-r-m għarram tgħarram għ-r-q¹ għereq għarraq tgħarraq għ-r-q² għereq għarraq għ-r-q³ għarraq għ-r-s¹ għarras tgħarras għ-r-s² għarras *tgħarras għ-r-x¹ għarax għarrax tgħarrax tgħarax ngħarax għ-r-x² għarrex tgħarrex għ-s-d għassed tgħassed għ-s-l għassel tgħassel ngħasar, għ-s-rgħasar *għassar ntgħasar ngħass, għ-s-s għass għasses tgħasses ntgħass għ-t-b¹ għatteb tgħatteb għ-t-b² għeteb għ-t-b³ għotob għattab tgħattab għ-t-l għattel tgħattel b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għ-t-n għattan tgħattan għ-t-q għattaq tgħattaq għ-t-r għotor għattar tgħattar għ-t-s għatas għattas għ-t-x għatax għ-x-q għaxxaq tgħaxxaq għ-x-r għaxxar tgħaxxar għ-x-x għaxxex tgħaxxex għ-ż-b għażżeb ngħażel, għ-ż-l¹ għażel ntgħażel għ-ż-l² għażel għażżel tgħażżel ngħażel għ-ż-l³ għażżel għażżen, għ-ż-n tgħażżen għeżżeż ngħażaq, għ-ż-q għażaq *għażżaq ntgħażaq ngħażż, għ-ż-ż¹ għażż għażżeż tgħażżeż ntgħażż għ-ż-ż² għażżaż tgħażżaż għ-ż-ż³ *għeżż għeżżeż weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għ-b-j għabba tgħabba għ-d-j għadda tgħadda għ-għ-j għagħa għajjeb, tgħajjeb, għ-j-b¹ għeb għawweb tgħawweb għ-j-b² għajjeb tgħajjeb għ-j-b³ għajjeb għad ngħad, għ-j-d¹ għawwed (qal) ntgħad għ-j-d² għajjed għ-j-j għeja għejja tgħajja għ-j-l għajjel għ-j-n¹ għajjen tgħajjen għ-j-n² għen stgħan għ-j-r¹ għajjar tgħajjar għ-j-r² għer għ-j-t¹ għat għajjet ngħat stgħat għ-j-t² għajjet tgħajjet għ-j-t³ għajjat tgħajjat ngħex, għ-j-x għex għajjex tgħajjex ntgħex għela, għ-l-j¹ għalla tgħalla għala għ-l-j² għola għolla tgħolla b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X għ-l-j³ għola għalla għ-m-j għama għamma għ-n-j¹ għana stgħana għ-n-j² għanna tgħanna għ-r-j għera għ-t-j għatta tgħatta ngħatta għ-x-j¹ tgħaxxa għaxxa, għ-x-j² tgħaxxa għaxxex għ-x-j³ għoxa għaxxa tgħaxxa għ-w-d għawed għawwed tgħawwed għ-w-ġ għawweġ tgħawweġ ntgħaweġ għ-w-j¹ għawa ngħewa, għ-w-j² għewa ntgħewa għ-w-l għawwel *tgħawwel għ-w-m għam għawwem tgħawwem għ-w-r¹ għajjar tgħajjar stgħar għ-w-r² għawar għawwar għ-w-r³ għawwar tgħawwar għ-w-s għawwas għ-ż-j għażża għ-ż-w għaża għażża tgħażża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X h-d-b heddeb theddeb h-d-d hedded thedded h-d-m heddem theddem h-l-l hellel thellel h-m-m hemmem themmem h-m-r nhamar h-m-ż hemeż hemmeż themmeż nhemeż h-r-ż hereż herreż therreż nhereż heżżeż, theżżeż, h-ż-ż heżheż theżheż

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X h-d-w heda hedda thedda h-n-j henna thenna h-r-j herra therra h-w-j sthewwa h-ż-j heżża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħ-b-b¹ ħabb ħabbeb tħabbeb nħabb ħtebb ħebb, ħ-b-b² hebb tħabbel, ħ-b-l¹ ħabbel tħambel ħebel, ħ-b-l² ħebbel *tħebbel hebel ħ-b-l³ ħobol ħabbel tħabbel ħabar, ħ-b-r¹ ħabbar tħabbar ħeber ħeber, ħ-b-r² ħabar ħ-b-s ħabes *ħabbes tħabbes nħabes ħ-b-t ħabat ħabbat tħabbat tħabat ħtabat ħ-b-ż¹ ħabeż ħabbeż tħabbeż nħabeż ħebeż, ħebbeż, ħ-b-ż² *tħabbeż hebeż ħabbeż ħaddet, ħ-d-t ħadded ħeddel, tħeddel, ħ-d-l ħedel ħdiel ħaddel tħaddel ħ-d-m ħadem ħaddem tħaddem nħadem ħ-d-n ħaddan tħaddan ħ-d-r¹ ħaddar tħaddar ħdar b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħ-d-r² ħadar ħaddar tħaddar tħaddet, ħ-d-t ħadded tħadded ħaff, tħaffef, ħ-f-f ħaffef ħfief ħeff tħeffef ħ-f-n ħafen nħafen ħ-f-r¹ ħafer nħafer ħ-f-r² ħafer ħaffer tħaffer ħ-ġ-b ħaġeb *ħaġġeb *tħaġġeb nħaġeb ħ-ġ-ġ¹ *ħaġġ *ħġaġ ħ-ġ-ġ² ħaġġeġ ħeġġeġ, tħeġġeġ, ħ-ġ-ġ³ ħaġġeġ tħaġġeġ heġem, ħeġem, ħ-ġ-m nheġem ħaġem, haġem ħ-ġ-r ħaġġar tħaġġar ħakkek, tħakkek, ħ-k-k ħakk nħakk ħtekk ħakwek tħakwek ħ-k-m ħakem ħakkem nħakem ħ-l-b ħaleb ħalleb tħalleb nħaleb ħ-l-f¹ ħalef ħallef nħalef b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħellef, ħ-l-f² tħellef ħallef ħ-l-ġ ħaleġ *ħalleġ nħaleġ ħtileġ ħ-l-l¹ ħall nħall ħ-l-l² ħallel tħallel ħ-l-l³ ħallel *tħallel ħolom ħallem tħallem nħolom nħalaq, ħ-l-q¹ħalaq ħallaq nħoloq ħ-l-q² ħallaq tħallaq ħ-l-q³ *ħallaq tħalaq ħ-l-s¹ ħallas tħallas ħ-l-s² ħeles ħelles tħelles nħeles ħ-l-t ħalat ħallat tħallat tħalat ħtalat ħemmed, ħ-m-d ħemed tħemmed ħammed ħ-m-ġ ħammeġ tħammeġ ħammel, tħammel, ħ-m-l ħamel tħamel nħamel ħemmel tħemmel ħ-m-m¹ ħemm ħemħem ħ-m-m² ħammem ħ-m-r¹ ħammar tħammar nħamar ħmar ħ-m-r² ħemer ħemmer *tħemmer b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħammes, ħ-m-s¹ tħammes ħemmes ħ-m-s² ħemmes ħ-m-ż *ħameż ħ-n-k ħannek tħannek ħenn, ħ-n-n¹ ħannen tħannen ħann ħ-n-n² ħinn ħanaq, ħ-n-q ħannaq tħannaq nħanaq ħtanaq ħonoq ħ-n-s *ħanes tħannes ħ-n-x ħannex tħannex ħ-q-q ħaqq ħaqqaq tħaqqaq stħaqq ħ-q-r ħaqar *ħaqqar nħaqar ħarab, ħ-r-b¹ ħarrab tħarrab nħarab harab ħ-r-b² ħarreb ħareb tħarreb tħareb ħ-r-f¹ ħarref tħarref ħorof, ħ-r-f² ħeref ħ-r-f³ ħorof stħarreġ, ħ-r-ġ ħareġ ħarreġ tħarreġ nħareġ staħreġ ħ-r-m ħorom *ħarrem b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħarraq, ħ-r-q¹ ħaraq tħarraq nħaraq ħteraq ħerraq ħ-r-q² ħarraq *tħarraq ħ-r-q³ ħarraq ħ-r-r ħarrar ħarar tħarrar ħ-r-s ħares tħares ħarat, ħ-r-t¹ ħarrat nħarat horot ħorot, ħ-r-t² ħarat ħ-r-x ħarrax tħarrax ħrax nħaseb, ħ-s-b ħaseb ħasseb tħasseb tħaseb ntħaseb ħ-s-d ħasad *ħassad nħasad ħ-s-l¹ ħasel ħassel nħasel ħesel, ħ-s-l²ħasel, ħassel tħassel ħasal ħ-s-r¹ ħassar tħassar ħ-s-r² *ħassar tħassar ħiser, ħ-s-r³ ħeser ħ-s-r⁴ ħeser ħ-s-s ħass ħasses tħasses nħass b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħ-t-b¹ ħattab tħattab ħ-t-b² ħatteb tħatteb ħotob, ħ-t-b³ nħotob ħatab ħataf, ħettef, nħataf, ħ-t-f *tħettef tħataf ħotof hettef nħotof ħattam, ħ-t-m ħatam tħattam ħattem ħ-t-n ħaten ħatten tħatten tħaten nħaten ħ-t-r¹ ħatar ħattar nħatar ħtar ħ-t-r² ħattar tħattar ħ-t-r³ ħater tħater ħ-t-t¹ ħatt ħattet nħatt ħ-t-t² ħattet tħattet ħ-x-b ħaxxeb tħaxxeb ħaxxem, tħaxxem, ħ-x-m¹ ħaxkem, tħaxxeb ħaxxeb ħ-x-m² *ħaxem ħaxxem ħ-x-n ħaxxen tħaxxen ħxien ħ-x-x ħaxxex tħaxxex ħ-ż-m ħażżem tħażżem nħażem ħ-ż-n¹ ħażen nħażen ħ-ż-n² ħażen *ħażżen tħażżen nħażen b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħ-ż-n³ ħażżen tħażżen ħżien ħ-ż-q ħażaq ħażżaq tħażżaq nħażaq ħ-ż-r *ħażar tħażżar tħażar nħażż, ħ-ż-ż ħażż ħażżeż tħażżeż ntħażż

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħeba, ħ-b-j¹ *ħebba nħeba staħba ħaba ħ-b-j² ħaba ħefa, ħ-f-j ħafa ħ-ħ-j ħaħa ħ-j-b¹ ħajjeb ħ-j-b² ħejjeb tħejjeb ħieb, ħ-j-b³ hieb ħ-j-ġ / ħ-w-ġ nħtieġ ħtieġ ħ-j-j¹ ħeja nħeja staħja ħejja, ħ-j-j² tħejja hejja b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħiel, stħajjel, ħ-j-l ħajjel tħajjel ħjiel staħjel ħajjem, tħajjem, ħ-j-m hejjem thejjem ħan, ħ-j-n ħajjen tħajjen nħan ħjien ħien ħar, ħ-j-r¹ ħier ħ-j-r² ħajjar tħajjar ħtar ħiet, nħiet ħ-j-t¹ħat, ħajjat nħât ħet ħ-j-t² ħajjet tħajjet ħ-l-j¹ ħela nħela ħ-l-j² ħalla tħalla ntħalla ħela, ħliel ħ-l-w ħalla tħalla nħela ħala nħema, ħ-m-j¹ ħema nħima ħ-m-w ħama tħama ħ-r-j¹ ħara ħarra nħara ħ-r-j² ħarra nħesa, ħ-s-j ħesa nħasa b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ħ-t-j¹ ħata ħ-t-j² tħata ħ-w-d ħawwad tħawwad ħ-w-f ħaf ħawwef tħawwef ħ-w-ħ¹ *ħawwaħ ħ-w-ħ² ħawwaħ ħ-w-j ħewa ħawwaħ ħ-w-l¹ ħawwel tħawwel ntħawwel ħ-w-l² ħwiel ħ-w-r ħawwar tħawwar ħ-x-j ħexa *ħaxxa nħexa weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X j-b-s webbes twebbes ibbies j-q-r stejqer j-s-r jassar tjassar strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X k-b-b kebb kebbeb tkebbeb k-b-r kiber kabbar tkabbar k-b-s kibes kebbes tkebbes k-b-x kebbex tkebbex k-b-z kabbaz tkabbaz k-ċ-ċ kaċċaċ k-d-d kedd kedded tkedded nkedd k-f-f keff keffef tkeffef nkeff k-f-n keffen tkeffen k-f-r kafar *kaffar tkeffer k-f-s kifes nkifes k-għ-b kagħab tkagħak, k-għ-k tkagħkagħ k-ħ-l kaħħal tkaħħal kħal k-l-b kileb nkileb k-l-l *kelel kellel tkellel k-l-m kellem tkellem k-m-d kemmed tkemmed k-m-m *kemmem k-m-x kemmex tkemmex k-n-k kannak kennen, k-n-n kenn tkennen stkenn gennen b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X k-n-s¹ kines *kennes nkines k-n-s² *kennes k-n-z kannaz karab, k-r-b karrab tkarrab nkarab korob kerrah, tkerrah, stkerrah, k-r-h krieh kerreh tkerreh stkerreh korom, k-r-m *karrem *tkarrem nkaram karam k-r-r *kerr stkerr k-r-t karrat k-s-b kiseb kisseb tkisseb nkiseb k-s-ħ kesaħ kessaħ tkessaħ k-s-r kiser kisser tkisser nkiser k-s-s kess k-t-b kiteb *kitteb tkieteb nkiteb k-t-f kittef tkittef k-t-n kitten tkitten kotor, k-t-r kattar tkattar katar k-x-f kixef kixxef tkixxef nkixef k-x-n *kexxen k-x-x kaxxax k-z-z kazzaz tkazzaz weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X k-ċ-j keċċa tkeċċa k-j-b kieb tkejjeb k-j-d kejjed k-j-f¹ kief kejjef tkejjef k-j-f² *kejjef tkejjef k-j-l¹ kejjel tkejjel

k-j-l² kiel wikkel *twikkel ttiekel

k-j-n kejjen tkejjen k-j-t kejjet tkejjet k-k-j kakka tkakka k-l-j kala tkala k-p-j kapa k-r-j¹ kera nkera k-r-j² korra k-s-j kesa nkesa k-v-j kava k-w-d kad k-w-j kewa tkiewa nkewa k-w-n kien kewwen *tkewwen k-w-s kewwes tkewwes k-ż-j¹ tkaża k-ż-j² kuża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X l-b-r labbar tlabbar l-b-s libes libbes tlibbes ntlibes libbet, l-b-r libet tlebbet ntlibet lebbet l-ċ-ċ laċċaċ l-f-f leff leffef tleffef ntleff lteff l-f-q lefaq leffaq leġġeġ, l-ġ-ġ laċċaċ l-ġ-m *liġem liġġem tliġġem tliegħeb, l-għ-b¹ lagħab ntlagħab tliegħab l-għ-b² liegħeb lagħan, l-għ-n legħen liegħaq, tliegħaq l-għ-q lagħaq *lagħgħaq ntlagħaq liegħeq tliegħeq l-h-b leheb lieheġ, l-h-ġ leheġ tlieheġ lieħeġ l-h-m lehem liehem laħħ, l-ħ-ħ¹ leħħ l-ħ-ħ² leħħ leħħeħ b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X l-ħ-m laħħam tlaħħam l-ħ-n leħħen tleħħen l-ħ-q laħaq laħħaq tlaħħaq ntlaħaq ltaħaq l-k-m likkem tlikkem l-l-x lellex tlellex l-m-ħ lemaħ lemmaħ tlemmaħ ntlemaħ ltemaħ l-m-q lemaq l-q-għ laqa' laqqa' tlaqqa' ntlaqa' ltaqa' l-q-m¹ laqqam tlaqqam l-q-m² laqqam tlaqqam leqq, l-q-q laqq l-q-t laqat laqqat tlaqqat ntlaqat ltaqat l-q-x laqqax tlaqqax l-s-n lisen lissen tlissen l-t-m litem l-x-n lexxen tlexxen l-ż-m liżem liżżem weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X l-j-l lejjel l-j-m lajjem tlajjem lema, l-m-j lama l-s- lissa l-w-ħ lewwaħ tlewwaħ l-w-j lewa ntlewa ltewa l-w-m liem lewwem tlewwem tliewen l-w-n lewwen tlewwen l-w-q lewwaq tlewwaq l-w-t lewwet tlewwet l-w-ż lewweż tlewweż laza, l-z-j alza strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X m-ċ-ċ meċċeċ tmedd, tmedd, m-d-dmedd *medded ntmedd mtedd *tmedded *tmedded m-għ-d magħad *magħgħad miegħed tmiegħed ntmagħad mtagħad *megħek, m-għ-k miegħek tmiegħek magħak m-għ-r megħer miegħer tmiegħer m-għ-x miegħex tmiegħex m-għ-ż miegħeż tmiegħeż m-h-l miehel tmiehel m-ħ-ħ maħħaħ m-ħ-r meħħer m-ħ-t maħat maħħat tmaħħat ntmaħat m-k-ħ mekkaħ tmekkaħ m-k-k mekkek tmekkek m-l-ħ¹ mellaħ tmellaħ mlieħ m-l-ħ² moloħ stmell, m-l-l mell tmell stmerr m-l-q mellaq m-l-s miles melles tmelles mlies m-l-t mallat tmallat m-n-għ mana' ntmana' m-n-s mannas tmannas b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X m-q-t maqqat tmaqqat mqat m-r-d¹ marad marrad tmarrad m-r-d² tmarrad m-r-ħ¹ meraħ tmieraħ m-r-ħ² merraħ merrek, m-r-k tmerrek merkek m-r-q merraq tmerraq m-r-r marrar tmarrar mrar m-r-x mirex m-s-ħ mesaħ messaħ tmessaħ ntmesaħ messes, tmess, m-s-s mess mesmes tmesses, ntmess mtess tmesmes matal, m-t-l matal mattal tmâtal mtal motal m-t-r¹ mattar tmattar tmatar m-t-r² mattar m-x-t maxat maxxat tmaxxat ntmaxat mtaxat m-ż-ġ mażaġ m-ż-ż miżżeż tmiżżeż m-z-z mazzaz tmazzaz weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X maħħa tmaħħa miel mejjel tmejjel mela *mella mtela miera tmiera mewweġ tmewweġ mar mawwar tmawwar mewwes tmewwes miet mewwet tmewwet ntmexx, mexa mexxa tmexxa ntmexxa strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X nebbaħ, tnebbaħ, ntebaħ, ntebaħ, stenbaħ, n-b-ħ¹ nebbeh tnebbeh ntnebeh ntebeh stenbeh n-b-ħ² nebaħ nebbaħ n-b-r *nabar nabbar tnabbar nebbet, n-b-t nibet tnebbet nibbet n-b-x nibex nibbex tnibbex ntnibex n-d-f naddaf tnaddaf ndaf n-d-m nidem niddem tniddem n-d-r¹ naddar tnaddar n-d-r² nadar ntadar n-d-r³ nider nafad, naffad, tnaffad, n-f-d¹ nafat naffat tnaffat n-f-d² nifed niffed tniffed ntnifed nefaħ, n-f-ħ neffaħ tneffaħ ntnefaħ ntefaħ nafaħ n-f-q nefaq neffaq tneffaq ntnefaq ntefaq n-f-r nafar naffar tnaffar n-f-s niffes tniffes n-f-x nifex niffex tniffex ntifex n-ġ-m niġġem tniġġem n-ġ-r naġar naġġar *tnaġġar ntnaġar ntaġar n-ġ-s niġġes tniġġes b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X n-ġ-ż niggeż tniggeż n-għ-ġ niegħeġ tniegħeġ n-għ-l nagħal *nagħgħal niegħel tniegħel n-għ-m niegħem n-għ-s nagħas niegħes tniegħes *nahad, n-h-d niehed tniehed *nehed n-h-r nahar n-ħ-l naħħal n-ħ-q naħaq *naħħaq n-ħ-r¹ naħar *naħħar n-ħ-r² naħar naħħar ntaħar n-ħ-s tnieħes n-k-b nikeb *nikkeb tnikkeb n-k-ħ nekaħ n-k-r nikker tnikker n-k-t¹ nikket tnikket n-k-t² nikket tnikket n-l-ħ nellaħ tnellaħ n-m-l nemmel tnemmel n-m-r nammar n-m-s nemmes tnemmes n-m-x nemmex tnemmex n-q-b naqab naqqab tnaqqab ntnaqab b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X n-q-għ naqa' n-q-l naqal n-q-r naqar naqqar tnaqqar n-q-s naqas naqqas tnaqqas nqâs n-q-x naqax naqqax tnaqqax ntnaqax ntaqax n-s-b nasab nassab tnassab ntnasab ntasab n-s-ġ niseġ nisseġ tnisseġ ntiseġ n-s-l nissel tnissel n-s-r nasar nassar tnassar n-s-s nassas tnassas n-t-f nitef nittef tnittef ntnitef nataħ, n-t-ħ nadaħ, netaħ n-t-n niten nitten tnitten n-t-q nataq n-x-f nixef nixxef tnixxef n-x-r naxar *naxxar ntnaxar ntaxar naża', nażża', tnażża', n-ż-għ ntnaża' ntaża' neża neżża' tneżża' n-ż-l niżel niżżel tniżżel weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X n-b-j nabba n-d-j¹ nidda tnidda n-d-j² *nedda nieda tnieda nafa, n-f-j naffa tnaffa nefa nagħa, n-għ-j negħa, tniegħa niegħa neħħa, n-ħ-j tneħħa naħħa n-j-b nejjeb n-j-ħ nieħ nejjek, n-j-k niek tnejjek nejjes nan, n-j-n *newwen nien n-j-q *naq tnejjeq n-j-r najjar tnajjar n-j-s nejjes tnikka, n-k-j neka tnekka n-n-j ninna n-q-j naqa naqqa tnaqqa b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X nessa, tnessa, ntnesa, n-s-j nesa ntesa nissa tnissa tnesa n-w-b newweb tnewweb naħ, n-w-ħ newwaħ nieħ n-w-l¹ newwel n-w-l² newwel tnewwel newwem, nawwam, n-w-m nam tnewwem nejjem, namnam n-w-n nawwan n-w-r nawwar tnawwar n-w-x newwex n-x-j nixxa tnixxa naża, n-ż-j / n-ż-għ naża' strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X p-l-m pallam p-n-t¹ pennet p-n-t² pannat perreċ, p-r-ċ tperreċ perreġ p-t-r pitter tpitter pexxex, p-x-x pexpex

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X p-ċ-j paċa tpaċa p-ġ-j poġġa tpoġġa p-j-p pejjep tpejjep pejjez, p-j-z pezpez p-p-j pappa p-s-j passa p-t-j patta tpatta p-x-j¹ paxxa tpaxxa p-x-j² pixxa strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X q-b-d qabad qabbad tqabbad tqabad nqabad q-b-l qabel qabbel tqabbel tqabel q-b-r *qabbar qb-s qabbas q-b-x *qebbex tqebbex q-b-ż qabeż qabbeż tqabbeż tqabeż nqabeż q-ċ-t qaċċat tqaċċat qaddeb, q-d-b tqaddeb qaddab q-d-d qadded tqadded q-d-f qadef qaddef nqadef q-d-ħ qadaħ q-d-m¹ qaddem tqaddem qdiem qaddam, q-d-m² tqaddam qaddem q-d-m³ qaddam q-d-r qadar q-d-s qaddes tqaddes q-f-l qafel qaffel nqafel q-f-s qaffas q-għ-d qagħad *qagħgħad qiegħed tqiegħed q-ħ-b qoħob qaħħab tqaħħab q-ħ-t qaħħat tqaħħat q-l-b qaleb qalleb tqalleb nqaleb b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X q-l-għ qala' qalla' tqalla' nqala' qall, qallel, tqellel qliel q-l-l¹ qell qellel q-l-l² qallel qileq, q-l-q qeleq, qoloq q-l-t qallat *tqallat q-m-l qammel tqammel q-m-r qammar tqammar qamas, qammas, tqammas, tqamas, q-m-s qomos, qammes tqammes tqames qames q-m-t qammat tqammat q-n-għ qanna' tqanna' qorob, qarreb, tqarreb, q-r-b tqareb qrab qarab qarrab tqarrab qarad, nqarad, q-r-d¹ qarrad tqarrad qorod nqorod qered, q-r-d² qerred nqered qired q-r-f qarraf tqarraf q-r-għ qara' qarra' tqarra' qriegħ b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X qaraħ, q-r-ħ qarraħ tqarraħ nqaraħ qeraħ qarram, tqarram, q-r-m qarrem tqarrem q-r-n qarran tqarran q-r-q¹ qarraq tqarraq qireq, q-r-q² qereq q-r-r¹ qarr qarqar q-r-r² qarr qarrer qarar tqarar stqarr q-r-r³ qerr q-r-s¹ qaras nqaras q-r-s² qarras tqarras qras q-s-b¹ qassab tqassab q-s-b² qassab q-s-m qasam qassam tqassam nqasam q-s-r qassar tqassar qsar qassas, tqassas, q-s-s¹ qass qasqas tqasqas, ntqass tqasses q-s-s² qasses tqasses q-s-t qassat tqassat *qata' q-t-għqata' qatta' tqatta'tqata' nqata' *qietgħa b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X q-t-l qatel qattel tqattel tqatel nqatel q-t-n qattan qatar, q-t-r qattar tqattar qotor qattet, q-t-t tqattet qattat q-x-r qaxxar tqaxxar q-ż-ż¹ qażż q-ż-ż² qażżeż tqażżeż weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X q-b-j qeba q-d-j qeda nqeda q-j-d qajjed tqajjed qal, q-j-l qajjel qiel q-j-m qiem qejjem tqejjem q-j-r qajjar tqajjar b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ntqies, qies, qejjes, tqejjes, tqies, q-j-s qas qajjes tqajjes nqies, nqas q-j-t qiet q-l-j qela qalla tqalla nqela q-q-j qaqa q-r-j qara qarra qara tqarra nqara q-w-j / q-w-w qawwa tqawwa nqal q-w-l qal qawwel tqawel ntqal qajjem, tqajjem, q-w-m qam qawwem tqawwem q-w-r qawwar tqawwar qawwas, tqawwas, q-w-s qawwes tqawwes strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X r-b-b rabbab r-b-għ rabba' r-b-ħ rebaħ rebbaħ trebbaħ ntrebaħ rtebaħ r-b-k rebbek r-b-t rabat rabbat trabbat ntrabat rtabat r-ċ-p reċċep r-d-d radd raddad traddad ntradd rtadd rada', radda', tradda', ntrada', r-d-għ reda' redda' tredda' ntreda' r-d-m radam ntradam rtadam redden, r-d-n tradden radden r-d-q redaq r-d-s reddes r-f-d rifed riffed triffed ntrifed rtifed refa', ntrefa', rtefa', r-f-għ rafa' ntrafa' rtafa' r-f-s rifes riffes triffes ntrifes rtifes reġa', reġġa', treġġa', r-ġ-għ raġa' raġġa' traġġa' r-ġ-l¹ riġġel triġġel traġġel, r-ġ-l² itraġġel r-ġ-n rigen ntrigen b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X r-għ-b regħeb riegħeb triegħeb rtiegħed, r-għ-d *ragħgħad riegħed triegħed rtogħod ragħem, r-għ-m regħem, *riegħem ragħan regħex, r-għ-x riegħex triegħex ragħax r-h-b rahab rahhab trabbah trabah r-h-n rahan rahhan râhan rtahan r-ħ-l raħħal traħħal r-ħ-m¹ raħħam traħħam r-ħ-m² raħħam traħħam r-ħ-s¹ roħos raħħas traħħas rħas r-ħ-s² raħħas traħħas r-k-b rikeb rikkeb trikkeb ntrikeb rtikeb rekken, r-k-n riken trekken rakkan r-k-s rakas rakkas rakat, ntrakat, r-k-t *rakkat rtakat rakad trakat rammel, trammel, r-m-l¹ rammal trammal r-m-l² romol rammal b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X r-q-d raqad raqqad traqqad trieqed rtaqad r-q-għ raqqa' traqqa' raqam, rtaqam, roqom, rtoqom, r-q-m riqem, rteqem reqem raqq, raqqaq, r-q-q traqqaq rqaq reqq reqqeq r-q-s raqas raqqas traqqas r-s-q resaq ressaq tressaq trass, r-s-s rass rassas ntrass rtass trassas r-s-t rassat r-t-b rattab trattab rtab r-t-ħ retaħ *rettaħ r-t-l rattal r-x-q rexaq *rexxaq r-x-x raxx raxxax traxxax ntraxx r-ż-ħ reżaħ reżżaħ treżżaħ trieżaħ r-ż-n rażan rażżan trażżan weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ra, ntara, r-'-jwera werra ntwera, nwera r-b-j rabba trabba r-ġ-j rieġa trieġa r-għ-j ragħa triegħa rtagħa ragħa, r-għ-w rigħa r-ħ-j reħa trieħa ntreħa rteħa r-j-d ried ntried r-j-ħ rieħ rejjaħ trejjaħ r-j-q raq rejjaq trejjaq r-j-x riex rejjex trejjex rama, r-m-j¹ ntrama rtama rama' r-m-j² rema ntrema rtema raqa, r-q-j riqa r-w-ħ¹ rieħ rewwaħ trewwaħ r-w-ħ² raħ rawwem, r-w-m ram trawwem rawwam r-w-n ran r-w-t rawwat trawwat strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X s-b-b *sabb *sabbab s-b-ħ¹ sebaħ sebbaħ ssebbaħ sbieħ s-b-ħ² sbieħ s-b-k sibek nsibek s-b-l sibel sebbel ssebbel s-b-m sebben ssebben s-b-q sebaq *sebbaq ssebbaq nsebaq s-b-r¹ sabar sabbar ssabbar stabar s-b-r² siber nsabat, s-b-t sabat sabbat ssabbat stabat nstabat sadd, nsadd, s-d-d¹ stadd sedd nstadd s-d-d² saddad ssaddad seddaq, s-d-q sseddaq saddaq s-d-r *sidder s-f-d seffed sseffed s-f-f¹ saffaf ssaffaf seff, sefsef, nseff, s-f-f² żiff, *seffef nsaff saff s-f-ħ sefaħ seffaħ ssiefaħ nsefaħ b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X safa', s-f-għ safa saffel, ssaffel, s-f-l *safal saffal ssaffal sfieq, s-f-q sefaq seffaq sseffaq sfaq s-f-r¹ saffar siefer s-f-r² saffar sfar s-f-r³ saffar ssaffar s-ġ-r saġġar ssaġġar s-għ-b sogħob sogħol, s-għ-l sagħal, siegħel segħel sagħan, s-għ-n ssiegħen sogħon s-għ-q sagħaq s-h-m siehem s-h-r sahar s-ħ-b¹ saħħab ssaħħab s-ħ-b² *saħħab sieħeb ssieħeb s-ħ-ħ¹ saħħ saħħaħ ssaħħaħ seħħ, s-ħ-ħ² saħħ b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X s-ħ-l seħel s-ħ-m / ħ-m-j² saħħam ssaħħam staħam s-ħ-n saħan saħħan ssaħħan ssaħan s-ħ-q saħaq saħħaq nsaħaq *seħer, s-ħ-r saħħar ssaħħar nseħer *siħer seħet, nseħet, s-ħ-t seħħet ssieħet steħet siħet nsteħet s-k-k sekkek sikem, nsikem, s-k-m *sikkem ssikkem żeqem nżeqem s-k-n sekken s-k-r¹ sakkar ssakkar s-k-r² siker sakkar ssakar s-k-t siket sikket ssikket nsalab, s-l-b salab sallab ssallab stalab nstalab nsilef, s-l-f silef sellef ssellef stilef nstilef nselaħ s-l-ħ selaħ *sellaħ stelaħ nstelaħ s-l-k silek sellek nsilek s-l-m silem sellem ssellem ssielem s-l-t¹ silet sellet ssellet ssielet nsilet stilet b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X s-l-t² salat s-m-d semed sama', samma', ssamma', nsama', s-m-għ sema' semma' ssemma' nstama', stama' nstema' s-m-m¹ sammam ssammam s-m-m² semmem ssemmem s-m-n¹ simen semmen ssemmen smien s-m-n² semmen s-m-r¹ smar s-m-r² sammar ssammar s-m-r³ sammar s-m-t samat nstamat stamat senneġ, ssenneġ, s-n-ġ sineġ żenneġ ssenċeċ sanagħ, s-n-għ senagħ, *sana' s-n-ħ snieħ sann, *sannan nsann, s-n-n senn *sannen nsenn s-n-r sannar ssannar s-p-p sappap ssappap s-q-f saqqaf ssaqqaf b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X sseqqer, s-q-r seqqer ssaqqar s-r-d sired serred s-r-f¹ sarraf ssarraf sorof, s-r-f² saraf s-r-ġ¹ sireġ serreġ s-r-ġ² sarraġ ssarraġ straħ, s-r-ħ serraħ sserraħ ssieraħ strieħ s-r-k sirek serrek nsaram s-r-m¹saram *sarram ssarram staram nstaram s-r-m² sarram s-r-p serrep sserrep nseraq s-r-q seraq steraq nsteraq s-r-r sarr sarrar ssarrar nsarr sata', s-t-għ seta' s-t-ħ settaħ ssettaħ s-t-r¹ satar *sattar nsatar s-t-r² *setter weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X siba, s-b-j seba safa, s-f-j saffa ssaffa sefa s-h-j seha saħa, s-ħ-j seħa sab, nsab, s-j-b¹ sieb nstab s-j-b² sejjeb ssejjeb s-j-d *sad *sajjad stad s-j-f¹ sejjef ssejjef *sief, ttiesef, s-j-f² sejjef *saf ttiefes saħ, s-j-ħ sejjaħ ssejjaħ nsieħ sieħ s-j-j sijja s-j-l siel s-j-r¹ sar sajjar ssajjar s-j-r² *sar *sies, s-j-s sejjes ssejjes *sejes s-j-t sajjat s-l-j¹ sala b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X s-l-j² salla sela, s-l-j³ sala s-l-j⁴ sella s-l-j⁵ sila s-m-j semma issemma saqa, ssaqqa, nsaqa, s-q-jsiqa, saqqa ssewwaq nseqa seqa s-r-j¹ sara ssara s-r-j² (ssargħa) sawwab, ssawwab, s-w-b sawweb ssawweb s-w-d sewwed ssewwed swied sawwaf, ssawwaf, s-w-f swaf sewwef ssewwef s-w-j sewa sewwa ssewwa s-w-l sal sawwal ssawwem, s-w-msam sawwam ssawwam s-w-n san sawwan s-w-q saq sewwaq ssewwaq nsaq s-w-r sawwar ssawwar b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X s-w-s¹ sewwes ssewwes s-w-s² sies s-w-t sawwat ssawwat strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X t-b-b tabbab teba', tabba', ttabba', ntabagħ, t-b-għ tebagħ tebba' ttebba' ntebagħ t-b-ħ tebaħ ntebaħ t-b-k tibek tebbak ntibek tabbal, t-b-l tabbar t-b-n tebben tebaq, t-b-q tebbaq ttebbaq ntebaq tbieq tabaq tafa', ntafa', t-f-għ ttafa' tefa' ntefa' t-f-l¹ taffal ttaffal teffel, tteffel, t-f-l² taffal ttaffal, ttiffel t-għ-b tiegħeb ttiegħeb t-għ-l tgħal t-għ-m tagħem tiegħem ttiegħem ntagħam tagħan, t-għ-n tgħan t-ħ-n taħan taħħan ntaħan taħar, t-ħ-r taħħar tahar b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X t-ħ-t taħħat ttaħħat tikkek, t-k-k ttekkek tektek t-k-l tiekel ttiekel t-k-n takkan t-l-b talab *tallab ttallab ntalab t-l-f tilef tellef ttellef ntilef tela', tella', ttella', ntela', t-l-għ tala' talla' ttalla' ntala' t-l-q telaq tellaq ttellaq ntelaq t-l-t tellet ttellet tamma, t-m-' tama ttama tamma' tama', tamma', t-m-għ ttamma' tema' temma' ttemm, t-m-m temm temmem ttemmem t-m-r tammar tammas, t-m-s ttammas tammes t-n-s¹ tines t-n-s² tanas t-p-n tappan ttappan t-p-p tappap b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z root I II III V VI VII VIII IX X t-q-b taqab taqqab ttaqqab ntaqab teqel, taqqal, teqal, teqqel ttaqqal tqal toqol t-r-b tarrab tarad, t-r-d *tarrad ntarad tarat t-r-f tarraf ttarraf t-r-ġ¹ taraġ t-r-ġ² tarraġ ttarraġ t-r-ħ teraħ terraħ tterraħ ttieraħ nteraħ t-r-q¹ teraq terraq t-r-q² terraq t-r-r tarr ntarr t-r-t terret t-r-x¹ tarax t-r-x² tarrax ttarrax ntarax tarraż t-r-ż ttarraż tarras ttiefes, t-s-f ttiesef weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X tefa, t-f-j¹ ntefa tafa t-f-j² taffa ttaffa t-f-j³ ttoffa t-j-b tieb tejjeb ttejjeb tjieb t-j-ġ tejjeġ tajjan, t-j-n ttajjan tajjen t-j-r tar tajjar ttajjar ttajar t-j-ż tejjeż t-k-j ttika itka, t-k-w *itteka tila, ntila, t-l-j¹ tela ntela t-l-j² tala t-m-j tamma ttama t-n-j tena tenna ttenna ntena tara, tarra, ttarra, t-r-j tira terra tterra t-t-j tetta t-w-b¹ tewweb ttewweb tawwab, t-w-b² ttawwab tawweb b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X t-w-j¹ tewa ntewa tewa, t-w-j² tiwa t-w-l tal tawwal ttawwal twal t-w-m¹ tewwem t-w-m² tewwem taq, t-w-q¹ tieq tewwak, t-w-q² ttewwaq tewwaq t-w-q³ tewwaq t-w-r tawwar twar t-w-s ttawwas t-w-t¹ tawwat t-w-t² tewwet weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X v-g-j vaga v-r-j¹ vara tvara v-r-j² vira vaża, v-ż-j avża weak-inital b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X webbel, twebbel, w-b-l wibbel twibbel w-b-n webben twebben w-ċ-ċ wiċċeċ twiċċeċ w-d-b¹ widdeb twiddeb waddab, w-d-b² twaddab ntwaddab waddaf widden, w-d-n twidden wedden w-ġ-b *wiġġeb wieġeb twieġeb ntwieġeb waġa', waġġa', twaġġa', w-ġ-għ weġa' weġġa' tweġġa' weġġeħ, w-ġ-ħ tweġġaħ weġġah w-għ-d *wegħed wiegħed twiegħed w-għ-r wiegħer twiegħer w-ħ-d waħħad twaħħad w-ħ-l weħel waħħal twaħħal w-ħ-m waħħam twaħħam w-ħ-r waħħar twaħħar w-ħ-x waħħax twaħħax w-l-d wiled welled wieled twelled twieled w-m-n emmen wemmen twemmen w-n-s wennes twennes b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X w-q-f waqaf waqqaf twaqqaf w-q-għ waqa' waqqa' twaqqa' w-q-t waqqat twaqqat w-r-b warrab twarrab w-r-ċ werreċ twerreċ w-r-d warrad twarrad w-r-k werrek twerrek w-r-q werraq twerraq w-r-t wiret werret *twerret ntiret w-r-x werrex twerrex wasa', wassa', twassa', usiegħ, w-s-għ wesa' wessa' twessa' wsiegħ w-s-l wasal wassal twassal w-s-q wessaq twessaq w-s-t wassat twassat w-t-q wettaq twettaq w-ż-b ważab w-ż-n wiżen *wiżżen wieżen twieżen ntiżen weak-final b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X wikka, w-k-j twekka wekka wella, w-l-j¹ twella willa w-l-j² *wela wissa, w-s-j twessa wessa w-t-j¹ witta wieta twitta wata, w-t-j² *weta wieta weża, w-ż-j ntuża uża strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X x-b-b *xebb xebbeb xxebbeb xaba', xabba', xxabba', x-b-għ xeba' xebba' xxebba' xebah, xebbah, xxebbah, x-b-h xiebah xxiebah xtedd xebeh xebbeh xxebbeh x-b-k xebbek xxebbek x-b-n xebben x-b-t xabbat xxabbat x-d-d¹ xedd xedded nxedd, x-d-d² xedd xedded xtedd nxtedd x-f-f xeffef xxeffef xaffar, x-f-r xeffer xiegħeb, x-għ-b xegħeb xxiegħeb xiegħab x-għ-f xogħof xiegħef nxegħel, x-għ-l¹ xegħel xiegħel xtegħel nxtegħel x-għ-l² xiegħel xxiegħel x-għ-r xiegħer xxiegħer xgħar xiegħet, xxiegħet, x-għ-t *xagħat xiegħat xxiegħat x-h-b xieheb xxieheb xhieb b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X nxehed, x-h-d xehed xiehed xtehed nxtehed x-h-r xeher xieher xxieher nxeher xteher xaħħ, x-ħ-ħ xaħħaħ xxaħħaħ xħaħ xeħħ x-ħ-m xaħħam xxaħħam x-ħ-t¹ xaħat xaħħat xxaħħat nxeħet, x-ħ-t² xeħet *xeħħet xteħed nxteħet x-k-k xekk xekkek xxekkek nxekk xxekkel, x-k-l xekkel xxikkel x-l-f xellef ixxellef x-l-g xelleg x-l-l xell xellel ixxellel x-m-għ xamma' ixxamma' x-m-k ximek xammem, xxammam, nxamm, x-m-m xamm xemmem xxammem nxtamm x-m-q xemaq x-m-r xammar xxammar x-m-x xemmex xxemmex xina', x-n-għ *xinna' xana' b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X x-n-q¹ xenaq x-n-q² xennaq xxennaq x-p-p xappap xxappap x-q-f xaqqaf x-q-q¹ xaqq xaqqaq xxaqqaq nxaqq x-q-q² xaqqaq x-q-r xxaqqar xqar nxorob, x-r-b xorobxarrab xxarrab xtorob nxtorob nxered, x-r-d xerred xxerred xtered nxtered x-r-f¹ xiref xerref xxerref x-r-f² xarraf xraf x-r-ħ xeraħ xerraħ xxerraħ nxeraħ x-r-k¹ xirek xerrek xierek xxerrek xxierek x-r-k² xarrak x-r-q¹ xeraq xerraq xxerraq x-r-q² xeraq x-r-r¹ nxtarr xtarr x-r-r² *xarr nxarr x-r-x xarrax x-t-b xattab xxattab x-t-ħ xxettaħ b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X xettel, xxettel, x-t-l xittel xxittel x-t-r *xatar xattar xxattar x-x-m *xixxem xxixxem weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X x-h-j xeha nxteha xteha x-j-b xieb xejjeb x-j-f xejjef xejjaħ, xjieħ, x-j-ħ xxejjaħ xejjeħ xjaħ x-j-n xejjen xxejjen x-j-r xejjer xxejjer x-j-t xejjet xxejjet xela, x-l-j¹ nxela xila x-l-j² *xala xxala xena xara, nxara, x-r-j xtara xera nxtara x-t-w xita xitta b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X x-w-f xaf nxewa, x-w-j xewa xtewa nxtewa xal, xawwel, x-w-l xxewwel xxiewel xiel xewwel xaq, xtaq, x-w-q xewwaq xxewwaq nxtaq xieq xtieq x-w-r xawwar x-w-t xawwat xxawwat x-w-x¹ xewwex xxewwex x-w-x² *xewwex x-x-j xoxxa strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ż-b-b¹ żabbab ż-b-b² żżebbeb ż-b-d żebbed ż-b-ġ¹ żżebbeġ ż-b-ġ² żebbeġ żżebbeġ żaba', nżaba', stebagħ ż-b-għ żeba', żebbagħ nżeba', żteba' żebagħ nżebagħ ż-b-l żebbel żżebbel ż-b-r żabar *żabbar nżabar ż-d-d żadd ż-d-m *żadam żaddam żżaddam żiden, ż-d-n żodon ż-f-n żifen żeffen żżeffen nżifen żeffet, ż-f-t żżeffet żiffet ż-ġ-ġ żeġġ żeġġeġ żżeġġeġ ż-g-g żeggeg ż-għ-b żżiegħeb żegħed, ż-għ-d żiegħed żżiegħed nżegħed żogħod żegħel, żiegħel, ż-għ-l żżiegħel żehel żiehel b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X żgħar, ż-għ-r żagħgħar sgħar ż-h-m żehem ż-h-r¹ żeher ż-h-r² żahar *żahhar ż-k-m żikkem ż-k-r żakkar żżakkar ż-l-ġ żileġ żelleġ żżelleġ ż-l-m żilem żellem żżellem ż-l-q żelaq żellaq żżellaq żżielaq ż-m-l żemmel żżemmel żżiemel żammam, ż-m-m żamm nżamm żammem ż-m-r żammar ż-n-d żenned żżanak, ż-n-k *żanak żżanika ż-n-n żennen ż-n-q żżennaq ż-q-q¹ żaqq żaqqaq żżaqqaq żqaq ż-q-q² żaqq żaqqaq nżaqq żarab, ż-r-b *żarrab żorob ż-r-d żarad żarrad żżarrad nżarad b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X żara', żarra', nżara', ż-r-għ żera' żerra' nżera' ż-r-k żirek *żerrek ż-r-q *żeraq *żerraq żrieq ż-t-t żattat żżattat weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ż-j-d żied nżied żdied ż-j-f żief żżief nżief ż-j-n¹ żejjen żżejjen ż-j-n² żien ż-j-t żejjet ż-j-ż żejjeż ż-n-j żena *żenna ż-w-f żaf ż-w-ġ żewweġ żżewweġ żawwal, ż-w-l żal żżawwal żewwel ż-w-q żewwaq żżewwaq ż-w-r żar żajjar nżar strong b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X z-k-k *zakkak zzakkak z-k-r zzakkar z-p-n zappan zappap, z-p-p zzappap zapzap

weak b ċ d f ġ g għ h ħ j k l m n p q r s t v w x ż z

root I II III V VI VII VIII IX X ziek, z-j-k nziek ċiek z-j-z zejjez z-k-j¹ zikka z-k-j² zzika z-w-q zewwaq irregular

root I II III V VI VII VIII IX X ('-d-j idda ('-d-n¹ idden ('-d-n² stieden ('-'-f af ntaf ('-ħ-d ħa ttieħed ('-ħ-j staħa ('-m-r amar amba, ('-m-b ambi stenna, ('-n-j ttenna ('-n-s stienes ('-ż-j iżża ('-ż-m ażżem

Appendix II Quadri-consonantal roots and patterns

quad b

root I II root I II root I II b-d-b-d badbad tbadbad b-r-b-r barbar bażbaż, b-ż-b-ż¹/b-s-b-s tbażbaż b-għ-b-s bagħbas tbagħbas b-r-b-x berbex tberbex basbas b-għ-t-r bagħtar tbagħtar b-r-f-l berfel tberfel b-ż-b-ż² beżbeż tbeżbeż bahrad, bergħed, b-ż-l-ħ beżlaħ b-h-r-d/b-h-r-ġ tbahrad bahraġ bergħet, b-ż-l-k beżlek tbeżlak b-r-għ-d tbergħed b-ħ-b-ħ¹ baħbaħ tbaħbaħ bargħad, b-ħ-b-ħ² baħbaħ bergħat b-ħ-n-n baħnan tbaħnan b-r-għ-n bergħen tbergħen b-j-t-r tbajtar b-r-k-l berkel tberkel bekbek, b-r-m-ċ bermeċ b-k-b-k tbekbek begbeg b-r-q-m barqam b-l-b-l¹ balbal b-r-w-n berwen b-l-b-l² belbel tbelbel b-r-w-ż barważ tbarważ b-l-j-j balja b-s-k-t baskat tbaskat b-l-k-j balka b-s-t-j basta b-m-b-j bimba b-t-b-t¹ betbet b-m-b-l bambal b-t-b-t² batbat b-n-d-j banda tbanda b-x-b-x bexbex tbexbex b-n-d-l bandal tbandal b-x-l-k bexlek b-n-ċ-l benċel tbenċel bixkel, b-x-k-l¹ tbixkel b-n-ġ-l benġel tbenġel bexkel b-q-b-q baqbaq tbaqbaq b-x-k-l² bexkel b-r-d-l berdel tberdel b-x-r-q *baxraq tbaxraq b-r-b-q berbaq tberbaq b-ż-w-r bażwar tbażwar ċ d

root I II root I II root I II ċ-f-ċ-f ċafċaf ċċafċaf ċarċar, dagħdagħ, ddagħdagħ, ċ-r-ċ-r¹ ċċarċar d-għ-d-għ / d-ħ-d-ħ ċ-f-ċ-q ċefċaq ċċefċaq xarxar daħdaħ ddaħdaħ ċ-f-l-q ċeflaq ċċeflaq ċ-r-ċ-r² ċerċer ċċerċer d-m-d-m damdam ddamdam ċ-f-l-s ċaflas ċċaflas ċ-r-k-k ċerkek d-n-d-l dendel ddendel ċ-f-t-r ċaftar ċċaftar ċerkes d-n-d-n dandan ddandan ċ-r-k-s ċ-j-p-r ċajpar ċċajpar (kerkes) d-q-d-q *daqdaq ddaqdaq ċ-k-ċ-k ċekċek ċċekċek ċ-w-l-ħ ċewlaħ ċċewlaħ d-r-d-r dardar ddardar ċ-m-b-s ċċambas d-r-w-x derwex dderwex ċ-m-ċ-m ċamċam ċ-m-ċ-q ċamċaq ċ-m-p-l ċempel ċ-m-p-r ċampar ċ-n-ċ-l ċenċel ċ-n-f-r ċanfar ċċanfar ċ-n-s-r ċenser ċ-n-t-j ċanta ċċanta ċ-n-t-r ċantar ċ-p-ċ-p ċapċap ċ-q-ċ-q ċaqċaq ċċaqċaq ċaqlem, ċċaqlem, ċ-q-l-m/ċ-k-l-m ċeklem ċċeklem ċ-q-l-ċ ċaqlaq ċċaqlaq ċ-r-ċ-q ċerċaq f ġ

root I II root I II root I II f-j-t-m fajtem f-r-t-s fartas tfartas ġagġag, ġ-g-ġ-g f-k-r-n fekren f-r-w-d tfarwad ġegġeg f-l-f-l¹ felfel tfelfel f-r-ż-n ferżen ġ-ħ-ġ-ħ ġaħġaħ f-l-f-l² felfel tfelfel f-s-d-q fesdaq tfesdaq ġ-l-b-n ġelben ġġelben f-n-d-j¹ fenda tfenda f-s-f-s fesfes tfesfes ġ-m-ġ-m ġemġem ġġemġem f-n-d-j² fonda f-s-q-j fisqa tfisqa ġ-n-b-l ġenbel f-n-d-j³ fonda f-t-f-t fetfet ġ-n-d-r ġandar f-n-d-q fandaq f-x-f-x fexfex ġ-n-ġ-j ġonġa ġġonġa f-n-f-r tfanfar f-x-k-l fixkel tfixkel ġenġen, ġ-n-ġ-l ġġelġel f-n-ġ-j finġa ġelġel f-n-q-l tfonqol ġ-r-b-n ġerben f-n-t-s fantas tfantas ġ-r-d-m ġardam f-r-ċ-ħ ferċaħ tferċaħ f-r-f-r¹ farfar tfarfar f-r-f-r² ferfer tferfer f-r-f-x ferfex tferfex f-r-għ-n fergħen tfergħen f-r-k-n ferken tferken f-r-k-x ferkex tferkex f-r-n-j forna tforna f-r-n-q fernaq f-r-n-ż ferneż tferneż f-r-t-l fertel f-r-t-n fartan *tfartan g għ

root I II root I II root I II g-d-w-d gedwed g-ż-g-ż¹ gażgaż għ-f-l-ġ għafleġ tgħafleġ g-ġ-w-ġ geġweġ g-ż-g-ż² geżegeż għ-f-l-s għaflas tgħaflas g-l-b-l gelbel g-ż-w-r geżwer tgeżwer għ-j-n-s għajnas tgħajnas g-l-g-l gelgel tgelgel għ-k-r-k għakrek tgħakrek g-m-g-m gemgem għ-k-r-x għakrex g-n-d-j ginda tginda għ-n-ċ-ċ għanċeċ g-r-b-b gerbeb tgerbeb għ-n-q-d għanqad tgħanqad g-r-b-ġ gerbeġ tgerbeġ għ-n-q-t għanqbat g-r-b-l gerbel għ-n-s-r għansar tgħansal g-r-f-x gerfex tgerfex għ-r-b-l għarbel (n)tgħarbel g-r-w-l gerwel għ-r-b-n għarben g-r-g-r¹ gargar għ-r-għ-r għargħar tgħargħar g-r-g-r² gerger tgerger għ-r-q-b għarqeb g-r-g-s gerges tgerges għ-r-w-l għarwel g-r-g-x gergex tgergex għ-r-w-n għarwen tgħarwen g-r-m-d germed tgermed għ-s-f-r għosfor tgħasfar g-r-n-s gernes għ-s-l-ġ għasleġ tgħasleġ g-r-w-l gerwel għ-x-w-x għaxwex g-r-w-x gerwex għ-ż-ż-r għażżar g-r-ż-m garżam g-s-g-s gesges gawda, g-w-d-j tgawda goda g-x-t-r tgexter h ħ

root I II root I II root I II h-m-d-m hendem thendem ħebħeb, ħawsel, ħ-b-ħ-b ħ-w-s-l tħawsel h-r-w-l herwel therwel hebheb ħawsal h-w-d-n hewden thewden ħ-b-r-k ħabrek ħ-w-t-l ħawtel ħ-l-ħ-l ħalħal tħalħal ħ-x-ħ-x ħaxħax ħ-m-b-q ħambaq tħambaq ħ-x-k-m ħaxkem ħ-m-ħ-m ħamħam tħamħam ħ-x-k-n ħaxken tħaxken ħ-n-d-b ħendeb ħ-x-l-f ħaxlef tħaxlef ħ-n-d-m¹ ħandem ħ-x-w-x ħaxwex tħaxwex ħ-n-d-m² ħendem tħendem ħ-ż-d-q ħażdaq ħ-n-d-q tħandaq ħ-n-f-s ħanfes tħanfes ħ-n-ħ-n ħanħan ħ-n-x-l ħanxel tħanxel ħ-n-x-r ħanxar tħanxar ħ-n-ż-r ħanżer tħanżer ħ-r-b-t ħarbat tħarbat ħ-r-b-x ħarbex tħarbex ħ-r-f-n ħarfen ħ-r-f-x ħarfex tħarfex ħ-r-ħ-r¹ ħarħar tħarħar ħarħar, ħ-r-ħ-r²/h-r-h-r harhar ħ-r-k-n ħarken ħ-r-t-m ħartam k l

root I II root I II root I II k-b-r-s kabras tkabras kasbar, l-b-l-b¹ lablab k-s-b-r/k-ż-b-r tkasbar keċner, każbar l-b-l-b² lebleb tlebleb k-ċ-n-r kaċnar k-s-k-s¹ keskes tkeskes l-f-l-f leflef tleflef k-ċ-w-n keċwen tkeċwen k-s-k-s² kaskas l-g-l-g legleg tlegleg k-f-k-f *kafkaf tkafkaf k-s-k-s³ keskes l-għ-l-għ lagħlagħ k-għ-b-r kagħbar tkagħbar ketket, l-ħ-l-ħ laħlaħ tlaħlaħ k-għ-w-ġ kagħweġ tkagħweġk-t-k-t katkat, lembet, l-m-b-t/l-n-b-t k-l-j-j kolja kitkit lenbet k-m-p-j kampa k-t-l-t ketlet lenbeb l-n-b-b/l-m-b-b tlenbeb k-m-pl -j kompl a tkompl a k-w-k-b kewkeb tkewkeb lembeb k-m-r-d kamrad k-w-l-t kawlat tkawlat l-n-ġ-s lanġas k-n-ġ-j kanġa k-w-t-l kewtel l-n-ż-t lanżat tlanżat k-n-t-j kanta tkanta kaxkar, l-q-l-q laqlaq k-x-k-r tkaxkar k-p-r-j kopra tkopra kaxkax l-s-t-j lesta tlesta k-r-ċ-ħ kerċaħ tkerċaħ k-x-k-x kexkex tkexkex l-t-l-t letlet k-r-f-s karfas tkarfas k-x-w-x kexwex k-r-k-r¹ karkar tkarkar k-r-k-r² kerker k-r-m-d karmad tkarmas, k-r-m-s karmas tkarmes k-r-n-j korna k-r-t-b kartab k-r-w-t karwat tkarwat m n

root I II root I II root I II m-ċ-l-q meċlaq marżeb, naħnaħ, m-r-ż-b m-ċ-m-ċ meċmeċ merżeb n-ħ-n-ħ nahnah, m-ġ-n-n meġnen m-r-ż-q merżaq tmerżaq nagħnagħ m-għ-d-r magħdar m-s-ħ-n masħan tmasħan nemnem, n-m-n-m/n-w-n-m m-għ-m-għ magħmagħ m-s-ħ-r masħar tmasħar newnem m-għ-r-k miegħrek m-s-k-n mesken tmesken n-q-n-q naqnaq m-ħ-m-ħ maħmaħ tmaħmaħ m-s-l-ħ meslaħ tmeslaħ n-r-v-ż nerveż tnerveż m-j-l-q mejlaq tmejlaq m-s-r-k mesrek n-ż-n-ż nażnaż m-j-n-j majna m-x-m-x mexmex tmexmex m-k-m-k mekmek m-x-t-r maxtar tmaxtar m-n-d-r¹ mander mażlaġ, m-ż-l-ġ m-n-d-r² tmandar meżleġ m-n-għ-l mengħel m-ż-m-ż meżmeż tmeżmeż m-n-tn -j mantn a tmantn a m-n-t-r mantar tmantar m-n-w-l manwal m-n-z-l menzel tmenzel m-q-d-r maqdar tmaqdar m-q-m-q maqmaq m-r-għ-n mergħen tmergħen m-r-k-k merkek m-r-m-r¹ marmar m-r-m-r² mermer tmermer m-r-t-l martel tmartel p q

root I II root I II root I II p-ċ-l-q peċlaq p-t-p-t¹ patpat q-b-r-s qabras tqabras p-ċ-p-ċ¹ paċpaċ p-t-p-t² petpet q-f-q-f qafqaf p-ċ-p-ċ² peċpeċ tpeċpeċ p-x-k-j pixka tpixka q-ħ-q-ħ qaħqaħ p-ħ-p-ħ paħpaħ tpaħpaħ p-x-p-x¹ tpaxpax q-l-f-t qalfat tqalfat p-k-p-k pakpak pexpex, qamqam, p-x-p-x² q-m-q-m p-l-p-l palpal pexxex qomqom p-n-ġ-j pinġa tpinġa pożda, q-n-ċ-ċ qanċeċ tqanċeċ p-ż-d-j/p-s-d-j penpen, posda q-n-d-l qandel tqandel p-n-p-n pennen p-z-p-z pezpez q-n-f-d qanfed tqanfed p-n-z-l penzel q-n-p-n qanpen p-n-z-r panzar qanqal, q-n-q-l/q-l-q-l tqanqal p-q-p-q paqpaq qelqel p-r-ċ-ħ perċaħ tperċaħ q-n-s-r qansar p-r-l-j parla q-n-t-r qantar p-r-p-n parpan q-n-ż-ħ qanżaħ tqanżaħ p-r-r-r¹ parpar q-r-b-n qarben tqarben p-r-p-r² perper tperper q-r-ċ-l qarċel tqarċel p-r-t-j parta q-r-d-n qarden tqarden p-r-t-t partat tpartat q-r-d-x qardax tqardax p-r-t-x partax tpartax q-r-m-ċ qarmeċ tqarmeċ p-s-p-r paspar tpaspar q-r-n-s qarnas tqarnas p-s-p-s pespes tpespes q-r-q-ċ qarqaċ tqarqaċ p-s-t-r pastar q-r-t-f qartaf tqartaf p-s-t-ż pastaż tpastaż q-r-t-m qartam tqartam /q r s

root I II root I II root I II q-r-t-s qartas tqartas r-ċ-p-l reċpel s-f-s-f sefsef ssefsef q-r-w-ż qarweż tqarweż rambal, trembel, s-f-t-r sefter qasdar, r-m-b-l rembel, trambel s-għ-t-r ssagħtar q-s-d-r/q-ż-d-r tqasdar qażdar rambel s-k-s-f seksef qawqeb, tqawqab, rampel, s-k-r-n ssakran q-w-q-b r-m-p-l trampel qawqab tqawqeb rampal s-k-s-k seksek sseksek q-x-l-f qaxlef tqaxlef r-n-d-j renda trenda s-l-s-l selsel q-x-q-x qaxqax tqaxqax r-n-ġ-j tranġa s-l-p-j salpa qażqaż, r-n-ġ-t ranġat tranġat s-l-t-n saltan q-ż-q-ż tqażqaż qażqeż r-s-t-t rastat s-l-v-j¹ salva ssalva q-ż-w-t qażwat raxka, s-l-v-j² solva ssolva r-x-k-j raska s-m-l-ħ ssemlaħ r-x-t-l rixtel trixtel s-m-p-l sempel s-m-s-m samsam s-m-s-r samsar ssamsar s-n-d-q sendaq s-n-d-r sandar s-n-ġ-t sangat s-n-s-l sensel ssensel s-n-s-r sensar ssapna, s-p-n-j nsapna saqsa, s-q-s-j staqsa /s t v

root I II root I II root I II s-r-b-t sarbat ssarbat t-f-t-f teftef tteftef v-l-ġ-j volġa s-r-d-n sardan ssardan t-għ-t-għ tagħtagħ v-l-j-j¹ vilja sserdaq, t-ħ-t-ħ taħtaħ v-l-j-j² vilja s-r-d-q/s-r-d-k serdaq sserdek t-k-t-k tektek ttektek v-n-ċ-j vinċa tvinċa s-r-ġ-j sorġa t-m-t-m temtem ttemtem v-n-v-n venven tvenven s-r-ġ-t sarġat t-n-b-r tanbar ttanbar v-r-j-j varja tvarja s-r-s-j ssarsa t-n-f-x tenfex ttenfex v-r-v-r verver sarsar, t-n-k-l tenkel s-r-s-r¹ ssarsa t-n-t-n tenten s-r-s-r² serser t-n-t-x tentex ttentex s-r-v-j serva sserva taptap, t-p-t-p¹/t-b-t-b sarwal, tabtab s-r-w-l/s-r-w-n ssarwal sarwan t-p-t-p² teptep t-q-t-q taqtaq t-r-b-ċ tarbaċ ttarbaċ t-r-n-j¹ torna t-r-n-j² *torna t-r-t-q tertaq ttertaq t-r-t-r¹ terter tterter t-r-t-r² tartar t-r-t-x tertex t-x-t-x textex w x ż

root I II root I II root I II w-d-w-d wedwed x-b-l-k xeblek xxeblek ż-b-l-ħ żeblaħ żżeblaħ weġweġ, x-ħ-x-ħ xaħxaħ xxaħxaħ ż-f-ż-f żafżaf żżafżaf weżweż, xajtan, ż-m-b-ħ żżembaħ w-ġ-w-ġ x-j-t-n xxajtan veġveġ, xajten ż-g-l-g żegleg żżegleg vexvex x-k-x-k xekxek ż-g-ż-g żagrag żżagrag w-ħ-w-ħ waħwaħ x-m-n-q xemnaq żegżeg, ż-g-ż-g w-q-w-q waqwaq x-n-d-r xandar xxandar żagżag w-r-d-n werden x-n-g-l xengel xxengel żżagħbel, w-r-w-r werwer twerwer xenxel, ż-għ-b-l/ż-għ-b-r żagħbel żżegħber, x-n-x-l xxenxel w-r-ż-q werżaq xelxel żżieber w-s-w-s weswes x-q-l-b xaqleb xxaqleb ż-għ-f-r żagħfar w-t-w-t¹ wetwet x-r-m-d xermed xxermed ż-j-b-r żajbar żżajbar watwat, x-w-l-ħ xewlaħ xxewlaħ ż-k-ż-k żekżek w-t-w-t² wetwet ż-l-ż-l żelzel w-ż-w-ż weżweż tweżweż ż-m-b-ħ *żembaħ żżembaħ ż-m-b-r¹ żambar żżambar ż-m-b-r² żambar żemżem, ż-m-ż-m żimżem, żamżam ż-n-b-l żenbel ż-n-g-l żengel żżengel ż-n-g-r żenger /ż z

root I II root I II żanqar, z-k-z-k zekzek ż-n-q-r/ż-n-ġ-r żanġar z-m-z-m zamzam żanżan, z-n-z-n zenzen ż-n-ż-n¹ żżanżan żenżen zeplet, z-p-l-t/z-p-z-p¹ żenżen, zepzep ż-n-ż-n²/ż-n-ż-l żenżel z-p-z-p² zapzap ż-p-ż-p żepżep z-r-z-r zerzer ż-r-b-n żarban z-r-ġ-ġ żarġaġ ż-r-ġ-n żarġan żżarġan ż-r-m-j żarma żżarma ż-r-n-q żernaq żerżaq, ż-r-ż-q/ż-l-ż-q żelżaq ż-r-ż-r¹ żarżar ż-r-ż-r² żerżer ż-q-ż-q żaqżaq