This book could not have been written without the support of many people. Numerous colleagues have provided critical comments at various stages of this project, some of them even reading subsequent versions of the same chapter. Their feedback was simply invaluable and has made me reformulate my ideas and arguments over and over again. Whoever commented on the text contributed very special insights and taught me that a text can be read and understood (and, yes, also misread and misunderstood) in many different ways. The following friends and colleagues have generously put in their time and energy: Birgit Alber, Harald Baayen, Maria Braun, Hartmut Gembries, Christiane Dalton-Puffer, Sabine Lappe, Martin Neef, and Jörg Meibauer. Needless to say, they are not to blame for the remaining flaws and inadequacies.

I also have to thank my student assistants Guido Bongard and Karina Lückoff, who read chapters, worked on the exercises and helped in various ways in the preparation of the manuscript. Special thanks go to Maria Braun for her meticulous work on the exercises, which made clear to me where things could go wrong and needed revision. Gisela Schwung made life much easier for me during the past two years by her efficient handling of all kinds of organizational matters.

Outline

This chapter provides an overview of the affixational word-formation processes of English. First, it discusses how affixes can be distinguished from other entities. This is followed by an introduction to the methodological problems of data gathering for the study of affixation through dictionaries and electronic corpora. Then some general properties that characterize the system of English affixation are introduced, and a survey of a wide range of suffixes and prefixes is presented. Finally, we investigate cases of infixation.

What is an affix?

In chapter 1 we defined ‘affix’ as a bound morpheme that attaches to bases. Although this seems like a clear definition, there are at least two major problems. First, it is not always easy to say whether something is a bound morpheme or a free morpheme, and second, it is not always obvious whether something should be regarded as a base or an affix. We will discuss each problem in turn.

Consider the data in (1) through (4), which show the putative affixes -free, -less, -like, and -wise in a number of derivatives, illustrated with quotations from the BNC:

1. (1) There was never an error-free text, Cropper said.

2. (2) Now the lanes were carless, lawless.

3. (3) Arriving on her broomstick at the prison-like school gates, Mildred peered through the railings into the misty playground.

4. (4) She had been a teacher, and made sure the girl went to a good school: “my granny had more influence on me education-wise.”

In this book we ask how ‘word’ should be defined. What are the criteria for ‘word’? Is ‘word’, as the term is generally understood, an appropriate unit to recognise for every type of language?

This introductory chapter first looks at what scholars have said about ‘word’, and then discusses the categories and distinctions which need to be examined. Chapter 2 suggests a number of typological parameters for the study of clitics. Following chapters then provide detailed examination of the notion of ‘word’ in a selection of spoken languages from Africa, North and South America, Australia, the Caucasus and Greece, together with a discussion of words in sign languages. The final chapter, by P. H. Matthews, asks what has been learnt from these general and particular studies.

This introduction begins by surveying the criteria that have been put forward for ‘word’, and suggests that one should sensibly keep apart phonological criteria, which define ‘phonological word’, and grammatical criteria, which define ‘grammatical word’. In some languages the two types of word coincide and one can then felicitously talk of a single unit ‘word’, which has a place both in the hierarchy of phonological units and in the hierarchy of grammatical units. In other languages phonological word and grammatical word generally coincide, but do not always do so. We may have a grammatical word consisting of a whole number of phonological words, or a phonological word consisting of a whole number of grammatical words.

Georgian is a language of the Kartvelian (South Caucasian) language family; this family appears to be unrelated to North East Caucasian and North West Caucasian language families. Georgian is spoken by some five million speakers in the Republic of Georgia, formerly part of the USSR, but now independent.

The notion ‘word’ (Georgian sit'q'va) is very much recognised in Georgian culture. The society is deeply literary, with the Georgian alphabet having been invented in the fourth or fifth century AD, and with texts dating from that time. Georgians have strong intuitions about what is a word, and this corresponds to orthographic practice, with few exceptions.

Summary of the typology

Like other languages of the Caucasus, Georgian has a large number of consonants (twenty-eight) and a modest number of vowels (five). Georgian is famous for its consonant clusters, which may contain up to seven consonants, e.g. mc'vrtneli ‘trainer’, vprckvni ‘I peel it’; clusters are not punctuated by epenthetic vowels. Traditional work in Georgian has identified so-called harmonic consonant clusters, which are characterised by (a) shared laryngeal properties (roughly, all voiceless, all voiced or all ejectives), (b) a structure in which each successive consonant is articulated further back in the mouth (e.g. pt, pk, tk are harmonic clusters, but kp and tp, while allowed, are not harmonic) and (c) a first segment that is a stop or affricate and a second that is a stop, affricate or fricative; see Vogt (1958) and Mač'avariani (1965).

Words and signs: on psychological and cultural validity

The question whether all languages have words may look like a nonsense question to many people, the universal existence of words being regarded as a truism in itself. Even though it is widely acknowledged that finding a strictly satisfying definition of ‘word’ is as difficult as defining similarly universal terms such as ‘sentence’ or ‘language’, the existence of words in all languages is not usually questioned.

As with all putative language universals, probing the validity of the claim depends crucially on looking at languages that are as ‘different’ as possible. If many otherwise very ‘different’ languages share a certain feature, it is more likely that this feature is a true universal than if only ‘similar’ languages are considered. The motivation for looking at the concept of ‘word’ in sign languages lies exactly here: for what could be more ‘different’ than a sign language? As Anderson (1982: 91) puts it: ‘Comparison of spoken and signed languages can be especially valuable because the parallels are so surprising at first, and seem so automatic and natural after we have worked with them. The challenge of finding these parallels produces important insights into the nature of human language in general. So we can often learn more by studying a sign language than by studying one more spoken language.’ This is of course not to ignore that modality-related differences between signed and spoken language can be just as revealing as the parallels between the two.

Jarawara has an eclectic verbal morphology in which some suffixes cannot have other suffixes added directly to them, and some suffixes cannot be directly added to what precedes. However, phonological word and grammatical word can be straightforwardly defined, and almost always coincide. There are just four circumstances in which one type of word involves two instances of the other. In providing a grammatical description for Jarawara – as for other languages – it is important to distinguish between the structure of a predicate and the structure of a verbal word.

Introduction

The small Arawá family of southern Amazonia (quite distinct from Arawak) consists of five extant languages – Dení, Kulina, Sorowahá, Paumarí and Madi (see Dixon 1999). The Madi language consists of three closely related dialects: Jamamadí (with about 190 speakers), Banawá (about 80 speakers) and Jarawara (about 150 speakers, spread over eight jungle villages). The description of Jarawara given here is based on materials gathered during six field trips, during 1991–99.

Jarawara is a highly synthetic language, basically agglutinative but with developing fusion (particularly in the gender-marking forms of inalienably possessed nouns – see Dixon 1995). There is a closed class of about fourteen adjectives, which only function as modifiers within an NP or as copula complements (other adjectival-type concepts are coded as verbs). Jarawara is head-marking, with the predicate including obligatory markers for S or A, and for O in a transitive clause.

Introduction

Ideally, the word is a transcendent unit of linguistic organisation serving at once as the maximal domain of lexical phonology, the maximal domain of morphology and the atom of syntax. Our everyday notion of ‘word’, I think it is fair to say, implies just this convergence, as do such scientific theories as the lexicalist hypothesis (of which Lapointe 1980 and DiSciullo and Williams 1987 are among the strongest formulations). This ideal is upheld empirically for a given language when, in the internal organisation of phonology, morphology and syntax, many logically independent processes each refer crucially, in some way, to the same unit or its boundaries, and when that unit is clearly comparable across languages. But to the extent that the ideal may not be upheld in each language, it is important for a general theory to be able to specify the ways in which phonology, morphology, syntax, and perhaps other subsystems may resolve the stream of speech in different, non-convergent ways.

I begin by introducing Cup'ik (§ 2), its typology (§ 3), and the word for ‘word’ in Cup'ik (§ 4). I then turn to the grammatical word (§ 5), the question of clitics and their grammatical status (§ 6), the phonological word (§ 7). I then summarise the mismatching that occurs between the grammatical and the phonological word (§ 8) and end with a brief conclusion (§ 9).

This volume includes a typological introduction, plus revised versions of ten of the sixteen presentations at the International Workshop on ‘Word’, held at the Research Centre for Linguistic Typology, La Trobe University, 7–12 August 2000. An earlier version of chapter 1 had been circulated to contributors, to ensure that the detailed studies of ‘word’ in individual languages were cast in terms of the same typological parameters.

All of the authors have pursued intensive investigations of languages, some of them little-known in the literature. They were asked to write in terms of basic linguistic theory – the cumulative framework in which most descriptive grammars are cast – and to avoid formalisms (which come and go with such frequency that any statement made in terms of them will soon become dated and inaccessible).

This volume comes from the third International Workshop organised by the Research Centre for Linguistic Typology. The first, held in 1997, gave rise to the book Changing valency: case studies in transitivity (edited by Dixon and Aikhenvald), published by Cambridge University Press in 2000. The second, held in 1998, resulted in the volume Areal diffusion and genetic relationship: problems in comparative linguistics (edited by Aikhenvald and Dixon), published by Oxford University Press in 2001.

Siouan languages were spoken in aboriginal North America in a broad belt stretching from what is now Arkansas northward and westward across the western prairies and eastern great plains to Alberta. In addition there were Siouan languages scattered in the East in what is now Virginia, Alabama and Mississippi. The languages we will be discussing in this chapter are primarily Crow and Hidatsa, of the Missouri River subgroup, and Dakotan (including Dakota and Lakota dialects) and Dhegihan (including Omaha, Ponca, Kansa, Osage and Quapaw) of the Mississippi Valley subgroup. Data from other languages will be introduced as necessary.

Typology

Typologically, Siouan languages are primarily head-marking, active–stative, AOV, SV languages of moderate morphological complexity. Verbs are the most highly inflected category, and they may optionally include numerous and productive instances of noun incorporation. There are also many lexicalised noun–verb compounds. Sapir (1921: 142) characterised Dakota Sioux as ‘complex pure-relational’ in basic type with derivational concepts signalled by agglutinating elements and pure relational (mostly inflectional) concepts somewhat fused. Dakota's overall morphological technique he characterised as ‘agglutinative-fusional’ and degree of synthesis he characterised as ‘synthetic (mildly polysynthetic)’.

Incorporation

Compared with languages like Inuit or Iroquoian we would probably not want to characterise Dakotan, or the rest of Mississippi Valley Siouan, as productively incorporating, but Siouan verbs do incorporate a number of notions that speakers of most European languages would convey with separate words. And these include pronominal, instrumental, locative and directional concepts.