This book is designed for students of linguistics who want to learn more about tone. It assumes a basic knowledge of phonological theory such as might be acquired in a year-long phonology course, but it does not assume any particular prior exposure to work on tone. The theoretical chapters of the book are framed in Optimality Theory (OT), but should be intelligible to students with no previous background in OT.

The book is suitable for a semester-long course on tonal phonology at the advanced undergraduate or graduate level. The theoretical chapters include some simple exercises, and the answers are given at the end of each chapter. It is also hoped that the book will be a useful reference work on the fundamentals of tone, and to this end it includes extensive references to both primary fieldwork sources and to theoretical works. An effort has been made to give broad coverage of tone languages, both typologically and geographically.

Having described the range of tonal contrasts in the previous chapter, this chapter proposes a feature system for capturing all and only the observed inventories. A good feature system has several desiderata, which are outlined in section 3.1. The following sections deal first with how to represent the number of levels. Since the clearly described systems have no more than a four-way contrast, the system will be based on two binary features, Register and Pitch. How to handle five-tone systems if they exist will also be covered. Next we move on to how to deal with contour tones, and conclude that they should be analysed as sequences of level tones. The reasons for choosing this over a unitary contour alternative will be explained and discussed. This section will be followed by a section on the geometric relation between Register and Pitch (comparing the approaches of, among others, Bao, Clements, Duanmu, Hyman, and Yip), and what sort of data bear on the choice. The penultimate section asks whether tonal features are entirely disjoint from segmental features, focussing mainly on Laryngeal features, but with some discussion of Pharyngeal and Tongue Root features. The chapter will end with the question of whether features are binary or unary, and how we deal with unmarked tones.

Desiderata for a feature system

It has been known for years that the smallest units of phonological structure are not phonemes, or indeed ‘individual sounds’ (if such things exist), but the properties that make up those sounds, stated in terms of distinctive features (or elements in Government Phonology, see Harris 1990).

Among the sounds of languages, consonants and vowels need no explanation to the lay person, but tones are another matter entirely. Tell someone you are writing a book about ‘Tone’, and they look blank, and yet by some estimates as much as 60–70 per cent of the world's languages are tonal. Begin to explain that you are interested in languages that use the pitch of the voice to convey meaning, and more often than not you will be interrupted with a remark such as ‘Oh, that must be really interesting: those emotions and nuances and subtleties are so important when we're speaking!’ Politely explain that actually you are interested in languages that use pitch to distinguish one word from another, not just to convey subtleties, and most people will assume that such languages are rare, and probably spoken only by isolated communities in less developed countries – until you point out that Mandarin Chinese (885,000,000 speakers), Yoruba (20,000,000) and Swedish (9,000,000) are all tonal.

Perhaps because of these misapprehensions (particularly prevalent in Western cultures), even among linguists tone is sometimes seen as a specialized topic that the general linguist can largely ignore. Undergraduate courses often pay it only cursory attention, and even graduate courses may devote no more than a class or two to the topic. The goal of this book is to fill that gap. It assumes a basic knowledge of phonological theory, but no prior acquaintance with the phonology of tone.

In the introduction I talked a little about how tones are produced, but of course this is only half the picture. For tones to be used in communication, they must also be perceived by the hearer, and not just as musical pitch, but as linguistic objects. Successful perception is also the first prerequisite for tonal acquisition, and that is the reason for combining in this chapter two apparently unrelated topics, perception and acquisition. I will begin with adult perception and then move on to the child. On adult tone perception, one of the most useful references is the excellent survey in Gandour 1978, still useful as I write this text.

Adult tone perception

Fundamental frequency, pitch and tone

I begin by reminding the reader of the difference between fundamental frequency (F0), pitch, and tone, already discussed in chapter one. In this order, the terms move from a purely phonetic term, F0, to a truly linguistic one, tone. F0 is an acoustic term referring to the signal itself: how many pulses per second does the signal contain, measured in Hertz (Hz) where one Hz is one cycle per second. The next term, pitch, is a perceptual term. What is the hearer's perception of this signal: is it heard as high or low, the same pitch as the previous portion of the signal or different? The mere existence of F0 differences may not be enough to result in the perception of pitch differences.

Introduction

By and large up until now I have emphasized the independence of tone from the rest of the phonology, with the major exception of its interaction with laryngeal features, especially voicing. However, tone bears a close relation to stress, and to intonation, and this interaction is the topic of this chapter. Tonal features can enter the phonology from a number of sources. Here there is a real difference of degree between tonal and segmental features, with segmental features being (almost) entirely associated with lexical items, rather than such things as prosodic heads, phonological phrases, or declarative intonation. Imagine what a language would sound like if the right edge of every phonological phrase ended in a voiced uvular fricative – we would have something very like the comedian Victor Borge's famous ‘Phonetic Punctuation’, where all commas are pronounced with a sort of gargle!

In tone languages, although the obvious source is lexical tone, it is by no means the only source. The overall picture that emerges is that the tonal phonology must accept not only input direct from the lexicon, but also from insertion during the lexical phonology, and from the utterance after lexical insertion. At this stage in addition to lexical items the representation may include certain abstract markers such as [FOCUS] and also syntactic structure, which in turn influences prosodic structure. The lexical entries that contain tones include not only the traditional lexical items such as nouns, verbs, roots, and affixes, but also functional elements and operators such as [WH], [FOCUS], [TENSE], prosodic phrases such as Phonological Phrase or Intonational Phrase, and pragmatic tunes such as Declarative.

The focus of this chapter is on the appropriate tools for the analysis of tonal phonology. Before we study the behaviour of tone in any detail, we need a common framework in which to couch our analyses. In section 4.1, I will begin by identifying five properties of tone that any analytical framework must be able to capture. Section 4.2 shows that an ‘autosegmental’ representation nicely deals with these properties. Section 4.3 introduces the Optimality Theory framework, and section 4.4 extends it to tonology. Section 4.5 applies the framework to the five central properties of tone, and sections 4.6–10 flesh out a full working model of tonal phonology within Optimality Theory.

Tone differs from many other phonological features in the following ways, rarely or never observed in more familiar consonantal or vocalic features:

1. a. Mobility: Movement away from point of origin

2. b. Stability: Survival after loss of original host segment

3. c. One-to-many: A single tonal feature shared by two or more segments

4. d. Many-to-one: Multiple tonal features surfacing on a single host segment

5. e. Toneless segments: Potentially tone-bearing segments that never acquire phonological tone

The only non-tonal features that regularly exhibit any of these characteristics are harmonizing features like nasality and certain vowel features, which may spread over large domains from their source segments. Features like consonantal place features, or laryngeal features like aspiration and voice, do not usually show these behaviours.

Classification

The languages of Africa display an astonishing variety of complex tone systems, and a correspondingly large body of literature has grown up in an attempt to understand them. Together with East Asia, Africa has probably the highest ratio of tonal to non-tonal languages, with tonal languages found in just about every language family with the exception of Semitic and Berber. In particular, the Niger-Congo family, which subsumes most of the sub-Saharan languages, including the Bantu group, is almost entirely tonal (or occasionally accentual). Table 6.1, based largely on Bendor-Samuel 1989, shows the affiliation of the major tonal languages of Africa, including all those discussed in this chapter.

Common or striking characteristics of African tone languages

By most estimates Africa contains well over a thousand languages, with the Niger-Congo group alone having more than 900; among the Bantu languages, at least fifteen are spoken by more than three million people, and at least six are spoken by more than five million. As one might expect in such a huge group of languages covering such an enormous geographical area, there is great diversity in their tonal systems. Nonetheless, one can identify some characteristics that are particularly widespread in Africa, and much less common in Asia or Central America, which we shall look at in later chapters.

The theoretical literature on tone has been dominated by studies of African languages, and as a result an understanding of African systems is a foundation for a study of other areas from a theoretical perspective.

The goal of this chapter is a descriptive summary of the types of tones and tonal inventories found around the world. It begins with a discussion of notational and fieldwork issues, then moves on to consider level tones and contour tones, the interaction of tone with segmental phenomena, and the origins of tonal systems (tonogenesis). It lays the groundwork for the following chapter, in which I will develop a theory of distinctive features for the analysis of tonal systems.

Which languages are tonal?

Although those of us whose first language is English think of tone as exotic, it is in fact extremely widespread. At a very rough estimate as many as 60–70 per cent of the world's languages may be tonal. Three areas of the world have so many tone languages that I have given them a chapter each later in this book: Africa (chapter six), East and South-East Asia and the Pacific (chapter seven), and the Americas (chapter eight). In parts of these areas (such as Meso-America), tonal languages are so common that they can almost be considered the norm. The vast majority of the world's tone languages are found in these areas. Conversely, there are parts of the world where tone languages are rare or perhaps non-existent, most notably Australia and New Zealand. Many other parts of the world show a more mixed picture. Let me go into a little more detail. In what follows, bear in mind that, as explained in chapter one, languages termed ‘accentual’ are often analysed as a sub-type of tone language having a small number of sparsely distributed tones.