1
Introduction

Eric Raimy and Charles E. Cairns

1 Scope and background

The segment is an elusive entity. On the one hand it appears so intuitively obvious that it might appear strange to some that it warrants the attention of an entire volume, yet on the other hand it is not at all clear what it is, where it comes from, or whether or not the concept is entirely chimerical and thus a hindrance in our attempts to understand human language. This volume takes a deliberately eclectic approach to the study of the segment. The editors believe that theoretical, empirical, and methodological heterogeneity provides the most effective strategy to elucidate any entity that is difficult to pin down to the satisfaction of all researchers. This is the approach we took in Cairns and Raimy (2011a) in the investigation of the syllable, an entity with a similarly gauzy existence; the current volume parallels that study, mutatis mutandis.

To the extent that the segment has been explicitly discussed at all, there has never been consensus on what the segment is, or even if it represents a cognitively real entity of any kind. What level does it exist on, if it exists at all? Is it merely part of a more basic unit like the syllable? Most scholars agree that segments should be broken down into more basic units of some sort. But if subphonemic units are the elementary atoms of phonological description, are they reliably coterminous, producing an emergent segment? That is, is the speech stream neatly segmented syntagmatically, each segment defined as a bundle of distinctive features? These are questions that have not been resolved to the satisfaction of all scholars in the field.

One reason the segment continues to be such an elusive entity is that its boundaries lack any uniform and direct phonetic correlates. Hockett's Easter egg analogy is particularly appropriate here. Hockett (1955: 210) compares the relationship between the segment source and the speech stream as that between a series of raw, but decorated Easter eggs and the resulting smeared mess obtained by running the eggs on a belt through a wringer. The original eggs end up being mashed together with only the presence of different colors reflecting the original source. Hockett says that the task for the inspector is only to identify the eggs and not to put the eggs back together. Making the analogy to phonology and phonetics, are there any tasks in phonology that require the identification of the eggs in the smashed output? Or, can phonology proceed without caring about how many and what colors of eggs there were, and where one ended and the other began? Or, do eggs even exist in the first place?

The remainder of this chapter is divided into three more sections. Section 2 provides a short overview of the history of the segment in the language sciences in order to identify the main issues at play. The third section explores some issues in contemporary views of the segment, and the final section provides a brief overview of the chapters in this book.

2 A short history of the segment

The assumption that the speech stream is divided into segments goes back to the earliest serious investigation into human language. From Pan?ani's rules of the sixth century BCE, which operated on discrete segments of speech, through most contemporary theories of phonology and phonetics, this assumption has prevailed consistently, sometimes implicitly and sometimes explicitly. This assumption has by no means been uncontested, as is reflected in this volume. There are multiple tensions that can be identified in the question as to whether or not the segment exists.

One tension was that between viewing the segment as defined by the features of which it is composed versus viewing the segment as an atomic whole. The former view appears to date back to Johann Conrad Amman (1692, see Ohala 2010) who proposed a hierarchically arrayed set of features. Erasmus Darwin (1803), grandfather of Charles Darwin, proposed that speech sounds are to be defined in terms of their distinguishing characteristics, thus subordinating the segment to the feature.

The most visible and productive work in nineteenth-century phonetics and phonology assumed atomic segments. For example, since its first publication by Passy in 1888, the International Phonetic Alphabet to this day conceptualizes segments as atomic elements; the "features" that do appear in the IPA are meant as attributes of segments and classes of segments, not as primitive constituent parts that can occur independently of one another.

The Neogrammarians made significant scientific progress using the atomic segment. The Neogrammarians discovered that sound change was based on segments; every occurrence of a segment, or class of segments, obeys the rules of change in that language. Verner's law was a particularly compelling example of the power of the segment. Verner and the other linguists at the time were attempting to explore the idea that all historical change is segment-based, that is, that sounds changed as classes of segments, not as a result of idiosyncratic changes in individual words. The existence of apparent exceptions to basic generalizations such as Grimm's Law was an incentive to push the concept further. It was Verner's (1877) insight, expressed in "Eine Ausnahme der ersten Lautverschiebung," that a particular sub-class of segments escaped Grimm's Law, namely the class of PIE voiceless stops when intervocalic and following an unstressed syllable in the same word. This was a major achievement of enlightenment science and it rested entirely on the assumption that laws operate at the level of phonology, which is autonomous from syntax and semantics, and on classes of segments viewed as unanalyzable wholes.

Significantly, the compositional view of the segment did influence important developments during the nineteenth and early twentieth centuries, despite the dominance of the atomic segment concept. For example, Alexander Melville Bell furthered the compositional idea in his 1867 publication of Visible Speech, a system of symbols each of which is composed of depictions of the activities of the relevant speech organs. Bloomfield (1914: 34) describes "nine typical tongue positions, three along the horizontal plane, front, mixed, and back, and three along the vertical, high, mid, and low," strongly suggesting a componential analysis of segments. The concept assumed a central role in Prague School phonology, where its revival was inspired by Saussure's central idea that language may be analyzed as a formal system based on the notion of distinctive oppositions. These oppositions became conceptualized as features which were the fundamental components of phonemes, eventually morphing into the concept put forth by Chomsky and Halle in the Sound Pattern of English (1968), where the phonetic string is neatly sliced into a sequence of segments, each of which consists of a bundle of binary features.

A second tension concerning the history of the segment was that between viewing rhythmic properties of speech, notably the syllable, as a fundamental unit and the segment is derived from it, or the other way around; this tension is well represented in the current volume. The former view can be traced back to Joshua Steele, who published Prosodia Rationalis in 1779; Steele took as his point of departure not segments, but rhythmic properties of speech such as tone, pitch, and cadence. This idea gained some traction during the nineteenth century, as witnessed in Sweet's 1877 A Handbook of Phonetics. The idea that the fundamental units of speech are in its rhythmic properties, not segments, reached its fullest expression in the work of Herbert Raymond Stetson in the first half of the twentieth century. We will return to Stetson's contributions below. But note for now that the concept of the segment as primary was dominant in the nineteenth century.

3 Contemporary issues concerning the segment

The tension that perhaps most shapes current debate about the segment concerns methodology and philosophy of science, and it is based on how one interprets empiricism in scientific inquiry. Put in a simplified form, the tension is based around observable vs. covert structure in the data collected from the speech stream. A strong empiricism will require the existence of the segment to be based on strictly observable measurements in acoustic or articulatory data, while weaker versions of empiricism are more tolerant of qualitative observations of behavior or introspective judgments that point to covert structure. The stronger forms of empiricism seem to have flourished in recent decades as more and better technology for measuring acoustic and articulatory aspects of speech have become available. Ohala (2010) is representative of contemporary views on the question of how speech is represented in the brain:

The first, most candid, answer to this question is: we don't know. Even such a fundamental issue as whether phoneme-sized segments are employed - or employed at all stages of encoding and decoding - has not been settled. There is an abundance of candidate answers given to the question as to how speech is represented mentally but until they have been properly evaluated they are just speculations. Within phonology the basic criterion applied in evaluating claims about mental representation of language is simplicity and such related notions as naturalness and elegance. But these are...