0 Introduction

It is well known that in some cases English stress performs a grammatical function in encoding the syntactic categories of the words concerned. For instance, this phenomenon can be observed with disyllabic words in -ate, with create and negate being backstressed ([01]) because they are verbs, while climate and senate are forestressed ([10]) because they are nouns. The difference holds for those which belong to several syntactic categories, as with filˈtrate and dicˈtate (verbs) vs ˈfiltrate and ˈdictate (nouns). The same phenomenon affects words with separable prefixes, like ˌupˈdate and ˌoverˈdose (verbs) vs ˈupdate and ˈoverdose (nouns).

One class of words is especially affected by this type of stress alternation, namely disyllabic noun–verb pairs starting with an etymological, inseparable Latin prefix, such as obˈject and reˈcord (verbs) vs ˈobject and ˈrecord (nouns). Interestingly, this alternation is far from being a general rule, since two other tendencies exist: pairs where both members are backstressed (conˈcern, deˈfeat) and pairs where both members are forestressed (ˈcomment, ˈconvoy). Using their own corpus, Abasq et al. (2014) indicate that the isotonic CONˈCERN type represents 59% of all pairs, the isotonic ˈCOMMENT type 13%, while the alternating OBJECT type fits in-between with 28%.

This broad description obviously raises the problem of understanding why this threefold distribution exists, specifically the alternating type. Yet it conceals at least two even more intricate problems. The first – and more immediate – is the existence of synchronic stress variation, as can be seen with dispute (noun) and increase (noun or verb), which may be stressed on either syllable. The second problem is diachronic variation. For instance, perfect (verb), now backstressed, used to be forestressed. Conversely, convoy (verb) changed from [01] to [10]. Of course, both problems are linked. For example, while discount (verb), alloy (noun) and research (noun and verb) used to be only backstressed, they may now be stressed on either syllable.

1 Description in the literature

This change in stress placement, which dates back to at least the sixteenth century (Minkova 2014: 311), has often been described, but most studies refer to Sherman (1975) as being the first to examine the phenomenon in detail.

0 Introduction

Reference to orthographic information is almost absent from phonological research, and one might ask why this should be. Is the question of the relationship between orthography and phonology considered as being ‘solved’ or is it simply ignored? Can phonology completely do away with orthographic information? This chapter argues that it should not and brings forward arguments which I hope will arouse interest for the links between these two representations of the signifiant.

Interest in the relationship between orthography and phonology has come mainly from psycholinguistics and reading acquisition. Several experiments have shown that orthographic information can influence tasks which do not involve any visual stimuli, such as rhyme judgement, lexical decision or auditory word recognition (Chéreau et al. 2007; Peereman et al. 2009; Perre and Ziegler 2008; Perre et al. 2009; Seidenberg and Tanenhaus 1979; Taft 2006; Taft and Hambly 1985). Brewer (2008) also found that the number of letters representing word-final voiceless obstruents had a positive influence of on the phonetic duration of these segments in both experimental conditions and spontaneous speech.1 Zamuner and Ohala (1999) found orthographic effects on syllable segmentation in preliterate children, for example cabin [kæ] – [bɪn] vs cabbage [kæb] – [bɪdʒ], although their experiment was replicated by Treiman et al. (2002), who did not confirm the results and only found orthographic effects in literate speakers. Significantly, an experiment on French speakers (Bürki et al. 2012) showed that a single orthographic presentation of a nonce-word is sufficient to induce a significant change in the phonological representation of that same nonce-word, which was constructed through 25 oral presentations of that word over several days. These results have been interpreted as signs that phonological knowledge is restructured when orthography is learned (Bürki et al. 2012; Ehri and Wilce 1980; Jaeger 1986; Perre and Ziegler 2008; Taft and Hambly 1985; Wang and Derwing 1986) and that this modification of the cognitive system is permanent. It might seem trivial to mention it but, as underlined by Jaeger (1986) and Laks (2005), learning how to read and how to write is a very long and tedious process and constitutes a powerful force to restructure phonological knowledge.

0 Introduction

Australian English has received much attention with a view to describe and model the components of its lexicon and phonology while assessing its role and status in the English-speaking world. Recently, possibly because the Australian science of linguistics thoughtfully relies on national programmes of spoken corpus-making as well as on a broad landscape of laboratory phonology, some perceptual and acoustic analyses of stress and pitch accent have regularly been published, notably about High Rising Terminal (HRT) (for example Guy and Vonwiller 1989; Fletcher and Harrington 2001; Fletcher et al. 2002; Horvath 2004; Jespersen 2016). Other linguistic works have investigated the issue of lexical stress in Australian English phonology (Tabain et al. 2004).

Australian English is no homogeneous variety. The demonstration that multiple British dialectal origins at the root of its endogenous development led to a common systemic core (one transported English) under a determined evolutionary process of native Englishes must be balanced with the contemporary reality of phonological variation in Australian English (Przewozny and Viollain 2016). Furthermore, while language contact between British dialects of English has constantly played a crucial role from the early days of Proto-Early Australian English (for example Mitchell 1940; Turner 1966; Hill 1967; Bernard 1969; Fielding and Ramson 1973; Horvath 1985; Gunn 1992; Przewozny 2016), the influence of the Aboriginal input has to be considered too, both from the point of view of lexical stress in Aboriginal loanwords in Standard Australian English (Dixon 2002; Dixon et al. 2006; Martin 2011b) and that of the lexicon of Standard Aboriginal English (Malcolm and Kaldor 1991; Eades 1993; Vinson 2008 among others).

The scope of this chapter is (i) to analyse the nature of lexical stress patterns in Aboriginal loanwords in Standard Australian English (SAusE), (ii) to present a systemic account of lexical stress in SAusE based on dictionary data, and (iii) to examine the common hypothesis that speakers of Standard Australian Aboriginal English (SAbE) have a tendency to stress words initially in contemporary spoken data.

Our theoretical stance is a Guierrian one (see Chapters 2 and 5 of this volume) inasmuch as dictionary data, morphological and orthographic information can support the systematic study of the Australian lexicon.

0 Introduction

In this chapter, we investigate the hypothesis that exceptional stress properties of English are under-represented in a systematic fashion.

What does under-representation mean here? It means that some phonological property occurs less often than expected. For example, in a sample of 19,528 English words, we find 3949 words with trochaic stress (σ´ σ˘ ) and 1416 words with final stress (σ˘σ´). On the assumption that both patterns should be equally likely, the latter class is clearly under-represented with respect to the former and a statistical comparison confirms this: χ2(1, N = 5365) = 1195.916, p < 0.05, Cramér's V = 0.472.

If we assume feet in English are generally trochaic, as argued, for example, by Hayes (1981), and assume the general framework of Optimality Theory (OT) (Prince and Smolensky 1993; McCarthy and Prince 1993b), then in the case of a disyllable with final stress, its markedness could stem from the fact that the initial syllable is unfooted (Parse-σ) or it could stem from the fact that a monosyllabic foot is built (FtBin): σ˘ [σ´ ]. Both constraints would be violated by the winning candidate.

If we assume that it is avoidance of unfooted syllables that makes iambs marked and less common, then this predicts that forms like σ˘σ´σ˘ should also be under-represented; if what is driving the under-representation of σ˘σ´ is avoidance of monosyllabic feet, then no such prediction is made about forms like σ˘σ´σ˘. Interestingly, when we turn to three-syllables words, the facts confirm that it is monosyllabic feet that are avoided, not unparsed syllables. Forms like σ˘σ´σ˘ are much more frequent (1233) than words like σ`σ´σ˘ (159).

This is consistent with the general claim of Input Optimization (Hammond 2013, 2014, 2016, 2017) and thus provides an argument for that framework. The claim there is that input representations are skewed so as to reduce phonological complexity, the rank and number of violations exhibited by winning candidates. This chapter offers a statistical analysis of the distribution of exceptional stress in English, demonstrating that the distribution can be unified by assigning a penalty to constraint violations commensurate with the rankings of the constraints.

The idea for this book emerged from conversations among participants at the 2015 conference in France entitled ‘PAC 2015: Variation, Change and Spoken Corpora – Advances in the Phonology and Phonetics of Contemporary English’. Jacques Durand made the initial call for the book, devised the plan for its implementation, and set it in motion. Anne Przewozny and Eiji Yamada then took over the editing of the book but, due to the busy schedule of the editors, the pace of progress was slow. Later, Nicolas Ballier and Jean-Michel Fournier joined us as editors. In the end, it took six years after the original plan to publish the book, which is an incredibly long incubation period. However, the significance of publishing this book has never wavered in spite of all the difficulties, including the COVID pandemic that began at the end of 2019 and is not yet under control as of this writing.

The reason our plan did not vanish is that we believe stress is an important subject. It has been one of the central issues in research since the early days of linguistics. Its study became more active in the twentieth century, especially with the work of Daniel Jones and Leonard Bloomfield, and was taken over by the generative phonology of Chomsky and Halle (1968) (hereafter SPE). Since then, no study of word stress in English has been possible without reference to this work, whether one accepts the SPE view or not. Despite the importance of the subject, although a number of papers have been written, in more than fifty years there have been (to the editors’ knowledge) only a few monographs in this tradition that have analysed word stress in English as a central issue. These include Fudge (1984), Burzio (1994), Hammond (1999), Wenszky (2004), Yamada (2010) and Zamma (2013).

Research on word stress in English since SPE can be roughly divided into ‘pre-Optimality Theory (pre-OT)’ and OT. Significant theories of ‘pre-OT’ include Segmental Phonology in SPE, Metrical Phonology in Liberman (1975), Liberman and Prince (1977), Hayes (1980, 1982), Prince (1983) and others, Lexical Phonology in Mohanan (1982) and Kiparsky (1982a, 1982b) and others, and later theory in Halle and Vergnaud (1987).

0 Introduction

There is no doubt that with the publication of The Sound Pattern of English by Chomsky and Halle in 1968 a new era began for phonology. This was by no means the first important publication in generative phonology (see for example Chomsky 1964, 1967; Halle 1959, 1962), but it was the first extensive application to English, yielding basic principles for a novel approach to the nature of phonological representations and processes. The cornerstone of The Sound Pattern of English (hereafter SPE) is its treatment of word stress. Whether for or against, specialists now had to contend with an original way of envisaging English word stress that underlined its regularity if one was ready to delve under the surface. Within the generative tradition itself there emerged a rich tapestry of revisions and counter-positions leading eventually to a number of radical overhauls of phonological theory. Our aim here is to survey some of these main revisions and to provide a route through a complex technical literature.

Of course, SPE was by no means the first work to deal with English word stress: the topic had been discussed by a wealth of publications, descriptive, pedagogical and theoretical. Although we focus here solely on the generative tradition, we emphasise that the background to SPE cannot be neglected. We have chosen to cover some of this ‘prehistory’ by dividing it into a British and an American tradition, and by selecting a few major figures that paved the way for SPE. In the British camp, we focus on the contrasting views of Daniel Jones and Roger Kingdon. In the United States, we consider the work of Leonard Bloomfield and John Samuel Kenyon, as well as some of the conclusions reached by various post-Bloomfieldian phonologists. Of course, each tradition is not monolithic in its treatment of stress: these studies have influenced one another, and similar ideas were defended on both sides of the Atlantic. It is against this backdrop that SPE came into the world.

In SPE, sets of ordered rules are postulated for English stress, linking underlying phonological representations to phonetic representations. Rules are applied cyclically by means of a universal transformational cycle, from the innermost constituents to the outermost.

0 Introduction

This chapter compares the two main groups of theories of word stress assignment in American English (General American, hereafter English), highlighting certain exceptional treatments in their accounts and identifying their limitations. It then briefly introduces a new approach that attempts to explain both exceptional and core (i.e. non-exceptional) examples in a unified way, together with an attempt to refine some of the new theory's limitations.

In section 1, we will look at certain types of exceptions to the theories. We will discuss one group of exceptional words to word stress treatments in The Sound Pattern of English (Chomsky and Halle 1968) and subsequent studies (section 1.1), followed by discussion of a converse group of exceptional words to treatments in classic Optimality Theory (section 1.2). In section 2, by applying the recently developed Positional Function Theory (PFT) of Yamada (2010, 2013a), we will suggest a solution to these two groups of exceptional words relating to subsidiary stress assignment (section 2.1), followed by a brief introduction to the treatment of main stress assignment within the PFT framework (section 2.2). In section 3, we will consider a number of remaining issues in PFT: first, we will solve a problematic word type by employing the concept of Coordinate Axis Shift (section 3.1); second, we will look at the treatment of another problematic word type using the concept of Stress Domain in tandem with Stress Retraction (section 3.2); and third, we will examine the motivations for the two types of ‘primary–secondary’ word-stress treatments discussed in 3.1 and 3.2 (section 3.3). The advantages of PFT will be outlined in section 4. The chapter ends with conclusions in section 5.

1 Exceptions

1.1 One type of exception to treatments in SPE and subsequent studies

First to be mentioned here is The Sound Pattern of English (Chomsky and Halle 1968, hereafter SPE). In SPE, main and subsidiary stress assignment are comprehensively accounted for by the Main Stress Rule (MSR) and auxiliary rules with the help of the cyclic application of stress rules.

Take the còndênsátion word type, (1a) below, for example. This is a noun derived from the verb condénse.

0 Introduction

At the end of the sixties, Lionel Guierre had been studying English stress for a number of years in view of a French doctoral dissertation when Chomsky and Halle published The Sound Pattern of English (1968, hereafter SPE), a truly revolutionary view of the workings of English stress, though its actual concern was more likely to have been linguistics and phonological theory than English stress per se. In view of what he considered a momentous scientific event, Guierre took the rare decision to re-start his analysis from scratch by resorting to computer analysis, which was becoming more easily available to scientists at the time. Let us recall that at the time, the challenge was daunting: each entry of Jones's English Pronouncing Dictionary (12th edition) had to be transferred on to a punched card, with an appropriate and efficient coding transfer, even before attempting any kind of systematic analysis. Guierre finally succeeded in completing his analysis of the very first computerised corpus of English pronunciation by the end of the seventies, when he presented his doctoral dissertation which allowed him to become the first Professor of English pronunciation in the history of French university.

Except for a few specialists around the world, his dissertation remained mostly unknown, however, as it was never published as such: unlike Chomsky and Halle's book, its scope was strictly limited to English stress and pronunciation, with no linguistic theoretical ambitions. Guierre's work is best known through his Drills in English Stress-Patterns (1984a), a book aimed at students of English as a foreign language, but with a foreword by A. C. Gimson himself:

0 Introduction

Most studies about English verb stress, however interesting from a theoretical point of view, seem to be based on limited sets of examples, leading their authors to assertions about the stress behaviour of English that are not actually vindicated through comprehensive data. Following Lionel Guierre's teachings, this study sets out to analyse as comprehensive a corpus as possible, based on dictionary data: though limited in some extent by the very nature of the source, it can still be seen as a first step towards a reliable description of the category. The chapter is structured as follows: first, we review the literature on English stress and distinguish two generalisations which were proposed in Chomsky and Halle's Sound Pattern of English (section 1). One generalisation refers to the weight of the final syllable of verbs while the other refers to the presence of a semantically opaque prefix. We then set out to evaluate the empirical validity of these two generalisations, as well as to propose a global overview of how English verbs are stressed, all morphological categories included. In section 2, we present how the data were collected, cleaned and classified. In the following section, we detail the stress distributions found in all the morphological categories found in the data (section 3). At the end of the presentation of our results, we measure the empirical validity of the two generalisations in the relevant morphological categories and show that they have similar efficiencies but different theoretical costs. Crucially, we find that the data support an analysis including semantically opaque morphological constituents among the possible determiners of stress.

1 English verbs: stress, syllable structure and morphology

1.1 The two generalisations

In Chomsky and Halle (1968, hereafter SPE), there are two parameters determining the position of primary stress in verbs that many morphologists would now analyse as monomorphemic. The first parameter is the segmental make-up of the final syllable. In their Main Stress Rule, a verb gets penultimate stress if the final syllable contains a ‘weak cluster’ (V̆C; for example astónish, édit, consíder, imágine, intérpret) and final stress otherwise (for example maintáin, eráse, appéar, collápse, exháust, tormént, usúrp). Since Kahn (1976), the terms ‘weak cluster’ or ‘strong cluster’ have been abandoned and replaced by the notion of syllable weight.