Central Alaskan Yup'ik is a language of the Eskimo-Aleut family, spoken in southwestern Alaska by over 10,000 people. Like other languages in the family, Yup'ik has much to contribute to the study of valency-changing derivation, particularly because of its explicit specification of grammatical relations and its wealth of valency-changing devices. The roles of participants in events and states are distinguished both by case suffixes on nouns and by pronominal suffixes on verbs. The language is highly polysynthetic, with hundreds of derivational suffixes, many of which affect argument structure. The rich inventory of valency-changing devices provides a fruitful basis for cross-linguistic comparison, showing us ways in which such devices can vary in their semantic, syntactic and discourse effects.

The basic grammatical structures of the Eskimoan languages are well understood, thanks to pioneering work on Greenlandic by Egede (1750, 1760), Kleinschmidt (1851), and many others working with Eskimo-Aleut languages since that time. Fine descriptions of Yup'ik are now available, especially Woodbury (1981), Jacobson (1984, 1995), Miyaoka (1984, 1987, 1996 and 1997) and Reed, Miyaoka, Jacobson, Afcan and Krauss (1977). These works have proven invaluable in the investigation of the structures discussed here. Additional studies are in Mithun (1996). Material cited in the present work comes primarily from conversations among members of the Charles family and their friends of Bethel, Alaska, especially Nick Charles (NC), Elena Charles (EC), George Charles (GC) and Elizabeth Charles Ali (EA).

Basic morphological structure

Yup'ik words are classified as either uninflected (particles) or inflected (nouns and verbs).

Further failings of affix-driven stratification

The single major point made in the preceding chapter was that, in the face of the dual stratum-affiliation displayed by a substantial number of English affixes, the hypothesis of affix-driven lexical stratification cannot be sustained. But while the failure of a large number of affixes to serve unambiguously in the definition of lexical strata suffices alone to discredit the affix-driven model, that model has a number of further, apparently unrelated, weaknesses on the morphological side that are worth noting – especially given that the replacement model to be proposed in this chapter must of course seek to avoid all the weaknesses of the current model. I hope to show in this section that those further weaknesses of the affix-driven model all amount to a systematic failure to explain certain stratification-related facts; further below I shall demonstrate that these facts are amenable to a unified explanation that follows automatically from a stratification model whose morphological diagnostics arise from characteristics of the affixation base (along the lines sketched in Section 2.2.2 above) rather than from the diacritic marking of affixes. I shall refer to the replacement model that will be presented in this chapter as one of ‘base-driven stratification’.

Stratification, stacking restrictions and etymology

One of the outcomes of Chapter 2 has been the recognition that any significance attached to the perceived effects of the AOG (Siegel 1974; Selkirk 1982b) is less straightforward regarding lexical stratification than has been assumed in the literature.

Shortly after the appearance of the first main-stream book on Lexical Phonology (Mohanan 1986), Gussmann published an incisive and detailed review (1988), which – as is strangely more apparent now than it was then – captured the mood of the time. In it, he attacked not just the book under review but the entire programme of Lexical Phonology, meticulously dismantling Mohanan (1986) chapter by chapter and concluding: ‘If the critical assessment of lexicalism presented here and elsewhere were to be accepted, then Mohanan's book would very likely come to stand as a requiem for Lexical Phonology’ (Gussmann 1988: 239). As at that time phonologists were beginning to abandon in droves not only derivationalist theories but also English – one of Mohanan's main concerns – Gussmann's review could not have come at a better time for some, and at a worse time for others. Such was the mood of the time.

The title of Mohanan (1986), The theory of lexical phonology, misleadingly suggested that the book reported, and indeed was, the state-of-theart. It wasn't anything like that; but the misled reviewer can be forgiven for responding in kind. Mohanan (1986) was an easy target not only for a reviewer hostile to the programme but, perhaps even more so, for the theory-internal and therefore constructively minded critic.

Introduction

This chapter, concerned with the predictions regarding syllable structure that are made by the present model of stratified phonology-morphology interaction, will focus on two kinds of issues. The first of these is the interaction of syllabification with the processes of the morphology. This is both a theory-internal matter, dealing with the mechanics of the lexical and postlexical derivation, and an empirical matter, concerning the generalisations that are available regarding syllable structure and the position of syllable boundaries in the vicinity of morphological boundaries. And it is of course connected with the issue of liaison: the various sandhi phenomena – [r],[j w], prevocalic [?] – and more generally the effect of morphological boundaries on the organisation of the speech continuum into syllables. The second issue concerns the possible existence of stratum-specific characteristics of syllable structure. While it is not inconceivable that syllabification may follow a uniform pattern throughout the derivation (in that case, the relevant mechanisms would simply not be stratum-sensitive), the opposite case whereby the syllable patterns found on different strata have different characteristics would lend valuable support to the stratification model proposed in this study. I shall show here that syllabification does indeed have a number of stratum-specific features, although the basic mechanisms are the same throughout the lexical derivation.

Obviously, stratum-specific and morphology-sensitive syllable patterns can only be defined against the background of a specific model of syllable structure, developed in turn with regard to the present derivational framework.

The origins

Let us assume that the English lexicon is divided into two strata. This is not only the position that appears to have met with broad consensus in recent research; it will also be extensively argued for in later chapters. Moreover, it happens to be the position most closely associated with Siegel's (1974) original observations and claims, which were to prove seminal to the framework while in turn harking back at least to SPE. Such origins are worth investigating, especially when – as we shall see – they are also the origins of a major flaw in most current stratification models.

At the root of the two-strata model lies the familiar generalisation, dating back to SPE and beyond (for example Bloomfield 1933), and related to the more general ‘close-juncture’ vs. ‘open-juncture’ distinction found in the American structuralist tradition (for example Trager and Smith 1951), that the derivational morphology of English has two types of affixation processes, distinguished from each other empirically by a syndrome of differences in terms of morphological and phonological behaviour that will be discussed in some detail below. The well-known ‘stress-shifting’ vs. 'stress-neutral’ effect on the affixation base is one such difference in behaviour. In formal terms, SPE expresses the distinction by associating affixes with different boundary symbols, ‘#’ and ‘+’, where the former ‘word boundary’ serves, among other things, to block the cyclic application of stress rules – #-affixes therefore lie outwith the domain of the stress rules, their presence having no effect on the stress pattern of the base – while the latter ‘morpheme boundary’ does not block the cyclic (re-)application of stress rules. Witness the stress shifts caused by the addition of +-affixes in átom – atóm+ic – àtom+íc+ity, and the absence of such shifts in átom – átom#less – átom#less#ness.

Explananda

[r]-sandhi: linking and intrusion

My main concern in this chapter and the following will be [r]-sandhi: the phenomenon of ‘linking [r]’ and ‘intrusive [r]’ in RP and its synchronic derivation within Lexical Phonology. The singular form, ‘phenomenon’, is appropriate here. While the descriptive accounts (for example Wells 1982; Giegerich 1992a; Gimson 1994) and many of the more formal analyses found in the literature (Kahn 1976; Mohanan 1984, 1985; Nespor and Vogel 1986; Broadbent 1991; McCarthy 1991, 1993; Scobbie 1992; McMahon, Foulkes and Tollfree 1994; Harris 1994; Kamińska 1995; McMahon 1996) tend to draw the well-known distinction between ‘linking [r]’ and ‘intrusive [r]’ (see the examples in (1) below), reporting that RP speakers freely use the former but often shun the latter, it is also clear from the descriptions that the avoidance of [r]-intrusion does not come naturally to those speakers: it is brought about and maintained only thanks to continuous enforcement by a strong intrusion stigma (Gimson 1994: 263f.). It may well be more natural for the RP speaker to have both linking and intrusion than it is to have the former but not the latter. The deliberateness of intrusion-avoidance alone suggests that linking and intrusion may, in purely synchronic-phonological terms, be nondistinct (and stronger arguments to this effect will be given below); but any account dealing with the phenomenon also has to address the question of how it is that at least some speakers (especially those of ‘speech-conscious adoptive RP’ Wells 1982: 284f.) succeed in implementing the intrusion stigma with remarkable reliability.