B. Elan Dresher

The contrastive hierarchy in phonology

I will show that phonologists have vacillated between two different and incompatible approaches to determining whether a feature is contrastive in any particular phoneme. One approach involves extracting contrastive features from fully-specified minimal pairs. I will show that this approach is provably untenable. A second approach arrives at contrastive specifications by ordering features into a hierarchy, and splitting up the inventory by successive divisions until all phonemes have been distinguished. I will show that this hierarchical approach solves the problems encountered by the minimal-pairs method. Moreover, a hierarchical approach to contrastiveness is implicit in much descriptive phonological practice, and can be found even in the work of theorists who argue against it. Given the centrality of the issue, it is remarkable that it has received almost no attention in the literature. Recovering this missing chapter of phonological theory sheds new light on a number of controversies over contrast in phonology.

On some supposed contributions of artificial intelligence to the scientific study of language

The earliest attempts to apply computers to the problem of understanding natural language, although entered into with great optimism, did not meet with much success. The ultimate failure of the various machine translation projects, chronicled by Bar Hillel (1964) showed that the basic approach of these projects was incorrect. This approach viewed translation between natural languages as a fairly trivial process: it was hoped that if the computer were only supplied with dictionaries of the two languages and a certain amount of grammatical information then it would be able to produce fairly high quality translations. The nature of this failure led to the belief that a more sophisticated view of language, involving not only a dictionary and grammar but also full semantics and total knowledge of the world, was required if computers were to deal with natural language with any success at all. The problem of applying computers to natural language, which had formerly seemed quite tractable, if not trivial, now appeared to be all but impossible. Thus it was to the considerable surprise and delight of many that Joseph Weizenbaum’s ELIZA program was able to simulate a sort of Rogerian psychiatrist with a degree of success which surpassed all expectations and which even aroused a certain amount of interest in psychoanalytical circles. Since then, less outwardly striking but more influential work, such as Terry Winograd’s SHRDLU program, have contributed to a renewed feeling of optimism that computers can be made to understand human language’.

Head-dependent asymmetries in phonology: complexity and visibility *

Developments in phonological theory have led to the recognition that phonological representations have a layered constituent structure. Many, perhaps all, of these constituents contain elements which can be identified as heads. Heads enter into various kinds of relations with their dependents. In this article, we identify a phenomenon which is quite pervasive in every part of phonology which has heads and dependents, namely, the existence of head–dependent asymmetries (henceforth HDAs). While various particular manifestations of these asymmetries are well known and have been much studied, this is the first attempt, to our knowledge, to unite a broad range of seemingly different phenomena under one heading. We identify various types of HDAs, and propose constraints on possible HDAs. Most importantly, we distinguish between HDAs that involve complexity, and those that involve visibility. These have properties which potentially contradict each other. We propose that they apply in fundamentally different types of cases: unlike complexity HDAs, visibility HDAs are limited to mappings from one phonological plane to another, and so are related to the notion of projection (cf. Vergnaud 1977). We also wish to show that an understanding of HDAs reveals general structural principles that play a role in diverse phenomena at various levels of the phonological hierarchy. For example, the fact noted in the Optimality Theory literature that certain positions tend to be more ‘faithful’ to underlying specifications (Beckman 1998) is a consequence of the fact that heads allow more complexity. These principles act as constraints on possible constraints, and on possible mappings from one plane to another.

Open Syllable Lengthening in West Germanic.

Vowels in stressed syllables in the West Germanic languages-e.g. Middle English, Middle Dutch and Middle High German-were lengthened under certain circumstances. There have been two different explanations for this change. The traditional assumption is that a process of open syllable lengthening (OSL) was introduced to standardize the quantity of stressed syllables (Prokosch 1939, among others). The second, quite different, approach assumes that the lengthening process (at least in Middle English) is not OSL but some sort of compensatory lengthening caused by the loss of a final schwa (Minkova 1982, 1985, Lass 1985, Hayes 1989, Kim 1993). We attempt to show that OSL was part of the grammar of all three languages, but that the motivation depended on the local contexts. We claim that all three languages endeavored to maintain and maximize the Germanic foot (Dresher & Lahiri 1991), and OSL contributed in different ways to do so.

Palatalization and "strong i" across Inuit dialects

Inuit dialects with palatalization all distinguish between “strong i” and “weak i”: instances of surface [i] that cause palatalization and those that do not, respectively. All dialects that have completely lost this contrast also lack palatalization. Why are there no /i, a, u/ dialects in which all instances of surface [i] trigger palatalization? We propose that this typological gap can be explained using a contrastivist analysis whereby only contrastive features can be phonologically active, palatalization is triggered by [coronal], and contrastive features are assigned in an order placing [low] and [labial] ahead of [coronal]. In a three-vowel inventory only [low] and [labial] are contrastive, while in the four-vowel inventory [coronal] must also be contrastive to distinguish strong and weak i. It follows from these assumptions that [i] can trigger palatalization only if it is in contrast with a fourth vowel.

THE CHRONOLOGY AND STATUS OF ANGLIAN SMOOTHING

Publisher Summary This chapter discusses the chronology and status of Anglian smoothing. Smoothing is a name given to the process whereby, in certain dialects, diphthongs were monophthongized before the velar consonants k, g, and x, when these followed either directly or with an intervening r or l. The rule is somewhat unexpected, because some of these contexts were ones that had formerly caused diphthongization. Hence, smoothing appears to reverse the effects of earlier rules, and this leads to the suspicion that one or the other is not a natural process. The main concern lies with the results of the interaction of smoothing with other processes and the extent to which these interactions bear on the theory of lexical phonology and vice versa. Moreover, the rules limited to the earlier level all have in common that they can be treated as persistent constraints. The facts suggest that certain types of rules are susceptible to be reinterpreted as constraints, thereby exhibiting complex interactions with other processes.

Reply to Winograd

In our article “On some supposed contributions of artificial intelligence to the scientific study of language” (Cognition, 4. 32 l-398) we argued that much recent AI research into language is concerned primarily with the task of developing programs that can “understand” language in some limited domain and not contrary to the claims made by AI researchers with developing explanatory theories of language. We supported this contention with a fairly detailed review of some recent work by prominent AI researchers in language such as Winograd, Minsky, and Schank’. We showed that whatever the technological merits of this research, it does not yield explanatory principles of the sort which are necessary to the development of any scientific theory. Our point in this regard is a general one: we nowhere criticize this research on the grounds that it is directed at explaining phenomena which are inherently unworthy of scientific investigation. Rather, we maintain that the purported explanations being advanced are not really explanations at all. More specifically, in practically every case, the accounts given are, at best, based on programs or plans for programs which simply presuppose the very phenomena which they are meant to explain. Thus, at the very heart of these programs lie the most idiosyncratic facts concerning the limited domains dealt with. It is this feature that makes these programs inextendable in any principled way to problems lying beyond the very restricted domain for which they were originally designed. Hence, whatever successes these programs enjoy is a function of the limited domains to which they were tailored and not to any general principles which could serve as the basis of an explanation of the phenomena in these domains. Lacking general principles, this work cannot contribute to the articulation of a scientific theory

Quasi-phonemic contrast and the fuzzy inventory: Examples from Scottish English

In this article we propose that contrast must be treated as a gradient phenomenon at the phonological level, with membership of a phonemic inventory being a matter of degree. This is because, though minimal pairs provide simple and strong evidence of contrast, things are not always so straightforward. Defining “minimal” is one challenge; as is determining which aspects of a contrast are distinctive and which redundant. Non-phonological information is sometimes a necessary consideration. These complications are usually thought to affect the analysis of a phenomenon in a discrete way, tipping the binary balance held by the phonologist towards either one analysis or another. We, on the other hand, see the necessity of evaluating contrastive evidence and of taking other linguistic information into account as being an indication that contrastiveness is a scalar property. We address some patterns in the sound system of Scottish English; ones which provide less than clear evidence of phonemicity — or, as we think, evidence of less than clear phonemicity.

The influence of loanwords on Norwegian and English stress

Rice (2006) presents a unified analysis of Norwegian word stress that applies equally to native words and to loanwords. In this analysis, stress is oriented to the right edge of the word, which suggests that the loanwords were responsible for changing what was originally a left-oriented grammar of stress. In this paper I consider a similar reorientation that took place in the history of English, also under the influence of Romance loanwords. Closer examination shows that the two cases appear to be different. Many loanwords of the sort that caused a change in Norwegian entered Middle English without causing any significant change in English stress. It was only in the Early Modern English period that the loanwords were able to impose a right-oriented stress pattern on English. Rice (2006) observes that the loanwords were able to change the Norwegian stress pattern without overtly contradicting the native words; that is, the loanwords could make a change only in aspects of the grammar where the native words were ambiguous. I argue that this principle also accounts for the English case.

Contrastive feature hierarchies as a new lens on typology

We propose a way of looking at phonological typology that is based on a fundamental distinction between a phonetic and phonological analysis of the sound systems of languages. We build on approaches to phonology pioneered by Sapir and the Prague School (Jakobson and Trubetzkoy), instantiated within a generative grammar. We view phonemes as being composed of contrastive features that are themselves organized into language-particular hierarchies. We propose that these contrastive feature hierarchies shed light on synchronic and diachronic phonological patterns, and therefore offer a new lens on phonological typology. Thus, on this view the subject matter for typological investigation is not a phonetic sound (e.g., [i]) or a phoneme (/i/), or even a phonemic inventory (/i, a, u/), but an inventory generated by a feature hierarchy: for example, /i, a, u/ generated by the hierarchy [low] > [round]. This yields a different set of representations from the same terminal symbols generated by the hierarchy [round] > [low]. We will illustrate this approach to phonological representations with a synchronic analysis of Classical Manchu, and then show how it accounts for the results of typological surveys of rounding harmony in Manchu-Tungusic, Eastern Mongolian, and Turkic, and for the distribution of palatalization in Yupik-Inuit dialects. We will then propose that contrast shift should be recognized as a type of phonological change, and show how it applies to diachronic developments of the Algonquian and ObUgric vowel systems. We find that feature hierarchies can be relatively stable, but contrast shifts do occur, for various reasons, and these can result in dramatic differences in patterning. Harvey’s analysis of Ob-Ugric also shows that elements of feature hierarchies can spread and be borrowed, like other aspects of linguistic structure. As Sapir (1925) proposed, languages whose phonemes line up in similar ways (i.e., have similar contrastive feature hierarchies) show similar phonological patterning, though they may differ considerably in their phonetic realizations. We conclude that contrastive feature hierarchies provide an interesting level of representation for typological research.