Cognitive Linguistics - A Survey of Linguistic Subfields

Ewa Dąbrowska, Dagmar Divjak

• Publisher: De Gruyter Mouton
• Year: 2019
• ISBN:
  9783110623154
• Edition: 1
• Topic: Filología
• Subtopic: Lingüística

The chapters provide comprehensive surveys of the major subfields of Cognitive Linguistics. Apart from phonology, construction grammar and lexical semantics, the areas of language use, language acquisition and literary discourse are comprehensively presented.

Geoffrey S. Nathan

Chapter 1: Phonology

Note: This paper has benefited from comments by Patricia Donegan, José Mompeán-González and three anonymous reviewers.

1 Introduction

Phonology is a branch of Cognitive Linguistics which has developed more slowly than fields such as lexical semantics or morphology (Lakoff 1987; Brugman 1981; Lindner 1981; Langacker 1990; Rubba 1993). In the history of the journal Cognitive Linguistics, only eleven articles on the topic have been published, an average of fewer than one per year. This author made some preliminary suggestions in an early article (Nathan 1986), and has since expanded these explorations in various directions (Nathan 2007, 2008; Donegan and Nathan 2014). A number of others have also written Cognitive Linguistics-based works on phonology including Nesset (2008), Bybee (1999, 2001), and Pierrehumbert (2002), as well as the articles in Mompeán-González (2006).

The question of how speech sounds are stored, perceived and produced, which is the essence of phonology, is an old one. The earliest theorizing in what we would now call synchronic phonology can be traced back to the late nineteenth century, where two different strands emerged. One, originating with the work of Baudouin de Courtenay (1972), and to some extent Sapir ([1933] 1972), exemplifies the active processing, psychologically oriented view still current in Natural Phonology and in one version of Cognitive Phonology, while another, founded by Saussure ([1916] 1974) emphasizes the structuralist, oppositional/contrastive view of how speech sounds are stored, with little or no interest in how sounds are actually produced. Generative Phonology, classically defined in Chomsky and Halle (1968) constituted somewhat of a compromise between the contrast-only storage model introduced by Saussure and the active production model championed by Baudouin, and, to some extent, Sapir. In recent years generative phonology has mostly been succeeded by a computationally-based model, Optimality Theory (OT), that is somewhat neutral on the actual psychological mechanisms of production and perception and storage. OT, however, takes other aspects of how phonology works very seriously (particularly the tension between “naturalness” and minimization of divergence between underlying form and surface form − “faithfulness”) (Prince and Smolensky 2004; Kager 1999; McCarthy 2008).

This chapter will examine the active theoretical issues that arise in trying to understand how phonology would work within a cognitive linguistics worldview. We will discuss the question of the nature of the entities stored (the “underlying form” or “phoneme” question), the nature of the production and perception mechanism (how psychologically real are “rules”?) and the extent to which physiology and perceptual mechanisms constrain or influence the answers to the preceding questions (the “naturalism” question).

I will focus my discussion primarily on the question of why there might be a field of “phonology”. That is, why there might be general principles governing how the phonologies of individual languages are constrained, rather than being simply a massive list of individual variants of lexical items. The task, from a Cognitive Grammar (CG) point of view, is to find governing principles that are not attributable to an innate language organ, since CG has no analog to the generative notion of Universal Grammar to fall back on. Hence, every explanatory principle must be based on pre-existing cognitive principles. I will argue that such principles as categorization, embodied perception and motoric organization account for all that is “universal” in (phonological) grammar. I should briefly mention that I will not be discussing questions of “phonetics”, which some earlier scholars did not distinguish from “phonology”. I will assume that phonetics deals with “raw materials” (anatomy, physiology, acoustics, perceptual psychology) and can be a source of explanation for phonological facts, but is not itself part of “the structure of English (or any other language)”. I consider it the same as the contrast between materials science and architecture − the former constrains the latter, but is not identical to it.

2 Invariance, segmentation and storage of units: The fundamental issues

2.1 The problem of invariance and variation

It has been known since the beginning of linguistics that the same word is pronounced differently every time it is uttered. To take a concrete example, the English word banana might be pronounced [bə̃.ˈnæ̃.nə̃] or it can be pronounced [ˈbnæ̃.nə̃], and there are a number of intermediate possibilities as well. And, of course, it can be pronounced by men, women, boys, girls, speakers from Detroit − [ˈbnẽə̃nə̃]) or RP − [bə̃ˈnɑ̃nə], and each one is different. Yet, to all intents and purposes, these variations are not even perceived by naive native speakers. Yes, anyone can recognize that Suzy or Daddy or Pierre said banana, but a child acquiring the language never thinks that the words refer to different objects, and it would be beyond imagination to try to spell each version differently − nobody (other than trained phoneticians) could perceive enough of a difference to know which spelling to use when.

It has also been known since very early in linguistics that the same sound is not the same either, but instead varies according to strict rules or principles based on such parameters as the nature of the preceding or following sound, or the prosodic status of the sound in question. The variation that has come to be known as “allophony” is classically illustrated by cases such as the American English phoneme /t/, which can be pronounced as a voiceless aspirated alveolar stop in tall, an unaspirated equivalent in stall, a voiced alveolar tap in flatter, a voiceless glottal stop in flatten and nothing at all in an allegro pronunciation of Saturday. Not only do native speakers not normally perceive this variation, it takes many months of phonetic training to get them to be able to transcribe it reliably. Similar facts pertain to Parisian French /e/, whose allophones [e] in open syllables and [ɛ] in closed syllables are completely opaque to most younger speakers (as noted by their inability to differentiate which way to spell the sound, among the traditional spellings for [e], namely <é> or <et> versus the traditional spellings for [ɛ]: <ais, ait, aient> etc.). To take one further example, Mandarin allophones [h] (as in hao ‘good’) and [x] (as in heng ‘very’) are indistinguishable to even moderately phonetically aware native speakers. The fact that speakers of a language are unable to hear the difference between sounds that are not only phonetically distinguishable, but are contrastive in other languages I consider to be significant evidence that phonemic perception is real.

2.2 The nature of units and storage

The vast majority of phonologists (of any stripe) accept that human beings segment the speech stream into roughly phoneme-sized segments. A few British phonologists in the nineteen-forties and fifties suggested that there might be some “segments” that were longer than a single phoneme (so that contrastive features such as nasalization might span a longer stretch of speech), but even contemporary autosegmental-based phonological models assume that the initial (“underlying”, phonemic) string consists of segment-sized units (X-nodes or root-nodes), even if subsequent processing operates on larger stretches.1 The fact that the vast majority of even minimally phonetically oriented writing systems are segment-based,2 with a minority being syllable-based with segmental additions (Japanese Hiragana, Cuneiform Hittite) suggests that segment-sized entities are readily accessible to speakers of most languages.3

Some researchers over the years have suggested that the notion of speech as consisting of phoneme-sized units is an illusion brought on by the high level of literacy in (at least) the western world (see, e.g., Port 2010) and suggests that the existence of segments is not justified by the physically continuous nature of the speech signal, the absence of invariant cues for phonemes, and the fact that certain kinds of speech error do not seem to occur. Some of these issues will be dealt with below. Note however that storage of units larger than the segment does not preclude the notion that those units are made up of smaller units. The existence of puns, spoonerisms and taboo avoidance behaviors in some cultures suggest strongly that speakers have access to identities at the segmental level, and, as Stampe (1979) noted, this identity is not at the level of fine phonetic detail but rather at a more abstract, schematic level that is equivalent to the phoneme of the “psychological realist” school of phonology that includes Baudouin, Sapir and others. Just to take a simple case, Batman is an illustration of assonance, despite the fact that the two vowels are enormously different − one is short and oral, the other long and nasalized, yet this difference is a revelation to students in an introductory phonetics course.

The history of phonology has ranged widely over the possible answers to this question. Baudouin held that phonemes are mentally-stored sound images, but his colleague Saussure argued that they are systems of oppositions. European Structuralists such as Trubetzkoy ([1939] 1969) argued that phonemes were defined solely by their system of oppositions, so that if some sound was lacking in an opposition in some context, it wasn’t even a real phoneme, but an archiphoneme. So, for example, English [ŋ] in thank was different from the [ŋ] in thing because there was an opposition thing:thin but none in thank:*thamk. (So think would have been represented phonemically as /θɪNk/ but thing would be “spelled” /θɪŋ/.

Most American Structuralists were wary of archiphonemes. Hockett (1942: 10) for example, says they “confuse the facts without adding anything”, but within various streams of Generative Phonology underspecified segments have become quite popular (there are various flavors of “underspecification theory” that go beyond the scope of the current chapter).

Natural Phonology...

Table of contents

• Cover
• Title Page
• Copyright
• Contents
• Chapter 1: Phonology
• Chapter 2: Lexical semantics
• Chapter 3: Usage-based construction grammar
• Chapter 4: Discourse
• Chapter 5: Historical linguistics
• Chapter 6: Variationist linguistics
• Chapter 7: First language acquisition
• Chapter 8: Second language acquisition
• Chapter 9: Poetics
• Index