Input-based Phonological Acquisition

Tania Zamuner

• Publisher: Routledge
• Year: 2004
• Print ISBN: 9780415967891
• eBook ISBN: 9781135885090
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This book provides an analysis of two theories of language acquisition: the theory that acquisition is primarily mediated by innate properties of language provided by universal grammar, and the opposing theory that language is acquired based on the patterns in the ambient language. A problem not often considered is that these two theories are confounded because the structures that are frequent across languages are also typically the most frequent within a specific language. In addition, the innate theory of language acquisition is difficult to quantify and qualify. Using cross-linguistic, corpus and experimental approaches, this book attempts to contrast these theories through an examination of the acquisition of word-final consonants in English.

CHAPTER 1
Accounts of acquisition

Universal Grammar and the input

“The Simpsons”

Herb: It’s a “Baby Translator.” It measures the pitch, the frequency and the urgency of a baby’s cries. Then it tells whoever’s around, in plain English, exactly what the baby’s trying to say. Everything from “Change me,” to “Turn off that damn Raffi record.”
* * * * * *

Herb: Maggie, who brought me my fortune. Just name anything you want in this world.

Maggie: Blah blah blah.

Translator: I want what the dog’s eating.

Homer: D’oh!

1.1.
Focus of research

The invention of a baby translator presupposes that children make systematic errors in their speech that reflect a system of organisation. Although this dissertation does not provide a blueprint for a baby translator, it does ask how children’s productions can help us understand how children interact with the ambient language, and it thereby explores the mechanisms that underlie children’s ability to successfully acquire language.

The primary question asked in this dissertation concerns the nature of language acquisition. In particular, I contrast two views. One view, which has been dominant in generative linguistics, is that language acquisition is best characterised as the unfolding of highly specified innate abilities (Chomsky, 1981). On this view, children come to the language-learning task with innate knowledge of how language is structured, and the input serves only to trigger innately specified options. The opposing view is that language is acquired based on the patterns in the ambient language or the input (Olmsted, 1966, 1971). On this account, language acquisition is characterised as a domain-neutral learning process in which frequently occurring properties of the input serve to organise children’s linguistic representations. I contrast these theories by examining young children’s acquisition of coda consonants.

This chapter focuses on nature versus nurture accounts for the observation that children’s productions mirror cross-linguistic markedness (to be defined later), and describes the case employed here to test these hypotheses. To begin, I describe the noted parallel between cross-linguistic markedness and child language. The primary account of this relationship relies on the Chomskian notion of innate universal grammar (UG). I define this theory of grammar acquisition as the Universal Grammar Hypothesis (UGH). I demonstrate that the data taken to support this account are often confounded, given that the distributions of sound patterns in a single language often encode what is argued to be innately provided by UG. I further describe a subset of research that illustrates children are sensitive to the frequency of sound patterns in the input language. The view that children acquire language according to the input is introduced as an alternative to the UGH, and defined as the Specific Language Grammar Hypothesis (SLGH). The acquisition of coda consonants by children acquiring English is presented as a means for distinguishing between the two hypotheses. Subsequently, the latter half of Chapter 1 introduces this dissertation’s focus: the test case of word-final codas, which allows for the examination of the relative roles of innate knowledge versus input frequency in phonological acquisition.

1.2.
Universal Grammar

One problem in discussing UG is that there is little clarity in the term’s use and meaning. Moreover, the definition and interpretation of “universal” and “universal grammar” differs across researchers and across the sub-fields of linguistics, such as in phonology and syntax. The following discussion will centre on two different characterisations of UG.

1.2.1.
Establishing Universal Grammar

There are two central interpretations of UG. The first interpretation is that UG contains the set of properties present in all languages. This set comprises unrestricted or absolute universals. For example, all languages have CV syllables (Spencer, 1996, p.82). Unrestricted universals refer exclusively to the necessary properties that define all languages or what are also called the “unmarked” properties of languages. This is formally represented in (1) (adapted from Hammond, Moravcsik & Wirth, 1988), and defined as: for all languages, Y.

See Table

All languages have CV syllables (Type 1); moreover, there are no languages without CV syllables (Type 2). Unrestricted universals are by definition based on cross-linguistic research.

The second interpretation of UG is that it defines the limits of human language and the extent to which languages can vary. Thus, UG contains and is defined by what is unmarked and marked in language. The term “unmarked” refers to the properties of language that are common and frequent (which can be different from what is frequent in any one language), whereas “marked” defines properties that are less common and less frequent. Various methods are used to determine the extent to which languages can vary, or to determine what is unmarked and marked in languages. This can be based on restricted universals, language processes, language change, patterns seen in child language and aphasia, and on the distribution of sounds in languages, all of which are discussed below. Moreover, the cross-linguistic markedness factors on which UG is defined differ according to different theoretical perspectives. For the purposes here, patterns established by all methods are taken as evidence for cross-linguistic markedness, and in turn, as evidence for UG.

Some argue that markedness is necessarily based solely on restricted universals (also called implicational universals or typological universals). These are defined as: for all languages, if X, then Y (Greenberg, 1963). For instance, if a language possesses obstruent codas, then it possesses sonorant codas. The logic is schematised in (2) (adapted from Hammond et al. 1988).

See Table

In other words, the presence of obstruent codas entails the presence of sonorant codas; no languages exist with only obstruent codas (Type 4 languages do not occur). The unmarked property is implied by the marked one (Greenberg, 1966); thus, sonorant codas are unmarked and obstruent codas are marked.

Markedness claims are also based on language processes and language change (Greenberg, 1974). For example, in neutralisation, one feature will be neutralised more often than the other feature, such as in final devoicing (Greenberg, 1966). In this case, laryngeal features are restricted in coda position (see Lombardi, 1991). In Catalan (3) and Turkish (4), voiced obstruents become voiceless in word-final position. For example, in Catalan, the /b/ in /ʎobə/ becomes [p] word-finally.

See Table

The feature that is most likely to be neutralised is argued to be marked, and the resulting feature is unmarked. With syllable-final devoicing, voiced stops are marked, subsequently becoming voiceless, which are unmarked.

Markedness claims are also based on patterns seen in child language acquisition and pathology, where children and aphasic adults are argued to produce the least marked structures in language (Blumstein, 1973; Jakobson, 1941/1968). Children and aphasics delete codas to produce CV syllables, which are also the least marked syllables cross-linguistically. This adds circularity to arguments of language acquisition that are based on UG.

Lastly, there are statistical, substantive or distributional universals (Greenberg, 1954, 1966, 1974; Hammond et al. 1988; Trubetzkoy, 1939/1969; Zipf, 1935/1965). These types of generalisations are not absolute, but rather, they describe languages’ tendencies to pattern in a particular way. This originated with the Prague school of linguistics, which first identified the relationship between markedness and frequency in phonology (Trubetzkoy, 1931, cited in Greenberg, 1966, p.11). This school stressed the importance of statistical information in language from a variety of sources, such as dictionaries and texts. The rationale for considering frequency with respect to UG is that frequency is generally related to markedness across languages (Greenberg, 1974), and markedness is translated as defining the properties of UG. Not only is markedness connected to frequency across languages, but it is also encoded in the frequency of sound patterns within a single language. (Frequency has also been argued to be not formally related to markedness, Leuninger, 1983, p.35.)

To summarise, there are two primary interpretations of UG. The first is based on unrestricted universals and refers solely to the unmarked properties of language. The second definition for UG is that it defines the possible limits of human language and refers to both the unmarked and marked properties of language. Markedness is established by a variety of means, such as by restricted universals and frequency. The characterisation of UG that is relevant for language acquisition is subject to debate; however, constant across all theories is that UG minimally provides to children the unmarked properties or settings for language, as part of children’s innate genetic endowment.

1.2.2.
Universal Grammar and child language: The parallel between cross-linguistic markedness and child language

The view that language acquisition is mediated by UG stems from Jakobson’s theory of acquisition, within the framework of the Prague School of Linguistics. Jakobson’s theory illustrates how adult language patterns can be extended to acquisition. For Jakobson, acquisition is a process of acquiring contrasts in meaningful productions. In this sense he makes a distinction between babbling and communication—word use marks the beginning of the development of a phonological system. Children’s acquisition of contrasts is governed by the same “general laws of irreversible solidarity” that govern adult language or the laws that govern the phonemic systems of languages. Thus, Jakobson makes the important observation that there are similarities between child language and adult language. He further notes that the order of acquisition can be predicted from cross-linguistic implicational universals. This order is invariant across children arguably because the knowledge is innate, however, the speed at which children acquire contrasts can vary. Roughly, Jakobson predicts that the first opposition to be acquired is between a consonant and a vowel. From there, finer distinctions are made and further contrasts are built.

To illustrate the parallel between adult language and child language, take the example of syllable structure. Languages of the world exhibit certain universal patterns. For example, the universally unmarked syllable shape is CV, such that all languages permit CV syllables (Blevins, 1995; Clements & Keyser, 1983; Greenberg, 1978). The least marked languages permit only this syllable type (CV) and do not permit coda consonants, e.g., Arabela, Cayuvava (5), Hawaiian (6), Hua, Kikuyu, Mazateco, Sesotho, Swahili, Tongan and Twi.

See Table

In these languages, coda consonants are not permitted, consequently all syllables are of a CV (or V) shape. In Hawaiian this effect is mirrored in English loan words. Words with syllables that do not conform to CV syllables shapes are modified (7).

See Table

For example, the English word “croquet” begins with two consecutive consonants /kr/. In Hawaiian, which does not allow consonant clusters, /kr/ is modified through epenthesis of [o] between the illicit /kr/ sequence, producing [koloke:]. (The change of /r/ to [l] results from other processes.)

Interestingly, children’s early word productions are largely characterised by this same open syllable shape. Young children initially delete or modify their productions of codas or other marked syllable structures to conform to CV shapes as in (8) and (9). For example, a child attempting to produce the final / n/ in “man,” might delete /n/ and produce [mæ], [mε], or[ ].

See Table

From these examples, we can see that children initially delete coda consonants; thus, their productions conform to an onset-peak or CV shape. Another strategy children use to avoid producing codas is to epenthesize a vowel after final consonants, e.g., CVC ➔CVCV, /tʃɪk/ ‘chick’ is produced as [dέdi] (PJ 23 months, taken from Demuth & Fee, 1995). Also see Fee and Ingram (1982). One interpretation of this phenomenon is that final consonant deletion results from innate pressures to conform words to the preferred CV shape. Thus, these errors have been traditionally interpreted as a reflection of innate UG (e.g., Demuth & Fee, 1995; Faingold, 1990; Jakobson, 1941/1968; Levelt, Schiller & Levelt, 2000; Levelt & Vijver, in press; Mowrer & Burger, 1991; Ohala, 1996, Vihman & Ferguson, 1987 and others).¹

The observation that children’s productions mirror cross-linguistic markedness has fuelled the hypothesis that substantial parts of linguistic structure are innate and that language acquisition is mediated by UG. The 6 Input-Based Phonological Acquisition standard linguistic argument is that the parallel between cross-linguistic markedness and child language is the effect of UG on both systems.

1.2.3.
Universal Grammar Hypothesis of language acquisition

Within generative theories of linguistics, the unmarked properties of language or the core grammar (Chomsky, 1981) comprise children’s initial hypotheses about language. Thus, children’s initial CV productions are explained by appealing to the CV syllable as the unmarked syllable type or state. As children’s grammars develop, they move away from the core grammar towards a language-specific grammar. Consequently, children produce more marked syllable structures.

Although markedness and UG are hypothesised to motivate patterns seen in child language, patterns in acquisition do not always follow UG. Children do not necessarily initially produce what is unmarked and acquire marked structures last (Hyams, 1986). For instance, although /l/ and /r/ are preferred consonants cross-linguistically, these sounds are difficult for children to perceive and produce (Poole, 1934; Welleman, Case, Mengert & Bradbury, 1931). Given that children exhibit perceptual and production constraints that differ from adults (as with /l/ and /r/), this raises an interesting question of how to reconcile these differences, because factors other than UG are evidenced in acquisition, and domains can interact with one another. These are complex issues that need to be addressed. As a starting point, however, I take the standard approach assumed in the acquisition literature, where adult language patterns are seen as reflecting the properties of UG.

In summary, the hypothesis that language acquisition is guided by innate principles of language structure, which are provided by UG, is defined as the Universal Grammar Hypothesis (UGH) (10):

(10) Universal Grammar Hypothesis (UGH): Children acquire language based on innate properties of language, which guide children’s acquisition of phonology.

Recall that UG is evidenced through cross-linguistic markedness effects. Thus, parallels between child language and cross-linguistic markedness can be represented as the product of UG. Because UG is innate, it will be represented in all human languages (as seen in cross-linguistic markedness). Similarly, UG is innately provided to children; thus, children’s productions reflect this genetic knowledge. Interestingly, markedness is seen as both as evidence for UG and as the product of UG.

1.3.
The input

This section centres on the alternative to the UG hypothesis for language acquisition, which is that language acquisition is best described with respect to the patterns in the input or ambient language. This hypothesis is in part motivated by the large body of research that illustrates children are sensitive to the frequency of patterns in the ambient language.

1.3.1.
Child language reflects the language specific input

There is considerable evidence that child language reflects the language-specific input at an early age. For example, nine-month-old infants show a preference for their language’s more common sound patterns (Jusczyk, Luce & Charles-Luce, 1994), and around 10 months, infants begin to ignore phonetic contrasts that are not phonemic in the language they are acquiring (Werker & Tees, 1984). This sensitivity to the input is also evident in young children’s pre-linguistic vocalisations, where infants’ babbling reflects the ambient language (see review by Vihman & Boysson-Bardies, 1994).

The effects of the input or the ambient language are further mirrored in children’s linguistic productions, at both the segmental and syllable structure levels. For example, high frequency sounds in the ambient language are more likely to be produced by children than are low frequency sounds. The palato-alveolar affricate /t∫/, which has a low frequency in English, is acquired late; however, in Spanish, /t∫/ has a high frequency and is acquired earlier (Macken, 1995). Similarly, results from Pye, Ingram and List (1987) showed that children’s initial consonant productions in Quiché and English were significantly correlated with the frequency of initial consonants in their respective languages. As in Spanish, the affricate /t∫/ is acquired early in Quiché, and the early acquisition of /t∫/ is argued to be linked to its frequency in child directed speech. In Finnish /d/ is acquired late, unlike in other languages, and this is hypothesised to reflect the phoneme’s low frequency in Finnish (Macken, 1995). Word initial /v/ is acquired earlier in Spanish, Estonian, and Bulgarian than in English, arguably because of the phoneme’s high frequency in the former languages (Ingram, 1988).

At the prosodic level, Levelt, et al. (2000) and Levelt and Vijver’s (in press) examinations of Dutch children’s acquisition of syllable types illustrate that children who learn languages with a complex or marked syllable structure must choose how to build syllable structure beyond the initial core CV syllables (arguably children choose between different learning paths given by UG). Faced with choice, children first acquire the syllable structure that is most frequent in Dutch. Another example, from the acquisition of coda consonants, demonstrates that children learning English produce codas before children learning Spanish (Demuth, 2001; Gennari & Demuth, 1997). It is suggested that this can be attributed to differences in coda frequencies in the two languages; codas are more frequent in English than in Spanish (Roark & Demuth, ...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Outstanding Dissertations in Linguistics
• Acknowledgements
• Preface
• Abstract
• Chapter 1 Accounts of acquisition
• Chapter 2 Cross-linguistic codas
• Chapter 3 English codas
• Chapter 4 Child language codas
• Chapter 5 Experiment 1
• Chapter 6 Experiment 2
• Chapter 7 Coda acquisition
• Appendix I Languages used in cross-linguistic analyses organised by language
• Appendix II Languages used in cross-linguistic analyses organised by language family
• Appendix III EFA for cross-linguistic place of articulation in coda position
• Appendix IV AFA for cross-linguistic place of articulation in coda position
• Appendix V EFA for cross-linguistic sonority in coda position
• Appendix VI AFA for cross-linguistic sonority in coda position
• Appendix VII EFA for coronal place of articulation in English
• Appendix VIII AFA for coronal place of articulation in English
• Appendix IX EFA for sonority in English
• Appendix X AFA for sonority in English
• References