Early Word Learning

Gert Westermann, Nivedita Mani

• Publisher: Routledge
• Year: 2017
• ISBN:
  9781317550587
• Edition: 1
• Topic: Psychology
• Subtopic: History & Theory in Psychology
• Index: Psychology

Early Word Learning explores the processes leading to a young child learning words and their meanings. Word learning is here understood as the outcome of overlapping and interacting processes, starting with an infant's learning of native speech sounds to segmenting proto-words from fluent speech, mapping individual words to meanings in the face of natural variability and uncertainty, and developing a structured mental lexicon.

Experts in the field review the development of early lexical acquisition from empirical, computational and theoretical perspectives to examine the development of skilled word learning as the outcome of a process that begins even before birth and spans the first two years of life. Drawing on cutting-edge research in infant eye-tracking, neuroimaging techniques and computational modelling, this book surveys the field covering both established results and the most recent advances in word learning research.

Featuring chapters from international experts whose research approaches the topic from these diverse perspectives using different methodologies, this book provides a comprehensive yet coherent and unified representation of early word learning. It will be invaluable for both undergraduate and postgraduate courses in early language development as well as being of interest to researchers interested in lexical development.

1
BEFORE THE WORD

Acquiring a phoneme inventory

Titia Benders

DEPARTMENT OF LINGUISTICS, MACQUARIE UNIVERSITY, AUSTRALIA

Nicole Altvater-Mackensen

DEPARTMENT OF PSYCHOLOGY, JOHANNES-GUTENBERG-UNIVERSITY MAINZ, GERMANY

Introduction

One of the benchmark questions parents are asked about their child’s language development is ‘Does she already say some words?’. While the first word is very tangible proof that a child’s language development has taken off, that first word is part of a developmental trajectory including babbling, word comprehension and—one of infants’ very first accomplishments on the way to the first word—the acquisition of language-specific sound perception. The current chapter reviews the main experimental findings and theoretical positions on infants’ acquisition of the native phoneme system. In particular, we will discuss how speech perception and production evolve in the early stages of language development and how speech perception shows continuity with later word learning. We first introduce general developments in speech perception and production in the first year of life and present the main theoretical views on the developmental relationship between perception, production and word learning. The second and third parts examine the acquisition of particular contrasts in speech perception, speech production and word learning, highlighting the parallels in development across domains and discussing their potential interplay. The chapter will conclude with avenues for future research.

Speech perception and production in the first year of life: The main developmental patterns

In one of the first studies investigating infant speech perception, Eimas et al. (1971) found that infants as young as 1 month of age perceive speech sounds categorically, i.e., they perceive acoustic differences that cross a category boundary better than within-category differences. This categorical perception of speech allows listeners to attend more to acoustic differences that are meaningful in the ambient language. Interestingly, monkeys and birds also perceive speech categorically (Dooling et al., 1989; Kuhl & Miller, 1975), suggesting a cognitive mechanism that partially results from the structure of the auditory system. Follow-up studies showed that infants are initially sensitive to both native and non-native sound contrasts but lose sensitivity to the latter between 6 and 12 months of age (Werker & Tees, 1984). The resulting view was that infants are born as “universal” listeners who categorically perceive all sound contrasts from the languages of the world, and then become “native” listeners in the first year of life who are only sensitive to the phonemic sound contrasts of their native language (Werker & Tees, 1999). Yet, not all sound contrasts follow this developmental path. Infants and adults retain their ability to discriminate non-native sound contrasts if the contrasting sounds are assimilated into distinct native phoneme categories or are very different from any native sound (Best et al., 2001; Best et al., 1995). In addition, some native contrasts are initially not discriminated optimally and infants’ sensitivity to these contrasts improves in the first year of life (Kuhl et al., 2006). Specifically contrasts that are acoustically less salient initially pose problems for younger infants. For instance, the discrimination of the (native) fricatives /f/, /∫/ and /θ/ is difficult for English babies (Eilers et al., 1977), and they have more difficulty discriminating differences in voicing for negative than positive voice onset times (Aslin et al., 1981). Changes in perception are further influenced by phoneme frequency, as sensitivity to sound contrasts that are initially difficult to perceive improves earlier for more frequent than less frequent sound contrasts (Anderson et al., 2003). This maintenance, facilitation and loss of sound contrasts in perception as a result of early experience with the native language is often referred to as perceptual attunement (Aslin & Pisoni, 1980).

In contrast to their elaborate perception capacities, infants are born with very limited speech production abilities. Their earliest vocalizations are vegetative sounds and crying, soon followed by an increasing use of vowel- and consonant-like sounds during cooing (Stark, 1980). This cumulates in canonical babbling, i.e., in the production of speech-like consonant-vowel syllables, starting around 6 months of age (Oller, 2000). Early babbles are repetitions of the same consonant-vowel syllables (e.g., bababa) and then evolve into variegated babbles with different syllables in one vocalization (e.g., wubade). Even though the prosody of pre-speech vocalizations is influenced by the native language (De Boysson-Bardies et al., 1984; Mampe et al., 2009), the segmental content and structure of early babbles are highly similar across languages (Wermke et al., 2013). The productive segmental repertoire is initially mainly restricted to labials, coronals and stops, and gradually expands along the lines of the native phonological system that is used in the production of early word forms (De Boysson-Bardies & Vihman, 1991, and references therein). The initial state is shaped by early speech motor control, and subsequent development is thought to be driven by the maturation of speech motor control (Davis & MacNeilage, 2000) and native language input (De Boysson-Bardies et al., 1984).

Speech perception and production thus appear to develop in opposite directions, in that infants are born with sophisticated perceptual but limited productive abilities and both are tuned to just the abilities that are required for their native language. Yet, infants’ babbling starts to show native-language characteristics around the same time that their speech perception attunes to the native language. This parallel development across domains has led to the proposal that productive and perceptive abilities develop in concert, as we will discuss next.

The interplay between speech perception and production in native language attunement

Many theories addressing the development of speech sound categories focus on perception rather than production (cf. Werker et al., 2012). The implicit assumption appears to be that speech production follows perception, as is often the case in language acquisition (Fenson et al., 1994). This assumption is explicit in Kuhl et al.’s (2008) Native Language Magnet Theory-expanded, which states that infants imitate the sounds they hear and in doing so create links between their perceptual and motor representations. Best’s (1991) Perceptual Assimilation Model envisions a more direct role for production, as it assumes that perception and production refer to the same gestural categories and are thus interdependent (see also motor theory of speech perception, Liberman et al., 1967, and direct realism, Fowler, 1986). Despite these theoretical stances about the relation between perception and production in phoneme acquisition, only a few studies so far provide experimental evidence that perception and production are related in development.

In a series of studies, Vihman and colleagues found that infants’ babbling patterns are related to their speech preferences (Majorano et al., 2014, and references therein). When infants initially produce word-like babbles, they prefer to listen to sound sequences that they master themselves; when their repertoire expands, their preference in perception shifts to sound sequences outside of their abilities. These findings comply with Vihman’s (1996) Articulatory Filter Hypothesis, which states that infants’ speech perception is influenced by motoric recall, i.e., that patterns from the infant’s own babbling stand out in the perceived speech stream.

Motoric experience also directly influences audio-visual speech perception in younger infants. Concurrently performed lip movements influence 4.5-month-olds’ ability to match auditory and visual speech cues (Yeung & Werker, 2013), and 8-month-olds are better able to map (non-native) sounds to the corresponding mouth gesture if these sounds occur often rather than rarely in infants’ early babbling (Mugitani et al., 2008). These results might merely reflect the impact of non-linguistic motoric experience on perception, but could also be indicative of a connection in development between the sound representations for production and perception. Such a link is, for example, suggested by Westermann and Miranda’s (2004) model that couples motoric (and visual) patterns with their auditory consequences during the babbling period. The resulting cross-modal sound representations are shaped by self-generated babbles as well as by ambient speech sounds. The constant coupling of auditory, visual and motoric patterns leads to a tight perception–production link in development.

Recent neurobiological evidence confirms that infants learn to recruit articulatory information during speech perception. Infants between 6 and 12 months of age increasingly activate motor areas in the brain during speech sound discrimination, whereas such activation is absent in newborns (Imada et al., 2006). And while 7-month-olds show similar activation in auditory and motor areas for native and non-native speech sounds, 11- to 12-month-olds show distinct activation patterns depending on the type of speech they hear (Kuhl et al., 2014). These results suggest that the perception of speech triggers activation of the motor system, but only after substantial (native) language experience.

The link between speech perception and early word learning

The literature agrees that the main factor driving native language attunement in perception and production is the ambient language input, but accounts largely differ when it comes to the role of words. Theories of phoneme acquisition originally entertained the idea that a lexicon is necessary to acquire which sounds contrast meaningfully in a language. According to this view, children represent their first words in a holistic, unsegmented fashion. The system of phonemic contrast results from the demand to learn and unambiguously represent similar-sounding words, and phoneme acquisition is a slow process relying on a crucial mass in the lexicon (Walley, 2005, for perception; Waterson, 1971, for production). The idea of roughly represented words was challenged by the finding that infants are sensitive to subtle phonological changes to words from early on (Swingley & Aslin, 2000, and subsequent studies). The findings discussed before showing that infants undergo native language attunement within the first year of life further suggests that at least some aspects of phoneme learning take place before the child has acquired a reasonably large lexicon. Theories now had to explain how phoneme acquisition takes place in the absence of a (large) lexicon.

One widely-accepted answer to this conundrum was statistical learning. Every realization of a phoneme sounds slightly different. The resulting phonetic variation reveals the underlying phoneme structure, as the distribution of speech sounds along a (relevant) phonetic continuum has one local maximum per phoneme (Allen & Miller, 1999). Such distributions of exemplars indeed modulate infants’ discrimination of speech sounds (Maye et al., 2002). The idea of distributional learning as the primary mechanism underlying infants’ early acquisition of language-specific phoneme perception has been embraced across theories of early speech perception (Werker & Curtin, 2005; Kuhl et al., 2008). However, computational modeling of phoneme acquisition suggests that sound categories can only be learned successfully when the word context of the phonemes is considered (Feldman et al., 2013). Behavioral evidence supports the modeling results, as familiarity with two different word forms that each contain a specific sound, such as /d/ in dog and /t/ in teddy, enhances infants’ perception of this sound contrast (Thiessen, 2011; Thiessen & Yee, 2010). Similarly, consistently pairing minimal pair syllables with visual objects or facial information fosters the perception of the phonological contrast involved (Mani & Schneider, 2013; Teinonen et al., 2008; Yeung & Werker, 2009). Since infants have basic knowledge about some word forms and their referents by 6 to 9 months of age (Bergelson & Swingley, 2012; Parise & Csibra, 2012; Tincoff & Jusczyk, 2012), the lexicon may indeed support native language attunement. This stance is reflected in interactive views on the development of speech perception and word learning (e.g., Werker & Curtin, 2005; Swingley, 2009). Note, however, that these accounts so far do not cover processes of speech production.

Vowels versus consonants

The perception of different sound contrasts is attuned to the native language at different rates, with the trajectories of attunement to vowels and consonants being the most obvious example. While the perception of vowels becomes language-specific around 5 to 6 months of age (Polka & Werker, 1994; Kuhl et al., 1992), the development of language-specific consonant per...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Dedication
• Table of Contents
• List of contributors
• Preface
• 1 Before the word: Acquiring a phoneme inventory
• 2 The proto-lexicon: Segmenting word-like units from the speech stream
• 3 Intrinsic and extrinsic cues to word learning
• 4 Mapping words to objects
• 5 Building a lexical network
• 6 Verbs: Learning how speakers use words to refer to actions
• 7 Listening to (and listening through) variability during word learning
• 8 Individual differences in early word learning
• 9 Early bilingual word learning
• 10 ERP indices of word learning: What do they reflect and what do they tell us about the neural representations of early words?
• 11 Computational models of word learning
• Index