Language-Specific Phonetic Perception in Infants - Literature review Example

Literature Review: Language-specific phonetic perception in infants. Investigating infant speech perception It is very difficult to understand exactly how human infants learn their first language and there are many theories which attempt to explain this phenomenon. The main difficulty is that researchers in this field have to devise methods for experimentation which expose infants to the particular linguistic and acoustic phenomena under investigation and they also have to devise sophisticated methods of inferring what the infant perceives without the kind of evidence that would be present in older individuals. Very young infants cannot speak, for example, and so production of particular sounds cannot be used as evidence to use in working out what sounds are perceived. Other methods such as the length of attention an infant pays to a sound, or variation in sucking rate, or head-turning are used. Despite this methodological challenge, considerable progress has been made in recent years in working out how the speech perception of infants develops over the first year of life. This literature review examines the most recent findings which have been made in relation to infant speech perception. Houston (2011) explains in a very clear and useful introductory chapter on this topic that a theoretical breakthrough arrived in the middle of the twentieth century through the seminal work of Chomsky (1968, 1975). Chomsky (1968, 1975) postulated that there must be an innate and universal device within the developing infant brain which enabled the child to make sense of the speech that he or she was exposed to in the family environment. Language involves not only perception of the sounds that are in the environment but also some kind of processing that makes sense of these sounds. Studies on the speech perception skills that infants have usually accept that there is some sort of innate capacity to acquire a language but there is still a matter of some debate over the relative importance of a Chomskian language acquisition device and other, more general learning mechanisms in infant speech perception and factors such as exposure to the native language, different kinds of affect, and the role of caregivers. In order to explore these related questions in more detail, Houston (2011) breaks down the speech perception skill into smaller parts, starting first of all the kind of perception that a foetus has in the womb. Speech sounds are attenuated at this stage, but it is clear that even before birth, “the sound that does reach fetuses provides sufficient information for perception of suprasegmental aspects of speech” (Houston, 2011, p. 48). Linguists differentiate between the segmental features of speech, such as phonemes which are generally defined using the International Phonetic Alphabet, and suprasegmental features such as rhythm, intonation and various subtle modifications of speech such as whispering. The suprasegmental aspects are particularly useful for conveying affect, allowing a speaker to differentiate between certain statements and uncertain statements or questions, or to convey anger, emphasis, sadness or any number of other attitudes and emotions. Phonemes on the other hand tend to be the bearers of meaning, and they are essential for basic intelligibility. It is thought that early encounters with suprasegmental language features in the womb encourage the foetus to encode at least some of these characteristics in the memory, and that this in turn influences later language learning after birth. Evidence for this very early ability to notice suprasegmental language features has been found in the apparent ability of infants to distinguish their mother’s voice from other voices, and their native language from other languages (Houston, 2011, p. 48). It appears that over time infants gradually attune to their native language and they learn to filter out sounds from their perception repertoire which are not required in their native language. This explains why older learners of second languages often have difficulty learning new phonemes which are not present in their native language. Experiments on infants in the first four to six months of life have confirmed that pitch, and especially intonation is the feature that is most relevant to their early perception of language. There is a preference for Infant-directed speech (IDS) over adult-directed speech (ADS) in this period of life. IDS is the structurally simplified, acoustically slower and usually very clearly enunciated kind of baby language that mothers and other adults use when speaking to very young children. IDS makes no allowances for infant perceptual limitations and is generally much faster, running segments together and presenting considerable difficulty in terms of perception. Linguists believe that that the reason why infants prefer IDS may not be due so much to the slower pace and greater clarity of IDS but that “infants prefer IDS because it generally conveys a positive affect, which is carried primarily through intonation” (Houston, 2011, p. 48). Positive affect means feelings of being cared for, loved and protected and this is a feature of the normal language that takes place when a mother speaks to her child. Further tests on infants up to the age of 8 months which used rate of sucking on a pacifier or head movements to detect infant ability to discriminate between different phonemes and syllables have established that “while infants may have some initial auditory sensitivities to particular acoustic-phonetic cues, these sensitivities are not rigid and can be influenced by linguistic input” (Houston, 2011, p. 50). This is evidence that innate capacity as well as social and linguistic both play a role in infant speech perception and it also suggests that it may be difficult for researchers to separate out cognitive, linguistic, social and emotional dimensions in this evolving phenomenon. Studies on infant speech discrimination across languages have produced the important finding that infants are able to distinguish a very wide range of segmental features from birth, but that at around the age of ten months or so, this range narrows down to those which are found in the native language. By the end of the first year of life, the ability to perceive sounds which do not occur in the native language is markedly reduced. Houston (2011, p. 52) qualifies this finding, however, with evidence which shows that this should be seen as a general tendency and not as an absolute outcome, due to the significant variations in speech perception between six and twelve months that have been observed in experimental work. Further work on infant ability to distinguish the beginnings and ends of words, phrases and clauses through phonotactic rules shows that the patterns of the native language are imprinted early in the memory. Suprasegmental features are already perceived in the womb, but by the age of six months, the infant begins to acquire the ability to perceive segmental features also. By nine months the evolving cognitive ability of the brain enables the infant to start using other means such as statistical and associative learning in order to make sense of speech (Houston, 2011, p. 58). At around six months it is suggested that infants could segment words from fluent speech “but only when preceded by a word they already knew” (Houston, 2011, p. 59). At later ages other cognitive aids come into play and the child processes fluent speech in his or her native language much more effectively. In summary, therefore, Houston’s (2011) review article explains what is currently known about how, and at exactly what age, and in what order the hearing instrument of the very young infant gradually becomes attuned, first to suprasegmental and then increasingly also to segmental features of speech. The universal listener hypothesis One of the questions which has proved problematic for researchers has been the need for a theory to explain exactly how and when young children start to focus on the phonetic distinctions which are required for their native language, and conversely how and when they lose their ability to process sounds which are not relevant for their native language. Kuhl et al. (2006) explore some of the theories that have been advanced to explain this process. The selection theory posits that infants are born with an innate ability to learn all human languages, but that at 6-8 months, or according to some of the empirical literature, 8-10 months, they select those features which are most relevant to their native language. This theory is also known as the universal listener hypothesis. As the infant becomes attuned to the native language, his or her perception of non-native segments declines during this first year of life. Another possible model is the so-called “Neural Language Magnet (NLM) or Neural Commitment (NC) model which “argues that, between 6 and 12 months of age, learning of the acoustic and statistical regularities of ambient speech alters neural tissue; neural connections responding to ambient regularities are strengthened” (Kuhl et al., 2006, p. F14). This theory implies that there is a biological component which facilitates the encoding and remembering of native language sounds. In order to test which model best explains infant behaviour, a Kuhl et al. (2006) conducted a cross language study with infants who had Japanese or English as their native language. The distinction between the phonemes /l/ and /r/ was chosen because distinction between the two is a feature which is essential to English language but not to Japanese. The results show, as predicted, that the American children improve in their performance of this distinction at 6-8 months whereas the Japanese speaking children exhibit a decline. The American children need to make this distinction in order to distinguish the meanings of minimal pairs such as lap and rap or look and rook, whereas Japanese does not have minimal pairs for these two sounds /l/ and /r/. The decline in the Japanese-learning infants takes place because there is no need to use this distinction. They lose the ability to make the distinction because it is not relevant to their needs. Similar results were achieved with Chinese/English and Taiwanese/English pair groups with different sounds (Kuhl et al., 2006, p. F18). The authors conclude that there is “a double dissociation between native and non-native speech perception. According to NLM/NC, native-language phonetic learning enables word learning; moreover, native-phonetic facilitation and non-native decline are linked (Kuhl et al., 2006, p. F18). This study (Kuhl et al., 2006) has some limitations, including the lack of consideration of variables such as the effect of social interaction or of varying amounts of infant and/or adult directed speech but it does appear to prove directional asymmetries in the acquisition of native and non-native consonant contrasts, and it suggests that there may be some biological component in the development of speech perception in the form of neural changes. Whether these neural changes are a cause or an effect of the variations observed in this work is, however, still an open question and it is entirely possible that they could represent both cause and effect, depending upon the timing of observations and some mutual reinforcement of biological and social factors. It seems likely that innate ability, exposure to native language speech, social and/or affective factors and neural changes over time all impact upon on infant speech perception and that there may be local and/or individual differences in the timing and extent of these effects. A recent study by Narayan et al. (2010) builds on this evidence relating to the reorganisation of speech perception during the first year of life and considers the hypothesis that “acoustic salience also affects the perceptual reorganization that takes place in infancy (Narayan et al., 2010, p. 407). This was tested in a cross language study with English-learning infants and Filipino-learning infants using two nasal place distinctions, namely the labial-alveolar [mɑ] and [nɑ] distinction and the alveolar-velar [nɑ] and [ŋɑ] distinction. Differences in the extent to which the infants could distinguish these pairs, and in the timing of changes in their ability to do this, were found. This is in line with previous studies on the subject and contributes more evidence which allows linguists to trace how an infant’s speech perception becomes attuned to the native language. It suggests that at least two factors influence discrimination capabilities, namely acoustic salience and exposure to the native language. A further dissertation length study by Krieger (2012) examined the ability of English-learning 6-10 month old infants to discriminate between voiced and voiceless stops, which are lexically contrastive in English, and between voiced and pre-voiced stops, which are not lexically contrastive in English. The results were rather surprising. The six-month-old infants could discriminate between voiced and voiceless stops, but not between voiced and pre-voiced stops, which is to be expected according to the rule discussed above that infants attune to their native language. If the universal listening hypothesis is true, then it might be expected that these infants at this early age should be able to discriminate between both pairs. The fact that the infants did not discriminate suggests that some exposure to language is necessary before some sound distinctions can be perceived. Even more surprisingly, perhaps, the ten-month-olds could not distinguish either contrast. One might expect that they would fail to discriminate the non-contrastive pair, because this is not a requirement in their native language. In theory, however, they should not have lost the ability to discriminate the pair that are salient in their native language. One might think that they should have either retained this earlier ability, or alternatively have facilitated it through exposure to the native language, at least when it comes to discriminating a lexically contrastive distinction. It is possible that this puzzling result is due to methodological flaws since “what may be an effective testing tool for one age group may not necessarily be appropriate for another” (Krieger, 2012, p. 22). This finding suggests that there is still some way to go before researchers in this field establish reliable and universal methodologies which can replicate results across different studies. A slightly different approach was adopted by Skoruppa et al. (2011) to establish whether or not English-learning infants at 6 months and twelve months are sensitive to lexical stress patterns. Some languages such as English and Spanish have stress which varies freely, for example the word discount can have stress on either the first syllable or the second syllable, and this change signals a different meaning (Skoruppa et al., 2011, p. EL50). Other languages have a more fixed stress rules which determine that the stress must always fall on the same position in a word, either on the initial, penultimate or some other syllable. Suprasegmental cues also play a role in English. A test was devised, using head-turn preference using familiarisation with a list of stress-initial or stress-final disyllabic nonwords and it was found that there was a preference for novel nonwords with the opposite stress pattern to the familiarisation list. The nonwords were spoken by a speaker of Spanish in order to ensure that no English language specific stress cues would be given. This result suggests that “infants learning English are able to discriminate stress patterns in segmentally varied nonwords” (Skoruppa et al., 2011, p. EL54). This result is different from the results of a similar study carried out previously with French-learning infants (Skoruppa et al, 2009), which is explicable by the fact that the stress rules for French are different from those of English. Spanish and English infants perform in a very similar way in this kind of test, and there is some discussion about whether any difference might be detectable using more sensitive measuring procedures such as electrophysiological measurements of auditory evoked potentials (Skoruppa et al., 2011, p. EL54). Once again methodological issues render the results of this research somewhat questionable since it is impossible to say exactly how infant preferences, in the form of head turning, equates to infants’ ability to discriminate stress patterns. Clearly more cross language research needs to be done before any certainty on these issues can be found. One further study (Skoruppa et al., 2013) addresses this same issue of when exactly infants with different native languages discriminate stress patterns. The authors subjected both Spanish and French-learning infants at six months and at 9 months to tests which are similar to those described in the previous paragraph above. The results show that both groups could discriminate between stress-initial and stress-final disyllabic nonsense words at six months but that “only Spanish infants learn a variable stress language, get better at tracking stress patterns in segmentally varied words between the ages of 6 and 9 months” (Skoruppa et al., 2013, p. 88). All of these studies provide interesting clues to the way infants learn to discriminate segmental and suprasegmental language features but it must be said that the universal listener hypothesis only paints an incomplete picture of infants’ perceptual abilities. It appears that young infants can discriminate stress pattern in segmentally identical stimuli. Multiple word stimuli were not used in these experiments, and this leaves open a large number of questions about how infants with different native languages and at different stages throughout the first year of life perceive and process segmental and suprasegmental linguistic data in a natural language environment. The perception of non-native lexical tone in English infants It has been pointed out by Mattock and Burnham (2006) that over half of the world’s population speaks a language which distinguishes between tones and yet most of the research in linguistics on infant speech perception concentrates on consonants and vowels, with very little consideration of tones. This is due to a longstanding dominance of English and other European scholars in this field, but this is now beginning to change thanks to international collaboration and the contribution in experts on Asian, African and other languages. An ingenious experiment was devised by Mattock and Burnham (2006) which tested Chinese and English infant perception of tone and non-tone speech, and also of tone in the non-speech form of violin sounds. The results of this test were that “Overall, Chinese infants performed equally well at 6 and 9 months for both speech and non-speech tone discrimination. Conversely, English infants’ discrimination of lexical tone declined between 6 and 9 months of age, while their non-speech tone discrimination remained constant (Mattock and Burnham, 2006, p. 241).. This is conclusive evidence that “the reorganization of tone perception is a function of the native language environment, and that this reorganization is linguistically based” (Mattock and Burnham, 2006, p. 241). Such results appear to be consistent with some brain imaging data which reveals how left hemisphere parts of the brain light up when Thai-speaking listeners hear lexical tones, but do not light up in the brains of older English speaking listeners. The additional contribution of Mattock and Burnham (2006) is to observe such differences in the perception of tones occurring during infancy. The discussion in this paper (Mattock and Burnham, 2006) points out that tone is a particularly interesting feature because it can be considered as both a segmental and a suprasegmental feature of speech. It can occur within and across syllables and words and it conveys a range of subtle messages which differ across languages, situations and speakers. There appears also to be some evidence that even short exposure to a tonal language can help English speaking infant listeners to improve their tonal discrimination capability. A comparative study of English and French infants at four, six and nine months by Mattock et al. (2008) explored the way that infants perceive vowels and consonants. As discussed above, it was demonstrated in the earlier article by Mattock and Burnham (2006) that infants learn to perceive suprasegmental features such as tones before segmental features of language. There is a close relationship between vowels and tones, but not such a close relationship between consonants and tones. It was hypothesised therefore that infants’ perceptual reorganisation for tone begins earlier than other reorganisation categories, and the authors aimed to examine exactly how early this occurs across different languages. A Thai language low vs. rising lexical tone contrast was used to test this hypothesis. It was noted that “tone has a lexical function like consonants and vowels, but acoustically tones, consonants, and vowels are very different” (Mattock et al., 2008, p. 1369). There is little, if any, coupling of consonants with tones, but there is with vowels, for example, and the differences among vowel classes are continuous in nature, whereas the differences between consonants are much more precise (Mattock et al., 2008, p. 1369). This means that in adults at least, vowels are perceived as part of a continuum and consonants as being arranged in distinct categories. As predicted, this study (Mattock et al., 2008) found that there was a decline in non-native tone discrimination performance at nine months. The authors note however that contrary to their hypothesis, this change was not detectable at four months or at six months. This means that “the disparity in the age at which perceptual reorganisation for vowels and tones emerges suggests that infants’ phonetic representations for vowels and tones are independent” (Mattock et al., 2008, p. 1377). A somewhat different and more nuanced conclusion was reached by a study involving English-, Mandarin- and Cantonese-learning infants by Yeung et al. (2013). A combination of findings from study (Yeung et al., 2013) and others leads the authors to conclude that all three groups show distinct tone preferences at ages 4, 6, 9 and 12 months but that there are differences which are clearly language-specific. This result echoes that of Narayan et al. (2010) mentioned above, showing that “acoustic salience likely plays an important role in determining the timing of phonetic development” (Yeung et. al., 2013, p. 123). The timing is not universal, therefore, but will vary according to the tonal or non-tonal nature of the native language that the infant is learning. In other words, “language-specific speech perception emerges over a more complex and extended schedule than previously thought: first for lexical stress and tone ( Read More