Notes On Werker "Exploring Developmental Changes In Cross-Language Speech Perception"

A considerable amount of research has been focussed on language acquisition. Emphasis of this chapter is on one component of this, development of speech perception, particularly of the perception of phonemes. "A phoneme is a subsyllabic unit, and the difference in a single phoneme can convey different meanings (for example, big versus dig or bat versus bet).

Different languages use different sets of phonemes. "Clearly, then, English or Hindi children -- like all children in the world -- require some working knowledge of the phonology (the sound structure) of their native language in order to map sound on to meaning." Issue: How does this working knowledge develop? Key point to get from this chapter: "long before many infants have enve acquired their first word, speech perception capacities have been modified to match the properties of the sound structure o fthe native language."

Sppech Perception Sensitivities In Very Young Infants

Extremely young children can distinguish phonemic categories, but cannot distinguish same-sized differences *within* the category. Further, phonemic categories that have never been heard can also be discriminated; this is not generally true for adults. In other words, infants are "universal listeners."

Early Effects Of The Language Environment

"If children are `universal listeners' but adults sometimes have difficulty discriminating nonnative phonemic contrasts, then there must be a decline across age in cross-language speech perception performance." Data supporting this, for example, comes from study comparing perception of phonemic distinctions, some of which occur in both English and Hindi, others which only occur in Hindi. Conditioned head turn procedure is used -- "the subject has to monitor a continuous background of syllables from one phonetic category (for example /ba/), and signla when the stimuli change to a contrasting phonetic category (for example /da/)." (Older subjects press a button!). Results were consistent with the universal listener perspective -- very young infants correctly make all the distinctions, as do Hindi adults, but English speaking adults have problems with the Hindi phonemic categories.

At what age does language experience first influence phonetic perception? Earlier studies, buying into Lenneberg's hypothesis of puberty marking the offset of critical period, compared 12 and 8 year olds -- no soap, though, they were both bad on distinctions not in their language. 4 year olds were even worse (NB: This is a key point later on!) Other tests reveal changes in speech perception essentically occur between 6 and 12 months of age!

"What can we conclude from these findings? First, infants stop being `universal listeners' for phoneme distinctions by the end of the first year of life. Second, it appears that the role of experience is to `maintain' those perceptual senstivities that are already evident in the young infant. Without such exposure, initial abilities will be lost." This is called the maintenance-loss model. "The central assumption of this model when applied to speech perception is that the neonate is equipped with sensitivity to all possible phonemes of all languages. The role of experience is to maintain sensitivity to only those ponemes that the infants hear. Sensitivity to nonexperienced (nonnative) contrasts declines and/or disapears."

Problems For The Maintenance View<

But there are fundamental problems for the maintenance-loss model:

  1. Sometimes, with the appropriate lab tests, continuing adult sensitivity to nonnative phonemic distinctions can be found.
  2. Model does not predict "details" correctly. For example, not a perfect correlation between age and performance -- remember now that those 4 year olds were *poorer* than the 8 and 12 year olds! In other words, evidence points to there being a decline in ability, followed by recovery.
  3. Model does not explain how "chidlren far older than 10-12 months of age can move to a new country and acquire a new language with no trace of an accent.
  4. Some nonnative contrasts are easy for adults to discriminate, even with no training at all.

"It is not just experience or lack thereof that leads to a change in speech perception performance; instead, the experience needs to map on to the phonology of the language of input. Phones taht are similar to those used in the native language will be susceptible to reorganization, but phones that are completely unlike the ponetic repertoire of the language of input will not be assimilated to native-language pones and will remain discriminable."

A New Approach: Functional Reorganization

An alternative approach is clearly required; Werker explores "functional reorganization", where "software" changes occur instead of "hardware" changes. "The central assumptin of this Functional Reorganization Hypothesis is that children's changing sensitivities reflect not absolute, hardware changes in the auditory system, but rather the needs of the developing linguistic community. According to this explanation, the `universal' sensitiviites seen in the newborn infant may continue to be present across the life span, and thus may show up in adults under certain kinds of testing conditions."

From this view, young infants respond to any detectable phonetic variation in speech. Adults have a bias to respond to phones that distinguish meanings, but can change strategy and listen for other phones if required (e.g., in the proper experiment). But between 10 mo and 4 years children "seem constrained to listen only to that information that might be used to distinguish meaning." Why? To aid mapping of sound to meaning during language acquisition phase of word production! Advantage of this is that it accounts for adult ability to discriminate, under proper conditions, phonemes from other languages.

Residual Problems For The Functional Reorganization Hypothesis

  1. Does not take into account the finding of a critical period, e.g., the period during which "accentless speech" can occur.
  2. 2-5 year olds have most difficulty discriminating words that differ in only a single phoneme, which seems contrary to this approach.

Werker's move is towards an altered version of the FRH. "Experience and cognitive developoment allow children to attend selectively to only the phonetic detail that is important for defining the phonemic categories of their native language. Children, however, need use only parts of that available information in word-learning situations, representing no more detail than is required to distinguish items in their lexicon. Finally, once children have a working knowledge of the phonology of the native language, test of word discrimination are no longer simple perceptual tasks. They are now tasks that require children to use the detail available in their representations of those words."


"It appears that in order to fully understand age-related changes in speech perception we have to consider not just the capabilities given by nature and the role the input plays in shaping those capabilities, but also the particular challenges the child faces at each juncture in langauge acquisition and the potential contribution of developing cognitive skills."

Pearl Street | "An Invitation To Cognitive Science" Home Page | Dawson Home Page |