Phonetic Detail in Developing Lexicon

1
Phonetic Detail in Developing Lexicon

• Daniel Swingley

2
The issue

• Do infants/children store every piece of phonetic
  details of words in the lexicon in the very
  beginning stage of phonological acquisition?

3
Continuous debates

• 6-8 mo Infants are broadly sensitive to different
  sound categories, and are gradually declined to
  perceive only their native categories.Perceptual
  Assimilation Model (Best 1994)Native Language
  Magnet (Kuhl 1995)

4
Continuous debates

• However
• The stimuli used in the experiments do not match
  the natural variability of speech.
• Translating an acoustic sequence into segments
  does not entail retaining (all of) them.

5
Continuous debates

• Jusczyk Aslin (1995) 7.5 mo infants can encode
  initial segment (e.g. /k/ in cup vs. /t/ in tup)
  by listening just a few tokens.
• Halle de Boysson-Bardies (1996) 11 mo French
  infants have the preference of ponjour and
  vonjour after listening to bonjour. (A global
  representation of labial)

6
Continuous debates

• Naming task fails for some chirldren ranged from
  10 to 28 mo (e.g. Matthews 1997).
• Dehabituation task fails for 14 mo children
  (Stager Werker 1997).
• Walley (1993) assumes that children with small
  vocabularies might represent cat as abrupt
  onset, or cap as labial.

7
Continuous debates

• Theoretical phonologists suggest the infants
  lexical representation is more or less identical
  to the adult form.
• Lexicon Optimization in Optimality Theory assumes
  all of the adult forms are stored into lexicon
  faithfully.

8
Authors view

• The results in the previous studies are not
  entirely reliable
• Metalinguistic responses are required it is just
  difficult to get any overt response
• Production tasks are not ideal to test childrens
  receptive forms in the lexicon

9
Experiment Design

• Visual fixation no overt response is needed.
• If children can discriminate phonetic details,
  the visual fixation on the target picture will be
  different.
• Participants Dutch learning children ranged from
  1807 (monthsdays) to 2017.

10
Hypothesis I

• At first, children have only global
  representations of words (e.g. labial for
  bonjour), unless they also learn some
  minimal-pair words (e.g. vonjour) that help them
  to distinguish one sound from the other.

11
Experiment I

• StimuliCP - bal ball, beer bearMP onset
  substitution of /g/ and /d/(Only dal valley is
  a real word in Dutch)
• If the children only have global representations
  of the two words, they should not be able to
  distinguish, for example, beer from geer or
  deer.

12
Experiment I

• TaskPictures of ball and bear are presented
  respectively children look at the pictures and
  listen to the stimuli of Waas is de target
  (Where is the target).

13
Experiment I

• Result

14
Experiment I

• ResultsChildren fixate the targets longer in
  the CP condition (p.03).
• Children do not need to know any minimal-pair
  word to distinguish sounds the phonetic details
  of a word are stored in the lexicon, which can be
  used for the sound discrimination.

15
Hypothesis II

• Children might learn a minimal-pair word due to
  misparsing of word boundaries. For example, peer
  can be learned from a longer word ___peer as a
  possible minimal-pair word of beer.

16
Experiment II

• StimuliCP baby baby, beker cupMP
  Word-medial substitution with /d/ and /g/ (e.g.
  bady or bagy).

17
Experiment II

• 2 minimal-pair words for beker are zeker and
  beter.
• Yet there is no possible minimal-pair word for
  baby, even with misparsinge.g. No instance of
  bagy or ___bagy

18
Experiment II

• Word-medial substitution rules out the
  possibility that children discriminate sounds
  solely by the very first part of a word.
• For example, children might know the target is
  baby instead of car by just hearing ba

19
Experiment II

• Result

20
Experiment II

• ResultsThe children again fixate longer on the
  target picture in the CP condition! (p.03)
• No correlation with the childrens receptive and
  productive vocabularies reported by their parents
  (r.06, r.07).

21
Authors Conclusions

• Children store every phonetic detail in the
  lexicon acquiring minimal-pairs is not a
  prerequisite for them to distinguish sounds.
• Why do some children fail to demonstrate the
  sound discrimination? They do encode the details,
  but they may forget which form corresponds to
  which word.

22
Follow-up Question I

• Is it enough to make the conclusion by testing
  just one single segment in each experiment? If
  children retains phonetic details completely,
  shouldnt we test the discrimination of every
  segment in the stimuli?

23
Follow-up Question II

• Does the phonotactic distribution help the
  children to identify the sound contrasts? For
  example, in English, none of baby, pig, beckon,
  happy, brush, and pickles can form minimal-pairs,
  but they show a /b/-/p/-/k/ onset contrast.

24
Follow-up Question III

• Is it possible that these children watch too
  many English TV shows and learn/hear some
  minimal-pair words from them??