Listening and the L2 mental lexicon: similarity neighbourhoods for Cantonese-speaking learners of English

 

Richard Pemberton

Hong Kong University of Science & Technology

Introduction

In previous studies, I have found that adult Cantonese-speakers at Intermediate levels of English are only able to recognise about two in every three content words from Level 1 vocabulary (the 1,000 most common words of English) when listening to a BBC news recording. This has been true when the tape has been played at the original speed (close to 200 wpm) and delivered at a slower speed (close to 150 wpm).

Words at this level of vocabulary are mostly monosyllables and can often only be uniquely identified after word offset. The words are also likely to be stored in dense esimilarity neighbourhoodsf in the lexicon – i.e. alongside similar-sounding words. These phonological neighbours are also likely to be frequent. Luce and colleagues (e.g. Luce et al. 1990; Luce & Pisoni 1998), in proposing the Neighbourhood Activation Model of spoken word recognition, have shown that English L1 listeners process words with many neighbours more slowly than they do words with few neighbours.

  This study aimed to:

bullet

investigate the effect of phonological neighbourhoods on the ability of L2 listeners to recognise spoken words (in isolation)

bullet

trial ways of gathering insights into the L2 mental lexicon

Method

Case study investigation of:

4 learners at Intermediate level

1 learner at Upper Intermediate level

2 learners at Advanced level

3 native speakers

 

Procedure

10K/Levels Tests

Spoken Word Recognition Test (target items: ePlosive-Vowel-Plosivef (PVP) monosyllables)

Spoken Word Association Test

Story-retelling

Spoken Associate Pairs Yes/No Reaction Test

Transcription of (simulated) news item

Follow-up questions after each test

 

Results

Table 1: % ePVPf word recognition rates in the Spoken Word Recognition Test (n = 50)

S1

S2

S3

S4

S5

S6

S7

NS1

NS2

NS3

56

76

68

68

76

90

90

100

100

100

Table 2: False alarm rate in the Spoken Word Recognition Test as a % of the number of ePVPf pseudowords (n = 25)

S1

S2

S3

S4

S5

S6

S7

NS1

NS2

NS3

72

44

60

64

80

72

48

0

4

0

  Table 3: Recognition rates for words with the highest neighbourhood density

Word

No. of neighbours

Recognition rate (%)

fight

8

86

boat

7

100

back

6

100

fade

6

43

  Table 4: Neighbourhood density of words with the lowest recognition rates

Target word

Recognition rate (%)

No. of phonological neighbours

No. of Ss not recognising

Most often identified as c

(no. of times)

bag

21

3

6

back (10)

dug

24

2

7

duck (12)

poured

25

3

6

port (4)

cared

38

2

5

cat (5)

toured

43

2

7

taught (4)

fade

43

6

4

fate (2)

debt

43

4

4

death/d (4)

dealt

43

1

4

doubt (3)

Discussion

·    Unexpectedly low recognition rates for isolated ePVPf words

·    Unexpectedly high false alarm rate for ePVPf pseudowords

·    Likely fuzzy specification in lexicon of Cantonese-speakers for plosives at end of words, where 2/3 of misidentifications occur

   However

·    Initial misidentifications often corrected in connected speech

·    Recognition rates for non-plosives much higher

Conclusions

·    ePhonetic featuref rather than ephonemef difference a more useful criterion for phonological neighbourhood for Intermediate L2 learners

·    Dense phonological neighbourhoods may not be problematic for L2 learners

·    Implications for lexicons of Cantonese-speakers: words may be loosely specified for final consonant; voicing of final consonant likely to be eopenf; competing candidates likely to be prioritised according to frequency/familiarity

·    Further testing of recognition of frequent monosyllables required, comparing recognition rates for plosive/non-plosive final consonants, low/high frequency neighbours, words from dense/sparse neighbourhoods

References

Luce, P.A., Pisoni, D.B. & Goldinger, S.D. (1990) eSimilarity neighbourhoods of spoken words.f In Altmann, G.T.M. (ed.) Cognitive Models of Speech Processing, pp. 122-147. Cambridge, MA: MIT Press.

Luce, P.A. & Pisoni, D.B. (1998) eRecognizing spoken words: the neighbourhood activation model.f Ear & Hearing 19: 1-36.