Some phonetic characteristics of Lebanese Australian English

Some phonetic
characteristics of Lebanese
Australian English
Felicity Cox and Sallyanne Palethorpe
Macquarie University
Australian Linguistic Society Conference 2006
Background
• Children of immigrant families typically adopt the
majority speech patterns of the adopted country rather
than those of their parents (Sankoff, 2001)
• Holmes (1992) suggests that some effect of the
parents’ language could be seen among the Australian
born children from various ethnic groups.
• This apparent change in the dialect of English spoken
by some Australian born children of migrants was
probably a response to significant social change that
occurred in the 1980s (e.g. multiculturalism).
Studies of Ethnic Accent Varieties
There are few studies of speech patterns of
Australian born
members
of
various
ethnocultural groups.
•
Horvath’s (1985) Sydney study did not
suggest transfer features.
•
Warren’s (1999) “wogspeak” study explored
a new variety used by children of migrants.
•
Kiesling’s (2001) acoustic analysis compared
two vowel variables in “Anglo” and “nonAnglo” speech.
Clyne, Eisikovits and Tollfree (2001) :
• refer to varieties used by Australian born to
mark ethnicity as “ethnolects”
• discuss “stabilised transference” as the
process of creating a new local dialect based
on transfer of features from the substratum
language or variety
• suggest that the ethnolect will contain
phonemes and allophones transferred from
the parents’ first language giving rise to a
“non-native accent”
Ethnolects are not necessarily the result of
second language learning as many speakers of
Australian ethnolect have English as their L1.
However, the transfer effects present in the
ethnolect are presumed to be the result of L1 to
L2 feature transfer that occurred in the parents’
or grandparents’ generation.
The phonetic characteristics of the ethnolect
may therefore be explained in terms of Flege’s
Speech Learning Model (SLM) (Flege, 1995).
SLM predicts that, if the phonology of the
parent language (L1) does not exploit a phonetic
feature, that feature will be difficult for the
learner to perceive and produce in the L2.
• One well recognised new Australian English
(AusE) dialect/variety (ethnolect) is that used by
Australian born speakers of Lebanese
background.
• We will call this variety Lebanese AusE
(LAusE).
• It is sometimes referred to as Lebspeak.
“Alan took a ham sandwich to school”
“Helen picked a good spot near the water and spent
the morning surfing and relaxing in the sun”
AIMS
• to examine some characteristics of LAusE
as a first step towards a description of this
variety
• to investigate some characteristics that may
reveal the kind of transfer features that have
been adopted from the original language of the
migrant group (i.e. Arabic)
Crowther and Mann (1994); Flege & Port (1981); Mitleb
(1984, 1985); Munro, (1993); Port & Mitleb
(1983) have found that speakers of Arabic accented
English (i.e. foreign accented) display characteristics
that differ from native speakers of English as a result of
L1 to L2 transfer.
•
•
•
•
reduced short / long vowel contrast
spectral influence from Arabic in vowel space
reduction of diphthongs
reduced cues to final stop voicing
RESEARCH QUESTIONS
• Are there phonetic differences between Standard
Australian English speakers and Lebanese Australian
English speakers which can be explained by SLM?
• Does LAusE display the same characteristics as
Arabic foreign accented English reported in the
literature?
SPEAKERS
Lebanese (LAusE)
• 5 male university students who were born in
Australia (Sydney) to Lebanese born parents.
• All were in their early twenties and bilingual
in English and Arabic.
Standard (SAusE)
• 6 male university students in their early
twenties who were born in Australia (Sydney)
and have SAusE as their first language.
SPEECH DATA
Speakers read a standard elicitation list 4 times
containing in random order :
•
18 stressed vowels of AusE
/hVd/, /hVl/, hVn/, /hV/, /hVt/
For this paper, /hVd/, /hVt/ and /hV/ will be
examined
•
10 sentences including
“Pete sat in the little boat with his bait and his
fishing boots on the floor beside him waiting for
the trout to bite.”
ANALYSIS
• The acoustic analysis was carried out using standard
procedures established at SHLRC (Harrington, Palethorpe
and Watson, 2000).
• The centring diphthongs have been omitted from the
database.
• Vowel targets were hand labelled according to criteria
described in Harrington, Cox and Evans, 1997.
• Data was normalised (Lobanov, 1971)
• Used a mixed model ANOVA with Accent and Vowel
as main effects (post-hoc error of p<.01).
PARAMETERS
•
•
•
•
•
F1 and F2 at vowel target(s)
F1 offset
vowel duration
closure duration
burst duration
EXPERIMENT 1
Does LAusE differs from SAusE in vowel space
and diphthong trajectories?
Munro (1993) found such spectral differences
for Arabic accented English.
/hVd/
LAusE
SAusE
/hVt/
“hid, head, had”
“hit, het, hat”
No significant differences (p<.05) between accent for either /hVt/ , /hVd/ or /hV/.
/hVd/ Diphthong Trajectories
No significant differences between accent groups except for:
F2 target 1 of /əʉ/ in “hoed”
/hVt/ Diphthong Trajectories
No significant difference between accent groups except for:
F1 target 1 of /oɪ/ in “hoyt”
F2 target 1 of /əʉ/ in “hote” & “hoe” (hV)
• No indication of transfer from Lebanese
Arabic which makes use of short/long pairs
and has fewer diphthongs.
• These findings are contrary to Munro
(1993) who found significant vowel spectral
L1 influence in Arabic accented English.
EXPERIMENT 2
Do LAusE speakers have shorter vowels than
SAusE speakers.
Munro (1993) found that speakers of Arabic
accented English had shorter vowels than
native speakers of English?
/hVd/
trend but
sig /iː, ɐː, oː/
/hVt/
trend but
sig /iː, ɪ, ʊ/
/hV/
Vowel Duration
• This result suggests that vowel length is
different between the two groups in certain
contexts.
• Final stop voicing increases the difference
between the accent groups and there is
also a vowel quantity effect.
• But could there be a speaking rate effect?
• Increased speech rate can lead to decreased
vowel length (e.g. Erickson, 2000; Smith, 2002).
• We measured the duration of the first phrase
of the “Pete” sentence to assess rate.
• Mann Whitney U statistical test showed no
significant difference between the two accent
groups.
•We therefore conclude that vowel differences
are not the result of speaking rate differences.
EXPERIMENT 3
What is the relationship between long and
short vowels in LAusE?
Do LAusE speakers have larger ratios of short
to long vowels as suggested by Munro (1993)
for Arabic accented English.
Is there a reduced length contrast?
Both accent groups differentiate long and short
vowels for /hVd/ and /hVt/.
The short/long ratios are greater for LAusE
indicating a reduced length contrast.
• /hVd/ short/long vowel ratios
LAusE
SAusE
= .65
= .56
• /hVt/ short/long vowel ratios
LAusE
SAusE
= .74
= .62
Vowel Duration
Lebanese
Vowel Duration (ms.)
Standard
300
D
D
200
D
D
D
D
D
D
100
0
HVD
.56
HVT
.62
HVD
.65
HVT
.74
Vowel Duration
/hVt/
Vowel Duration (ms.)
/hVd/
300.00
D
*
D
200.00
*
D
D
D
D
D
100.00
0.00
D
LONG
SHORT
LONG
SHORT
The results or Experiment 3 show a larger ratio of
short to long vowels in both /hVd/ and /hVt/
contexts for LAusE speakers.
This is similar to Munro’s (1983) findings for
Arabic accented English.
EXPERIMENT 4
Do typical English vowel + stop rhyme temporal
patterns differ between LAusE and SAusE?
Arabic accented English shows different vowel
+ stop rhyme patterns than English (Port and
Mitleb, 1983).
Well known final stop voicing cues for English
are:
•
•
•
•
long duration of preceding vowel
lower F1 offset
shorter closure duration
shorter burst duration
(e.g. Raphael, 1972; Walsh and Parker, 1983)
Consonant Closure Duration
/hVd/
post hoc sig./æɔ, ɪ, oː/ < .01
/æ, ɐː, e, æɪ, iː, ɔ, oɪ, ʊ, ʉː/
(.05 <p >.015)
/hVt/
no sig. differences
Vowel + Consonant Closure Duration
/hVd/
/hVt/
- no significant accent differences for either /hVd/ or /hVt/ words
Ratio Closure/(Vowel + Closure)
/hVd/
11/16 vowels significant
exceptions were /ɪ, e, ʊ, ɜː,
æɔ/ i.e. LAusE: longer closure
relative to total
/hVt/
no sig. difference
Burst Duration
All hvt > hvd significant
Burst Duration
All accent non-significant (/hvt/ short p<0.08)
Summary Between Groups
• LAusE
• LAusE
/hVd/
shorter long vowels
larger ratio of closure to vowel+closure
longer stop closure
/hVt/
• LAusE
longer short vowels
Summary Within Groups
Parameter
SAusE
LAusE
Vowel duration
hVd > hVt
Ratio=closure:
V+closure
F1 Offset
hVd < hVt
hVd < hVt except
/æ, e, ɪ, ɔ, ɐ, ɐː/
hVd < hVt
hVd < hVt
Burst duration
hVd < hVt
hVd < hVt
hVd > hVt – for long
vowels &/æ/
n.s. (sig. for /əʉ,
Closure duration hVd < hVt
Long vowels æɔ, oɪ/)
Vowel + closure hVd > hVt
hVd > hVt except
short V &/ɜː, ɐː,æɪ/
These results indicate that LAusE and SAusE
make use of different temporal patterns in VC
rhyme.
• vowel duration effects - /hVd/ and /hVt/
• ratio effects - /hVd/
There is an interaction between vowel quantity
and voicing.
Long Vowel Component Durations
*
#
*
*
*
*
*#*
#
“heart”
*
#
*
*
“hard”
“hard”
“heart”
Vowel-Closure-Burst
* within group significance # across group significance
Short Vowel Component Durations
*
*
*
*
“hud”
“hut”
*
“hud”
*
“hut”
Vowel Closure Burst
• Previous studies of the temporal characteristics
of Arabic accented English have shown
¾smaller vowel duration contrast
¾smaller closure duration contrast
¾reduced F1 offset contrast
•LAusE speakers distinguish voicing by
¾vowel duration in long vowel contexts
¾vowel+closure in long vowel contexts
¾F1 Offset
¾Burst duration
•The magnitude of the difference is reduced
compared with Standard AusE speakers.
SUMMARY
Four analyses have been conducted to examine:
¾
vowel space characteristics
¾
vowel duration
¾
long/short vowel differences
¾
component durations and phonetic voicing
rules for final stops
The results show:
•
•
•
no vowel space differences
minimal diphthongal differences
timing effects relating to voicing and
vowel length.
The timing differences appear to
be indicative of higher order
differences in metrical structure
between the accent groups rather
than simply the result of phonetic
voicing implementation rules.
• Broselow et al (1997) discuss Levantine
Arabic
as having a moraic structure where vowel
length determines whether a coda
consonant shares the mora with the
preceding vowel (is weightless) or occupies
its own mora in order to comply with the
bimoraicity constraint.
• Could the interaction between phonetic
implementation of voicing and moraic
structure be responsible for the observed
temporal differences?
• What does this tell us about the features that
Limitations and future directions.
•We acknowledge the small sample size and restricted
speaking style.
•Future work could
analyse a greater number of speakers with a
wider range of socio-demographic characteristics
in a broader context of use.
¾
¾
examine a wider range of phonetic variables
¾
further explore metrical structure and prosody
The End