View
219
Download
0
Category
Preview:
Citation preview
Using the MLF model to identify convergence in the speech of Welsh-English bilinguals
Peredur Davies & Margaret Deuchar(University of Wales, Bangor)
2nd June 2007, Hamburg
Outline of talkOutline of talk
Research aims Convergence Theoretical model Previous research Methodology and data Results Analysis Conclusions
Research aimsResearch aims
Studying language change in Welsh Using a theoretical model of code-
switching to identify possible sites of convergence in Welsh/English bilingual data
Ascertaining the usefulness of this method and exploring possible sites of convergence not highlighted by the model
ConvergenceConvergence
A contact-induced process of change The “enhancement of inherent structural
similarities found between two systems” (Bullock & Toribio, 2004)
Convergent structures “already present, but less prominent” (Thomason, 2001) than in the language before convergence occurred
Differences between English and Differences between English and WelshWelsh English: SVOSVO
Siôn caught the ballS V O
Welsh: VSOVSO
Daliodd Siôn y bêl
V S O English: Adjective+NounAdjective+Noun
Red wine
Adj N Welsh: Noun+AdjectiveNoun+Adjective
Gwin coch
N Adj
Similarities between English and Similarities between English and WelshWelsh But Welsh also makes some use of SVO
order, e.g. for emphasis:Fi ddaliodd y bêl
S V O
“I caught the ball” … and some adjectives precede the noun,
e.g.:hen ddyn
Adj N
“old man”
Convergence in Welsh?Convergence in Welsh?
Welsh seems to be extending the use of certain word orders from contact with English
This results in some unexpected word orders
We shall apply the MLF model to data to identify where word order does not match subject-verb agreement as expected, or where there is inherent word order discrepancy within a clause
Methodology: Methodology: the Matrix Language Frame modelthe Matrix Language Frame model
Proposed by Myers-Scotton (1993, 2002 etc.), based on earlier ideas, e.g. Joshi (1985)
Two principles identify the Matrix Language (ML) for any given clause: Morpheme Order Principle (MOP)
S/V (finite only) and/or head-modifier order System Morpheme Principle (SMP)
Subject/Verb agreement
Previous work: Deuchar (2006)Previous work: Deuchar (2006)
Applied MLF to 163 Welsh/English bilingual clauses to analyse “classic code-switching” patterns
99.39% clauses had an identifiable ML “very low frequency of problematic
data” (Deuchar 2006; p. 2009) Only 1 clause where the MLF could not
be applied
How we apply the principlesHow we apply the principles
For a clause, if both principles indicate the same language, that is the ML; e.g.
Mae o wneud rhywbethbe.3S.PRES 3SM do.NONFIN something
“He’s doing something”
MOP: Verb (mae) before Subject (o) SMP: morphology of S and V agree
= ML is identifiable (Welsh)
Unidentifiable MLUnidentifiable ML
Required to identify ML of a given clause: Finite verb with subject, or NP with head and modifier
Insufficient material results in that clause having an Unidentifiable ML: Verbless (“yeah”, “mm”, etc.) Non-finite Finite but no visible Subject
Dichotomous Matrix LanguageDichotomous Matrix Language
Sometimes one or both of the principles fails to identify the ML in a clause
2 languages providing structural information Manifests as word order discrepancy, e.g.
Ddaru ni gyfweld […] amhappen.PAST 1PL interview.NONFIN for
ddeg awr assistantten hour assistant“We interviewed for a ten-hour assistant”
Welsh V/S order but English head/modifier order (with morphemes from both languages)
We call this a Dichotomous ML
Our dataOur data
Bangor AHRC project Analysing 2 transcripts of conversations:
(A) two men in 20s (B) woman in 30s with woman in 50s
55’22” recorded speech (1,808 clausal units) The MLF model applied to all clausal units—
monolingual and bilingual Two main aims:
1) See if frequency of clausal units with identifiable ML is similar to previous study
2) Analyse Dichotomous clausal units for signs of convergence
Results: Matrix Language Results: Matrix Language distribution (all clausal units)distribution (all clausal units)
90450%
774.26%
351.94%
79043.69%
20.11%
0100200300400
500600700800
9001000
Matrix Language
Welsh English Either Unidentified Dichotomous
Results: Matrix Language Results: Matrix Language distribution (clausal units with verb)distribution (clausal units with verb)
• 62.44% of all clausal units have a verb
85275.47%
494.34% 5
0.44%
22219.66%
10.09%
0
200
400
600
800
1000
Matrix language
Welsh English Either
Unidentified Dichotomous
Results: Matrix Language Results: Matrix Language distribution (finite clausal units)distribution (finite clausal units)
• 48.29% of all clausal units have a finite verb
82894.85%
445.04%
00%
00%
10.11%
0
200
400
600
800
1000
Matrix Language
Welsh English Either
Unidentified Dichotomous
Discussion of resultsDiscussion of results
ML identifiable in 56.19% of all clausal units 80.24% of clausal units with verbs 99.89% of finite clausal units
Very similar findings to Deuchar (2006): high rate of ML identifiability
In general, quite easy to apply MLF model—
—but this is bad news for identifying convergence!
Dichotomous clauses in the dataDichotomous clauses in the data
An anomalous head/modifier construction:Mi oedd drws-nesa # pobl yn wneud sloe ginPRT was door-next people PRT make sloe gin“The next-door people [=neighbours] made sloe gin”
This is a NP+NP modifier/head construction, where drws-nesa modifies the head, pobl
Order of NP+NP is English (cf. next-door people) but with Welsh morphemes
However, within the first NP (drws-nesa), the order is head+modifier = Welsh
Thus no one language supplies word order, and so the MOP cannot be applied
Interpreting this DichotomyInterpreting this Dichotomy
Is drws-nesa pobl a sign of Welsh order converging onto English?
Other N+Adj constructions in the data follow the expected order for the languages (and the MLF)
The sole occurrence of this phenomenon suggests it may be idiosyncratic
Application of the method to a larger sample may yield more examples of this Dichotomy
Convergence not highlighted by the Convergence not highlighted by the modelmodel There are some clausal units, which are
Unidentifiable ML according to the model, which yet seem to show convergence
Common ellipsis of finite auxiliary, resulting in a clause-initial Subject (paralleling English?)
NB: No identifiable ML because they lack a finite verb to show agreement or S/V order
Examples of auxiliary ellipsisExamples of auxiliary ellipsis
Ti ’n jocian!You PRT joke.NONFIN
S V
“You’re joking!”
Compare with ‘full’ (unellipted) form:
Wyt ti ’n jocian!be.2S.PRES 2S PRT joke.NONFIN
Aux S V
Ellipted versus unellipted formsEllipted versus unellipted forms
64 out of 67 (95.5%) examples of constructions using the 2nd singular pronoun ti ellipt the finite verb
All non-ellipted examples are interrogatives
The 2 younger speakers (20s) ellipt 98.2% (average) of the time
The 2 elder speakers (30s and 50s) ellipt 88.5% (average) of the time
Future researchFuture research
Analysis of a larger corpus will demonstrate the frequency of this construction and of other potentially convergent constructions
Consider evidence for possible convergence of Welsh VS → English SV
ConclusionsConclusions
The MLF model shows that, in a larger sample than that studied in Deuchar (2006), a ML is still identifiable in most cases
The method does identify one convergent clausal unit in the sample, but fails to identify other forms of convergence
Further research will help show the extent of convergence in Welsh towards English
Diolch yn fawr!Diolch yn fawr! Thank you!Thank you!
peredur@cantab.net
m.deuchar@bangor.ac.uk
© 2007
Recommended