Exploring register variation in learner lexis
description
Transcript of Exploring register variation in learner lexis
1
Exploring register variation in learner lexis
The high-frequency verb make in native and learner speech and writing
Claire HugonCECL
Louvain-la-Neuve 24. January 2008
2
Outline of the presentation
Background and aims of the study Methodology Setting the scene: make in the BNC Make in native and French-speaking learner speech
and writing Methodological implications and avenues for future
research
3
Background and aims of the study
Broader context: PhD research on the acquisition of high-frequency verbs
3 preliminary remarks: The influence of L1 as the « darling variable » of
learner corpus linguists Learner writing is frequently said to be speech-
like SLA variables are often studied in isolation
4
Background and aims of the study
Research questions: Does register have an influence on the use of
high-frequency verbs (HFVs) such as make in learner English?
Is the use of make in learner writing similar to native speech?
Can register differences be an alternative/ complementary explanation to features of non-nativeness attributed to L1?
5
Methodology
CIA
IL vs IL
registera
vs.
registerb
(CRIA)
L1a
vs
L1b
proficiencya
vs.
proficiencyb
taska
vs.
taskb
...
6
Methodology
Confrontation of native and learner data to detect similarities and differences and try to explain them (to-ing and fro-ing between the two components)
NS (writing)
LOCNESS
NS (speech)
LOCNEC
NNS (writing)
ICLE-FR
NNS (speech)
LINDSEI-FR
7
Implementing the methodology: the example of make
1. native language: make (and other HFVs) in the BNC see how HFVs behave in native language before looking for
differences in learner language BNC: wide-coverage corpus, much larger than LOCNESS better suited for broad, quantitative analysis
2. quantitative and qualitative analysis: make in native and learner speech and writing
native: LOCNESS and LOCNEC learner: ICLE-FR and LINDSEI-FR Comparison of the results
8
Top HFVs in the BNCWhole BNC Spoken BNC Written BNC
1. say 1. get 1. say
2. go 2. go 2. make (rel.freq. 2,190)
3. get 3. say 3. go
4. make (rel.freq.2,160) 4. know 4. take
5. seem 5. think 5. see
6. know 6. see 6. get
7. take 7. come 7. know
8. think 8. mean 8. come
9. come 9. want 9. give
10. give 10. take 10. use
11. look 11. look 11. think
12. use 12. make (rel.freq. 1,905) 12. look
9
Make in the BNC
spoken BNC
(/million words)
written BNC
(/million words)
chi-square value
1,905 2,190 *** 349.7
•Make is less frequent in speech than in writing•the difference is highly significant according to the chi-square test•atypical (most HFVs are more typical of speech)
10
Implementing the methodology: the exampe of make
1. native language: make (and other HFVs) in the BNC see how HFVs behave in native language before looking for
differences in learner language BNC: wide-coverage corpus, much larger than LOCNESS better suited for broad, quantitative analysis
2. quantitative and qualitative analysis: make in native and learner speech and writing
native: LOCNESS and LOCNEC learner: ICLE-FR and LINDSEI-FR Comparison of the results
11
Make in native and learner speech and writing: some findings
350.6 > 245146.8 ≥ 126.6NS vs NNS
126.6 < 245
146.8 < 350.6
Speech vs writing
245126.6NNS
350.6146.8NS
WritingSpeech
Overall frequency (/100,000 words):
highly significant (***) underuse of make in NNS writing brings frequency in NNS writing closer to NS speech
slight underuse of make in NNS speech, but not significant
make is significantly(***) less frequent in NNS speech than in NNS writing
make is significantly (***) less frequent in NS speech than in NS writing
Make is a polysemous verb qualitative analysis to explain the results
12
7 main semantic subdivisions
core meaning (produce, create) delexical uses
‘speech’ collocates other collocates
causative uses causative uses
make + adj make + verb make + noun
‘money’ make phrasal verbs other uses link verbs
13
020406080
100120140160
NS writing NNS writing
NS speech NNS speech
Distribution of the occurrences of make in the four corpora, by semantic category
14
Delexical uses of make
120.9 > 80.928.7 < 42.9NS vs NNS
42.9 < 80.9
28.7 < 120.9
Speech vs writing
80.942.9NNS
120.9 28.7NS
WritingSpeech
Overall frequency (/100,000 words):
highly significant (***) underuse of make in NNS writing
significant (*) overuse in NNS speech
significantly(***) less frequent in NNS speech than in NNS writing
significantly (***) less frequent in NS speech than in NS writing
15
Delexical uses of make
NNS writing: underuse of EAP delexical structures (make a case, make a statement) maybe register-related
NNS speech: overuse of delexical uses probably communication strategy (pressure, online
processing, make as default verb): especially one course we have to make erm . a kind of work when I go . eat em . with my master the: the cooking he
made for us is just er . about er .. an .. experience which I .. made when I was in
first candi
16
Causative uses of make
142.1 > 102.664.9 > 24.2NS vs NNS
24.2 < 102.6
64.9 < 142.1
Speech vs writing
102.624.2NNS
142.164.9NS
WritingSpeech
Overall frequency (/100,000 words):
significant (**) underuse in NNS writing
significant (***) underuse in NNS speech
significantly(***) less frequent in NNS speech than in NNS writing
significantly (***) less frequent in NS speech than in NS writing
17
Causative uses of make
underuse of causative structures as a whole in learner language (both in speech and in writing)
3 causative structures: • make + adjective (make sth easier)• make + verb (make someone feel bad)• make + noun (make someone an outcast)
18
The proportion of each category is remarkably similar for NS and NNS registers
NS writing NNS writing
NS speech NNS speech
Adjective 57% 57.6% 39% 40.9%
Verb 30% 32.7% 54.5% 50%
Noun 13% 9.7% 6.5% 9.1%
Total 100% 100% 100% 100%
19
Some previous findings about make:
1. French-speaking and Swedish-speaking learners underuse make in delexical structures (Altenberg & Granger 2001, Altenberg 2001)
2. Swedish-speaking learners overuse causative make + adj and make + verb (Altenberg 2002a, 2002b)
(Partially) L1-related explanations:1. delexical structures: avoidance strategy due to
arbitrary and L1-specific choice of the verb2. causative structures: transfer of frequency from L1 +
overgeneralisation
20
Plausible register-related explanation?1. delexical combinations:yes.
• Transfer and register have a similar impact. Underuse of delexical structures in NNS writing: much less frequent in NS speech than in NS writing: possible transfer of frequency from target language speech
2. causative structures: no (at least not for Swedish-speaking learners).
• Transfer and register seem to pull in opposite directions: • L1 Swedish causes overuse of L2 English ADJ and VERB
causative structures• English speech uses fewer causatives structures, so poor register
awareness is not a valid explanation for the Swedish-speaking NNS’observed overuse of causative structures.
21
To sum up:
Make is a multi-faceted verb with many meanings, functions, and patterns: a very interesting picture of scale of proficiency of advanced interlanguage emerges
from no knowledge at all (e.g. some phrasal verbs, link verb uses, ‘money’ make are nearly absent)
to near-perfect knowledge (e.g. proportions of 3 causative syntactic structures)
including various levels of partial knowledge (e.g. core uses, delexical uses, overall frequency of causative uses, etc.)
knowing a word is not an all-or-nothing matter
22
Methodological implications
The results can be partially skewed by one part of the interview:
e.g. for the core meaning of make (= produce, create), overuse in LINDSEI-FR due to picture description task NS: do/draw a portrait, do/paint a picture
he paints the picture of a beautiful woman NNS: make a portrait/a drawing/ a picture
there is a painter he’s making a portrait the portrait of a of a girl
23
Methodological implications
e.g. for the causative make + V structure, in LOCNEC 16 instances/42 involve look: he’s now repainting it making her look . much more attractive he makes her look . totally different makes her look very
glamorous clearly topic-induced by picture description which
elicits predictable patterns bears unduly on the overall results for that category not mirrored in LINDSEI-FR (1/11)
probably more appropriate to study the picture description (elicited) separately from the more spontaneous tasks
24
Where to from here? Possible avenues for further research
Complement quantitative analysis of native English HFVs by carrying out a similar analysis on learner data (requires preparation of the data, e.g. tagging of LINDSEI)
Combine corpus data with other types of data (e.g. elicitation) Complement qualitative analysis of make by carrying out similar
analyses of other HFVs reach better understanding of how these complex verbs are gradually
acquired in the interlanguage system Study other variables:
L1: Carry out transfer analysis on the same data + other learner populations Proficiency: longitudinal approach (data from other proficiency levels)
also help to understand the gradual evolution of the interlanguage system in time
25
Thank you!