Upload
doananh
View
229
Download
2
Embed Size (px)
Citation preview
MicroMicroMicroMicro---- and Macroand Macroand Macroand Macro----Perspectives Perspectives Perspectives Perspectives on the Rise of Englishon the Rise of Englishon the Rise of Englishon the Rise of Englishon the Rise of Englishon the Rise of Englishon the Rise of Englishon the Rise of English
Eline Zenner (FWO Flanders), Eline Zenner (FWO Flanders), Dirk Geeraerts & Dirk SpeelmanDirk Geeraerts & Dirk Speelman
University of LeuvenUniversity of LeuvenRU Quantitative Lexicology and Variational LinguisticsRU Quantitative Lexicology and Variational Linguistics
Overview
Background and Research Question
Methodology
Results
ICLCE 4, Osnabruck, Germany (July 2011)
Conclusion
Background: The Spread of English
Crystal (2003: 61), using Kachru's circles (Kachru 1986):- inner circle: 300 million – 500 million speakers of English- outer circle: 320 million – 380 million- expanding circle: 500 million – 1000 million
Focusing on Europe, the spread of the English language presentsitself at two separate levels, each studied in a specific paradigmFocusing on Europe, the spread of the English language presentsitself at two separate levels, each studied in a specific paradigm
MACRO-levelthe spread of English as language for (international) communication
MICRO-levelthe spread of English as resource for borrowing
ICLCE 4, Osnabruck, Germany (July 2011)
ELF paradigm
anglicismresearch
ELF vs. Anglicism Research
� typically, MACRO and MICRO spread are dealt with separately� lack of communication between both lines of research
BUT: (1) how different are the phenomena under scrutiny really?
~ how different are the mechanisms underlying the spread of English as a language for (international) communication and the English as a language for (international) communication and the spread of English as resource for borrowing? E.g.: diachronic evolution and contextual distribution
(2) methodological issues: how to make the comparison in a reliable way? Following the Cognitive Sociolinguistic approach, we advocate
- the use of large datasets- employing a multivariate approach- verifying the impact of effects with inferential statistics
ICLCE 4, Osnabruck, Germany (July 2011)
� TOPIC: language choice on MACRO and MICRO level
Research Questions
MICRO: what determines the choice for inserting English elements in ones mother tongue?
how comparable are the mechanisms underlying language choice on both levels and how can we
MACRO: what determines the choice for English as a language for communication?
language choice on both levels and how can we measure this in a reliable way?
ICLCE 4, Osnabruck, Germany (July 2011)
CASE STUDY: the use of English in two Dutch job ad magazines
Research Questions
MICRO: what determines the choice for inserting English elements in ones mother tongue?
MACRO: what determines the choice for English as a language for communication?
BODY: what determines the choice for using English as matrix language of an ad?
TITLE: what determines the choice for using English elements in job titles?
ICLCE 4, Osnabruck, Germany (July 2011)
Research Questions
BODY: what determines the choice for using English as matrix language of an ad?
TITLE: what determines the choice for using English elements in job titles?
MICRO: what determines the choice for inserting English elements in ones mother tongue?
MACRO: what determines the choice for English as a language for communication?
ICLCE 4, Osnabruck, Germany (July 2011)
how comparable are the mechanisms underlying language choice on BODY and TITLE level and how
can we measure this in a reliable way?
Overview
Background and Research Question
Methodology
Results
Conclusion
ICLCE 4, Osnabruck, Germany (July 2011)
METHOD
1. collect data
2. determine language choice
3. identify possible determinants of language choice (which apply to both levels of analysis)
why job ads?which job ads?
to both levels of analysis)
4. measure the impact of the different determinants at both levels
5. compare the impact of the determinants at both levels
ICLCE 4, Osnabruck, Germany (July 2011)
Why Job Ads?
1. as previous studies show: both BODY and TITLE
2. accessibility of data
ICLCE 4, Osnabruck, Germany (July 2011)
What Job Ads?
• 16 622 job ads
• period of 40 years (1970 to 2008)� asymmetrical collection
• 2 magazines:
JulyApril
1970
1989• 2 magazines:
Vacature (Belgian Dutch)Intermediair (Netherlandic Dutch)
NDNDNDND BDBDBDBD
2008
n=13508
ICLCE 4, Osnabruck, Germany (July 2011)
METHOD
1. collect data
2. determine language choice
3. identify possible determinants of language choice (which apply to both levels of analysis)
at BODY levelat TITLE level
to both levels of analysis)
4. measure the impact of the different determinants at both levels
5. compare the impact of the determinants at both levels
ICLCE 4, Osnabruck, Germany (July 2011)
Measuring the Spread of English: language choice on two levels
BODY: what is the matrix language of the ad?
ICLCE 4, Osnabruck, Germany (July 2011)
TITLE: is English used in the job title?
(note: only elements recognizable as English by naive Dutch NS)
METHOD
1. collect data
2. determine language choice
3. identify possible determinants of language choice (which apply to both levels of analysis)to both levels of analysis)
4. measure the impact of the different determinants at both levels
5. compare the impact of the determinants at both levels
ICLCE 4, Osnabruck, Germany (July 2011)
Possible Determinants for Language Choice
CULTURAL
CONTEXTUAL
INTERNALINTERNAL
ICLCE 4, Osnabruck, Germany (July 2011)
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
1989-1995
1995-2000
-Job Content
-HR Agency
-Size
1995-2000
2000-2005
2005-2008
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
-Job Content
-HR Agency
-SizeNetherlandic DutchBelgian Dutch (Flanders)
-Period
-Region
-Branch of industryDegree of foreign influence (Geeraerts 2004)
Belgian Dutch: - ambivalent relationship with French
Possible Determinants for Language Choice
-Job Content
-HR Agency
-Size
Belgian Dutch: - ambivalent relationship with Frenchoil-stain: purist tendency for English
- English L3 in education
Netherlandic D: - English L2 in education- no history of foreign rule
open attitude towards foreign languages� more English in Netherlandic Dutch?
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
IT company
-Job Content
-HR Agency
-Size
government
Information Technology Industry
Service & HR Food
Sales Health
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
Pharmaceutical Industry Education & Research
Finance Government
Energy & Safety unknown
-Job Content
-HR Agency
-Size
Information Technology Industry
Service & HR Food
Sales Health
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
modern, internationally oriented branches
Pharmaceutical Industry Education & Research
Finance Government
Energy & Safety unknown
-Job Content
-HR Agency
-Size"traditional" and locally
based branches
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
sales job
-Job Content
-HR Agency
-Sizefinancial job
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
Information Technology
Human Resource Management
Sales & Marketing
Financial Jobs
modern, internationally oriented branches
-Job Content
-HR Agency
-Size
Financial Jobs
Technical Jobs
Administration & Communication
Public Sector
(Other)
"traditional" branches
-Period
-Region
-Branch of industry
Possible Determinants for Language Choice
0.6
0.8
1.0
it
hr
sales
finance
techn
adm.com
public
other
admin
fin
techn techn techn techn
public
public
0.6
0.8
1.0
it
hr
sales
finance
techn
adm.com
public
other
admin
fin
techn techn techn techn
public
public
-Job Content
-HR Agency
-Sizeit srv.HR sales fin phrm indr erg.saf food edu.rch health govern.cult unknown
0.0
0.2
0.4
it
sales
sales sales sales sales sales
techn
sales
it srv.HR sales fin phrm indr erg.saf food edu.rch health govern.cult unknown
0.0
0.2
0.4
it
sales
sales sales sales sales sales
techn
sales
-Period
-Region
-Branch of industryno
Possible Determinants for Language Choice
-Job Content
-HR Agency
-Size
yes
with external agency: more English
(~ survey JobAt)
-Period
-Region
-Branch of industrysmall vs.
large
Possible Determinants for Language Choice
-Job Content
-HR Agency
-Size
larger add � bigger company? � more English?
METHOD
1. collect data
2. determine language choice
3. identify possible determinants of language choice (which apply to both levels of analysis)to both levels of analysis)
4. measure the impact of the different determinants at both levels
5. compare the impact of the determinants at both levels
ICLCE 4, Osnabruck, Germany (July 2011)
Statistical AnalysesWhat we wish to explain:
language choice (English or not?)
What features we think help explain this:period, region, branch of industry, job content, HR agency
Method to assess impact:Logistic Regression AnalysisLogistic Regression Analysis
BUT: assess impact both for BODY and for TITLE level
two regression models1. impact of predictors on language choice at BODY level2. impact of predictors on language choice at TITLE level
ICLCE 4, Osnabruck, Germany (July 2011)
METHOD
1. collect data
2. determine language choice
3. identify possible determinants of language choice (which apply to both levels of analysis)to both levels of analysis)
4. measure the impact of the different determinants at both levels
5. compare the impact of the determinants at both levels
ICLCE 4, Osnabruck, Germany (July 2011)
Comparison of Regression Analyses
RQ: how comparable are the mechanisms underlying language choice on both levels?
BODYBODYBODYBODY MODELMODELMODELMODEL TITLE TITLE TITLE TITLE MODELMODELMODELMODEL
COMPARE:• overall amount of English on each level (descriptive)• impact of each of the predictors• behaviour of each of the predictors
using graphs based on confidence intervals(the actual regression models aren't presented)
ICLCE 4, Osnabruck, Germany (July 2011)
Overview
Background and Research Question
Methodology
Resultsoverall amount of English on each level (descriptive)• overall amount of English on each level (descriptive)
• impact of each of the predictors• behaviour of each of the predictors
using graphs based on confidence intervals(the actual regression models aren't presented)
AttenuationConclusion
ICLCE 4, Osnabruck, Germany (July 2011)
Results (1): How much English is used?
BODY level TITLE level
English ads:1093
Titles with English:4765
Total number of ads:13 508
Total number of titles:1309413 508 13094
% English:8.1%
% English:36.4%
ICLCE 4, Osnabruck, Germany (July 2011)
� the spread of English as a language of communication is less wide-spread (so far) than the intrusion of English elements in given language
Overview
Background and Research Question
Methodology
Resultsoverall amount of English on each level (descriptive)• overall amount of English on each level (descriptive)
• impact of each of the predictors• behaviour of each of the predictors
using graphs based on confidence intervals(the actual regression models aren't presented)
AttenuationConclusion
ICLCE 4, Osnabruck, Germany (July 2011)
Results (2): Presenting the Models
BODY model TITLE model
selected all predictors all predictors
pseudo R² 15.7% 35.2%
C 0.76 0.81
ICLCE 4, Osnabruck, Germany (July 2011)
� the spread of English as a language of communication is less tied to our predictors than the intrusion of English elements in Dutch. This might be linked to the general sparseness of ads with English as a matrix language (8% vs. 36%).
yearGr
branch
funcGr
Results (3): Impact of Predictors in ANOVA
branch
yearGr
funcGr
BODYBODYBODYBODY MODELMODELMODELMODEL TITLE TITLE TITLE TITLE MODELMODELMODELMODEL
χ2 − df
0 200 400 600 800
funcGr
region
hrAgency_Gr
sizeGr
χ2 − df
0 50 100 150 200 250 300 350
funcGr
hrAgency_Gr
region
sizeGr
� The diachronic parameter is far stronger for the TITLE-model, where the effect of branch of industry is stronger for the BODY-model. A more detailed analysis of the actual patterns is needed:
Overview
Background and Research Question
Methodology
Resultsoverall amount of English on each level (descriptive)• overall amount of English on each level (descriptive)
• impact of each of the predictors• behaviour of each of the predictors
using graphs based on confidence intervals(the actual regression models aren't presented)
Conclusion
ICLCE 4, Osnabruck, Germany (July 2011)
Results (4): Impact of the PredictorsExplaining the method: the effect of Period on choice for English at BODY-level
<<
< n
oEng
en
g >
>> 2
3
reference value
behaviour of other value on the predictor
(with confint)
sign. more chance of English than reference
value
ICLCE 4, Osnabruck, Germany (July 2011)
<<
< n
oEng
en
g >
>>
-10
1
BODY
reference value(i.c. 1989-1995)
reference value1989-1995
Results (4): Impact of the Predictors<
<<
noE
ng
eng
>>
> 23
Explaining the method: the effect of Period on choice for English at BODY-level
reference value
behaviour of other value on the predictor
(with confint)
ICLCE 4, Osnabruck, Germany (July 2011)
<<
< n
oEng
en
g >
>>
-10
1
BODY
no sign difference with reference value
reference value1989-1995
reference value(i.c. 1989-1995)
Results (4): Impact of the Predictors<
<<
noE
ng
eng
>>
> 23
Explaining the method: the effect of Period on choice for English at BODY-level
reference value behaviour of other
ICLCE 4, Osnabruck, Germany (July 2011)
<<
< n
oEng
en
g >
>>
-10
1
BODY
reference value(i.c. 1989-1995)
behaviour of other value on the predictor
(with confint)
reference value1989-1995
sign less chance of English than reference
value
Results (4): Impact of the Predictors<
<<
noE
ng
eng
>>
> 23
Explaining the method: adding the TITLE-level
ICLCE 4, Osnabruck, Germany (July 2011)
<<
< n
oEng
en
g >
>>
-10
1
BODYTITLE
reference value1989-1995
if overlap � similar behaviour
if no overlap � difference in effect
Results (4): Impact of the Predictors<
<<
noE
ng
eng
>>
> 23
DIACHRONIC EVOLUTION
ICLCE 4, Osnabruck, Germany (July 2011)
<<
< n
oEng
en
g >
>>
-10
1
BODYTITLE
reference value1989-1995
Results (4): Impact of the PredictorsDIACHRONIC EVOLUTION
0.6
0.8
1.0
� the rise of English as a language for communication is less steep than the cline for the intrusion of English elements in Dutch. Macro-level change as more drastic, and hence slower?
ICLCE 4, Osnabruck, Germany (July 2011)
BODYTITLE
% E
ng
0.0
0.2
0.4
0.6
1989-1995 1995-2000 2000-2005 2005-2008
Results (4): Impact of the PredictorsREGIONAL VARIATION
<<
< n
oEng
en
g >
>>
0.5
1.0
1.5
reference valueBELGIAN DUTCH
<<
< n
oEng
en
g >
>>
-1.5
-1.0
-0.5
0.0
BODYTITLE
� there is a clear difference in behaviour: compared to Netherlandic Dutch, we see a purist tendency for Belgian Dutch at the TITLE-level, but an embracing of the BODY-level use of English
Results (4): Impact of the PredictorsBRANCH branch of industry
<<
< n
oEng
en
g >
>>
2-1
01
ICLCE 4, Osnabruck, Germany (July 2011)
reference valueIT
<<
< n
oEng
en
g >
>>
-5-4
-3-2
BODYTITLE
� very similar patterns, but notice how government and social profit agencies are more reluctant to switch to English as a language for communication than to use English in job descriptions
Results (4): Impact of the PredictorsJOB CONTENT
<<
< n
oEng
en
g >
>>
2-1
01
ICLCE 4, Osnabruck, Germany (July 2011)
reference valueIT
<<
< n
oEng
en
g >
>>
-5-4
-3-2
BODYTITLE
� similar behaviour for public jobs as for government as branch of industry (mind: attenuation below), but bigger differences between groups on TITLE-level
Results (4): Impact of the PredictorsHR AGENCY
<<
< n
oEng
en
g >
>>
0.5
1.0
ICLCE 4, Osnabruck, Germany (July 2011)
reference value"no external involvement"
<<
< n
oEng
en
g >
>>
-1.0
-0.5
0.0
BODYTITLE
� similar and expected behaviour
Results (4): Impact of the PredictorsSIZE
<<
< n
oEng
en
g >
>>
0.5
1.0
ICLCE 4, Osnabruck, Germany (July 2011)
reference valuesmall ads
<<
< n
oEng
en
g >
>>
-0.5
0.0
0.5
BODYTITLE
� similar and expected behaviour(but impure measure)
Overview
Background and Research Question
Methodology
Results
Conclusion
ICLCE 4, Osnabruck, Germany (July 2011)
ConclusionAmount of English used:
the use of English as language for communication (BODY; MACRO) is not as widely spread as the use of English elements in Dutch (TITLE; MICRO)
Impact of the predictors:bigger impact of period for TITLE vs. bigger impact of branch of industry for BODYindustry for BODY
Effect size:overall, results are very much in line with expectations, and there appears to be a large degree of overlap between both levels (micro and macro)
in our dataset, the mechanisms underlying the choice for English as a language of communication and the choice for English elements in Dutch are highly comparable
ICLCE 4, Osnabruck, Germany (July 2011)
For more information:For more information:http://wwwling.arts.kuleuven.be/qlvl