Variation in the I r a q Vowels: Conservatives vs. Liberals Lauren Hall-Lew, Elizabeth Coppock, & Rebecca Starr Stanford University { dialect, rlstarr, coppock } @ stanford . edu NWAV36 , University of Pennsylvania, October 13, 2007

How do YOU say it?

Observation Iraq varies according to the second vowel: r/æ/k vs. r/ah/k [1], [2] Iraq also varies according to the first vowel: /ay/ vs. /Ih/ vs. /iy/ [1] Shapiro 1997 [2] Boberg 1997; 1999

Iraq varies according to the second vowel:

r/æ/k vs. r/ah/k [1], [2]

Iraq also varies according to the first vowel:

/ay/ vs. /Ih/ vs. /iy/

Hypotheses Second Vowel: Ir a q /æ/ variant indexes political conservativism. /ah/ variant indexes political liberalism. First Vowel: I raq /ay/ variant indexes political conservatism. /iy/ and /Ih/ index political liberalism.

Second Vowel: Ir a q

/æ/ variant indexes political conservativism.

/ah/ variant indexes political liberalism.

First Vowel: I raq

/ay/ variant indexes political conservatism.

/iy/ and /Ih/ index political liberalism.

Motivation Identity & Political Psychology intersect: “ ... intergroup conflict, conformity to group norms, the effects of low group status ...” [3] Recent Work: American identity, Patriotism, & Ethnicity [4] Identity & Phonological Variables potential resource for the expression of a political identity [3] Tajfel & Turner 1979; Huddy 2001 [4] Citrin et al, 1990 & 2000, cited in Huddy 2001

Identity & Political Psychology intersect:

“ ... intergroup conflict, conformity to group norms, the effects of low group status ...” [3]

Recent Work:

American identity, Patriotism, & Ethnicity [4]

Identity & Phonological Variables

potential resource for the expression of a political identity

Motivation “ Foreign (a)” variable, as in plaza , is often realized as /ah/ in US English due to attitudinal factors rather than phonological factors. [2] US English speakers evaluate /ah/ to be “more correct, educated, and sophisticated than /æ/ as a nativization of foreign (a).” [2] “ Respect engenders imitation [retention of foreign sounds]; disrespect integration [full nativization]” [5] [5] Weinreich 1968, cited & comments added in Boberg 1999

“ Foreign (a)” variable, as in plaza , is often realized as /ah/ in US English due to attitudinal factors rather than phonological factors. [2]

US English speakers evaluate /ah/ to be “more correct, educated, and sophisticated than /æ/ as a nativization of foreign (a).” [2]

“ Respect engenders imitation [retention of foreign sounds]; disrespect integration [full nativization]” [5]

Pilot Study: Methods Two networks with different political leanings: FoxNews and KQED (Bay Area NPR station) television and radio. Collected all instances of Iraq occurring in top news stories at the time, as well as in all news stories within a 24-hour period. Classified speakers as publicly conservative, publicly liberal, or unknown.

Two networks with different political leanings: FoxNews and KQED (Bay Area NPR station) television and radio.

Collected all instances of Iraq occurring in top news stories at the time, as well as in all news stories within a 24-hour period.

Classified speakers as publicly conservative, publicly liberal, or unknown.

Pilot Study: Data 686 tokens of Iraq 248 speakers 44 publicly conservative 32 publicly liberal 172 unknown Distribution of conservatives and liberals not significantly different on KQED or FoxNews No significant differences in Iraq pronunciation between networks Network data was combined

686 tokens of Iraq

248 speakers

44 publicly conservative

32 publicly liberal

172 unknown

Distribution of conservatives and liberals not significantly different on KQED or FoxNews

No significant differences in Iraq pronunciation between networks

Network data was combined

Pilot Study: Results Significant finding: 83% of conservatives, but only 31% of liberals, had at least one occurrence of the /æ/ variant. ( F =17.7, df=1, p < 0.0001)

Significant finding:

83% of conservatives, but only 31% of liberals, had at least one occurrence of the /æ/ variant. ( F =17.7, df=1, p < 0.0001)

Pilot Study: Other Results Military personnel: All 15 military personnel used /æ/ 14 of 15 used /ay/ for the first vowel, while all non-military speakers used /Ih/ The 1 military member who used /Ih/ was explicitly an anti-war Democrat, General Wesley Clark.

Military personnel:

All 15 military personnel used /æ/

14 of 15 used /ay/ for the first vowel, while all non-military speakers used /Ih/

The 1 military member who used /Ih/ was explicitly an anti-war Democrat, General Wesley Clark.

Pilot Study: Other Results Virtually all news readers and news anchors use /ah/, confirming the prediction by Boberg (1999) of /ah/ in more formal styles

Virtually all news readers and news anchors use /ah/, confirming the prediction by Boberg (1999) of /ah/ in more formal styles

Pilot Study: Interpretation Indications that both vowels in Iraq may index political orientation.

Indications that both vowels in Iraq may index political orientation.

Limitations of Pilot Study Political leanings of speakers not always clear. Sampling was not highly systematic. Not controlled for: Register Audience Regional dialect variation

Political leanings of speakers not always clear.

Sampling was not highly systematic.

Not controlled for:

Register

Audience

Regional dialect variation

Main Study: Speakers Members of the US House of Representatives: Representative of all dialect regions of the US Clear political affiliations

Members of the US House of Representatives:

Representative of all dialect regions of the US

Clear political affiliations

Data: Iraq Troop Surge Debate Three days of debate in February 2007 Debate topic: Resolution 63, stating that the House disapproves of troop surge All speeches aired on C-SPAN, posted online at http://iraq.armyofone.org Speeches given by 304 out of 435 total members of the House

Three days of debate in February 2007

Debate topic: Resolution 63, stating that the House disapproves of troop surge

All speeches aired on C-SPAN, posted online at http://iraq.armyofone.org

Speeches given by 304 out of 435 total members of the House

Surge Debate Study: Methods Total number of speakers included: 259 Only included speakers who said Iraq 3 times or more. Total number of tokens coded: N=1959 Mean = 8 tokens per speaker Median = 6 tokens per speaker

Total number of speakers included: 259

Only included speakers who said Iraq 3 times or more.

Total number of tokens coded: N=1959

Mean = 8 tokens per speaker

Median = 6 tokens per speaker

Speaker Factors Coded: Political Stance measures Party affiliation Vote on anti-surge resolution (to capture war stance) Economic and Social Liberalism rating: percentage ratings from http://www.ontheissues.org based on voting record Conservatives have a higher economic % Liberals have a higher social %

Party affiliation

Vote on anti-surge resolution (to capture war stance)

Economic and Social Liberalism rating:

percentage ratings from http://www.ontheissues.org

based on voting record

Conservatives have a higher economic %

Liberals have a higher social %

Speaker Factors Coded: Speech variety measures [+/-] Southern Accent /ay/ monophthongization perceptual measures Regional accent / Speech variety subjective, with some inter-rater judgments State they represent Region of the US, based on US Census Red/Blue status, based on 2004 election

[+/-] Southern Accent

/ay/ monophthongization

perceptual measures

Regional accent / Speech variety

subjective, with some inter-rater judgments

State they represent

Region of the US, based on US Census

Red/Blue status, based on 2004 election

Speaker Factors Coded: Other demographic info Ethnicity Particular attention to AAE speakers Age Here, 4 age categories, approx same N per cat. Sex Class Subjective classification

Ethnicity

Particular attention to AAE speakers

Age

Here, 4 age categories, approx same N per cat.

Sex Class

Subjective classification

Surge Debate Study: Methods Logistic regression analysis based on binary dependent variable: over 50% use of /æ/ in Iraq for any given speaker Only tokens of Iraq included Iraqi and Iran also coded for, but not grouped with Iraq for analysis.

Logistic regression analysis based on binary dependent variable:

over 50% use of /æ/ in Iraq for any given speaker

Only tokens of Iraq included

Iraqi and Iran also coded for, but not grouped with Iraq for analysis.

%/æ/ use by number of speakers 85% of speakers consistently pronounce Iraq always with one vowel or the other. 15% of the speakers varied in their pronunciation of Iraq. This does not include Iraqi , which was more likely to be pronounced differently In contrast to news readers from pilot, Representatives favor /æ/ over /ah/

85% of speakers consistently pronounce Iraq always with one vowel or the other.

15% of the speakers varied in their pronunciation of Iraq.

This does not include Iraqi , which was more likely to be pronounced differently

In contrast to news readers from pilot, Representatives favor /æ/ over /ah/

%/æ/ use by number of speakers

Results by Party Political party is a highly significant predictor of the Iraq pronunciation correlation=7.29, p=0.025 Republicans are statistically more likely to use the /æ/ variant Political party significantly predicts Iraq pronunciation even when controlling for region, regional accent, age group, sex class, and ethnicity.

Political party is a highly significant predictor of the Iraq pronunciation

correlation=7.29, p=0.025

Republicans are statistically more likely to use the /æ/ variant

Political party significantly predicts Iraq pronunciation even when controlling for region, regional accent, age group, sex class, and ethnicity.

Vowel Use by Political Party

Economic & Social Liberalism http://www.ontheissues.org Economic & Social Liberalism scores based on voting record. High Economic Lib = more conservative High Social Lib = more liberal

http://www.ontheissues.org

Economic & Social Liberalism scores based on voting record.

High Economic Lib = more conservative

High Social Lib

= more liberal

Economic & Social Liberalism Economic Liberalism score: Speakers with high Econ % (conservatives) are significantly more likely to say /æ/ in Iraq, when controlling for all factors except party. consistent with our political party findings No significant effect of Social Liberalism score

Economic Liberalism score:

Speakers with high Econ % (conservatives) are significantly more likely to say /æ/ in Iraq, when controlling for all factors except party.

consistent with our political party findings

No significant effect of Social Liberalism score

Results by Age Age grouped into 4 categories, each including approximately 68 speakers: ≤ 1944, 1945-1949, 1950-1957, ≥ 1958 No significant effect of age, although older people appear to use less /æ/. [1] Political Party within each age category: consistent pattern, not significant. [1] cf. Shapiro 1997, that /æ/ is the earlier form

Age grouped into 4 categories, each including approximately 68 speakers:

≤ 1944, 1945-1949, 1950-1957, ≥ 1958

No significant effect of age, although older people appear to use less /æ/. [1]

Political Party within each age category: consistent pattern, not significant.

Results by Region Test according to: region of representation the presence of monophthongized /ay/ in closed syllables Potential interaction between region, surge stance, and political party: More pro-surge conservatives in Southern states

Test according to:

region of representation

the presence of monophthongized /ay/ in closed syllables

Potential interaction between region, surge stance, and political party:

More pro-surge conservatives in Southern states

Results by Region 4 Regions: South Midwest Northeast West Defined by the U.S. Census Bureau

4 Regions:

South

Midwest

Northeast

West

Defined by the U.S. Census Bureau

Results by Region No significant effect of Region: No effect of southern accent, broadly coded any presence of monophthong /ay/ No effect of southern accent, narrowly coded all /ay/ are monophthongized Standard vs. Regional accents: not significant

No significant effect of Region:

No effect of southern accent, broadly coded

any presence of monophthong /ay/

No effect of southern accent, narrowly coded

all /ay/ are monophthongized

Standard vs. Regional accents: not significant

Variation by Speaker Sex 40 of 271 (15%) Congresspeople in this sample are Congresswomen. Women tend to use /ah/ more than /æ/, but not significantly more than men. (p < .08 as a main effect) Party remains significant when controlling for speaker sex class.

40 of 271 (15%) Congresspeople in this sample are Congresswomen.

Women tend to use /ah/ more than /æ/, but not significantly more than men. (p < .08 as a main effect)

Party remains significant when controlling for speaker sex class.

What About The First Vowel? Variants we heard for the first vowel in Iraq : /Ih/ /iy/ /ay/ Grouped /Ih/ and /iy/ together as (i) in final analysis because they were difficult to distinguish perceptually, and /iy/ was rare.

Variants we heard for the first vowel in Iraq :

/Ih/

/iy/

/ay/

Grouped /Ih/ and /iy/ together as (i) in final analysis because they were difficult to distinguish perceptually, and /iy/ was rare.

The First Vowel in ‘Iraq’

The First Vowel in Iraq Same bimodal distribution as in the histogram for % /æ/: 46 Congresspeople have >50% /ay/ 141 Congresspeople have 50% or less /ay/ 18% of the speakers vary between /ay/ and (i) Within all second-vowel /æ/ tokens, 24% use first-vowel /ay/ 2 speakers say Iraq as /ay/r/ah/q categorically.

Same bimodal distribution as in the histogram for % /æ/:

46 Congresspeople have >50% /ay/

141 Congresspeople have 50% or less /ay/

18% of the speakers vary between /ay/ and (i)

Within all second-vowel /æ/ tokens, 24% use first-vowel /ay/

2 speakers say Iraq as /ay/r/ah/q categorically.

The First Vowel in Iraq There is no statistically significant predictive variable for the first vowel. Non-significant trends: Within people who have Southern accents, /ay/ correlates with anti-surge Omitting AAE speakers overall, within non - Southern accents, /ay/ correlates with pro-surge

There is no statistically significant predictive variable for the first vowel.

Non-significant trends:

Within people who have Southern accents, /ay/ correlates with anti-surge

Omitting AAE speakers overall, within non - Southern accents, /ay/ correlates with pro-surge

SUMMARY In the US House of Representatives, variation of the second vowel in Iraq correlates with the Political Party of the speaker. /ah/ correlates with Democrats /æ/ correlates with Republicans No factors significantly predict the pronunciation of the first vowel.

In the US House of Representatives, variation of the second vowel in Iraq correlates with the Political Party of the speaker.

/ah/ correlates with Democrats

/æ/ correlates with Republicans

No factors significantly predict the pronunciation of the first vowel.

Indexing Political Identity Implications for variation Potential complications Complex: How to operationalize? Political psychology [6] Political views may be a critical aspect of an individual’s identity Correlated with Political Identity: Age (old vs. young) Class Mobility & Social Networks Personality Type & Value System Religion [6] See, e.g.: Brewer 2001; Conover and Feldman 1981; De Graaf et al. 1995; Huddy 2001; Huddy and Khatib 2007; Mackenzie 1978; Kymlicka 2001, cited in Charney 2003

Implications for variation

Potential complications

Complex: How to operationalize?

Political psychology [6]

Political views may be a critical aspect of an individual’s identity

Correlated with Political Identity:

Age (old vs. young)

Class Mobility & Social Networks

Personality Type & Value System

Religion

Operationalizing Political Identity 3 measures here: Political Party Troop Surge Stance Econ/Social Liberalism All 3 correlated highly All 3 also differed in terms of predictive power.

3 measures here:

Political Party

Troop Surge Stance

Econ/Social Liberalism

All 3 correlated highly

All 3 also differed in terms of predictive power.

Operationalizing Political Identity Anti-Surge Republicans (N=10) Pro-Surge Democrats (N=1) 8 of the Republicans use /æ/ & the Democrat uses /ah/. 2 Republicans have less than 50% /æ/, so we can take a closer look at them.

Anti-Surge Republicans (N=10)

Pro-Surge Democrats (N=1)

8 of the Republicans use /æ/ & the Democrat uses /ah/.

2 Republicans have less than 50% /æ/, so we can take a closer look at them.

Future Directions People whose voting record diverges from their political party’s Synchronic patterns of variation within individuals speech communities What's going on with /ay/r/ah/k? Will Iraq data from Congress change as the war continues?

People whose voting record diverges from their political party’s

Synchronic patterns of variation

within individuals

speech communities

What's going on with /ay/r/ah/k?

Will Iraq data from Congress change as the war continues?

Future Directions Other Terms: Iraqi: more variable within speaker, what’s the pattern? Iran & Iranian: do their vowels pattern like Iraq ’s? Vietnam: consider data from the 60s/70s vs. now Work with non-politicians: Attitudinal Surveys ( cf. Boberg 1999) Production Experiment Perception Experiment

Other Terms:

Iraqi: more variable within speaker, what’s the pattern?

Iran & Iranian: do their vowels pattern like Iraq ’s?

Vietnam: consider data from the 60s/70s vs. now

Work with non-politicians:

Attitudinal Surveys ( cf. Boberg 1999)

Production Experiment

Perception Experiment

Thank You! We would like to thank: Ben Munson Laura Staum Casasanto www.youtube.com “ Eric,” Canadian software architect, age 28, a.k.a. the guy who assembled the videos on: iraq.armyofone.com

We would like to thank:

Ben Munson

Laura Staum Casasanto

www.youtube.com

“ Eric,” Canadian software architect, age 28, a.k.a. the guy who assembled the videos on: iraq.armyofone.com

References Boberg, C. (1997). Variation and change in the Nativization of Foreign (a) in English. PhD thesis, University of Pennsylvania. Boberg, C. (1999). The Attitudinal Component of variation in American English foreign (a) Nativization. Journal of Language and Social Psychology , 18:49–61. Brewer, M. B. (2001). The many faces of social identity: Implications for political psychology. Political Psychology , 22:115–25. Charney, E. (2003). Identity and Liberal Nationalism. American Political Science Review. 97(2): 295-310. Conover, P. J. and Feldman, S. (1981). The origins and meaning of liberal/ conservative self-identification. American Journal of Political Science , 25:617–45. 2 De Graaf, N. D., Nieuwbeerta, P., and Heath, A. (1995). Class mobility and political preferences: Individual and contextual effects. The American Journal of Sociology , 100:997–1027. Huddy, L. (2001). From social to political identity: A critical examination of social identity theory. Political Psychology , 22:127–56. Huddy, L. and Khatib, N. (2007). American patriotism, national identity, and political involvement. American Journal of Political Science , 51:63–77. Kymlicka, W. (2001). Politics in the Vernacular: Nationalism, Multiculturalism, and Citizenship. Oxford: Oxford University Press. Mackenzie, W. J. M. (1978). Political Identity . New York: St. Martin’s Press. The Pew Research Center (2004). Overview: News audiences increasingly politicized. http://people-press.org/reports/display.php3?ReportID=215 Rhodebeck, L. A. (1993). The Politics of Greed? Political Preferences among the Elderly. The Journal of Politics, 55(2):342-364. Shapiro, M. (1997). Broad and Flat A in Marked Words. American Speech. 72(4): 437-439.

Boberg, C. (1997). Variation and change in the Nativization of Foreign (a) in English. PhD thesis, University of Pennsylvania.

Boberg, C. (1999). The Attitudinal Component of variation in American English foreign (a) Nativization. Journal of Language and Social Psychology , 18:49–61.

Brewer, M. B. (2001). The many faces of social identity: Implications for political psychology. Political Psychology , 22:115–25.

Charney, E. (2003). Identity and Liberal Nationalism. American Political Science Review. 97(2): 295-310.

Conover, P. J. and Feldman, S. (1981). The origins and meaning of liberal/ conservative self-identification. American Journal of Political Science , 25:617–45. 2

De Graaf, N. D., Nieuwbeerta, P., and Heath, A. (1995). Class mobility and political preferences: Individual and contextual effects. The American Journal of Sociology , 100:997–1027.

Huddy, L. (2001). From social to political identity: A critical examination of social identity theory. Political Psychology , 22:127–56.

Huddy, L. and Khatib, N. (2007). American patriotism, national identity, and political involvement. American Journal of Political Science , 51:63–77.

Kymlicka, W. (2001). Politics in the Vernacular: Nationalism, Multiculturalism, and Citizenship. Oxford: Oxford University Press.

Mackenzie, W. J. M. (1978). Political Identity . New York: St. Martin’s Press. The Pew Research Center (2004). Overview: News audiences increasingly politicized. http://people-press.org/reports/display.php3?ReportID=215

Rhodebeck, L. A. (1993). The Politics of Greed? Political Preferences among the Elderly. The Journal of Politics, 55(2):342-364.

Shapiro, M. (1997). Broad and Flat A in Marked Words. American Speech. 72(4): 437-439.

Variation by Ethnicity 23 of 271 (8.5%) Congresspeople in this sample were identified as speakers of AAE. Given the low N, no significant effect was found for the /ah/ vs. /æ/ variable between AAE and non-AAE speakers. Impressionistically, AAE speakers tended to favor /ay/ for the first vowel more than non-AAE speakers, who appeared to favor /iy/ or /ih/. However, no significant effect was found.

23 of 271 (8.5%) Congresspeople in this sample were identified as speakers of AAE.

Given the low N, no significant effect was found for the /ah/ vs. /æ/ variable between AAE and non-AAE speakers.

Impressionistically, AAE speakers tended to favor /ay/ for the first vowel more than non-AAE speakers, who appeared to favor /iy/ or /ih/.

However, no significant effect was found.

What about Shrubby? George W. Bush is a little bit variable Based on data from Pilot Study ( NPR and FoxNewsRadio ): speeches, interviews, etc. From N=20 tokens, 2 = /ah/ & 18 = /æ/ Possible gradual switch from /æ/ to /ah/? Call for longitudinal analysis!!

George W. Bush is a little bit variable

Based on data from Pilot Study ( NPR and FoxNewsRadio ): speeches, interviews, etc.

From N=20 tokens, 2 = /ah/ & 18 = /æ/

Possible gradual switch from /æ/ to /ah/?

Call for longitudinal analysis!!

No Interactions Are there interactions between any of the variables? Checking for interactions: Geography is correlated with Party, Party is correlated with Vowel use, But Geography is not statistically correlated with Vowel use.

Are there interactions between any of the variables?

Checking for interactions:

Geography is correlated with Party,

Party is correlated with Vowel use,

But Geography is not statistically correlated with Vowel use.

No Interactions When controlling for all other factors, we still get a highly significant effect for political party. No interaction between factors is significant for predicting vowel use. In an additive statistical model there is only an effect from political party.

When controlling for all other factors, we still get a highly significant effect for political party.

No interaction between factors is significant for predicting vowel use.

In an additive statistical model there is only an effect from political party.