arabic accent perception and prejudice in the usa

ARABIC ACCENT PERCEPTION AND PREJUDICE IN THE USA
A Thesis
submitted to the Faculty of the
Graduate School of Arts and Sciences
of Georgetown University
in partial fulfillment of the requirements for the
degree of
Master of Arts
in Linguistics
By
Greg J. Niedt, B.A.
Washington, D.C.
April 26, 2011
Copyright 2011 by Greg J. Niedt
All Rights Reserved
ii
ARABIC ACCENT PERCEPTION AND PREJUDICE IN THE USA
Greg J. Niedt, B.A.
Thesis Advisor: Robert J. Podesva, Ph.D.
ABSTRACT
This paper reports the results of a study investigating the extent to which native speakers of
American English associate specific personal qualities to native speakers of Arabic based solely
on auditory information. 234 participants, recruited through online social networks, heard sound
clips of Arabic-accented English and were asked to rate each speaker from 1 to 5 in categories
such as ―assertive‖, ―intelligent‖, and ―religious‖. Responses were analyzed to determine which
demographic factors, such as sex, race, and political/religious self-identification, were
statistically significant. The results were found to be significant for some of the respondent
groups, in certain qualities: the paper illustrates which are seen as particularly related to Arabic
accent, as opposed to ―non-native‖. The survey results are combined with interviews with native
Arabic speakers on their experiences as speakers of accented English in the USA. The paper
attempts to construct a picture of the American language attitudes towards Arabic, highlight
stereotypes and prejudices that are drawn out through linguistic interaction, and offer directions
for further research beyond this initial study.
iii
The research and writing of this thesis would not have been possible
without the continuous support of my family, friends, and colleagues.
I dedicate the research and writing of this thesis to
Bruce and Kathleen Niedt, Angela Carothers, Corinne Seals, Anna Marie Trester,
and everyone else who participated in making this a reality.
I also dedicate it to anyone that might be better served
from the results presented herein.
My sincerest thanks,
Greg J. Niedt
iv
TABLE OF CONTENTS
Introduction............................................................................................................................................... 1
The Arab and Arabic-speaking community in the USA ..................................................................... 6
The Study: Qualitative Methodology .................................................................................................. 10
Qualitative Data ..................................................................................................................................... 11
The Study: Quantitative Methodology ............................................................................................... 17
Quantitative Data ................................................................................................................................... 21
Analysis ................................................................................................................................................... 22
Comparing accents for levels of significance .............................................................................. 22
Examining particular quality ratings across respondent groups ................................................ 24
Effects of different subcategories of sound clip .......................................................................... 29
Interpreting the data ........................................................................................................................ 32
Limitations and Solutions .................................................................................................................... 36
Conclusions ............................................................................................................................................ 39
Appendix A: Survey ............................................................................................................................. 42
Appendix B: Survey versions and order of sound clips .................................................................... 45
Appendix C: Respondent pool broken down by demographic category.......................................... 46
Appendix D: Demographic factors and factor groups in least-squares regression; graphs of nonstatistically significant responses for Arabic accent by demographic group ............................ 47
Bibliography ........................................................................................................................................... 52
v
Introduction
The question of foreign accent is one of great interest for the purpose of discussing how speakers
of one language acclimate to another in terms of both the way those speakers produce the new
language, and the way in which their accents are thereby perceived. Furthermore, from mastery
of a language comes easier integration into a new society. Dennis R. Preston (2004) points out,
―Language attitudes are a significant part of how we assess one another. Understanding folk
attitudes towards ways of speaking contributes to our knowledge of how speech can influence…
critically important matters for maintaining equality in a democratic society‖ (480). What are the
social and psychological effects of making judgments based on the sound of someone‘s voice?
Does this kind of linguistic gatekeeping prevent speakers with particular variations in their
language from getting jobs, finding houses, receiving services?
In the context of mainstream American language ideology (one that emphasizes
monolingual English use despite the diverse multiculturalism of the nation), native speakers are
often positioned above immigrant or non-native speakers. Stereotypes arise from a priori
―knowledge‖ about various groups, and when a non-native speaker‘s identity is determined
through symbols (which can include language), native speakers may place negative assumptions
upon that person. With the marked increase of media reporting on the Middle East and the Arab
world, and increased visibility of a Muslim minority in the United States, the Arabic-speaking
community deserves some attention in this regard. Some of the most severe stereotypes—
terrorist, extremist, misogynist—are reserved for members of this community. If language can be
linked to a particular group in an American listener‘s mind, are they hearing Arabic-influenced
features, and in turn evaluating speakers more negatively than they otherwise would? If so, how
and why is this occurring?
1
There has been extensive research conducted on the processes by which particular
features of a native language (or L1) become present in the speaker‘s newly-acquired language
(or L2), and on the phonetic features that constitute this or that accent. However, this study is
more concerned with the social perception of accented speakers by the natives of the language
they are using. More recent studies (see for example: Major 2010; Rindal 2010; Derwing and
Munro 2009; Ibrahim, Evatar, and Leikin 2008; Rahman 2008) have established the hardship of
overcoming the patterns of one‘s original language and the social or psychological pressures this
can create for speakers. Al-Issa (2003) discusses the ultimate impossibility—or at least,
improbability—of distancing oneself from the social norms of the L1 when attempting to
communicate in the L2. He defines the process of carrying over these norms as ―sociocultural
transfer‖, and quotes Thomas (1984) in saying this transfer results in ―sociopragmatic failures…
the mismatch which arises from cross-culturally different assessments within the social
parameters affecting linguistic choice, etc.‖ Similarly, the ways in which people evaluate their
own speech and that of others are ―culturally inherited‖ and ―consist of different values,
preferred communication styles, expectations, and interpretations‖ (al-Issa 2003: 581). The L2
speaker must undertake a supreme effort to overcome the choices in his or her speech that a
native speaker perceives as anomalous.
It is one thing to discuss whether a native speaker of that L2 senses these anomalies
phonologically, syntactically, or even idiomatically; but it is quite another to discuss what effect
that perception will have on L1 listeners with whom the speaker is interacting. In a multicultural
society with an unprecedented level of media involvement in everyday life such as we see in the
United States, we are constantly subject to the opinions of others as they match up with
backgrounds, ethnicities, and perhaps even languages. A great deal of linguistic research
2
(beginning with Labov 1972; see also for example Ochs and Capps 1996; Eckert and Rickford
2001) has been done on how [ethnic] identity is co-constructed in interaction, in a delicate ballet
between participants that relies on communicative cues to constantly re-evaluate self and others.
A linguist must concede, though, that the average person may not mediate their evaluative
process with this level of metalinguistic awareness, and holds set perceptions about identity in
general, particularly ethnicity. Rampton (1995) states that in a given interaction, ethnicity is
almost always marked as either neutral (as an in-group feature, where it is a non-issue) or
negative (as an intergroup feature, since the difference in speech style is salient). Two
complications of this concept are that even within ethnicities, there is plenty of demographic
variation that can cause speaker bias; and that there is often a visual component to the perception
of ethnicity. This latter point is especially important, as it is a significant factor in creating the
idea of ethnicity as an unalterable category which has a direct influence on interaction. ―Because
participants are assumed to experience ethnicity as an unchangeable inheritance…ethnicity is
liable to lead to a communicative breakdown‖ (Rampton 487). In this paper, I will show that the
accents of particular ethnicities can effect reactions in listeners on their own, without other
signaling cues.
We can presume that the features of a non-native speaker‘s speech pattern are noticeable
and that they lead to discontinuity in communication. If the role of ethnicity in creating this
discontinuity means it is given negative value by the native speaker, we should consider how
particular languages may be affected. The work of Jane H. Hill on Spanish, especially its use and
devalued caricature in the public sphere—what she calls ―Junk Spanish‖ (Hill 1995)—is
particularly illuminating in the example of that language. She discusses the role that native
speakers‘ language ideologies about their linguistic superiority play in interacting with non-
3
native speakers (particularly those resident in ―their‖ country), the great chain of association
from perceived characteristics of a group that is lower in the majority‘s view to their native
language and marked use of the L2, and the means by which ideology does this as a function of
―public‖ and ―private‖ rights by the native speakers. She cites van Dijk (1993) in calling this
―covert racism‖, a discourse ―that protects the positive self-image of the racist and in turn the
positive image of whites [in this case] more generally…permit[ting] racist discourse and its
negative and exclusionary functioning to proceed‖ (Hill 1995: 199). In order to expose and tear
down that ideology where it does exist, we must find a way to make it overt rather than covert.
Anderson (1991) has pointed out that in fact, a nation (and not so far by extension, an ethnicity)
is actually an ―imagined‖ community, demonstrating the fallacy of such ideologies in the first
place.
Of course, the result of language ideology that appears in everyday interaction is that L2
speakers suffer from value judgments L1 listeners place upon them based on a perceived lack of
mastery of the language. Rosina Lippi-Green (2004) states, ―we rely on language traits to judge
others. […] Language is – among other things – a flexible and constantly flexing tool for the
emblematic marking of social allegiances. […] Speakers choose among sociolinguistic variants
available…and their choices cluster together in ways that are obvious and interpretable to other
speakers‖ (291). The problem arises when the dominant members of the speech community
assign themselves the right to deem particular clusters of those variants (or the speakers
associated with them) as less in some way, and perform discriminatory acts as a result. LippiGreen frames her chapter with many anecdotes from workplaces, schools, and government
offices (including courts) where language was the salient feature that resulted in prejudice.
Researchers have documented many similar instances in the United States (such as Baugh 1999,
4
an example of his well-known work on linguistic profiling via telephone), while Gluszek and
Dovidio (2010) draw on several social psychology studies to state that ―speaking with an accent
can elicit considerable stigmatization‖ (28-29). The majority of these studies note the frequency
with which the subjects committing these acts of discrimination are so unaware of what they‘re
doing: these kinds of judgments have become so normalized in American society that when
quantitative data display the magnitude of the issue, people are shocked that there could be – and
that they might contribute to – such a situation (Lippi-Green 292).
There is a wealth of research on these perceptions that has been done for other languages
in the United States; the overwhelming majority focuses on Spanish, since the Hispanic
population is now the largest (collectively-identified ―ethnic‖) minority in the country. A great
deal of work besides Hill‘s has been written for the purpose of identifying and mitigating bias
towards Spanish speakers in the classroom (Valdés 2003), workplace (Barrett 2006), and media
(Santa Ana 2009). A number of other papers and studies have also taken place for the other
speech communities (including Arabic) in other countries, which often fall into two varieties:
small minority communities that are native to the region, or diaspora communities. However,
very little work has been done on the status of Arabic speakers in the USA specifically examining
how they are regarded based on their language use and accent. My goal in this paper is to address
this topic, and attempt to determine how American listeners react to hearing an Arabic accent;
what particular qualities do they associate with the accent? Do they immediately make strong
character judgments about the speakers? Are these judgments positive or negative in comparison
with an American accent, and with another non-native accent?
I will first discuss the status of the community and language as a whole. Afterwards, I
present samples from interviews with Arabic speakers who generously shared their stories and
5
opinions with me about their language use here. These qualitative data are combined with
quantitative: an online survey distributed to American native English speakers asking them to
make judgments about accented speakers using sound clips. The analysis of these data is then
presented, followed by my interpretation of them and a discussion of the study‘s limitations. I
conclude with suggestions for how this study, and further research, could be applied.
The Arab and Arabic-speaking community in the USA
The US Census Bureau‘s 2007 American Community Survey (ACS) estimated 767,319 speakers
of Arabic in the USA, representing 1.4% of the population who speaks a language other than
English at home. They are the 10th largest language minority in the country, though these figures
do not necessarily represent the total number of immigrants who may speak English at home, but
have Arabic accents (Shin and Kominski 2010). The ACS from the previous year reports that
1,466,874 foreign-born residents identify as being of Arab ancestry, while according to the Arab
American Institute (AAI), ―at least 3.5 million Americans are of Arab descent.‖ Their website
also notes that ―census ancestry data have historically undercounted the Arab American
population…by as much as two-thirds‖ (AAI 2011). Although this community is smaller than
other well-documented ones (Chinese, Vietnamese, Korean, etc.), and is a far cry from the 40million-strong Hispanic community, the numbers still justify attention.
Examining bias towards this community has become more important due to the
sociopolitical climate which recent history has bred. AAI released a report in October 2001, soon
after the attacks on the World Trade Center and Pentagon, outlining hundreds of criminal
incidents that had targeted Arab-Americans within just one month. (The report also noted that
there had been many acts of violence towards members of other religious or ethnic communities
6
who were perceived as being Muslim or Arab, including Persians, Sikhs, South Asians, and even
Hispanics.) Incidents included assaults, harassment, ―airplane profiling‖ including an Arab
Congressman being banned from a flight, and even murder. The document includes the results of
a poll stating that ―45% of students and 37% of Arab Americans of the Muslim faith report being
targeted by discrimination since September 11‖, and also bleakly states that such discrimination
following acts of terrorism or foreign interventions were something the community had ―come to
expect‖ over several decades (Zogby 2-3). Continued involvement in the Middle East, Northern
Africa, and Southwest Asia, along with an ongoing public dialogue about associating Islam with
extremism, remains constant in American media.
The presence of Arabs in the USA is nothing new, but with the development of the
Internet and the attention of mass media in reporting world affairs, non-Arab Americans are
taking notice of the Arabic language and its speakers in an unprecedented way. Sehlaoui (2009)
points out the rapid increase in Arabic schools and university programs in the past decade: from
1998 to 2002, the number of college enrollments in Arabic classes increased by 92 percent, and
then from 2002 to 2006, by another 126 percent, effectively quadrupling enrollment in eight
years, a higher rate than any other foreign language. He also notes that this sudden expansion has
―brought to light a number of problems associated with Arabic language instruction,‖ namely
unqualified instructors and inadequate teaching materials; the development of appropriate
language education cannot keep up with the keen interest students are displaying. Furthermore,
while Sehlaoui is primarily interested in education for heritage speakers, he speculates on the
motivation to learn Arabic for the majority of these non-Arab students: genuine curiosity or a
―know your enemy‖ strategy (Sehlaoui 281-282). Because of this increased visibility and
awareness of Arabic and its speakers in the past decade, the community has become more
7
solidified in the non-Arab American consciousness as a salient ethnic group, with its own
stereotypes, conceptions, misconceptions, and associated qualities.
The diglossic situation of colloquial and standard Arabic has resulted in a focus on that
interplay within Arabic-majority countries, while studies of Arabic-speaking diaspora
communities are much rarer, and almost invariably focus on how the speakers themselves
maintain Arabic in these contexts. Authors like Naficy (2003) and Sehlaoui discuss heritage
language schooling and mass media, respectively, as tools for native speakers to maintain Arabic
in the USA, while Aleya Rouchdy‘s comprehensive book The Arabic Language in America
presents several case studies of Arabic maintenance in the country prior to 2001. Clyne and Kipp
(1999) analyze pluricentric language (including Arabic) presence and maintenance in Australia,
while Al-Sahafi and Barkhuizen (2006) provide an illuminating case study of Arabic in
Auckland, New Zealand. They point out that like the USA, their city (and country) has become a
progressively more multicultural environment in which different languages either thrive or
wither. Their focus is on Arabic maintenance and perception in an immigrant context, and the
important private domains in which the language is strongest.
Al-Sahafi and Barkhuizen provide an excellent model for this paper, by combining a
survey to collect data on the frequency of Arabic immigrants‘ language use with sociolinguistic
interviews to gather narratives of the respondents‘ experiences. Rather than surveying the
speakers, the survey portion of this study was instead directed to non-Arab American listeners; I
am more interested here in perception of Arabic rather than production, and how those
perceptions play into a possible language ideology of American English in relation to Arabic.
Mills (2005), talking about identical issues for Urdu speakers in the UK, situates her research in
the ideological discourses of citizenship, participation in society, and media representation; the
8
two cases are rather similar. She discusses English being ―characterised as the way to be a full
citizen…[and] use of a heritage language as cutting off such engagement‖ (Mills 253). Her
methodology consists of interviews with ten Pakistani immigrants and ten of their (firstgeneration British) children, combined with analysis of ―official‖ pronouncements about
language in the UK. This pairing of speaker opinion and describing national ideology is useful,
but leaves out listener opinion. Taha (2006) provides an example of how to address this missing
piece by surveying native English speakers (in this case, college students) to uncover their
perceptions about Arabic words that have frequently appeared in media discourse about the Iraq
War. She argues that ―many of them appear to have gradually undergone a process of
degeneration of meaning‖ (90). Media and linguistic majority often appear to feed off of each
other to maintain an ideological assessment of Arabic, but uncovering the minority‘s selfperceptions through their experiences can demonstrate whether this effect is destructive or not.
The primary difficulty is separating linguistic perceptions and associated judgments with
a priori knowledge; impressions of ethnicity can be partially constructed from non-auditory
information (especially visual cues) just as easily as they can from language (Rampton 1995). A
relatively recent public example would be controversial pundit Juan Williams, who described his
strong reactions to individuals based on significant elements of appearance he thought of as
―Arab‖. Even information as seemingly innocent as an individual‘s name can play a part in
determining what a speaker ―knows‖ and assumes about a person. Immediately after the Fort
Hood shooting in 2009, Arab-American groups received death threats when the only public
information that had yet been released about the shooter was his name of Jordanian origin
(Associated Press). Such (assumed) knowledge of an individual‘s background can all play a part
in a speaker‘s conception of that individual, making it difficult to tease out the particular effect
9
of the accent component. In order to truly determine what the role of accent alone is, the auditory
information must be considered on its own.
The Study: Qualitative Methodology
This project grew out of a pilot study conducted for a course entitled ―Language and Identity‖ in
Fall 2009, which involved surveying Arabic speakers about their experiences with being nonnative English speakers in the USA. This study brought forth numerous stories of discrimination
during interviews with the speakers – fully half the respondents had a personal narrative, and
others had a friend or family member they could easily reference who did – giving rise to the
question of whether these were common experiences for all individuals with Arabic accents in
the USA. If so, what was the process happening in American listeners that caused them to make
and do prejudicial judgments and actions?
To expand upon the data from the original study, I conducted sociolinguistic interviews
with three Arabic speakers from the Washington, D.C. community that I knew, both prior to and
during the study. The interviews lasted about half an hour each and were very free-form in
nature. I asked questions centered on the speakers‘ linguistic backgrounds, how they ended up in
the US, and narratives they might have about using Arabic or having an Arabic accent. They
shared both their positive and negative experiences as speakers of accented English in the USA,
which portrayed the other side of the data. This study benefits from native speaker voices in the
same way that the Mills (2005) and al-Sahafi and Barkhuizen (2006) studies did, by showing the
relevance of the subject to real people. While providing numbers and statistics about the
judgments made by listeners is useful, without the human element, it would be too easy to
divorce those numbers from who is being affected by what they represent. Additionally, Anna
10
Marie Trester cites Coupland (2001) in saying ―quantitative approaches consistently fail to fully
theorize the interaction of language context‖ on their own, when social context can in fact be one
of the most important variables to consider (Trester 2008: 50).
Qualitative Data
Al-Sahafi and Barkhuizen‘s study demonstrated that Arabic is primarily used in private
domains, namely in the home and in religious contexts, by minority Arabic-speaking
communities (2006: 63). For non-Arabic-speaking Americans, the Arabic language is therefore
unfamiliar, and exposure to it is usually fleeting; this allows their limited knowledge, colored by
whatever qualities have been added to it (by media, for example) to remain incomplete. When
Arab-Americans have used Arabic around those who don‘t speak it, the reactions are mixed at
best, if the responses I received are any indication. I present here a number of quotes from the
original survey that speakers provided (all names have been changed to preserve anonymity):
“They usually tell Arabic sounds so strong. One of my American friends once told me that I
sound so gay when I speak in English but so straight when I speak in Arabic.”
--Abdul, 24-year-old student from Saudi Arabia
“I was at a US family house at Thanksgiving when my mother called. The lady at the hosting
house, when hearing me speak Arabic with my mother, said „Oh, I bet he is giving
instructions to buy a bomb.‟”
--Amir, 25-year-old banker from Saudi Arabia
“Whenever I speak Arabic in front of someone, the listener is often surprised. They usually
„can‟t believe‟ I speak Arabic because I „don‟t look Arab.‟”
--Osman, 25-year-old college graduate from Jordan
“Americans are sometimes impressed when I tell them that I speak Arabic…however, they do
not seem to enjoy the sound of it. I recall someone asking me once „how can such horrible
sounds come out of such a pretty girl?‟”
--Tahira, 25-year-old finance specialist from Tunisia
11
Each of these quotes has something to tell about the perception of the language. Abdul‘s
quote is especially relevant for this study, since it shows that ―masculinity‖ is a trait associated
with an Arabic man speaking Arabic, but a kind of ―de-masculinity‖ when he speaks English.
(This does seem to go against the statistically significant ratings that specifically show up in the
Arabic male speaker‘s ―assertive‖ and ―powerful‖ qualities, as noted above; perhaps Amir‘s
friend was just choosing what he thought would be an offensive jibe.) Amir‘s story echoes a
common idea bred from certain kinds of media: Arabic (especially when spoken by men) is
equated with terrorism. Moreover, it shows the disregard Americans can have for Arabic
speakers, whether joking or serious, by openly making such comparisons to them. Osman‘s
quote demonstrates the important role of visual cues in determining how a listener will react to a
speaker‘s accent or language use; listeners find it difficult to conceive of Arabic—or any other
language—separated from what they think a speaker should look like. (Osman is fair-skinned,
light-haired, and clean-shaven: not the stereotypical Arab man in appearance.) And Tahira‘s
story shows that beyond making character judgments about a person from their accent, the voice
itself (in Arabic‘s case, probably the phonetics of its pharyngeal and uvular consonants) can
receive negative evaluation that affects communication. The Bayard et al. 2001 study included
categories like ―pleasant voice‖ as mentioned before; these are not qualities that are determined
after the fact by the sound of the voice, but can be just as important in constructing a listener‘s
impression of the speaker.
In line with these responses are the longer interviews that were conducted more recently.
While the survey featured fairly brief responses, in these cases I was able to go more in-depth
and understand the more complicated beliefs and impressions of speakers. The first interviewee
12
was Aliya, a 24-year-old Palestinian student who shared a number of stories and thoughts with
me about her experiences as an Arabic speaker in the US:1
(on speaking Arabic in public)
A: Well, I was walking M Street [in Georgetown], and I was – I always talk on the phone to my
friends and family back home, and of course I always speak in Arabic. So a man out of
nowhere, just came and was like, ‗What the hell's this language? You sound like a cat! You
don't sound like a human being!‘ And I didn't even know how to respond, I was really
shocked, and I – I felt so humiliated that he said that my language sounded like a cat, like it
was meowing.
G: Was this like... a crazy homeless person, or – I mean, who says that…?
A: No no, just a guy, a normal guy with, you know, a T-shirt and pants...
(on the role of language versus appearance in people‟s reactions)
A: It's more the appearance, if you look like an Arab. But the language is... maybe 40 percent,
50 percent, but not the whole thing... but once you get that appearance, then the language
gets associated with it. I guess it's easier for females more than males. Males are more
associated with, um... you know how politics is, I don't want to say the word—!
G: Terrorism?
A: Yeah. But they are more, they are more associated with that, being male Arab, more than a
female even if she had a hijab on.
(on speaking with an Arabic accent in a university context)
A: In one of my classes in the summer, um, she's a woman, she's from here, and she was taking
a summer course with me, with Dr. ______ as well. And um, she said, ‗Oh, your English is
perfect, where are you from?‘ And I said, okay, I'm from Palestine. So, she was really
shocked, and she started speaking slowly to me, she was like, ‗O-kay... um, what - sounds do - you - use - in - Arabic?‘, and she would - speak - like - this. It's like, okay, I understand
English. I have a BA in English literature and I'm going for my master's, my English is not
that bad.
As a non-Muslim, Aliya does not wear the hijab or other articles of clothing that are
widely associated with women from the Arab world; in light of this lack of visual cues, her first
story especially shows an experience centered on her language use. Again, while this study asked
respondents to judge the speakers rather than the quality of their voices, as in the case of Tahira
above, these assessments of quality are certainly indicative of a negative attitude towards the
speaker. Furthermore, the fact that random passerby feel entitled to express these sentiments to
Since I was interested in the interviewees’ narratives, rather than specific features of their speech, the interviews are
presented here in block quotes instead of intonation units.
1
13
speakers is astonishing in and of itself; if only we could find out who, in fact, that man was, and
why he thought his comments were acceptable.
The second interview was with Salim, a 44-year-old café manager in Washington, D.C.
Salim has a somewhat complicated upbringing: born in Cairo to Lebanese Christian parents, he
came to the US at age five, and grew up speaking French and Arabic. He says that he made
efforts to be hyper-correct when learning English in school (to the point of adopting a British
accent), but that living abroad in Europe and Asia for a period of time brought his original accent
back to the foreground. He was kind enough to share his stories as well:
(on the change in people‟s reactions to his accent)
S: ―Now they‘re intrigued. Because they have no idea. Prior, growing up, it was, ―Oh, you‘re a
foreigner, you come from where? You sleep in what? You sleep with who?‖ Basically all the
crap you go through when you‘re growing up. And my brother especially had a hard time of
it.‖
G: ―It was harder for him than you?‖
S: ―Well, I‘m the youngest, when we came over – he had a hard time to the point where he lost
his Arabic and French because he was going through school at age 12 or 13. I didn‘t have
that…for him, now he can‘t even stand to hear them.‖
(on speaking Arabic as unifier and separator)
G: ―I heard you speaking Arabic to a woman the other day, she seemed happy about it…‖
S: ―Yes, there are a lot of Arabic people who come in here who are introverted because of all
the nonsense going on, so when they hear someone speaking Arabic, I have no shame in
telling people how it is. And there are lots of stereotypes, especially the Muslim customers
coming in, present-day Muslims, and they‘re young and female…they would not be
comfortable alone. But this –‖ [indicates his cross pendant]
G: ―The cross?‖
S: ―Like, even speaking Arabic, this one woman came in the other day…she could tell I was
Lebanese, and then she saw the cross, and she knew by whatever I wasn‘t from her… her end
of the world of religion. And I could not serve her. I could not serve her coffee except like,
putting it into two cups and giving her the one my hand didn‘t touch.‖
(on how Americans could overcome prejudice)
S: ―It‘s very simple, every single person, even if you don‘t speak the language, should be
thrown in another country for the year. Because you learn, you grow, and you see – what
we‘ve lost as Americans is…you learn how a different society functions. France, Italy,
Greece, Germany, they all forced to leave the country one year.‖
G: ―Or the UK, the gap year.‖
S: ―Exactly.‖
G: ―No, I agree, it would be wonderful if, um, everyone could. You think that‘s the best way?‖
14
S: ―Well you learn other ways. You‘re not snobby towards other people‘s ways. And as far as
accents, anybody who‘s family-educated, it‘s about the way you‘re raised, would not have a
problem with an accent. Your neighborhood…you need that nonwhite culture, you have to
associate with people outside your, your class.‖
Salim was a good resource because he could talk about how attitudes in his experience
had changed over the past forty years, especially after 9/11. His comments show two particular
ideas: the interconnectedness of identity symbols (language with religion, class, etc.) and the fact
that even other Arabic speakers can have particular reactions to accents. Following Silverstein‘s
(2003) concept of the indexical order, Salim‘s Lebanese dialect when speaking Arabic first
makes plain (or indexes) his background, which in turn carries social information that can be
interpreted as very relaxing and reassuring for some fellow Arabic speakers (such as the first
woman he mentioned), or completely off-putting for others (such as the second). The Arabicspeaking population in the US has become much more diverse throughout the last century (see
Rouchdy 1992), bringing these tensions within the speech community to the forefront. On the
other hand, the dialect variation and differences between the many groups that make up this
entity the ―Middle East‖ are rarely discussed by Americans. While native speakers have their
own strong opinions about particular Arabic accents and what they suggest, as Salim suggests,
many Americans are generally just ―intrigued‖ by the Arabic accent in general when they ask
about it, and have no deeper understanding of the internal dialect differences. Although the
speakers in the sentences elicited for the survey are native speakers of Levantine Arabic, the
judgments of respondents can still be applied to other Arabic speakers in general.
The third interview presented here is with Zahra, a 28-year-old analyst from Morocco:
(on Arabic accent humor)
Z: Everybody‘s familiar with the bark joke? ―Where can I bark?‖
G: No, I don‘t know that one.
Z: So, this guy is looking for parking – Egyptian guy, or Jordanian, place any nationality there.
And, uh, he sees a cop by a 7-11 or whatever, and is like, ―Can I bark here?‖ The American
15
cop is like, ―Well, it‘s America, it‘s a free country, you can bark anywhere.‖ So he parks and
then the next thing you know, he gets a ticket. ―How come I got a ticket?‖ ―Well, cause you
parked there.‖ ―But I asked you if I could bark here!‖
(on mispronunciation of the language)
Z: And nobody really knows what Arabic sounds like. Everybody‘s aware of the terrorist,
―AKHmed‖ the terrorist, which is not even pronounced properly, because Ahmed is with a
Haa‟. It‘s a softer H than ―khaa‟‖1… Everybody‘s like, ―Yeah, his name is AKHmed‖ and
I‘m like, ―No!‖ Because I have a coworker named Ahmed, and they keep repeating it, and
they just refuse to fix that. Even though the guy is from Pakistan!
(on the portion of American perception of Arabs that comes from media)
Z: Entirely! Well, actually, maybe eighty percent. The other twenty percent relies on…Arabic
people. I mean, we, as Arabs, need to speak up and do something about it. Nobody‘s really
making the point of saying, ―Okay, I‘m Arab, I look normal just like you.‖ But for the most
part it‘s the media. FOX channel, warmongering, war machine, among others. Um…I mean,
it‘s just, it‘s reached proportions that I never even thought were possible for the US. IT‘s
pretty sad.
G: And of course, a large part of it is 9/11.
Z: Yeah, 9/11 and then… it‘s not even 9/11 in itself because there are other terrorist attacks that
have occurred in the US.
G: Like Oklahoma City.
Z: Exactly, Timothy – MacVeigh? So I‘m going to just start singling out everybody called
MacVeigh or something? That‘s not how it works.
1
[ˈax mɛd] versus [ˈaħ mɛd]
Zahra, like Aliya, does not present an outward appearance that matches the majority idea
of Arab women (i.e. she does not wear the hijab). Having learned English and French throughout
her childhood in addition to Arabic, her accent carries more of these features, and so she says she
has had fewer personal negative experiences based on her accent; however, she also describes
her immigrant experience as atypical. As an analyst who deals with humanitarian crises in North
Africa, she finds Arabic useful in her everyday work, especially with the current political
situation. Still, her comments on the American perception of the Arabic language and the
representation of its speakers in the media were insightful, and she had numerous stories
(including the joke above, whose humor lies in the difficulty of pronouncing ―P‖ for Arabic
speakers) that reflected this perception.
16
Although several of these experiences seem to center on language, it is difficult to tell if
accent and/or Arabic use is playing a key role in affecting the speakers‘ interaction with
Americans. What these narratives establish is that such prejudice does exist, either stemming
from or picking out linguistic features, as part of a larger culture of discrimination. However,
quantifiable data must be gathered to demonstrate whether the linguistic aspect plays a
statistically significant role. To this end, in this next part of the study I will expand upon the
accounts above by constructing a survey meant to gather the reactions of Americans to hearing
the Arabic accent. Statistical analysis of these data will show the value judgments that listeners
are making about a speaker when presented with that accent.
The Study: Quantitative Methodology
This study will use an audio-only perception survey, such as those created by Niedzielski
(1999), Bayard et al. (2001), or Preston (2004). These were originally designed to determine how
listeners assigned character traits to speakers based on their (American English) dialect features,
the procedure can be extended to examine native speakers‘ judgments of non-native speakers.
Bayard et al. asked respondents to rate the qualities of speakers in sound clips using a Likert
scale (which measures a quality from ―less X‖ to ―more X‖) from 1 through 6. The participants
had to assign values based on personal qualities (such as ―authoritative‖, ―hardworking‖,
―warm‖, etc.), voice qualities (such as ―pleasant voice‖), and status indices (for example, if they
had higher or lower income) (Bayard et al. 34). By measuring the responses, the researchers were
able to determine patterns in the opinions of individual communities (whose members had, of
course, self-identified as being part of one community or another) regarding the others, and to
examine how these reflected established ideologies about each one. This method is useful for
17
testing the hypothesis that listeners self-identifying as members of one group or another will
assign qualities in a certain way. Moreover, it illustrates again how an accent indexes different
social meanings to different groups (in this case, ones outside the speech community), and
whether the results can be interpreted relative to the experiences of prejudice and selfperceptions Arabic speakers presented about themselves when interviewed.
In order to determine which qualities would be best to test for in such a perception study,
I gathered four informal focus groups, comprised of three individuals each, with a range of ages,
ethnic backgrounds, occupations, and political/religious views. Each group listened to clips of
native Arabic speakers, one male and one female, from the public domain Speech Accent
Archive, hosted through George Mason University (http://accent.gmu.edu), then was asked to
assign three positive and three negative qualities to each speaker they heard. While the vast
majority of participants focused on how ―foreign‖ the speakers sounded, enough character traits
were gathered that I was able to choose some of the most common:
approachable
intelligent
assertive
laid-back
educated
powerful
friendly
reliable
independent
Note that not all of these qualities were originally in a positive format; quite a few
respondents would define a speaker as ―weak‖ or ―unfriendly‖, for example. The nature of the
survey, as described below, required that these be changed to ―less‖ or ―more‖ of a particular
positive quality. In addition to these, a tenth quality, ―religious‖, was added to the list. I
discussed with the focus group participants the associations between religious extremism and
Islam, and between Islam and Arabic. They were nearly unanimous in their assumption that the
Arabic speakers were at least practicing Muslims, even though (despite their close historical
18
relationship) being an Arabic speaker is not predicated upon being Muslim, and vice versa (as
seen in two of the interviewees, for example).
Following the focus group sessions, I recruited speakers to record the sound clips for the
online survey. The speakers were all educated at the same American university, at the graduate
level, and are fluent English speakers. They consisted of one male and one female speaker each
in three native languages: English (Southern United States dialect), Arabic (Levantine dialect,
centered on the eastern Mediterranean area), and Spanish (Ecuadorian dialect). Although Arabic
vs. English was the key pair for this study, Spanish was included to differentiate where surveytakers were hearing general ―foreign-ness‖ instead of specific qualities associated with Arabic.
For the survey, the English speakers both elicited the following two sentences (a total of four
clips), while the other speakers elicited both sentences in English (eight clips), as well as in their
respective languages (four in Arabic, four in Spanish):
EN: ―My brother Peter is very clever; he was a student at Georgetown for just over three years.‖
―Charlie bought a ring to propose to his girlfriend; God willing, I think she‘ll accept.‖
AR:
(akhii Boutros dhaakii jiddan: kaana Taaliban fii jaami`at Georgetown aTwal qaliilan
min thalaathah sanawaat)
(ishtaraa Karlus kha‟iman li-khuTibah Habiibatahu: in sha‟ Allah aDHunn annahaa sataqbil)
ES: ―Mi hermano Pedro es muy inteligente; estudió en Georgetown un poquito más de tres
años.‖
―Carlos compró un anillo para declararse a su novía; ojalá que le acepte."
The sentences were constructed in this way for three reasons. First, the topics are rather
neutral: it was easy for all the speakers to elicit the sentences without too much inflection that
might sway a listener towards rating one quality higher or lower. Second, they contain several
19
phonemes that are not native to the dialect of Arabic spoken by the subjects: [p], [g], [v], [dʒ],
[tʃ], [ɹ], [ŋ], and several vowels, such as [ʌ], [o] and [ɔ]. (Some of these are also unrepresented in
Spanish.) The motivation behind this was to foreground non-native pronunciations and make the
accents more evident. Finally, Taha (2006) outlines a number of Arabic terms that have
permeated American vocabulary since the onset of U.S. intervention in Iraq; the best-known, and
the one with the most overwhelming reaction is Taliban (92). This word literally means
―student‖ in Arabic, and so when the first sentence was translated into Arabic, the term would
stand out to an American listener among words that might be otherwise unrecognizable. In the
second sentence, ―God willing‖ was translated to in sha‟ Allah, which would also be salient; this
was intended to test for perception of the ―religious‖ characteristic as well, and to make it
apparent that the person in those sound clips were speaking Arabic.
Once the sound clips were recorded, the anonymous survey was constructed online and
distributed using social media (particularly Facebook) and word of mouth. All participants were
required to be native speakers of American English, born and raised in the US, and at least 18
years of age. After reading through the consent information, they were asked to listen to each of
the twenty clips in succession and rate the speaker in each clip for all ten qualities. A Likert scale
was used from 1 to 5 to rate from ―less ---‖ to ―more ---‖ for each. After the survey was
complete, participants were asked to provide free-response demographic information for data
analysis, and results were sent through an anonymous mail service.
The complete survey can be found in Appendix A, while the order of sound clips in the
different versions can be found in Appendix B.
20
Quantitative Data
A total of 234 respondents took part in the survey portion of the study. (The sample size for each
of the demographic categories can be found in Appendix C.) Ten categories were considered for
analysis: sex, race, geographic origin (within the US), news viewing habits, news sources,
exposure to Arabic accent, exposure to other accents, religious identification, political
identification, and survey version. It is apparent that certain demographic groups are better
represented than the others, by virtue of the social networking method of distributing the survey.
I took this into account when considering the statistical significance of different groups, and as a
general rule, considered groups of fewer than 15 respondents to be too small for discussion here.
(I would be interested to see if a continuation of this study can gather more respondents in those
small groups which were statistically significant, and see if their patterns hold.)
An initial comparison of the means shows some immediate bias:
Average rating for female Arabic speaker:
Average rating for male Arabic speaker:
3.0626 }
3.1215
mean 3.0921
Average rating for female Spanish speaker: 3.1495 }
Average rating for male Spanish speaker:
3.0116
mean 3.0806
Average rating for female English speaker:
Average rating for male English speaker:
mean 3.1999
3.2168 }
3.1829
Clearly the English speakers have the advantage here, while Arabic and Spanish speakers are
fairly closely rated. This suggests the importance of finding which of the ten qualities are being
selected as higher-rated for English vs. non-English, and which ones are being selected as
particularly Arabic. I also divided the speakers by sex not expecting to find much of a difference,
and although the English speakers have similar means, Arabic and especially Spanish turned out
to have a large disparity. In addition to breaking down the results along lines of the respondents‘
21
demographics, it may also be important to investigating the differences in responses for each
speaker based on sex.
Even just seeing these numbers, there is clearly a great deal to be analyzed for all three
accents and their speakers. However, since I am concerned in this paper with specifically Arabic
accent and its perception, I will have to reserve a deeper examination of Spanish and English
speakers‘ accents for another occasion.
Analysis
1. Comparing accents for levels of significance
For the analysis, I first conducted a two-tailed t-test, a statistical test to determine if there is a
non-coincidental relationship between the independent and dependent variables (in this case, the
listener demographic categories and their ratings). The test examined the difference in means
between Arabic-accented English, Spanish-accented English, and Southern USA-accented
English, in order to determine which of the ten qualities had statistically significant differences
between their ratings. Figure 1 illustrates the means for each quality in each accent.
. These initial data reveal a number of interesting effects. First, it is notable that English
is, for almost every quality (with the glaring exception of ―religious‖), rated more highly overall
than the other two languages. The gap between means in each case is relatively narrow between
Arabic and Spanish, somewhat larger between Arabic and English, and (with a few exceptions)
largest of all between English and Spanish. This suggests that English is rated more highly than
the other two; Arabic is rated slightly higher than Spanish for most of the qualities, with the
exceptions of ―intelligent‖ and ―laid-back‖. A number of participants articulated in the freeresponse section of the survey their confusion at hearing the word ―God‖ in the elicited
22
Figure 1: Overall speaker ratings broken down by accent and quality (respondent pool n = 234)
An asterisk linking two or three of the means indicates that there was a statistically significant rating between the pair ( |t| < .05, for the purposes of this study). For
example, the first cluster shows that English native speakers were rated most approachable, then Spanish speakers, then Arabic; the differences between mean ratings
of Arabic/English and Spanish/English were found to be significant. Remember that means are on a scale of 1 to 5, so that a mean of 1 would very strongly suggest
the speakers are not associated with the appropriate quality, while 5 would suggest that they are; 3 is neutral. The range of the graph has been restricted to 2.5 – 3.5,
and the mean values to the second decimal place, for the sake of visibility.
23
sentences, yet clearly there is a (statistically) significant association between Arabic accent and
being ―religious‖ here. (Juxtaposing those English sentences with Arabic ones featuring the word
Allah may have reinforced the effect, as discussed previously in this paper.)
Beyond the initial finding that Arabic language, or at least the accent, is ranked between
Spanish and English, we must consider which qualities are being judged by foreign accent in
general, as opposed to Arabic specifically. If a quality was rated significantly different in English
than in both Arabic and Spanish, but not between Arabic and Spanish, then the respondents are
differentiating between English and non-English; this is the case for ―approachable‖, ―friendly‖,
and ―independent‖. Conversely, if a quality was rated significantly different in only Spanish or
Arabic, then the respondents are differentiating between English and that language; this is the
case for ―laid-back‖ and ―religious‖ (Arabic) and ―educated‖ and ―reliable‖ (Spanish). Finally,
―assertive‖ and ―powerful‖ had significantly different ratings for all three accents, suggesting all
three languages are being strongly associated with their particular ratings in the perception of
respondents.
Respondents didn‘t perceive any statistically significant difference that they would equate
with ―intelligent‖; it is the only quality that seems truly unaffected by accent as a whole, and is
higher than all the others for all three accents
2. Examining particular quality ratings across respondent groups
Now we turn our attention to the two qualities that stood out from the t-test: ―laid-back‖ and
―religious‖. It should be noted that respondents found the English speakers to be more laid-back
but less religious than Arabic-accented speakers. As there are a number of demographic
categories being tested in this study, the following graphs will show the mean ratings for these
24
two qualities for each group, within each category. Here I have included the graphs for sex and
race (as they were easily manageable groups with only two factors, where I expected to find
significance), as well as those that did in fact turn out to have statistically significant populations
within them. All other graphs for these two qualities can be found in Appendix C. Note that the
graphs are shown in close-up to more easily demonstrate the differences between the average
means, and their heights should not be considered statistically significant on their own. (As noted
above, the difference in means itself is still important.)
2.1
Laid-back
Figure 2.1.1: Mean ratings for Sex (F = female, M = male) and Race (N = nonwhite, W = white)
Figure 2.1.2: Mean ratings for Political affiliation (Key – f = libertarian, g = left, h = centre-left, j = moderate, k = centreright, l = right)
25
Figure 2.1.3: Mean ratings for Religious affiliation (Key – a = Catholic, A = Orthodox, d = Jewish, D = Buddhist, s =
Protestant, S = neo-pagan, y = atheist, Y = agnostic)
It should be noted that with a few exceptions, none of the means are wildly different; as
such, we are more interested in where the differences are statistically significant than where they
are large. Furthermore, not all of the graphed data fit with the general trend established in the
initial analysis: Arabic as least laid-back, then Spanish, then Southern. There are some occasions
when Arabic overtakes Spanish, and others in which Spanish overtakes Southern. (The only
group who completely reversed this trend was the handful of self-identified Buddhist
respondents.)
The significance becomes more apparent when examining Southern vs. Arabic accent, as
opposed to Spanish vs. Arabic accent. In most demographic categories, Arabic never overtakes
Southern (though they are roughly equal at some points); the rare exceptions are those
respondents who receive their news from multiple sources (which rated Southern remarkably
low), political conservatives, and in religion, self-identified Jews and Buddhists. All the groups
but the first have rather small sample sizes (see Appendix C), and so results should be interpreted
with caution; equally, some of the other very small samples (Alaskan residents, Orthodox
respondents, etc.) have comparatively large discrepancies that cannot necessarily be considered
26
representative. Spanish accent, on the other hand, varies widely in relation to both Arabic and
English.
A least-squares regression (see Appendix D) displays significance at a much more
detailed level than the initial data, taking into account both sample size and consistency across
the group. The results are somewhat surprising: of all the possible demographic groups, the only
ones who had statistically significant consistency to their ratings were those who identified as
politically liberal (left) and/or atheist. While the majority of respondents were liberal, not even
one-tenth identified as atheist; liberals rated Arabic speakers higher overall, while atheists rated
them lower. (These data are visible in the bar graphs above.) Finally, no one factor (including
religious beliefs and political leanings) was significant in its entirety, only individual sub-groups
within them; atheists will tend to rate in a particular way, but religious self-identification itself is
not necessarily significant. The overall difference between laid-back ratings for English accents
and Arabic accents cannot be attributed to any one demographic factor, even if members of
particular demographic sets tend towards rating Arabic accents higher or lower than the mean.
(As a final note, the multiple-news-source group‘s low rating of Southern accents was
statistically significant, as was the atheists‘ high rating of Spanish accents. The atheists, overall,
seem to have stronger and more consistent opinions.)
2.2
Religious
Figure 2.2.1: Mean ratings for Sex (F = female, M = male) and Race (N = nonwhite, W = white)
27
Figure 2.2.2: Mean ratings for Contact with Arabic (i = frequent/close contact, o = occasional contact, p = rare/no
contact) and Contact with other languages/accents (I = frequent/close contact, O = occasional contact, P = rare/no
contact)
It is immediately apparent that the ratings for ―religious‖ do not as obviously fit the trend
displayed by the initial analysis: Arabic as most religious, followed by Spanish, then English.
The ratings are different from group to group, and occasionally the trend reverses altogether; the
only place where it holds absolutely true is in the sex and race categories. Rather than concluding
Arabic is almost always the lowest-rated this time, we can see that English is almost never the
highest-rated, again with a few notable exceptions: a few small-sample geographic areas,
newspaper-readers, and Jews all rated the Southern USA accent as the most religious.
Again, a least-squares regression (see Appendix D) shows the significance of each
demographic category. This time, the categories were much different: both sexes showed
significance in their ratings, as did the group of respondents who had frequent or close contact
with Arabic speakers. Women tended to feel Arabic accents sounded more religious than the
mean (and more religious than either of the others), while men rated them slightly lower than the
mean (but also considered them more religious than the other two). Since both men and women
had significant patterns to their responses, and both sample sizes were rather large, we can say
with statistical confidence that sex is a determining factor in whether or not a listener hears an
Arabic accent as more or less ―religious‖. Additionally, if both groups rated Arabic as more
28
religious than either English or Spanish (and all respondents identified as one of the two), this
alludes to a general impression among all respondents that Arabic speakers are religious.
Analyzing the data by other demographic factors shows that there can be exceptions to this trend,
but very rarely are they statistically significant.
The groups that had strong patterns to their responses for the Spanish accent were
political moderates—who showed a strong inclination to rate slightly higher than the mean—and
curiously enough, respondents who took survey versions 1 and 3 (who rated higher and lower
than the mean, respectively). Religious ratings for the English accent were significant among
residents of the Mid-Atlantic region (another large sample) and lower than the mean, as well as
again, survey version 1 (for which they were lower than the mean). The fact that both of these
accents found significance in the survey version seriously opens the question of whether the
order of the questions themselves had an effect on respondents‘ answers. Version 1 began with
both Southern-accent speakers, while Version 3 immediately presented all three accents in quick
succession (see Appendix B); if there are more subtle effects taking place in the way sound clip
order affected perception, it will have to be determined by further research.
3.
Effects of different subcategories of sound clip
So far I have analyzed trends in listener response for Arabic and other accents in general;
however, the sound clips themselves can be sub-categorized by sex and language spoken. In
order to test if respondents were rating Arabic accent differently based on whether the speaker
was male or female, and if he or she were speaking Arabic or English, I conducted further two-
29
tailed t-tests. The results for speakers in general showed significant results. Figure 3.1 shows the
means for each speaker when using Arabic vs. English, and the differences between them.
Figure 3.1: Ratings for Arabic speakers broken
down by speaker sex and speaker language
(respondent pool n = 234)
An asterisk linking two of the means indicates
that there was a statistically significant rating
between the pair ( |t| < .05, for the purposes of this
study). Remember that means are on a scale of 1
to 5, so that a mean of 1 would very strongly
suggest the speakers are not associated with the
appropriate quality, while 5 would suggest that
they are; 3 is neutral. The range of the graph has
been restricted to 2.5 – 3.5 for the sake of
visibility.
The asterisks present differences in the means that were statistically significant. These
indicate that there was a strong correlation between a higher rating and an Arabic female
speaking Arabic, and an Arabic man speaking English. In terms of speaking Arabic versus
having Arabic-accented English, there was a correlation between a higher rating and the male
speaker; there was no statistically significant preference for male or female when the speakers
were using their native language.
The results become even more interesting when broken down by quality, as shown in
Figure 3.2. It is clear that the speaker‘s sex can have an effect on the listener‘s perceptions of his
or her individual qualities, as can the language he or she is speaking. The means all indicate that
listeners prefer the Arabic female speaker to speak Arabic instead of English, while they prefer
the Arabic male speaker to speak English instead of Arabic. The one notable inversion is that
listeners found the Arabic female speaking English to sound ―more religious‖. The preferences
are significant for the female speaker in ―educated‖ and ―intelligent‖, suggesting respondents
30
Figure 3.2: Ratings for Arabic speakers broken down by speaker sex, speaker language, and quality (respondent pool n = 234)
An asterisk linking two of the means indicates that there was a statistically significant rating between the pair ( |t| < .05, for the purposes of this study). For example, the
first cluster shows that male Arabic speakers using English were rated most approachable, then female Arabic speakers using Arabic, female speakers using English, and
male speakers using Arabic. The differences between the female speaker‘s mean ratings were found to be significant, as were the male speaker‘s. Remember that means
are on a scale of 1 to 5, so that a mean of 1 would very strongly suggest the speakers are not associated with the appropriate quality, while 5 would suggest that they are;
3 is neutral. The range of the graph has been restricted to 2.5 – 3.5, and the mean values to the second decimal place, for the sake of visibility.
31
strongly feel she sounds more so when speaking Arabic; the same is true for the male speaker
sounding more ―approachable‖ and ―friendly‖ in English. In terms of sex, respondents showed
significance in associating the male speaker with ―assertive‖ and ―powerful‖ in Arabic over the
female speaker, and several more (―approachable‖, ―educated‖, ―independent‖, ―intelligent‖,
―reliable‖) in English.
The two factors which were seen as generally significant for Arabic accent in general,
―laid-back‖ and ―religious‖, are the only two which have no significant differences when broken
down by sex and language spoken: all elicitations in the study were perceived to be in the
(statistical) same neighborhood, which was (significantly) lower than the other accents for the
former quality, and higher than the others for the latter.
4. Interpreting the data
There is clearly a large amount of data to be interpreted here, and these are just the first steps in
examining all the potential correlations between speaker and listener information. I have focused
on the most significant features that will address the research question: where are the strongest
correlations between the groups a listener is part of and their relative ratings of the speaker‘s
qualities? With the first analysis, picking out the qualities of ―laid-back‖ and ―religious‖ as the
ones which sound respectively least and most ―Arabic‖ sheds some light on general perceptions
of the populace. There is a consistent barrage of rhetoric in American media that portrays Arabs
and/or Arabic speakers as devout fanatics, the exact kind of person you would expect to fit those
ratings.
The second analysis pulled out the demographic groups who followed that trend most
strongly. For ―laid-back‖, liberals rated Arabic accent speakers slightly (but significantly) higher
32
than the mean, while atheists rated them lower. I would argue the atheist trend, at least, could
stem again from the association of ―Arabic accent‖ with ―religious extremist‖; a self-identified
atheist could imagine such an individual to be less laid-back and more uptight than average. For
―religious‖, in turn, besides the earlier mention that sex was a determining factor in the (overall)
high rating, close or frequent contact with Arabic speakers had a significant downward pull. This
may show that those who encounter Arabic speakers most frequently find the stereotypes about
their religion justified, or simply that the speakers they deal with present their faith more
outwardly. I know of no statistics indicating how many Muslim women wear the hijab or other
symbols of religion, but it is not difficult to imagine that a respondent who encountered a given
number of veiled women would believe he or she had more frequent contact than a respondent
who encountered a similar number of secular Arab women.
I should mention that in terms of news viewing habits and sources (as seen in Appendix
D), while almost all groups seem to follow the trend in ratings for the two qualities, none of their
ratings were statistically significant. This could mean that media have less of a consistent effect
on listener bias than thought, or that the sample size for each group isn‘t large enough to capture
a mean common to all of the speakers of the group.
Finally, examining the differences in ratings for male versus female speakers and Arabic
versus Arabic-accented English showed a preference for the Arabic female to speak Arabic and
the Arabic male to speak English. The fact that this latter preference was statistically significant
for ―assertive‖, ―powerful‖, and ―reliable‖ suggests that Arabic accent is a desirable quality if
one wants to sound more traditionally masculine when speaking English. It would be interesting
to further break down the data and see if there were correlations between particular demographic
groups and this trend: for example, do male respondents rate the Arabic male speaker more
33
highly for these qualities? What about white respondents, or liberal respondents? On the other
hand, the only significant effects for the female speaker were that respondents rated her lower on
―educated‖ and ―intelligent‖ (the quality that was untouched in the first analysis) when speaking
English. These qualities were still the highest rating-wise, but the difference between them
makes it crystal clear that respondents preferred her speaking Arabic in those two regards. (She
was, however, still rated more highly than the male speaker overall in certain qualities –
―friendly‖ and ―reliable‖ – and her Arabic was rated slightly higher.)
A tool that is useful to visualize how these data are arranged is the indexical field, what
Penelope Eckert calls ―a constellation of meanings that are ideologically linked…[it] is not a
static structure, but at every moment a representation of a continuous process of reinterpretation‖
(Eckert 2008: 464). In terms of the data given above, the indexical field allows for a particular
(usually phonetic, possibly also discursive) feature to be analyzed by centering on its main
interactive connotation, then moving outward in the direction of a particular speaker making the
choice to use the feature in the first place. Individual qualities, as we have seen above, are
associated more or less strongly with individual speakers. So, if we were to pick out phonetic
features that signaled Arabic accent—the voiced [p] that‘s not quite [b] referenced in the joke
back in Zahra‘s interview, or the pharyngeal approximant [ʕ] that is the hallmark of Arabic—and
place it within the indexical field, combined with what we know about perceived qualities for
male and female speakers, a listener might perceive something like this:
34
The one trait I was not able to represent well in this model (after Moore and Podesva
2009) was the numerical values of the means for each quality, for each of the four groups.
However, the graphic emphasizes where the quality is most strongly associated: an Arabic
female speaker is not perceived as powerful in (statistically significant) relation to the male
speaker, while both speakers are seen as more friendly when speaking English, etc. Clearly these
are only neutral or positive qualities; the accent can (and probably) does index negative qualities
as well in conjunction with these. As discussed previously, respondents were skewed heavily in
favor of the Arabic male speaking English, hence the cluster of qualities in his quadrant of the
field. Two other points must be made: first, this is obviously only the indexical field for the
limited sample of the survey respondent pool. The field might be better rendered with variables
(such as those Silverstein used in his cornerstone 2003 paper investigating indexicality: n, n+1,
etc., to show successive moves of perceived social meaning), whose content would be
determined by whomever was hearing the accent. And secondly, besides the lack of quantifiable
numbers here, the field may falsely imply that such indexed characteristics fall in a discrete way
in relation to the speakers, when in fact it is more of a continuum: a male Arabic speaker using
35
English is not perceived as being the same exact amount of ―friendly‖ every time. This effect is
easily demonstrable from the individual survey responses, some of which deviate greatly from
the mean.
I don‘t wish to make sweeping generalizations about Arabic male versus female speakers
and how they are being perceived from these initial data, but it is not difficult to pinpoint the
stereotypes they are reinforcing. Arabic men (as Aliya alluded to in her interview) are often
associated with terrorism, especially when they are speaking Arabic; therefore, listeners may
perceive the male speaker‘s English as less threatening, and rate him higher. Conversely, there is
a tradition of Orientalism in Western culture that exoticizes Arabic women; therefore, when the
female speaker is using Arabic, she is rated more highly than when she uses English. However,
in either language, she is rated as sounding less ―assertive‖ or ―powerful‖, calling to mind all
kinds of stereotypes about oppressed Arabic women, especially in a Muslim context. In future
research, I will have to ask respondents to share their thoughts on gender relations in the Arab
world, and see if those attitudes correspond with these ratings.
Limitations and Solutions
There are a number of limitations to this study that I must acknowledge. First and foremost is the
distribution of the survey through the Internet: while social networking is invaluable as a means
of quickly and efficiently getting large numbers of respondents, the pool tends to be skewed. As
a 20-something white agnostic liberal from New Jersey, who checks the news daily and has a
multicultural group of friends, the vast majority of my direct recruits to take the survey fit into
one or more of those categories. To truly get an accurate portrait of what ―American‖ thought
processes when hearing accents, it is important to have a diverse selection of respondents who
36
represent a cross-section of the population. There were a surprising number of conservatives,
Protestant Christians, Californians, and middle-aged respondents in the sample population—
mostly friends and family of classmates here at the university—but not enough to analyze those
groups‘ perceptions without hesitation. Additionally, the very nature of an optional survey
creates bias in the results: those who do not wish to take such a survey might have much stronger
opinions (or weaker) than those who do. A truly accurate description of this study would state
that it reflects the perceptions of only a narrow segment of the American population; however,
even within such a narrow segment, there are interesting results to be found.
The free-response comments illustrated a number of issues that people had personally
with the survey. Some responses had to be removed from the data because computer troubles
wouldn‘t allow the sound clips to play. A few others took genuine offense to the idea that accent
had an impact on their character judgments, and rated a ―3‖ for all speakers, in all qualities,
rather than simply not taking the survey. Beyond the technical complexity of getting an adequate
data set, there is also the possibility that respondents were overthinking their answers. The
survey specifically asked them to rate as quickly and instinctively as possible, but there was no
time limit on the questions. On the other hand, a number of respondents (notably those who
identified as linguistics or social science students) wrote long comments stating their ideals about
judging the person for who they are rather than their superficial attributes, and then proceeded to
have some of the most wide-ranging ratings of all.
The sound clips themselves must also be considered. The qualities assigned by the
respondents are reactions to only one male and one female speaker of each accent, eliciting only
two or four sentences. Additional speakers and elicitations would be helpful to determine
whether these initial samples were in fact representative of respondent populations‘ typical
37
reactions to an entire group. However, as stated before, it is important to separate the auditory
information from any other knowledge, and it wouldn‘t have been profitable to provide, for
example, a clip from a Friday service in a mosque, and then ask respondents to rate how
religious the speaker sounded. Purely linguistic bias must be distinguished from religious bias,
political bias, and every other kind that can be controlled for. The elicitors were very uninflected
in their speech, and did not stray too far in one direction or another in any of their clips. Still,
ideally the survey might have had numerous speakers with a range of sentences, both male and
female, from a variety of ages and backgrounds. Such a survey would have been much longer
than the projected 20 minutes it took respondents to get through this one, though, and it was
difficult enough to get enough respondents to find the time.
Finally, this study can by no means capture all the intricacies of accent perception for the
populations in question. Other studies of a similar nature have focused on much smaller
populations, with much more specific respondents being surveyed, and have still turned out
massive amounts of data. The process by which we index particular speech features and groups
of features to particular personal qualities is an immensely complicated one, and a study of this
size can only scratch the surface of the topic. Assigning qualities in this way is daunting for two
reasons: not only is it difficult for respondents to consciously make these choices, but not all
qualities are created equal. For example, the relative neutrality of ―intelligent‖ in the data
suggests that either we realize so quickly that it is socially inappropriate to make intelligence
judgments, that we suppress them before we can honestly report them, or that intelligence truly is
a quality that does not show through in audio clips. Additionally, different groups may have
different impressions of the qualities being examined: ―more religious‖ is not necessarily a
38
positive quality for atheists, and ―more laid-back‖ may not be palatable to Type A respondents
who work 80 hours a week.
Nevertheless, the study presents a great deal of data that can be analyzed and re-analyzed
to find significance. What it could primarily benefit from is expansion: a larger and more diverse
respondent pool, an analysis of the Spanish accent properties similar to that of the Arabic, more
qualities to be analyzed, etc. It is also important to bear in mind that these data capture just one
instant (or at least, one survey-length instant) in a listener‘s perception. Bucholtz and Hall (2005)
describe the process of stance accretion, the slow build-up of features into an identity that is both
constructed by a speaker and assigned by a listener. In a given interaction, the listener does not
assign a value judgment and then stop; he or she continues to do throughout, and each interaction
builds on the next. The combination of a priori knowledge about a speaker and the listener‘s
personal experience could play a powerful role in determining the perception of an accent (or
indeed, any speech variation), necessitating a more expansive diachronic study. In that regard, it
would be fascinating to trace the complete process of a listener‘s attribution of qualities to a
particular accent, from first contact onward (although I can‘t imagine how such a study would be
possible, given the magnitude of that process).
Conclusions
This study is intended as an initial step in addressing the purely linguistic prejudice faced by
Arab-Americans based on their language and accent that exists. By gathering relevant data about
the opinions of the majority population, analyzing it statistically, and comparing the results with
the common stereotypes and misconceptions present in mainstream culture, it is possible to see
the direction in which listeners‘ attitudes lie. The data support the hypothesis that there are
39
statistically significant correlations between particular categories of respondents or speakers
(along demographic lines), and a higher or lower rating of the speakers‘ qualities (overall and
individually), but I want to stress that the statistical significance does not entail the degree of
prejudice. This study merely illustrates that listener bias can, in certain instances, be predicated
on particular demographic groups the listener belongs to. Although individual respondents may
have liberally assigned 1‘s and 5‘s to the speakers, none of the means for an entire group‘s rating
of a speaker fell outside the central range of 2.5 to 3.5.
The differences between a Southern or Spanish accent and an Arabic accent level out as
the sample sizes increase, implying that the overall prejudice based on language alone is rather
small. (Given the Likert scale, how far does one have to go from ―neutral‖ before a response can
be declared overwhelmingly positive or negative?) It is encouraging that judgments of this kind
may not be as severe or overt as those based on visual cues or a priori knowledge, as in the cases
outlined by Zogby 2001. However, even a comparatively small prejudice can build upon or index
others, so even the smaller components of overall bias must be addressed. My hope is that
explicitly stating where that bias lies, and which groups have more significantly negative
attitudes overall, will enable people to find ways to stop themselves from making those
judgments. If they realize what specific effects a particular accent is having on their attitudes
towards a speaker, they will be at least be able to acknowledge it. A number of respondents from
the survey ended their participation with free-response comments that displayed an impact the
survey had on their self-perception:
- “Filling this out has made me feel really stereotyped/biased.” (white female, 24)
- “It seems to be a bit stereotyped. Made me feel like I was judging people about their
personal virtues just by their voice. I did not like deciding these things about them.”
(white female, 25)
40
- “Interesting. I was surprized [sic] by how easily I drew conclusions about people based
solely on the recording. One impression is that I viewed people more positively when they
spoke more rapidly...drew negative conclusions from hesitation or slow speech. Another
impression is that I viewed more favorable individuals who were not speaking English at
all compared to someone who was speaking heavily-accented English.” (white male, 54)
- “I never thought I profiled in any way but was a bit surprised at my answers, which I
believe show prejudice and some discomfort with non-English speaking persons. Thank
you for the opportunity to participate. I learned something about myself.” (white female, 63)
If nothing else, these respondents are now more aware of the attitudes they hold, however
subconsciously. They will be able to recognize when they are making these linguistic judgments,
and to give them more thought to determine why they are happening and how to overcome them.
Examining language attitudes and opinions is also important because it uses linguistic
techniques to make a difference outside the walls of the university. While I‘ll leave it to other
researchers to determine how best to apply these data (e.g. refining curricula for students to
increase their exposure to accents, providing intercultural trainings in the workplace designed to
demonstrate any biased feelings one speaker may have towards another, presenting more positive
role models in the media, etc.), it is apparent that a difference in how an individual‘s personal
qualities are perceived based on their voice can affect a listener‘s interactions with him or her.
The interviews with Arabic speakers showed just how negative those interactions can become;
these data can help mitigate that prejudice. True freedom from accent bias would mean that
everyone would rate every speaker equally for every quality, or at least have no bias predicated
on the way they sound. We are capable of overcoming our own wariness of others: bearing the
results of this study in mind, we can address and help put an end to the expression of negative
attitudes that keep the world from being a more understanding place.
41
APPENDIX A: Survey (Page One)
Anonymous Survey: Linguistic Perception of Accented English in the USA
You are invited to participate in a research study titled ―Linguistic Perception of Accented English in
the USA‖. This study is being conducted by Greg Niedt, a graduate student at Georgetown, for a
thesis on how accent affects the way listeners perceive speech.
Participation in this study is entirely voluntary at all times. You can choose not to participate at all or
to leave the study at any time. Regardless of your decision, there will be no effect on your
relationship with the researcher or any other consequences.
You are being asked to take part in this study only if you are a native English-speaking American;
the study focuses on the perceptions of your population.
If you agree to participate, you will be asked to fill out a survey in which you will rate what you feel
are the personal qualities of different speakers based solely on sound clips. There are two sentences
being analyzed both in English and in translation:


"My brother Peter is very clever; he was a student at Georgetown for just over three years."
AND
"Charlie bought a ring to propose to his girlfriend; I think she'll accept, God willing."
The survey should take around 20 minutes to complete, and you will be asked to enter demographic
information at the end. The survey will be completed entirely online, and the results will be sent only
to the researcher.
All of your responses to this survey will remain anonymous and cannot be linked to you in any way.
No identifying information about you will be collected at any point during the study, and your survey
will be identified only with a random number. Once you submit your completed survey, there will be
no way to withdraw your responses from the study because the survey contains no identifying
information.
Study data will be kept in digital format on the researcher‘s computer. Access to digital data will be
password protected. Only the researcher will have access to the data.
There are no risks associated with this study. While you will not experience any direct benefits from
participation, information collected in this study may benefit others in the future by helping the
researcher to collect information on speech perception that may eventually benefit others.
If you have any questions regarding the survey or this research project in general, please contact the
principal investigator, Greg Niedt, at gjn5 (at) georgetown (dot) edu or his faculty advisor, Dr.
Robert Podesva, at rjp39 (at) georgetown (dot) edu. If you have any questions about your rights as a
research participant, please contact the Georgetown University IRB at (202) 687-6553 or irboard (at)
georgetown (dot) edu.
By completing and submitting this survey, you are indicating your consent to participate in this
study.
42
Survey (Page Two)
There are twenty questions to this survey. Each sound clip is a matter of seconds in length.
Listen to the sound clip, then rate the speaker's qualities based on his or her voice. Don't think
about it too carefully; there is no "right" or "wrong" answer for these questions, and first
impressions are everything! Please fill out each ranking for each question.
(This format was repeated for all twenty sound clips.)
43
Survey (Page Three)
Thank you for completing the questions! Please enter the demographic information and any
comments below before submitting the survey. This information is necessary for analysis of the
data; although the answers are free response, please indicate the following in a way that will be
clear to the researcher.
Age:
Gender:
Racial or ethnic background:
Hometown (city, state):
Current town (city, state):
Length of residence in current town:
Occupation:
- How often do you read news? What are your primary sources for the news?
- Do you have contact with any speakers of foreign languages, particularly Arabic or Spanish, or
with regional accents, particularly Southern USA? Are they acquaintances, friends, significant
others? How often do you interact with them?
- How do you identify politically (conservative, liberal, etc.)? Religiously?
- Please share any other comments you may have about the study:
Thanks again! Please click the button below to submit the survey. Note that when you click the
survey, the form provider ("Mail Maniac") will ask if the fields have been entirely filled in, and
then may ask you to pass a spam filter.
44
APPENDIX B: Survey Versions and Order of Sound Clips
Clip
1
2
3
4
5
6
7
8
9
10
Version 1
EN♂ Charlie
EN♀ Peter
SP♂ Charlie
AR♀ Peter
AR♂ Charlie
SP♀ Peter
SP♂ Carlos
AR♀ Boutros
AR♂ Karlus
SP♀ Pedro
Version 2
AR♀ Karlus
SP♂ Pedro
SP♀ Charlie
AR♂ Peter
EN♂ Charlie
EN♀ Peter
SP♂ Charlie
AR♀ Peter
AR♂ Charlie
SP♀ Peter
Version 3
EN♀ Charlie
SP♂ Peter
AR♀ Charlie
EN♂ Peter
SP♀ Carlos
AR♂ Boutros
AR♀ Karlus
SP♂ Pedro
SP♀ Charlie
AR♂ Peter
Version 4
AR♂ Charlie
SP♀ Peter
SP♂ Carlos
AR♀ Boutros
AR♂ Karlus
SP♀ Pedro
SP♂ Peter
AR♀ Charlie
EN♂ Peter
SP♀ Carlos
11
12
13
14
15
16
17
18
19
20
EN♀ Charlie
SP♂ Peter
AR♀ Charlie
EN♂ Peter
SP♀ Carlos
AR♂ Boutros
AR♀ Karlus
SP♂ Pedro
SP♀ Charlie
AR♂ Peter
SP♂ Carlos
AR♀ Boutros
AR♂ Karlus
SP♀ Pedro
EN♀ Charlie
SP♂ Peter
AR♀ Charlie
EN♂ Peter
SP♀ Carlos
AR♂ Boutros
EN♂ Charlie
EN♀ Peter
SP♂ Charlie
AR♀ Peter
AR♂ Charlie
SP♀ Peter
SP♂ Carlos
AR♀ Boutros
AR♂ Karlus
SP♀ Pedro
EN♀ Charlie
AR♂ Boutros
AR♀ Karlus
SP♂ Pedro
SP♀ Charlie
AR♂ Peter
EN♂ Charlie
EN♀ Peter
SP♂ Charlie
AR♀ Peter
This table shows the order of sound clips in each version of the survey. EN, SP, and AR
represent English, Spanish, and Arabic speaker, respectively; the male or female symbol indicate
speaker‘s sex. The name indicates which sentence was being elicited, either the ―Charlie‖
sentence or the ―Peter‖ sentence. ―Carlos‖ and ―Pedro‖ indicate when those sentences were
elicited in Spanish, while ―Karlus‖ and ―Boutros‖ indicate when they were elicited in Arabic.
All Arabic-accented English sentences have been boldfaced.
45
APPENDIX C: Respondent Pool (n = 234) broken down by Demographic Category
Category
Group
n
Sex
Female
Male
Nonwhite
White
Alaska/Hawaii
California
Great Lakes (IL, IN, MI, MN, western NY/PA, OH, WI)
Mid-Atlantic (DC, DE, MD, NJ, NYC, PA, northern VA)
Midwest (IA, KS, MO, ND, NK, SD)
New England (CT, MA, ME, NH, upstate NY, RI, VT)
Northwest (ID, MT, OR, WA, WY)
South (AL, AR, LA, MS, OK, TX)
Southeast (FL, GA, KY, NC, SC, TN, VA, WV)
Southwest (AZ, CO, NM, NV, UT)
Reads/watches news daily
Reads/watches news multiple times a week
Reads/watches news weekly
Reads/watches news less than once a week
Unknown
Internet
Newspapers
Radio/television
Multiple sources
Unknown
Does not know/rarely interacts with Arabic speakers
Has some acquaintance with Arabic speakers
Frequent/close interaction with Arabic speakers
Does not know/rarely interacts with accented speakers
Has some acquaintance with accented speakers
Frequent/close interaction with accented speakers
Libertarian
Left /liberal
Centre-left
Centre
Centre-right
Right/conservative
Unknown
Agnostic/non-religious/spiritual
Atheist
Buddhist
Eastern Orthodox
Jewish
Neo-pagan/Wiccan
Protestant Christian
Roman Catholic
Unknown
Version 1
Version 2
Version 3
Version 4
171
63
34
200
2
28
20
129
0
17
2
8
25
3
137
39
17
28
13
126
23
30
46
9
170
31
33
26
59
149
9
129
29
38
10
13
5
75
26
3
1
11
5
39
46
28
34
78
51
71
Race1
US Origin
News habits
Main news sources
Arabic exposure
Other accents2
Political identification
Religious identification
Survey version
1: Due to the small sample size of individual self-reported nonwhite races, these have been grouped into one
nonwhite category. It should be noted that no survey respondent self-identified as Arab-American.
2: Exposure to Arabic and other accents/languages were treated as one free response question, and separated in this
manner during data analysis. This was done so that the focus on Arabic would not be apparent to the survey taker.
46
APPENDIX D: Demographic factors and factor groups in least-squares regression; graphs of
non-statistically significant responses for Arabic accent by demographic group
Parameter Estimates: laidback
Effect Tests: laidback
Mean ratings for US Origin: (0 = Alaska/Hawaii, 1 = New England, 2 = Mid-Atlantic, 3 = Southeast, 4 = South,
5 = Great Lakes, 7 = Southwest, 8 = California, 9 = Northwest)
47
Mean ratings for News habits (c = weekly, v = less than weekly, x = multiple times a week, z = daily)
News sources: (b = Internet, B = multiple sources, m = mass media (radio/TV), n = newspapers)
Mean ratings for Contact with Arabic (i = frequent/close contact, o = occasional contact, p = rare/no contact)
Contact with other languages/accents (I = frequent/close contact, O = occasional contact, P = rare/no
contact)
Mean ratings for Survey version
48
Parameter Estimates: religious
Effect Tests: religious
Mean ratings for US Origin: (0 = Alaska/Hawaii, 1 = New England, 2 = Mid-Atlantic, 3 = Southeast, 4 = South,
5 = Great Lakes, 7 = Southwest, 8 = California, 9 = Northwest)
49
Mean ratings for Political affiliation: (f = libertarian, g = left, h = centre-left, j = moderate, k = centre-right, l = right)
Mean ratings for Religious affiliation: (a = Catholic, d = Jewish, D = Buddhist, s = Protestant, S = neo-pagan, y = atheist,
Y = agnostic)
Mean ratings for News habits: (c = weekly, v = less than weekly, x = multiple times a week, z = daily)
News sources: (b = Internet, B = multiple sources, m = mass media (radio/TV), n = newspapers)
50
Mean ratings for Survey version
51
Bibliography
al-Batal, Mahmoud (2007). ―Arabic and National Language Educational Policy.‖ The Modern
Language Journal 91, ii: 22-25.
al-Issa (2003). ―Sociocultural transfer in L2 speech behaviors: evidence and motivating factors.‖
International Journal of Intercultural Relations 27: 581-601.
al-Sahafi, Morad A. and Gary Barkhuizen (2006). ―Language use in an immigrant context: The
case of Arabic in Auckland.‖ New Zealand Studies in Applied Linguistics 12(1): 51-69.
Arab American Institute (accessed April 5, 2011). ―Demographics.‖ Retrieved from
http://www.aaiusa.org/pages/demographics/.
Associated Press (11/7/2009; accessed April 10, 2011). ―U.S. Muslims fear backlash after base
rampage.‖ Retrieved from http://www.msnbc.msn.com/id/33748670/ns/us_news-faith/.
Barrett, Rusty (2006). ―Language ideology and racial inequality: Competing functions of
Spanish in an Anglo-owned Mexican restaurant.‖ Language in Society 35: 163-204.
Baugh, John, William Idsardi, and Thomas Purnell (1999). ―Perceptual and Phonetic
Experiments on American Dialect Identification.‖ Journal of Language and Social
Psychology 18(1): 10-30.
Bayard, Donn et al. (2001). ―Pax Americana? Accent attitudinal evaluations in New Zealand,
Australia, and America.‖ Journal of Sociolinguistics 5/1: 22-49.
Bucholtz, Mary and Kira Hall (2005). ―Identity and interaction: a sociocultural linguistic
approach.‖ Discourse Studies 7(4-5): 585-614.
Clyne, Michael and Sandra Kipp (1999). Pluricentric Languages in an Immigrant Context:
Spanish, Arabic, and Chinese. New York: Mouton de Gruyter.
Collins, James and Stef Slembrouck (2005). ―Multilingualism and diasporic populations:
Spatializing practices, institutional processes, and social hierarchies.‖ Language &
Communication 25: 189-195.
Coupland, Nikolas (2001). ―Language, situation, and the relational self: Theorising dialect-style
in sociolinguistics.‖ In Penelope Eckert and John R. Rickford (eds.) Style and
Sociolinguistic Variation. New York: Cambridge University Press. 185-210.
Derwing, Tracey and Murray J. Munro. (2009) ―Putting accent in its place: Rethinking obstacles
to communication.‖ Language Teaching 42: 476-490.
Eckert, Penelope (2008). ―Variation and the indexical field.‖ Journal of Sociolinguistics 12/4:
453-476.
52
eds. Eckert, Penelope and John R. Rickford (2001). Style and Sociolinguistic Variation. New
York: Cambridge University Press.
Gluszek, Agata and John F. Dovidio (2010). ―Speaking with a Nonnative Accent: Perceptions of
Bias, Communication Difficulties, and Belonging in the United States.‖ Journal of
Language and Social Psychology, 29(2): 224-234.
Gluszek, Agata and Anna-Kaisa Newheiser (2011). ―Social Psychological Orientations and
Accent Strength.‖ Journal of Language and Social Psychology, 30(1): 28-45.
Hill, Jane H. (1995). ―Junk Spanish, Covert Racism, and the (Leaky) Boundary between Public
and Private Spheres.‖ Pragmatics 5(2): 197-212.
Ibrahim, Raphiq, Zohar Eviatar and Mark Leikin (2008). ―Speaking Hebrew with an accent:
Empathic capacity or other nonpersonal factors.‖ International Journal of Bilingualism,
12(3): 195-207.
Labov, William (1972). ―The social motivation of a sound change.‖ In William Labov,
Sociolinguistic Patterns, Philadelphia: University of Pennsylvania Press. 1-42.
Lippi-Green, Rosina. (2004) ―Language ideology and language prejudice.‖ In Language in the
USA, eds. Edward Finegan and John Rickford, Cambridge: Cambridge University Press.
Major, Roy (2010). ―First language attrition in foreign accent perception.‖ International Journal
of Bilingualism, 14(2): 163-183.
Mills, Jean (2005). ―Connecting Communities: Identity, Language, and Diaspora.‖ The
International Journal of Bilingual Education and Bilingualism 8(4): 253-274.
Moore, Emma and Robert J. Podesva (2009). ―Style, indexicality, and the social meaning of tag
questions.‖ Language in Society, 38(4): 447-485.
Naficy, Hamid (2003). ―Narrowcasting in Diaspora: Middle Eastern Television in Los Angeles.‖
In The Media of Diaspora, ed. Karim H. Karim. London: Routledge.
Niedzelski, Nancy (1999). ―The Effect of Social Information on the Perception of Sociolinguistic
Variables.‖ Journal of Language and Social Psychology, 18(1): 62-85.
Ochs, Elinor and Lisa Capps (1996). ―Narrating the Self.‖ Annual Review of Anthropology 25:
19-43.
Preston, Dennis R. (2004). ―Language attitudes to speech.‖ In Language in the USA, eds. Edward
Finegan and John Rickford, Cambridge: Cambridge University Press.
Rahman, Jacquelyn (2008). ―Middle-Class African Americans: Reactions and Attitudes Toward
African American English.‖ American Speech 83(2): 141-176.
53
Rampton, Ben (1995). ―Language Crossing and the Problematisation of Ethnicity and
Socialisation.‖ Pragmatics 5(4): 485-513.
Rindal, Ulrikke (2010). ―Constructing identity with L2: Pronunciation and attitudes among
Norwegian learners of English.‖ Journal of Sociolinguistics 14/2: 240-261.
Rouchdy, Aleya (1992). The Arabic Language in America. Detroit: Wayne State University
Press.
Santa Ana, Otto (2009). ―Did You Call in Mexican? The Racial Politics of Jay Leno Immigrant
Jokes.‖ Language and Society 38(1): 23-45.
Sehlaoui, Abdedilah Salim (2009). ―Language Learning, Heritage, and Literacy in the USA: The
Case of -Arabic.‖ Language, Culture and Curriculum 21(3): 280-291.
Sharma, Devyani (2005). ―Dialect stabilization and speaker awareness in non-native varieties of
English.‖ Journal of Sociolinguistics 9/2: 194-224.
Shin, Hyon B. and Robert A. Kominski (2010). ―Language Use in the United States: 2007.‖
American Community Service Reports. U.S. Census Bureau.
Silverstein, Michael (2003). ―Indexical order and the dialects of sociolinguistic life.‖ Language
and Communication 23: 193-229.
Taha, T.A. (2006). ―College students‘ evaluative reactions to Arabic loanwords used in the
context of the Iraq War.‖ Education 127(1): 86-99.
Thomas, J. (1984). ―Cross-cultural discourse as ‗unequal encounter‘: Towards a pragmatic
analysis.‖ Applied Linguistics 5: 226-235.
Trester, Anna Marie (2008). ―Improvising onstage and off: Combining variationist, discourse,
and anthropological approaches to style.‖ (Doctoral dissertation.) Retrieved from
ProQuest Dissertations and Theses. (Publication No. 3339913)
Valdés, Guadalupe (2003). ―Expanding definitions of giftedness: the case of young interpreters
from immigrant communities.‖ New Jersey: Lawrence Erlbaum Associates.
van Dijk, Teun A. (1993). Elite Discourse and Racism. Newbury Park: Sage Publications.
Zogby, James (2001, accessed April 5, 2011). ―Submission to the United States Commission on
Civil Rights.‖ Retrieved from: http://www.aaiusa.org/dr-zogby/entry/submission-to-theunited-states-commission-on-civil-rights/.
54