Speaker Evaluation Measures of Language Attitudes: Evidence of

Speaker Evaluation Measures of Language
Attitudes: Evidence of Information-Processing
Effects
Aaron Castelan Cargile
Department of Communication Studies, California State University, Long Beach,
CA 90840-2407, USA
Language attitudes are often inferred from hearers’ evaluative reactions to speech variations. However, the study of information processing suggests that speaker evaluations may also be sensitive to the conditions under which hearers develop impressions
of speakers. If so, even greater caution should be exercised in the use of such evaluations as a measure of language attitudes. In order to test the potential effect of information processing on speaker evaluations, an experiment was conducted in which hearers
evaluated speakers under a condition of time constraint, and again under a condition
representative of a typical language attitude study. Results indicated that evaluations
of a female African-AmericanVernacular English speaker were affected.A subsequent
experiment provides evidence regarding the likelihood of an alternative explanation
for these results.
Over the past 40 years, the social scientific study of language attitudes has relied
mainly on three investigative approaches: content analyses, direct measures, and
speaker evaluations (see Ryan et al., 1988). Of these three, the speaker evaluation
approach has been used most widely. It consists of listeners evaluating one or
more tape-recorded speakers and treating those evaluations as a measure of the
listeners’ attitude toward the tested language variety. This approach was first
practised with Lambert’s development of the ‘matched-guise’ technique
(Lambert et al., 1960; Lambert et al., 1965), and has since been used regularly (e.g.
Cargile, 1997; Dailey-O’Cain, 2000; Hopper, 1977; Johnson & Buttny, 1982;
Luhman, 1990; Seligman et al., 1972). Because they afford experimental control
and are believed to minimise biases of social desirability, speaker evaluations
have become foundational to our current understanding of language attitudes.
Despite its widespread use, the speaker evaluation approach is not without
criticism. Giles and Coupland (1991) recognised that because syntactic structures
and semantic features have evaluative consequences, what a speaker says may
influence evaluations as much as how it is said (e.g. the spoken language, accent,
or vocal pitch). In addition, contextual features, such as situational formality,
have also been found to affect speaker evaluations (Bradac et al., 1976; Creber &
Giles, 1983; Street & Brady, 1982), thus there is reason to question the assumption
implicit in this methodology – namely, that manifest evaluations reflect only
latent language attitudes. A researcher may conclude that an evaluation reflects
an attitude toward a tested accent or form of speech, but it might instead reflect a
reaction to a combination of the speaker’s accent, voice, message and situation
(among other features of a given study).
Because this criticism strikes at the heart of many language-attitude studies,
Cargile, Bradac, and others (Bradac et al., 2001; Cargile et al., 1994; Cargile &
Bradac, 2001) responded by developing a processual model that, among other
0965-8416/02/03 0178-14 $20.00/0
LANGUAGE AWARENESS
© 2002 A.C. Cargile
Vol. 11, No.3, 2002
Information-Processing Effects
179
things, considers explicitly the relationship between language attitudes and
speaker evaluations. In its most recent form, this model proposes that beyond the
problem of confounding combinations of language and non-language stimuli,
individuating information and information-processing variables also serve to
compromise the ability of manifest speaker evaluations to reflect validly given
latent language attitudes. In other words, listener evaluations are likely to be the
product of individually-specific knowledge, the listeners’ motivation/capacity
to engage in information processing, as well as several evoked language
attitudes.
Although the processual model challenges several features of the foundation
upon which speaker-evaluation studies rest, only the role of information
processing in language-attitude measurement is investigated herein as an initial
test of the model. Information processing has received considerable theoretical
and empirical attention in other domains of inquiry (e.g. stereotyping and
persuasion) and these discussions serve as a basis for the current investigation.
Review of the Literature
There are two types of information that strongly influence person evaluations:
individuating information and stereotypes (Brewer, 1988; Fiske & Neuberg,
1990; Locksley et al., 1982). Stereotypes are beliefs about a group of people
(Leyens et al., 1994) and they are cognitive in nature. Attitudes, however, are
cognitive, affective and behavioural in nature (Edwards, 1982), therefore they
comprehend stereotypes as well as both emotional responses and behavioural
predispositions. A language-based stereotype (e.g. ‘I think a New York accent
makes people sound rude’) is thus the cognitive component of a more general
language attitude, and it is the part of an attitude that is subject to information
processing.
Although the distinction between individuating information and stereotypes
may be ambiguous on occasion, the idea is that some of our information about a
given individual is person-specific (i.e. based on idiosyncratic behaviours and
unique characteristics), whereas other information reflects general and
pre-existing associations between nonverbal or linguistic stimuli (e.g. style of
dress or accent) and traits (e.g. socio-intellectual status or attractiveness). After
attending to certain traits and behaviours of a speaker, several (language) attitudes, along with other individuating information, may be activated. Consequently, a hearer must somehow synthesise both types of information as he or
she works toward developing a response. Although no-one can say definitively
how this process occurs, Kunda and Thagard’s (1996) parallelconstraint-satisfaction model (PCSM) offers some important insights.
To review this briefly, the model indicates that stereotypes (i.e. cognitive attitudinal features), traits, and behaviours can be represented as interconnected
nodes in a network of activation whose spread is constrained by both the positive
or negative associations between the nodes and by the relative strength of these
associations.Nodes in the network are first activated when either a stereotype or
individuating information is made salient. For example, hearers may attend to a
speaker’s high level of lexical diversity and stereotypically associate it with the
trait of ‘intelligence’. Simultaneously, they may also observe that the speaker
180
Language Awareness
received poor grades in college. These three nodes (i.e. ‘lexically diverse’, ‘intelligent’, and ‘poor grades’) would be initially and automatically activated (Bradac
et al., 2001).
Unlike other, similar models (Brewer, 1988; Fiske & Neuberg, 1990), the PCSM
gives priority to neither stereotypes nor individuating information, instead
giving them equal weight in the process of impression formation. Inferred traits,
activated by course of a stereotypical association, have the same impact on a
hearer’s impression as speaker-specific information, as the hearer interprets all
of the activated nodes in the network. Of course, the network could consist of
only attitudinally derived nodes, or of only person-specific information. In this
way, the model incorporates the entire attitude/nonattitude continuum
discussed by Fazio and his colleagues (e.g. Fazio et al., 1986; Roskos-Ewoldsen,
1997). Responses to attitude objects can be generated in three ways: without use
of a priori evaluations, through strict reliance on a highly accessible attitudes, or
based on some combination of both attitude and nonattitude information.
When a hearer’s nodal network is initially activated by some configuration of
attitudinally derived and/or person-specific information, the information
contained within the network may be quickly and automatically integrated for
impression formation purposes. However, the network may also spread, sometimes widely, by repeatedly updating the activation of interrelated nodes. Such
spread occurs over time and can require numerous iterations until a point is
reached at which the network ‘settles’. Although the PCSM attends primarily to
the automatic integration of information, Cargile and Bradac ( 2001) contend that
this model, when interpreted broadly, can also accommodate more elaborate
and controlled information integration processes.
The recognition that hearers can process information about the speaker in
either ‘automatic’ or ‘controlled’ fashions is important in order to account for
recent empirical findings. Devine (1989) suggests that automatic processes
‘involve the unintentional or spontaneous activation of some well-learned set of
associations,’ whereas controlled processes ‘are intentional and require the
active attention of the individual’. Many recent experimental investigations
provide evidence that, when forming impressions of social actors, persons may
make two different kinds of evaluations, reflecting either automatic processes or
controlled processes (e.g. Banaji & Greenwald, 1995; Blair & Banaji, 1996;
Dovidio et al., 1997; Fazio et al., 1995; Monteith, 1993). This suggests that information that is automatically activated in an initial iteration of the nodal network
may or may not affect a hearer’s evaluation of a speaker. As Dovidio et al. (1997:
511) describe it:
the presentation of an attitude object may automatically activate an associated
evaluation from memory which may influence subsequent judgments. However,
as Gilbert and Hixon (1991) argue, automatic activation ‘does not mandate such
use, nor does it determine the precise nature of its use. It is possible for (automatically) activated information to exert no effect on subsequent judgments or to have
a variety of different effects’ (due the impact of controlled processes).
Relating this in terms of the PCSM, automatic processing of information occurs
when few iterations occur in a hearer’s activated nodal network that lead to a
rapid response, e.g. an immediate evaluation of speaker status. However,
Information-Processing Effects
181
controlled processing entails numerous iterations or elaborations of the network
resulting in a relatively slow or delayed hearer response; in fact, in some cases
elaboration may be so extensive that initially activated information will have
little effect on this response. In even more controlled circumstances, the network
may be said to migrate as the information elements present in the first iteration
(and activated in response to observed stimuli) may be completely absent from
the finally settled network. This would be true in the case of socially desirable
responses, for example. A hearer may attend to the accent of the speaker which in
turn evokes an attitude represented in the initial nodal network as the activation
of several negative traits. However, as attention is brought to bear, the hearer
may realise the undesirability of responding based on the currently activated
network. Consequently, the network may migrate as other, more positive behaviours and traits are activated, while the nodes that were initially active are
deactivated.
The variability in processing described here should be regarded as a continuum such that sometimes hearer responses are based on the relatively automatic
integration of information; other times these responses are based on a nodal
network whose spread has been greatly controlled; and other times still these
responses are based on a network whose nodes have been activated in both automatic and controlled manners. The primary factors determining whether automatic or controlled processing occurs are motivation and ability (see Petty &
Cacioppo, 1986b). A hearer must be sufficiently motivated and possess the ability
to go beyond the initial state of the nodal network in order for controlled processing to occur. Otherwise, most perceivers are cognitively lazy to such a degree that
they will usually avoid the continued processing of information if that information has already been automatically and sufficiently integrated.
To be sure, the study of information processing has not yet led to a consensus
regarding many critical details. Despite this, even the widely-accepted generalities have important implications for language-attitude study. Namely, speaker
evaluations should be a reflection not only of evoked information (both individuating and
attitudinally-derived), but also of the process that integrates this information into a
response. Until now, researchers assumed that speaker evaluations measured
only (attitudinally-derived) information. However, the information-processing
literature suggests that these evaluations are not discriminating attitude
measures. For example, two hearers may hold the very same language attitude
yet provide different evaluations of the same speaker if the process through
which they integrate their responses differs (i.e. automatic versus controlled
processing). Researchers would claim to have observed two different attitudes,
but they would be wrong in their account. As this example illustrates, if information-processing functions in the language-attitude domain as herein proposed,
the implications for language-attitudes research are significant. Thus a critical
first step is to test whether conditions of processing do indeed affect speaker
evaluations.
As described earlier, controlled processing can occur only when hearers
possess both sufficient motivation and ability. Hearer motivation is a complex
construct comprised of at least three elements: need for cognition, involvement,
and diversity of argument (Petty & Cacioppo, 1986a). Hearer ability, however, is
less complex and more easily controlled; it is a product of available time and
182
Language Awareness
cognitive capacity (Devine, 1989). While some studies have limited perceivers’
abilities to engage in controlled processing by reducing their cognitive capacity
through non-task cognitive loads (e.g. Gilbert & Hixon, 1991; Shiffrin & Schneider, 1977, Expt 1), most studies have focused on the element of time. Either by
measuring perceivers’ response times (e.g. Dovidio et al., 1997, Expt 1), or by
controlling exposure to stimulus material (e.g. Neely, 1977), researchers have
accounted for the information processing (i.e. controlled or automatic) in which
study participants engage.
Considering this, one simple method of testing whether or not conditions of
processing affect speaker evaluations is to control the amount of time hearers
have available to record their judgements of the speaker. Traditionally, hearers
have had as much time as necessary to develop and mark their evaluations, thus
conditions of processing could best be described as ‘controlled’ (safely assuming,
of course, that these sorts of evaluations do not overly burden their cognitive
capacities). If hearers, however, were allowed only one second to record their
evaluation before being asked to make another evaluation of a different speaker,
their ability to iterate the nodal network activated by the tape recording would be
compromised. In this case, conditions of processing would best be described as
‘automatic’, or at least more automatic. Presuming that different conditions of
available time engender different sorts of information processing, a difference in
speaker evaluations across conditions should be expected.
As already suggested, controlled processing is often associated with prejudice-reducing responses. According to researchers (Devine, 1989; Monteith,
1993), stereotypes have a long history of activation,thus stereotyped-based (prejudiced) responses are easily and automatically accessible. However, sometimes
respondents may note that their automatic, stereotype-based response is unsuitable, because it is either discrepant with their personal beliefs or with norms of
social propriety. In these cases, the stereotype-based response can be modified to
produce a more appropriate response, but this takes longer as it occurs only
through controlled processing. Thus, only in cases where either personal beliefs
or social norms are discrepant with stereotype-based responses should speaker
evaluations change. In cases where beliefs, norms, and stereotypes are consistent, opportunity for additional nodal network iteration should not affect the
recorded evaluation.
Because the effect of information processing on speaker evaluations should
depend on the nature of the evoked language attitude, this study compared reactions to both standard and non-standard American accented speech. Of all the
varieties of non-standard American English, the most researched and perhaps
most stigmatised variety is African-American Vernacular English (AAVE). With
one exception (cf. Koch & Gross, 1997), AAVE speakers are always downgraded
on status-related traits (i.e. ‘education’ or ‘wealth’), and half of the time this is
combined with favourable attractiveness-related evaluations (e.g. ‘friendly’ or
‘kind’) (Buck, 1968; Garner & Rubin, 1986; Irwin, 1977; Johnson & Buttny, 1982;
Speicher & McMahan, 1992)and half of the time with unfavourable ones (Bishop,
1979; Doss & Gross, 1992; Doss & Gross, 1994; Larimer et al., 1988; White et al.,
1998). Thus, it appears that the stereotype regarding AAVE is generally negative
(at least in comparison to Standard American English).
Despite the culturally funded negative stereotype, an aversive racism frame-
Information-Processing Effects
183
work (see Dovidio & Gaertner, 1997; Gaertner & Dovidio, 1986) suggests that
some hearers may have personal beliefs about AAVE that may be more favourable. In addition, most hearers are likely to be sensitive to US American norms
that mark the overt expression of discrimination against African-Americans as
socially inappropriate. Consequently, evaluations of AAVE speakers produced
under conditions of time availability will probably be more sensitive to (positive)
personal beliefs and norms of social desirability, whereas evaluations produced
under conditions of time constraint will probably be more sensitive to the underlying (negative) stereotypes.
Although compelling, the above arguments are highly speculative given that
conditions of information processing have never before been tested in a speaker
evaluation context. Consequently, the present study will focus only on the
following simple, yet important research question:
Research Question: Will available time affect hearers’ evaluations of speakers, and will it affect all evaluations equally?
Experiment 1
Method
In order to detect differences in processing condition only, a within-subjects
design was selected. In this experiment, 49 undergraduates at a major western
urban university in the USA twice listened to and evaluated four tape-recorded
speakers. The sample had an average age of 22.72 years, consisted of 16 men and
32 women (one declined to state), 28 Anglos, 10 Hispanics, 5 African- Americans,
and 3 Asian-Americans (3 declined to state).
Of the four speakers, two African-Americans (one male and one female)
provided the AAVE recordings. The other two speakers were Anglo-American
(male) and Asian/Anglo-American (female) and they provided the Standard
American English recordings. Each of the four speakers recorded a brief passage
(51 words) from a children’s story about a fisherman (see Appendix). Although
the speakers sounded representative of their self-identified social groups to the
author, study participants were also asked to comment on the speakers’ accents.
On a Likert response item where 1 represented ‘very accented’ speech and 7
represented ‘unaccented (native-like)’ speech, both the African-American male
and female were rated as having an ‘accent’ (M = 2.80, SD = 1.40; M = 3.78, SD =
1.71, respectively), whereas the Anglo-American male and Asian/Anglo -American female were perceived as standard (i.e. ‘unaccented’) speakers (M = 6.08, SD
= 1.74; M = 6.14, SD = 1.58, respectively).
Respondents also judged each speaker’s perceived ethnicity on a
free-response item. For the African-American male, only 51% of the respondents
identified him as African-American; all other judgements (32.7%) categorised
him as belonging to some non-dominant social group e.g. ‘Jamaican’, ‘ghetto’, or
‘foreign’ (16.3% of the responses were blank). For the African-American female,
91.8% of the respondents identified her as African-American (8.2% of the
responses were blank). For the Anglo-American male, 67.3% of the respondents
identified him as Anglo, 6.1% as Hispanic, 2% as Asian, and 2% as African-American (22.4% of the responses were blank). Lastly, for the
184
Language Awareness
Asian/Anglo-American female, 77.6% of the respondents identified her as
Anglo and 2% as Asian (20.4% of the responses were blank).
Because of the difference in perceived degree of accentedness between the
African-American male and female, and because of the low rate of recognition
for the male, their recordings were analysed to ensure the presence of phonetic
features identifiable as AAVE.
African-American Female. This speaker’s reading of the text was 15.7s long and
included 14 phonological features of AAVE as identified by Bailey and Thomas
(1998) and Rickford (1999). This includes final cluster reduction for the words
‘and’, ‘fished’, and ‘last’, final consonant deletion in the word ‘fisherman’, final
obstruent devoicing in the word ‘lived’, stopping of interdental fricatives in the word
‘with’, monophthongisation of /aI/ in the words ‘shining’ and ‘line’, front stressing
in the phrase ‘at last’, glottalisation of /t/ in the words ‘sitting’ and ‘little’,
devoicing of /v/ in the word ‘lived’, and realisation of the final /ng/ as /n/ in the
word ‘sitting’.
African-American Male. This speaker’s reading of the text was 13.6s long and
included 18 phonological features of AAVE. This includes final cluster reduction
for the words ‘and’, ‘fished’, and ‘last’, monophthongisation of /aI/ in the words
‘shining’ and ‘line’, glottalisation of /t/ in the words ‘sitting’ and ‘little’, realisation
of the final /ng/ as /n/ in the words ‘sitting’ and ‘something’, tensing of lax vowels
in the word ‘lived’, stopping of interdental fricatives in the word ‘the’, medial
unstressed syllable deletion in the word ‘fisherman’, and rising intonation in the
phrases ‘little shack’ and ‘into the shining water’.
Once recorded, each of the four tapes was digitally edited to create both a
time-availability and a time-constraint condition. In the time-availability condition, each recording was played in its entirety. Afterwards, participants evaluated the speaker, taking as much time as necessary. In the time- constraint
condition, each of the four recordings was edited into eight segments, for a total
of 32 segments which were then randomly ordered onto a master tape. Each
segment was roughly 1-2s in length and separated by 1.1s of silence. During the
1.1s pause, participants quickly marked their evaluation of the speaker on only
one Likert-scaled item of the Speaker Evaluation Instrument (Zahn & Hopper,
1985) before preparing to evaluate the next speaker.
In both experimental conditions, hearers evaluated the exact same speakers,
saying the exact same thing, along the exact same evaluative items. There were,
however, two important differences between the conditions. First, of course, was
the designed difference in available time. Necessarily confounded with this
difference, though, was a second difference in the amount of tape-recorded stimulus material presented; speaker passages were played in their entirety in the
time-availability condition, but only segmentally in the time-constraint condition. Unfortunately, it was necessary to accept such a threat to the internal validity of the experiment given concern over its external validity. By choosing to
include a time-availability condition that closely matched those of typical
speaker evaluation language-attitude studies (i.e. listen to an entire
tape-recorded passage, then evaluate the speaker along a host of items), it was
impossible to design an identical condition of time-constraint; any opportunity
that study participants have to listen to a speaker’s passage in its entirety is also
an opportunity they have to integrate and reflect upon their evaluations. Thus
Information-Processing Effects
185
given that no experimental design could perfectly address all concerns of validity, this design errs in favour of external validity while reserving investigation of
the potential confound for a second experiment.
Because this study targeted four speakers in a within-subjects design, participants were required to make many different evaluations. In order to minimise
fatigue, only eight items from the original 30 item Speech Evaluation Instrument
were used here to measure both the status- and attractiveness-related evaluations (see Appendix B). The four status (‘intelligent–unintelligent’, ‘rich–poor’,
‘upper class–lower class’, ‘educated–uneducated’) and four social attractiveness
(‘kind–unkind’, ‘sweet–sour’, ‘likeable–unlikeable’, ‘friendly–unfriendly’) items
were selected on the basis of their consistently high factor loading scores
(Cargile, 1997; Cargile & Giles, 1997; Zahn, 1990; Zahn & Hopper, 1990). Despite
the chaotic nature of hurriedly marking evaluative items for 32 randomly
ordered speech segments representing four different speakers, the status and
attractiveness subscales demonstrated surprising reliability (Cronbach’s alpha
equalled 0.79 and 0.82 respectively). As expected, the scales exhibited even
greater reliability within the time-availability condition (Cronbach’s alpha
equalled 0.90 and 0.92 respectively).
Before participation, respondents were non-randomly assigned to one of four
different groups. The tape recordings were listened and responded to in a group
setting and each of the four groups heard the recordings in a different order. The
first (N = 11) and third (N = 10) groups responded to the speakers under the
time-constraint condition first, and then under the condition time availability.
For the second (N = 13) and fourth (N = 15) groups, this order was reversed. In
addition, the order of speaker presentation within the time-availability condition
was matched (i.e. each speaker appeared once in each ordered position), and the
order of the 32 speech segments within the time-constraint condition was reassigned randomly across the four group conditions. MANOVA analyses were
conducted to explore whether these presentation orders affected the speaker
ratings. In each case, the multivariate F-statistic was non-significant, thus no
order effect was found.
Results
The data were submitted to a 2 (time condition) ´ 2 (speaker accent) ´ 2
(speaker sex) within-subject , repeated measures MANOVA (mean scores are
reported in Table 1). The three-way interaction effect was significant, lambda =
0.73, F(2,46) = 8.58, p < 0.001.The two-way interaction effect between time condition and speaker sex was next tested while holding accent constant. This effect
was significant only for the AAVE speakers, lambda = 0.74, F(4,186) = 7.69, p <
0.001. Following this, the simple main effect of time condition was tested among
AAVE speakers and was found to be significant only for the female, lambda =
0.53, F(2,46) = 20.70, p < 0.001. Subsequent univariate tests indicated that time
condition affected attractiveness ratings of the female AAVE speaker, F(1,47) =
41.30, p < 0.001, eta-square = 0.47; the same hearers judged the same
non-standard speaker as more socially attractive under a condition of time availability (M = 5.62) than under a condition of time constraint (M = 4.79).
These results indicate an important finding – namely, that speaker evaluations can be influenced by conditions of available time believed to engender
Language Awareness
186
Table 1 Mean scores of evaluations observed in experiment 1
Status
Evaluations
Availability
Constraint
Attractiveness Availability
Evaluations
Constraint
AAVE Speakers
Male
Female
3.23
3.94
3.31
3.74
3.64
5.62
3.64
4.79
SAE Speakers
Male
Female
5.22
5.11
4.98
5.30
5.10
5.70
5.20
5.74
different sorts of information processing on the part of the hearer. However, as
discussed earlier, this experiment necessarily confounds a difference in available
time with a difference in the amount of tape-recorded stimulus material
presented. In order to determine the likelihood that stimulus material differences are responsible for these results, a second experiment was conducted.
Experiment 2
Method
Experiment 2 was conducted exactly as experiment 1, except that participants
were instructed to take as much time as necessary to complete all of their evaluations, including the evaluations for which they heard only a segment of a
speaker’s tape-recorded passage. A second sample of 57 participants was drawn
from the same undergraduate population used in experiment 1. The sample had
an average age of 23 years, consisted of 21 men and 36 women, 34 Anglos, nine
Hispanics, five African-Americans, and five Asian-Americans (four declined to
state). Participants were non-randomly assigned to one of four different listening
groups, each one of whom heard the speakers in a different order of presentation.
As in experiment 1, MANOVA analyses were conducted to explore whether
these presentation orders affected the speaker ratings. In each case, the
multivariate F-statistic was non-significant, thus again no order effect was
found.
Results
The data were submitted to a 2 (stimulus tape) ´ 2 (speaker accent) ´ 2 (speaker
sex) within-subject, repeated measures MANOVA. The three-way interaction
effect was significant, lambda = 0.876, F(2,55) = 3.90, p < 0.05, thus subsequent
two-way interaction effects were examined. Of these effects, only the interaction
between speaker accent and speaker sex was significant, lambda = 0.497, F(2,55)
= 27.86, p < 0.001. Neither the interaction effect between stimulus tape and
speaker accent, lambda = 0.925, F(2,55) = 2.23, p = 0.116, power = 0.437, or
between stimulus tape and speaker sex was significant, lambda = 0.985, F(2,55) =
0.41, p = 0.665, power = 0.113. In addition, no main effect for stimulus tape was
found, lambda = 0.961, F(2,55) = 1.12, p = 0.334, power = 0.237. As such, these
results demonstrate that the presented form of the speakers’ tape recordings (i.e.
played entirely or in segments) had no impact on their subsequent evaluation.
Apparently, in a prescriptive, experimental setting, people can make enough
Information-Processing Effects
187
sense of utterance segments to construct the same attributions about the speaker
as they do when they hear the speaker read the text in its entirety.
Discussion
Together, the results of these two experiments provide a reasonably clear
answer to this study’s research question. Available time can affect hearers’ evaluations of speakers, but in this study the effect appears to be very limited. In experiment 1, evaluations of Standard American English speakers made under a
condition of time constraint were not significantly different from those made
under a condition of time availability. In contrast, differences were found in the
evaluation of AAVE speakers across conditions, but this was limited to a single
female and affected only her social attractiveness ratings. Although limited, the
fact that this female was judged less socially attractive under atypical speaker
evaluation circumstances is most likely attributable to automatic information
processing engendered by limited time; experiment 2 demonstrated that differences in stimulus-tape presentation did not influence hearer evaluations.
Although these experiments provide the first evidence of informationprocessing effects in language-evaluation circumstances, many important questions remain regarding the size and direction of these effects for particular speakers and language behaviours. As discussed earlier, information processing
should affect speaker evaluations only in cases where either personal beliefs or
social norms are discrepant with stereotype-based responses. Viewing this, it
could be anticipated that evaluations of Standard American English speakers
would not change across conditions, as was observed here. However, why were
only attractiveness evaluations affected in the case of the female AAVE speaker?
Although it must be emphasised here that any response to this question has to be
considered speculative at this point, some clues may lie in the evaluative pattern
in which AAVE speakers are judged negatively more often on status-related
items than on those related to social attractiveness.
As discussed earlier, AAVE speakers are sometimes rated as more socially
attractive than Standard American English speakers. Ryan et al. (1984) theorise
that such an evaluative pattern represents a form of token appeasement on the
part of dominant group members. In the case of AAVE speakers, language
stereotypes more directly connect them with lower levels of socioeconomic
success. For example, it is widely perceived that the material success of African-Americans is tied to their ability to speak standard American English (see
Hewitt, 1995; Mydans, 1990). Evaluations of social attractiveness, however,
appear to be less central to the language stereotype, thus hearers may experience
more discretion when making these sorts of judgements. Because AAVE does
not debilitate the perceived social attractiveness of a speaker, it is possible that
hearers may inflate these ratings (when given enough time) as defence against
claims of prejudice that may be brought in light of their unfavourable ratings of
the speaker’s status. Such upgrading just might function as the corollary to
discursive claims such as ‘I’m not a racist, but…’ or ‘some of my best friends are
black’ (see van Dijk, 1987).
In addition to questions concerning the evaluative dimensions affected, this
study also poses questions concerning the interactive effect of information
188
Language Awareness
processing; namely, why weren’t evaluations of the male AAVE speaker influenced by time availability? Of course, it should be noted that the present results
represent only two instantiations of the variable ‘AAVE’, thus overgeneralisation
is an easy trap here (see Jackson, 1992). Even so, because the results of a second
experiment (reported in Cargile, 2001) confirm the present evaluative pattern, it
could be argued that a difference between stereotypes of African-American males
and females may be at work here.
In the USA, African-American males are often stereotyped as angry and criminal (e.g. Hall, 2001; Lombardo, 1978; Majors & Billson, 1992). Indeed, among a
sample of 23 undergraduates asked to provide open-ended comments about each
speaker in this study, 48% indicated that the male AAVE speaker sounded threatening (30% indicated that the second male AAVE speaker used in Cargile (2001)
also sounded threatening). US stereotypes of African-American females,
however, are often more varied. For example, Collins (1990) argues that the image
of the matriarch surrounds African-American women and carries with it both
positive (caretaking) and negative (neglectful) associations. Thus, one possibility
is that activation of such a powerfully negative stereotype in the case of African-American males drives unfavourable and unchanging speaker evaluations,
while a more selective stereotype in the case of African-American females
provides enough ground for hearers to amend their mixed evaluations through
controlled processing. Although this argument is provocative, additional research
is certainly needed to establish the generalisability of the present results, and to
help determine the theoretical mechanisms at work.
Conclusion
Despite important questions that remain regarding the specificity of the effect
of available time, the fact that an effect was observed at all provides some support
for a processual model of language attitudes (Cargile & Bradac, 2001) and its challenge to traditional speaker-evaluation research. These two studies suggest that
speaker evaluations are indicators not only of evoked language stereotypes, but
also of information processing. As a result, past conclusions about language attitudes based on speaker evaluations may be subject to re-evaluation. Perhaps the
evaluations reflect a latent stereotype, but perhaps they better reflect hearers’
personal beliefs regarding the speaker’s language behaviour or their desire to
conform to norms of appropriateness. Whatever the case, language-attitude
researchers need to recognise that speaker evaluations may be a problematic
measure and should continue to investigate ways of using them more
appropriately.
Acknowledgements
Many thanks to Malcolm Finney, Assistant Professor of Linguistics at California State University, Long Beach, for his assistance in describing the speakers in
this study.
Correspondence
Any correspondence should be directed to Dr A.C. Cargile, Department of
Information-Processing Effects
189
Communication Studies, California State University, Long Beach, CA 90840-2407,
USA ([email protected]).
References
Bailey, G. and Thomas, E. (1998) Some aspects of African-American Vernacular English
phonology. In S. Mufwene, J. Rickford, G. Bailey and J. Baugh (eds) African American
English: Structure, History, and Use (pp. 85–109). London: Routledge.
Banaji, M.R. and Greenwald, A.G. (1995) Implicit gender stereotyping in judgments of fame.
Journal of Personality and Social Psychology 68, 181–198.
Bishop, G.D. (1979) Perceived similarity in interracial attitudes and behaviors: The effects of
belief and dialect style. Journal of Applied Social Psychology 9, 446–465.
Blair, I. and Banaji, M.R. (1996) Automatic and controlled processes in gender stereotyping.
Journal of Personality and Social Psychology 70, 1142–1163.
Bradac, J.J., Cargile, A.C. and Halett, J. (2001) Language attitudes: Retrospect, conspect, and
prospect. In H. Giles and P. Robinson (eds) The New Handbook of Language and Social
Psychology (pp. 137–155). Chichester, England: John Wiley and Sons.
Bradac, J.J., Konsky, C.W. and Davies, R.A. (1976)Two studies of the effects of lexical diversity
upon judgments of communicator attributes and message effectiveness. Communication
Monographs 43, 70–79.
Brewer, M.B. (1988) A dual process model of impression formation. In T.K. Srull and R.S.
Wyer (eds) Advances in Social Cognition (pp. 1–36). Hillsdale, NJ: Erlbaum.
Buck, J.F. (1968) The effects of negro and white dialectal variations upon attitudes of college
students. Speech Monographs 2, 181–186.
Cargile, A.C. (1997) Attitudes toward Chinese-accented speech: An investigation in two
contexts. Journal of Language and Social Psychology 16, 434–443.
Cargile, A.C. (2001) Measures of language attitudes: Testing the validity of speaker
evaluations. Paper presented at the Western Speech Communication Association’s annual
convention, Coeur d’Alene, ID.
Cargile, A.C. and Bradac, J.J. (2001) Attitudes toward language: A review of
speaker-evaluation research and a general process model. In W.B. Gudykunst (ed.)
Communication Yearbook 25 (pp. 347–382). Mahwah, NJ: Lawrence Erlbaum Associates.
Cargile, A.C. and Giles, H. (1997)Understanding language attitudes: Exploring listener affect
and identity. Language and Communication 17, 195–217.
Cargile, A.C., Giles, H., Ryan, E.B. and Bradac, J.J. (1994) Language attitudes as a social
process: A conceptual model and new directions. Language and Communication 14, 211–236.
Collins, P.H. (1990) Black Feminist Thought: Knowledge, Consciousness, and the Politics of
Empowerment. New York: Routledge.
Creber, C. and Giles, H. (1983) Social context and language attitudes: The role of
formality-informality of the setting. Language Sciences 5, 155–162.
Dailey-O’Cain, J. (2000) The sociolinguistic distribution of and attitudes toward focuser like
and quotative like. Journal of Sociolinguistics 4, 60–80.
Devine, P.G. (1989) Stereotypes and prejudice: Their automatic and controlled components.
Journal of Personality and Social Psychology 56, 5–18.
Doss, R.C. and Gross, A.M. (1992) The effects of Black English on stereotyping in intraracial
perceptions. The Journal of Black Psychology 18, 47–58.
Doss, R.C. and Gross, A.M. (1994) The effects of Black English and code-switching on
intraracial perceptions. Journal of Black Psychology 20, 282–293.
Dovidio, J.F. and Gaertner, S.L. (1997) On the nature of contemporary prejudice: The causes,
consequences, and challenges of aversive racism. In J. Eberhardt and S.T. Fiske (eds)
Confronting Racism (pp. 3–32). Thousand Oaks, CA: Sage.
Dovidio, J.F., Kawakami, K., Johnson, C., Johnson, B. and Howard, A. (1997) On the nature of
prejudice: Automatic and controlled processes. Journal of ExperimentalSocial Psychology 33,
510–540.
Edwards, J. (1982) Language attitudes and their implications among English speakers. In E.
Ryan and H. Giles (eds) AttitudeTowards Language Variation:Social and Applied Contexts (pp.
20–33). London: Edward Arnold.
190
Language Awareness
Fazio, R.H., Jackson, J.R., Dunton, B.C. and Williams, C.J. (1995) Variability in automatic
activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of
Personality and Social Psychology 69, 1013–1027.
Fazio, R.H., Sanbonsatsu, D.M., Powell, M.C. and Kardes, F.R. (1986) On the automatic
activation of attitudes. Journal of Personality and Social Psychology 50, 229–238.
Fiske, S.T. and Neuberg, S.L. (1990) A continuum of impression formation, from
category-based to individuating processes: Influences of information and motivation on
attention and interpretation. In M. Zanna (ed.) Advances in Experimental Social Psychology
(pp. 1–74). San Diego, CA: Academic Press.
Gaertner, S.L. and Dovidio, J.F. (1986) The aversive form of racism. In J.F. Dovidio and S.L.
Gaertner (eds) Prejudice, Discrimination, and Racism (pp. 61–89). Orlando, FL: Academic
Press.
Garner, T. and Rubin, D.L. (1986)Middle class Blacks’ perceptions of dialect and style shifting:
The case of southern attorneys. Journal of Language and Social Psychology 5, 33–48.
Gilbert, D.T. and Hixon, J.G. (1991) The trouble of thinking: Activation and application of
stereotypic beliefs. Journal of Personality and Social Psychology 60, 509–517.
Giles, H. and Coupland, N. (1991) Language attitudes: Discursive, contextual, and
gerontological considerations. In A. Reynolds (ed.) Bilingualism, Multiculturalism, and
Second Language Learning (pp. 21–42). Hillsdale, NJ: Lawrence Erlbaum Associates.
Hall, R.E. (2001) The ball curve: Calculated racism and the stereotype of African American
men. Journal of Black Studies 32, 104–119.
Hewitt, D. (1995) The language factor. In D. Hewitt (Producer) 60 Minutes. New York: CBS
Inc.
Hopper, R. (1977) Language attitudes in the employment interview. Communication
Monographs 44, 346–351.
Irwin, R.B. (1977) Judgments of vocal quality, speech fluency, and confidence of southern
black and white speakers. Language and Speech 20, 261-266.
Jackson, S. (1992) Message Effects Research: Principles of Design and Analysis. New York: The
Guilford Press.
Johnson, F.L. and Buttny, R. (1982) White listeners’ responses to ‘sounding black’ and
‘sounding white’: The effects of message content on judgments about language.
Communication Monographs 49, 33–49.
Koch, L.M. and Gross, A. (1997) Children’s perceptions of Black English as a variable in
intraracial perception. Journal of Black Psychology 23, 215–226.
Kunda, Z. and Thagard, P. (1996) Forming impressions from stereotypes, traits, and
behaviors: A parallel-constraint-satisfaction theory. Psychological Review 103, 284–308.
Lambert, W., Anisfeld, M. and Yeni-Kosmshian, G. (1965) Evaluational reactions of Jewish
and Arab adolescents to dialect and language variations. Journal of Personality and Social
Psychology 2, 84–90.
Lambert, W., Hodgson, R., Gardner, R. and Fillenbaum, S. (1960) Evaluational reactions to
spoken languages. Journal of Abnormal and Social Psychology 60, 44–51.
Larimer, G.S., Beatty, E.D. and Broadus, A.C. (1988) Indirect assessment of interracial
prejudices. The Journal of Black Psychology 14, 47–56.
Leyens, J., Yzerbyt, V. and Schadron, G. (1994) Stereotypes and Social Cognition. Thousand
Oaks, CA: Sage.
Locksley, A., Hepburn, C. and Ortiz, V. (1982) Social stereotypes and judgments of
individuals. Journal of Experimental Social Psychology 18, 23–42.
Lombardo, B. (1978)The Harlem Globetrotters and the perception of the Black stereotype. The
Physical Educator 35, 60–63.
Luhman, R. (1990) Appalachian English stereotypes: Language attitudes in Kentucky.
Language in Society 19, 331–348.
Majors, R. and Billson, J.M. (1992) Cool Pose: The Dilemmas of Black Manhood in America. New
York: Touchstone.
Monteith, M. (1993) Self-regulation of prejudiced responses: Implications for progress in
prejudice-reduction efforts. Journal of Personality and Social Psychology 65, 469–485.
Mydans, S. (1990)Black identity vs. success and seeming ‘White’. The New York Times 89 (April
25 p. B9).
Information-Processing Effects
191
Neely, J.H. (1977) Semantic priming and retrieval from lexical memory: Roles of inhibitionless
spreading activation and limited-capacity attention. Journal of Experimental Psychology 106,
226–254.
Petty, R.E. and Cacioppo,J.T. (1986a)Communication and Persuasion:Central and PeripheralRoutes
to Attitude Change. New York: Springer.
Petty, R.E. and Cacioppo, J.T. (1986b) The elaboration likelihood model of persuasion. In L.
Berkowitz (ed.) Advances in Experimental Social Psychology (pp. 123–205). San Diego, CA:
Academic Press.
Rickford, J. (1999) African American Vernacular English. Malden, MS: Blackwell.
Roskos-Ewoldsen, D.R. (1997) Attitude accessibility and persuasion: Review and a transactive
model. In B. Burleson (ed.) Communication Yearbook 20 (pp. 185–225). Thousand Oaks, CA:
Sage.
Ryan, E.B., Giles, H. and Hewstone, M. (1988) The measurement of language attitudes. In U.
Ammon, N. Dittmar and K.J. Mattheier (eds) Sociolinguistics: An InternationalHandbook of the
Science of Language (pp. 1068–1081). Berlin: de Gruyter.
Ryan, E.B., Hewstone, M. and Giles, H. (1984) Language and intergroup attitudes. In J. Eiser
(ed.) Attitudinal Judgment (pp. 135–160). New York: Springer.
Seligman, C., Tucker, G.R. and Lambert, W.E. (1972) The effects of speech style and other
attributes on teachers’ attitudes toward pupils. Language in Society 1, 131–142.
Shiffrin, R.M. and Schneider, W. (1977) Controlled and automatic human information
processing: II. Perceptual learning, automatic attending, and a general theory. Psychological
Review 84, 127–190.
Speicher, B. and McMahan, S. (1992) Some African-American perspectives on Black English
Vernacular. Language in Society 21, 383–407.
Street, R.L. and Brady, R.M. (1982) Speech rate acceptance ranges as a function of evaluative
domain, listener speech rate, and communication context. Communication Monographs 49,
290–308.
van Dijk, T.A. (1987) Communicating Racism. Newbury Park, CA: Sage.
White, M.J., Vandiver, B.J., Becker, M.L., Overstreet, B.G., Teple, L.E., Hagan, K.L. and
Mandelbaum, E.P. (1998) African American evaluations of Black English and standard
American English. Journal of Black Psychology 24, 60–75.
Zahn, C.J. (1990)The construct validity of the Speech Evaluation Instrument. Paper presented at
the annual meeting of the Speech Communication Association, Chicago.
Zahn, C.J. and Hopper, R. (1985) Measuring language attitudes: The speech evaluation
instrument. Journal of Language and Social Psychology 4, 113–123.
Zahn, C.J. and Hopper, R. (1990) The Speech Evaluation Instrument: A User’s Manual. Miami:
Miami University.
Appendix A: Speaker Text
There was once a fisherman who lived with his wife in a miserable little shack close
to the sea. He went to fish everyday and he fished and fished and at last one day, as he
was sitting looking deep down into the shining water, he felt something on his line.
Appendix B: Speaker Evaluation Scales
1.
2.
3.
4.
5.
6.
7.
8.
Unkind
Intelligent
Sweet
Poor
Unlikeable
Upper class
Educated
Unfriendly
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
:____:____:____:____:____:____:____:
Kind
Unintelligent
Sour
Rich
Likeable
Lower class
Uneducated
Friendly