20
Language Power and Persuasion Charle s S. Areni University of Sydney John R. Spa rks University of Dayton  ABSTRACT Powerless language involves the use of various linguistic markers (i.e., hedges , intens ifiers, deicti c phras es, overly polite languag e, tag ques- tions, and verbal and nonv erbal hesit ations) , which sig nify relativ ely low social status in a given communication context. Power ful language suggests higher social status and is characterized largely by the absence of these markers. The results of a laboratory experiment indi- cated that language power influenced attitudes toward a hypothetical new consumer electronics product, regardless of whether the commu- nication was presented in print versus videotape. For both print and video modalities, speakers using p owerful language were more persua- sive than speaker s using powerles s language. Howev er , power less lan- guage had the additional effect of generating more thoughts about the speaker . These results are discussed in terms of the multiple roles pos- tulate of the elaboration-likeli hood mode. © 2005 Wiley Pe riodi cals, Inc. The idea that lan guage refl ects social powe r is a not a new one. Over three decades ago Lakoff (1975) explored language and gender , and pro- posed that certain patterns of speech constituted a “female register,” which was associated with low social power . Subsequent research on lan- guage and social power examined how particular language markers con-  vey the impression of speaker social power, and the subsequent effects Psychology & Marketing, Vol. 22(6): 50 7–525 (June 200 5) Published online in Wiley InterScience (www .interscience.wiley .com) © 2005 Wiley Periodicals, Inc. DOI: 10.1002/mar .20071 507

Language Power and Persuasion

Embed Size (px)

Citation preview

Page 1: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 1/20

Language Power andPersuasion

Charles S. AreniUniversity of Sydney

John R. SparksUniversity of Dayton

 ABSTRACT

Powerless language involves the use of various linguistic markers (i.e.,hedges, intensifiers, deictic phrases, overly polite language, tag ques-tions, and verbal and nonverbal hesitations), which signify relativelylow social status in a given communication context. Powerful languagesuggests higher social status and is characterized largely by theabsence of these markers. The results of a laboratory experiment indi-

cated that language power influenced attitudes toward a hypotheticalnew consumer electronics product, regardless of whether the commu-nication was presented in print versus videotape. For both print andvideo modalities, speakers using powerful language were more persua-sive than speakers using powerless language. However, powerless lan-guage had the additional effect of generating more thoughts about thespeaker. These results are discussed in terms of the multiple roles pos-tulate of the elaboration-likelihood mode. © 2005 Wiley Periodicals, Inc.

The idea that language reflects social power is a not a new one. Overthree decades ago Lakoff (1975) explored language and gender, and pro-posed that certain patterns of speech constituted a “female register,”which was associated with low social power. Subsequent research on lan-guage and social power examined how particular language markers con-

 vey the impression of speaker social power, and the subsequent effects

Psychology & Marketing, Vol. 22(6): 507–525 (June 2005)

Published online in Wiley InterScience (www.interscience.wiley.com)

© 2005 Wiley Periodicals, Inc. DOI: 10.1002/mar.20071507

Page 2: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 2/20

on message and speaker evaluation. O’Barr (1982) analyzed courtroomtranscripts and identified specific speech markers used by witnesseshigh versus low in social power. The testimonies of expert witnesses of both sexes contained several speech markers typical of high social power,and the testimonies of nonexpert witnesses of both sexes was more likelyto contain speech markers associated with low social power (see alsoHosman & Wright, 1987;O’Barr, 1982). Moreover, experimental researchfound that, regardless of speaker sex, speakers using language stylesassociated with high social power were evaluated more favorably thanspeakers using language associated with low social power (Erickson,Lind, Johnson, & O’Barr, 1978).

O’Barr identified “powerless” speech as characterized by the frequentuse of (a) nonverbal hesitations (e.g., “umm,” “ehh,” etc.), (b) verbal hes-itations (e.g., “you know,” “I mean,” etc.), (c) deictic phrases (e.g., “thatman there”), (d) formal language (e.g., “yes sir,” “therefore,” etc.), (e) tag questions (e.g., “that’s how it happened, isn’t it?”), (f) hedges (e.g., “Iguess,” “sort of,” etc.), and (g) intensifiers (e.g., “he really did,” “it was

 very, very dark,” etc.). On the other hand, high-status speakers werecharacterized largely by the absence of these speech markers. O’Barr(1982) called this speaking style powerful language.

O’Barr’s (1982) initial conception of powerless language was as a uni-fied group of language markers whose frequency of use collectively covar-ied with the status of the speaker. However, later research adopted amolecular view, examining the specific effects of individual power mark-ers. This stream of research identified tag questions, hedges, and verbaland nonverbal hesitations as the strongest indicators of low power speechin experimental settings (Bradac & Mulac, 1984a, 1984b). Subsequentmanipulations of powerful versus powerless language tended to focus

on this subset of powerless speech markers.Initial examinations of the effects of language power found that,although

it influenced perceptions of the source, it had little or no influence on atti-tudes for print communications (Gibbons, Busch, & Bradac, 1991). How-ever, subsequent research found that language power influenced persua-sion when communications were presented via audio or videotape (Sparks& Areni, 2002; Sparks, Areni, & Cox, 1998). As discussed in greater detailbelow, the elaboration-likelihood model (ELM) accounts for the moderat-ing role of communication modality on the effect of language power in per-suasion (Petty & Cacioppo,1986; Petty, Unnava,& Strathman, 1991). How-ever, language power has unique characteristics that produce a myriad of 

effects on persuasion. These effects are discussed below in terms of the“multiple-roles” postulate of the elaboration-likelihood model.

THE ELABORATION-LIKELIHOOD MODEL OF PERSUASION

The elaboration-likelihood model (ELM) refers to a continuum of per-suasion processes,anchored at one end by the peripheral route and at the

 ARENI AND SPARKS508

Page 3: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 3/20

other by the central route. When motivation and ability to think abouta given topic are both high, message recipients elaborate on relevantarguments, and persuasion follows the central route. Under these con-ditions, resulting attitudes are based on whether the arguments pre-sented make a convincing case for the position advocated. However, if either motivation or ability is limited, elaboration likelihood is relativelylow and persuasion follows the peripheral route. In these instances, atti-tudes may be influenced by simple cues such as the persuasion context(i.e., preexisting mood, uncomfortable temperature, pleasant surround-ings, etc.), irrelevant aspects of the communication (i.e., an unattractivespeaker, a credible spokesperson, pleasant visual elements, etc.), and/orthe use of simple heuristics (i.e., “experts are always right,”“you get whatyou pay for,” etc.) (Petty, Cacioppo, & Schumann, 1983).

Within this basic framework, communication variables can influencepersuasion in five ways: (1) as a peripheral cue, (2) as an elaborationenhancer, (3) as an elaboration inhibitor, (4) as a biasing influence onargument processing, and (5) as a relevant argument. Importantly, the

ELM posits that a given variable may influence persuasion in more thanone of these roles (Petty & Cacioppo,1986; Petty, Unnava, & Strathman,1991). Previous explications of the ELM have suggested that variablesoperate in the first role when elaboration likelihood is relatively low, inthe second and third roles when elaboration likelihood is at moderatelevels, and in the fourth and fifth roles when elaboration likelihood isrelatively high.

However, in many tests of the ELM, variables specified as peripheralcues have been temporally and/or spatially separated from the centralarguments in the communication.That is, cues are perceptually distinctfrom arguments, and this separation may overly simplify the role(s) of 

these variables in persuasion (cf. Munch & Swasy, 1988; Swasy & Munch,1985). The research reported below examines the effects of languagepower—a variable that is, by definition, intertwined with the text of pre-sented arguments—on elaboration and persuasion. It tests the predictionthat, although language power operates as a simple cue when elabora-tion likelihood is restricted, when elaboration likelihood is relativelyhigh, language power both biases and inhibits the processing of pre-sented arguments.

Language Power as a Peripheral Cue

Previous research has demonstrated that language power operates as aperipheral cue when elaboration likelihood is relatively low. More specif-ically, under these conditions language power influences an audience’s per-ception of the speaker or source of the communication, which then servesas a simple heuristic for forming or modifying attitudes toward the focaltopic. For example, Sparks et al. (1998) reported that audio and videomessages produced more favorable attitudes when the language usedwas powerful as opposed to powerless. However, language power had lit-

LANGUAGE POWER AND PERSUASION 509

Page 4: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 4/20

tle or no effect on persuasion when the message was presented in print.For print messages, the quality of presented arguments was the dominantinfluence on attitudes. Likewise, Sparks and Areni (2002) found that aspeaker was more persuasive when he used powerful rather than pow-erless language, but only when the communication was presented onaudiotape. Language power had little or no effect on attitudes when thecommunication was presented in the form of a printed transcript.

Presumably, the “real time” nature of the audio and video messages lim-ited the audience’s opportunity to respond to message content (MacIn-nis, Moorman,& Jaworski, 1991). Moreover, the audio and video formatsmay have increased the salience of the actual speaker (e.g., voice quali-ties, intonations, mannerisms, etc.), thus diverting attention away fromrelevant arguments (Chaiken & Eagly, 1983); hence language powerinfluenced persuasion via the peripheral route. By contrast, messagerecipients could process the print communications at their own pace,without inference from the speaker’s gestures, speech patterns, etc., solanguage power was inconsequential. As shown in Figure 1, this sug-

gests the following specific predictions for the effects of language poweras a peripheral cue:

H1: When a message is presented in a video format, powerless lan-guage operates as a negative peripheral cue.

H1(a): Under these conditions, message recipients form less favorableattitudes when the speaker uses powerless as opposed to pow-erful language.

H1(b): Under these conditions, speaker-related thoughts are more neg-

ative when the speaker uses powerless as opposed to powerfullanguage.

 ARENI AND SPARKS510

Figure 1. Predicted effects of language power on persuasion in the video modalitycondition.

Page 5: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 5/20

H1(c): Under these conditions, language power has little or no influ-ence on the valence of argument-related thoughts.

H1(d): Under these conditions, the effects of language power on atti-tudes are mediated by the valence of speaker-related thoughtsrather than argument-related thoughts.

Powerless Language as a Distracting Influence

Bradac and Street (1990) and Sparks and Areni (2002; Sparks et al.,1998)have discussed yet another effect of language power within the ELM frame-work.The various linguistic markers characteristic of powerless languagerepresent disruptions to the flow of the discourse (Bradac & Street, 1990).This may have the effect of distracting audiences from processing mes-sage content. For example, Gibbons et al. (1991) and Sparks and Areni(2002) found that audiences generated fewer message-related thoughtswhen a speaker used powerful rather than powerless speech.The disrup-

tions created by powerless language appear to redirect message recipi-ents’ thoughts toward the speaker (cf. Munch & Swasy, 1988).Along theselines, both Gibbons et al. (1991) and Sparks and Areni (2002) reported agreater number of speaker-related thoughts when the communicator usedpowerless as opposed to powerful language. Within the ELM, variableshave a distracting influence on argument-related processing when elabo-ration likelihood is otherwise high (Petty, Wells, & Brock, 1976; Munch &Swasy, 1988), as with print media (Sparks et al., 1998). Hence, as shownin Figure 2, the distraction effect suggests the following hypotheses:

H2: When a message is presented in a print format, powerless lan-

guage distracts message recipients from processing presentedarguments.

LANGUAGE POWER AND PERSUASION 511

Figure 2. Predicted effects of language power on persuasion in the print modalitycondition.

Page 6: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 6/20

H2(a): Under these conditions, message recipients generate morespeaker-related thoughts when the speaker uses powerless asopposed to powerful language.

H2(b): Under these conditions, message recipients generate fewerargument-related thoughts when the speaker uses powerless asopposed to powerful language.

Language Power as a Biasing Influence

 Additional research on language power suggests a third role within theELM framework. Numerous studies have found that powerless lan-guage creates the perception that the speaker is unsure or lacks confi-dence in what he or she is saying (Bradac & Mulac, 1984a, 1984b). Thismay have the effect of encouraging message recipients to doubt thespeaker and generate counterarguments attacking his or her position(Sparks & Areni, 2002). In addition, given that language style is embed-

ded in the text of presented arguments, it may be difficult for audiencesto separate the effects of one variable from another. For example, the useof hedges (e.g., “may,” “probably,” “could,” etc.) is an indicator of power-less language (i.e., a peripheral cue), but it may also be viewed as weak-ening the substance of key assertions (i.e., a weak argument) (Areni,2002).Together, these effects point to language power as a biasing influ-ence on the processing of relevant arguments. Hence, the following hypotheses are advanced:

H3: Language power acts as a biasing influence on the processing of relevant arguments when a message is presented in a print

format.

H3(a): Under these conditions, message recipients form less favorableattitudes toward the product when the speaker uses powerlessas opposed to powerful language.

H3(b): Under these conditions, speaker-related thoughts are more neg-ative when the speaker uses powerless as opposed to powerfullanguage.

H3(c): Under these conditions, argument-related thoughts are more

negative when the speaker uses powerless as opposed to pow-erful language.

H3(d): Under these conditions, the effect of language power on atti-tudes is mediated by the valence of argument-related thoughtsrather than speaker-related thoughts.

 ARENI AND SPARKS512

Page 7: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 7/20

METHOD

These hypotheses were tested via a laboratory experiment with a 2 (video-taped versus printed transcript) 2 (powerful versus powerless lan-guage) 2 (strong versus weak arguments) factorial design. One hun-dred twenty-one undergraduate students enrolled in a business courseat a midwestern university were recruited to participate in a study involv-ing the evaluation of a test ad.They received two percent credit towardtheir overall mark as an incentive to participate. Respondents were ran-domly assigned to experimental conditions and processed in groups of 10–12.

Procedure

For all conditions, the experimenter entered the room, distributed aquestionnaire, and asked respondents to leave it facedown. When allrespondents had received the questionnaire, they were asked to read

the instructions on the first page while the experimenter read themaloud. Each respondent was exposed to only one experimental stimulus.Respondents in the print conditions were told they would read a “ver-batim printed transcript of a person giving a testimonial about a com-puter product” and they were to evaluate the testimonial on a numberof different dimensions, including its suitability for use in an actualadvertisement. The hypothetical focal product was the Scann-X com-puter scanner. After the transcripts were distributed, respondents had5 minutes to read the transcripts, each of which was approximately 600words in length. The 5-minute exposure time was designed to be ade-quate for processing message content, allowing for the scrutiny of each

argument. After all respondents finished reading the transcripts, theywere instructed to complete their questionnaires. They had 10 minutesfor this task.

Respondents in the video conditions were told they would view a “1-to 2-minute videotape of a person giving a testimonial about a computerproduct”and they were to evaluate the testimonial on a number of dimen-sions, including its suitability for use in a television commercial.After thispoint, the procedures for print and video conditions were the same.Whentime had elapsed, the questionnaires were collected and respondentswere debriefed regarding the study’s purpose. There was no deception inany aspect of the experiment. No attempt was made to alter respondents’

motivation to process the content of the stimuli.

Independent Variables

Video Versus Print Modality. Communication modality was manipu-lated by varying whether the testimonial was presented as a written

LANGUAGE POWER AND PERSUASION 513

Page 8: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 8/20

transcript versus as an actual videotape (MacInnis et al., 1991). In orderto create the latter condition, a professional actor and film crew werehired.

 Language Power. Language power was manipulated according to thepresence versus absence of speech markers associated with powerlesslanguage.The language used in the powerless condition contained speechmarkers commonly associated with a powerless language, including sixtag questions (e.g., “. . . isn’t it?” “. . . right?”), 12 hedges (e.g., “”I guess,”“I think”), 18 verbal hesitations (e.g., “. . . you know . . . ,” “Let’s see now. . . ”), and 12 nonverbal hesitations (e.g., “um,” “ah”). The frequency of these indicators is consistent with those reported by O’Barr in his orig-inal work on language power (Ericksen et al., 1978; O’Barr, 1982). Thepowerful version of the testimonial was characterized by the absence of these indicators (see O’Barr, 1982).

 Argument Quality. Both the strong and weak arguments were based

on actual magazine advertisements for computer scanners. Briefly, thestrong arguments stated that the Scann-X (a) scans files in 5–10 sec-onds, (b) captures colors and details perfectly, (c) comes with five photo-editing software packages, (d) automatically adjusts for transparentmedia, (e) is easy to install in minutes, (f) has a 24-hour, 7-day-a-week,toll-free technical assistance telephone line, (g) costs less than $200, and(h) comes with a 5-year warranty. By contrast, the weak arguments statedthat the scanner (a) scans files in 5–10 minutes, (b) captures black-and-white graphics for simple diagrams, (c) comes with a photo-editing soft-ware package that costs $150 extra, (d) requires a special transparencycover to adjust for transparent media, (e) can be installed in about an hour,

(f) has a 900 technical assistance telephone line available from 9 to 5 onweekdays, (g) costs $300, and (h) comes with a 30-day warranty.

Dependent Variables

 Attitude Toward the Product. To measure attitude toward the prod-uct, respondents answered the question, “If you were in the market fora new scanner, how would you rate the purchase of the Scann-X” on six7-point scales anchored by undesirable desirable, harmful–beneficial,

bad–good, foolish–wise, unfavorable–favorable, and negative–positive.The results of an exploratory factor analysis revealed a single factor

solution; the first factor accounted for 85% of the variance, and no othereigenvalue exceeded 1. Factor loadings exceeded .86 for all six items, andinternal consistency was more than adequate (α   .96). The mean of the six items was the measure of attitude toward the product.

Valenced Thought Indices. Respondents listed “any and all thoughtsyou had while (reading versus viewing) the endorser’s testimonial. These

 ARENI AND SPARKS514

Page 9: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 9/20

can include things related to the speaker, the actual product, or anything else you might have thought about—including things completely unre-lated to the testimonial.” In order to foster distinct cognitive responses,respondents were instructed to “list each thought on a separate line.”Two coders, one a professional market researcher and the other a post-graduate research student, assigned individual cognitive responses tothe following categories: (a) positive argument-related thoughts, (b) neg-ative argument-related thoughts, (c) positive thoughts about the speaker,(d) negative thoughts about the speaker, (e) other positive thoughts,and(f) other negative thoughts. Both coders were blind to the experimentalhypotheses. Intercoder agreement was 81% across all categories, withdiscrepancies resolved via discussions between the experimenter andthe two coders. Following procedures adapted from Greenwald (1968),the valenced index for argument-related thinking was the differencebetween the number of positive argument-related thoughts and the num-ber of negative argument-related thoughts. Likewise, the valenced indexfor endorser-related thinking was the number of positive speaker-related

thoughts minus the number of negative speaker-related thoughts.

 Extent of Argument-Related Thinking. The extent of respondents’argument-related thinking was measured as the total number of argu-ment-related thoughts.

 Extent of Endorser-Related Thinking. Likewise, the extent of endorser-related thinking was measured as the total number of speaker-related thoughts.

RESULTS

H1(a) and H3(a) predict that language power influences attitude towardthe product when the message is presented in a video versus print modal-ity, respectively. These hypotheses were tested via a three-way ANOVA with message modality, argument quality, and language power as theindependent variables, and attitude toward the product as the depend-ent variable. Not surprisingly, there was a main effect of argument qual-ity ( F 1,11458.3, p .001, ω 2 .29). Respondents reported more positiveattitudes when they were exposed to strong ( M  4.9) rather than weak

( M  3.3) arguments. However, contrary to previous research, the argu-ment qualitymodality interaction was not significant ( F 1,114 1), indi-cating that argument quality influenced persuasion even when theendorsement was presented via videotape. As predicted, the main effectof language power attained significance ( F 1,114 18.3, p   .001,  ω 2 0.09). Attitudes toward the product were more favorable when theendorser used powerful ( M  4.6) as opposed to powerless ( M  3.7) lan-

LANGUAGE POWER AND PERSUASION 515

Page 10: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 10/20

guage.However, the language powermodality interaction effect was notsignificant ( F 1,114 1), indicating that this effect was apparent in boththe print and video conditions. Hence, both H1(a) and H3(a) were sup-ported.

H1(b) and H3(b) predict that language power affects the valence of speaker-related thoughts when the message is presented in video versusprint formats, respectively. These hypotheses were tested via a three-way ANOVA with modality, argument quality, and language power asthe independent variables, and the valence of speaker-related thoughtsas the dependent variable. The main effect of language power on the

 valence of speaker-related thoughts was highly significant ( F 1,114 73.7, p .001, ω 2 .36). However, the language powermodality interactiondid not achieve significance ( F 1,114 1). Further analyses indicated thatthe effect of language power on the valence of speaker-related thoughtswas significant in both the video and print conditions (video: F 1,54 37.7,

 p .001, ω 2 .38; print: F 1,60 35.4, p .001, ω 2 0.34). In both cases,respondents had fewer unfavorable thoughts when the speaker used

powerful rather than powerless language (video: M powerful

–.67, M power-

less –3.1; print M powerful –2.2, M powerless –.19). Hence, the data sup-port both H1(b) and H3(b).

H3(c) and 1(c) predict that language power influences the valence of argument-related thoughts when the message is presented in print, butnot when it is presented in video, respectively. These hypotheses weretested via a three-way ANOVA with modality, argument quality, andlanguage power as the independent variables, and the valence of argu-ment-related thoughts as the dependent variable. The results indicatea main effect of argument quality ( F 1,114 39.8, p .001, ω 2 .23). Notsurprisingly, respondents reported more favorable argument-related

thoughts when the speaker presented strong ( M  .40) rather thanweak arguments ( M  –1.7). However, the language power main effectwas nonsignificant ( F 1,114 1). Moreover, neither the argument quality modality ( F 1,114 1) nor the language powermodality ( F 1,114 1)interaction effect attained significance. Hence, neither H1(c) nor H3(c)is supported.

H1(d) posits that language power influences attitude toward the prod-uct via its effect on the valence of speaker-related thoughts when themessage is presented in a video format. The criteria for mediation werebased on Baron and Kenny (1986). Briefly, a variable mediates the effectof an experimental factor on a dependent variable if (a) the effect of theexperimental factor on the mediator is as large or larger than its effect onthe dependent variable, (b) the mediator is correlated with the depend-ent variable, and (c) when the mediator is introduced as a covariate, theeffect of the experimental factor on the dependent variable wanes or iseliminated altogether.

H1(d) was tested via three 2-way ANOVAs and two 2-way ANCOVAsperformed on the data from the video conditions. Language power and

 ARENI AND SPARKS516

Page 11: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 11/20

argument quality were the independent variables, and attitude towardthe product, valence of speaker-related thoughts, and valence of argu-ment-related thoughts were the dependent variables. The effect of lan-guage power on the valenced speaker-related thought index was larger( F 1,54 37.7, p .001,  ω 2 .38) than its effect on attitude toward theproduct ( F 1,54 14.9, p .001,  ω 2 .13). However, neither the secondnor the third criteria were met. When the valenced speaker-relatedthought index was included as a covariate, it was not predictive of atti-tude toward the product ( F 1,53 1.43, p .20). Moreover, with the covari-ate in the model, the effect of language power on attitude toward the pro-posal was approximately the same ( F 1,58 15.9, p   .001,  ω 2 .11).Hence, H1(d) was not supported.The valenced argument-related thoughtindex did not fare any better as a mediating variable in the video con-dition. The effect of language power on this index was not significant( F 1,54 1), failing a criterion necessary for mediation. Hence the effectof language power on attitude toward the proposal was not mediated byeither valenced thought index in the video condition.This overall result

is discussed in detail below.H3(d) predicts that language power influences attitude toward theproduct via its effect on the valence of argument-related thoughts whenthe message is presented in print. However, with the valenced argu-ment-related index as the dependent variable and argument qualityand language style as independent variables, the data did not producea significant language power main effect ( F 1,60 1). In the absence of a significant effect, argument-related thoughts do not appear to medi-ate the effect of language style on attitude. Therefore, the data do notsupport H3(d).

H2(a) and H2(b) posit that when the message is presented in print,

powerless language redirects thought toward the speaker and awayfrom relevant arguments, respectively. As a result, message recipientsgenerate fewer argument-related thoughts and more speaker-relatedthoughts when the speaker uses powerless as opposed to powerful lan-guage. Two 3-way ANOVAs were conducted with modality, argumentquality, and language power as the independent variables and numberof argument-related thoughts and number of speaker-related thoughtsas the two dependent variables. The three-way ANOVA for speaker-related thoughts revealed a main effect of language power ( F 1,114 58.0,

 p .001, ω 2 .31). However, the language powermodality interactioneffect was nonsignificant ( F 1,114 1). Additional analyses revealed the

effects of language power were significant regardless of modality (print: F 1,60 24.15, p .001, ω 2 .26; video: F 1,54 34.2, p .001, ω 2 .36).In the video condition, more speaker-related thoughts were generatedwhen powerless ( M  3.75) rather than powerful ( M  1.55) languagewas used. Likewise, for the print version respondents generated morespeaker-related thoughts in the powerless ( M  2.97) versus powerful( M  1.21) language condition. Hence, these results support H2(c), with

LANGUAGE POWER AND PERSUASION 517

Page 12: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 12/20

the qualification that a similar effect of language power also emergedin the video condition.

With respect to the number of argument-related thoughts, the maineffect of language power did not achieve significance ( F 1,114 1), nor didthe predicted language powermodality interaction ( F 1,114 1). Hence,H2(b) was not supported. The result of each hypothesis test is summa-rized in Table 1 (see also Table 2).

 ARENI AND SPARKS518

Table 1. Summary of Hypothesis Tests.

Research Hypothesis Status

Language Power as a Peripheral Cue

H1(a) When a message is presented in video format, message recipients

form less favorable attitudes toward the product when the Supportedspeaker uses powerless as opposed to powerful language.

H1(b) When a message is presented in video format, speaker-relatedthoughts are more negative when the speaker uses powerless as Supportedopposed to powerful language.

H1(c) When a message is presented in video format, language powerhas little or no influence on the valence of argument-related Not Supportedthoughts.

H1(d) When a message is presented in video format, the effect oflanguage power on attitudes is mediated by the valence of Not Supportedspeaker-related thoughts rather than argument-related thoughts.

Powerless Language as a Distracting Influence

H2(a) When a message is presented in print format, message recipientsgenerate more speaker-related thoughts when the speaker uses Supportedpowerless as opposed to powerful language.

H2(b) When a message is presented in print format, message recipientsgenerate fewer argument-related thoughts when the speaker Not Supporteduses powerless as opposed to powerful language.

Language Power as a Biasing Influence

H3(a) When a message is presented in print format, message recipientsform less favorable attitudes toward the product when the Supportedspeaker uses powerless as opposed to powerful language.

H3(b) When a message is presented in print format, speaker-relatedthoughts are more negative when the speaker uses powerless Supportedas opposed to powerful language.

H3(c) When a message is presented in print format, argument-relatedthoughts are more negative when the speaker uses powerless Not Supportedas opposed to powerful language.

H3(d) When a message is presented in print format, the effect of languagepower on attitudes is mediated by the valence of argument- Not Supportedrelated thoughts rather than speaker-related thoughts.

Page 13: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 13/20

LANGUAGE POWER AND PERSUASION 519

   T  a   b   l  e   2 .   M  e  a  n  s  a  n   d   S   t

  a  n   d  a  r   d   D  e  v   i  a   t   i  o  n  s   R  e   l  a   t  e   d   t  o

   H  y  p  o   t   h  e  s   i  s   T  e  s   t  s .

   V   i   d  e  o

   S   t   i  m  u   l   i

   V  a   l  e  n  c  e   d   S  p  e  a   k  e  r -

   V  a   l  e  n

  c  e   d   A  r  g  u  m  e  n   t -

   E  x   t  e  n   t  o   f

   A   t   t   i   t  u   d  e   t  o  w  a  r   d

   R  e   l  a   t  e   d   C  o  g  n   i   t   i  v  e

   R  e   l  a

   t  e   d   C  o  g  n   i   t   i  v  e

   E  x   t  e  n   t  o   f   S  p  e  a   k  e  r

 -

   A  r  g  u  m  e  n   t -

   C  o  n   d   i   t   i  o  n

   t   h  e   P  r  o  p  o  s  a   l

   R  e  s  p  o  n  s  e   I  n   d  e  x

   R  e  s  p  o  n  s  e   I  n   d  e  x

   R  e   l  a   t  e   d   T   h   i  n   k   i  n  g

   R  e   l  a   t  e   d   T   h   i  n   k   i  n  g

   A  r  g  u  m  e  n   t   Q  u  a   l   i   t  y

   L  a  n  g  u  a  g  e   S   t  y   l  e

   N

   M  e  a  n

   S

   D

   M  e  a  n

   S   D

   M  e  a

  n

   S   D

   M  e  a  n

   S   D

   M  e  a  n

   S   D

   S   t  r  o  n  g

   P  o  w  e  r   f  u   l

   1   6

   5 .   6

   9

   0 .   4   6

 -   0 .   3

   8

   0 .   9

   6

   0 .   8   6

   1 .   4

   5

   1 .   2

   5

   1 .   1   3

   1 .   3

   6

   1 .   3

   1

   S   t  r  o  n  g

   P  o  w  e  r   l  e  s  s

   1   4

   4 .   1

   5

   1 .   2   2

 -   3 .   0

   7

   1 .   8

   2

 -   0 .   2

   9

   1 .   9

   0

   3 .   5

   7

   1 .   5   0

   2 .   2

   1

   1 .   4

   8

   W  e  a   k

   P  o  w  e  r   f  u   l

   1   5

   3 .   5

   2

   1 .   4   8

 -   1 .   0

   0

   1 .   3

   1

 -   2 .   5

   3

   1 .   8

   5

   1 .   8

   7

   1 .   3   6

   3 .   1

   3

   1 .   5

   1

   W  e  a   k

   P  o  w  e  r   l  e  s  s

   1   3

   2 .   6

   3

   1 .   4   1

 -   3 .   0

   8

   1 .   7

   5

 -   1 .   9

   2

   1 .   7

   8

   3 .   9

   2

   1 .   7   1

   3 .   0

   8

   2 .   7

   8

   P  r   i  n   t   S   t   i  m  u   l   i

   V  a   l  e  n  c  e   d   S  p  e  a   k  e  r -

   V  a   l  e  n

  c  e   d   A  r  g  u  m  e  n   t -

   E  x   t  e  n   t  o   f

   A   t   t   i   t  u   d  e   t  o  w  a  r   d

   R  e   l  a   t  e   d   C  o  g  n   i   t   i  v  e

   R  e   l  a

   t  e   d   C  o  g  n   i   t   i  v  e

   E  x   t  e  n   t  o   f   S  p  e  a   k  e  r

 -

   A  r  g  u  m  e  n   t -

   C  o  n   d   i   t   i  o  n

   t   h  e   P  r  o  p  o  s  a   l

   R  e  s  p  o  n  s  e   I  n   d  e  x

   R  e  s  p  o  n  s  e   I  n   d  e  x

   R  e   l  a   t  e   d   T   h   i  n   k   i  n  g

   R  e   l  a   t  e   d   T   h   i  n   k   i  n  g

   A  r  g  u  m  e  n   t   Q  u  a   l   i   t  y

   L  a  n  g  u  a  g  e   S   t  y   l  e

   N

   M  e  a  n

   S

   D

   M  e  a  n

   S   D

   M  e  a

  n

   S   D

   M  e  a  n

   S   D

   M  e  a  n

   S   D

   S   t  r  o  n  g

   P  o  w  e  r   f  u   l

   1   7

   5 .   0

   9

   1 .   0   4

   0 .   0

   6

   0 .   7

   5

   0 .   4   7

   0 .   9

   4

   1 .   5

   9

   1 .   3   3

   1 .   0

   0

   1 .   1

   7

   S   t  r  o  n  g

   P  o  w  e  r   l  e  s  s

   1   6

   4 .   7

   1

   1 .   2   2

 -   2 .   5

   0

   1 .   7

   9

   0 .   4   4

   2 .   0

   0

   3 .   2

   5

   1 .   9   8

   1 .   3

   1

   1 .   5

   4

   W  e  a   k

   P  o  w  e  r   f  u   l

   1   5

   3 .   9

   2

   1 .   1   3

 -   0 .   4

   7

   1 .   1

   9

 -   1 .   4

   0

   1 .   8

   4

   0 .   8

   0

   1 .   0   1

   2 .   0

   0

   1 .   4

   1

   W  e  a   k

   P  o  w  e  r   l  e  s  s

   1   6

   3 .   1

   3

   1 .   1   4

 -   1 .   8

   1

   1 .   3

   3

 -   1 .   0

   0

   0 .   5

   9

   2 .   6

   9

   1 .   2   5

   1 .   9

   4

   1 .   6

   5

Page 14: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 14/20

DISCUSSION

The results reported above generally support the prediction that lan-guage power plays multiple roles in the persuasion process, depending on the modality in which the message is presented. But the actual effectsobserved differed from those predicted by H1(a)–3(d). As shown in Fig-ures 3 and 4, neither speaker-related thoughts nor argument-relatedthoughts mediated the effect of language power in persuasion, regardlessof whether the message was presented on video versus print. There areat least two explanations for this result.

First, it could be the case that language power influences persuasionat an unconscious level, such that message recipients are unaware of the specific language features influencing their attitude. Thought-listing tasks only capture conscious thinking subsequent to actual exposure tothe message. It is not obvious that such an approach can detect subtleeffects of language variables, and as a result, cognitive responses may notbe correlated with attitudes. Instead, the thoughts listed may be formed

well after exposure in order to complete the task requested by theresearcher. However, it should be noted that the results regarding theeffect of powerless language on speaker-related thoughts supported H2(a),suggesting that the thoughts listed captured concurrent cognitiveresponses as intended.

 A second and related explanation is that it was simply too difficult todistinguish between the powerless language markers versus the “qual-ity” of the arguments. Ancillary data provide evidence of confusionbetween the effects of language power versus argument quality. Respon-dents indicated their perception of the endorser on 21 seven-point seman-tic-differential scales. They were instructed to circle the number “that best

reflects your opinion of the speaker.” The positive anchors were intelli- gent, reliable, polished,assertive, honest, powerful, easy to understand, lik-

able, strong, active, effective, dependable, aggressive, influential, knowl-

 edgeable, arrogant, competent, convincing, professional, friendly, andcomposed, and the negative anchors were unintelligent, unreliable,unpol-

ished, timid, dishonest, powerless, hard to understand, unlikable, weak,

 passive, ineffective, undependable, not aggressive, not influential, igno-

rant, humble, incompetent, unconvincing, unprofessional, unfriendly, andnot composed, respectively. Exploratory factor analysis revealed a sin-gle factor solution, so speaker rating was defined as the mean score forthese 21 items. A subsequent three-way ANOVA revealed a main effect

of argument quality on speaker rating ( F 1,108 6.2, p .001).Again, sub- jects rated the speaker more favorably when he presented strong ( M 

3.9) rather than weak ( M  3.5) arguments.Respondents also reported their assessment of the reasons presented

for buying the product on five 7-point scales. The positive anchors werevalid arguments, persuasive, clear ideas, strong, and convincing, and thenegative anchors were invalid arguments, unpersuasive, unclear ideas,

 ARENI AND SPARKS520

Page 15: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 15/20

weak, and unconvincing, respectively. A second exploratory factor analy-sis revealed a single factor solution, so argument rating was defined asthe mean score on these five items. Another three-way interaction indi-cated a main effect of language power on respondents’ ratings of the pre-sented arguments ( F 1,108 24.5, p .0001). Respondents rated the argu-ments has being more convincing when the speaker used powerful ( M 

4.1) as opposed to powerless ( M  3.0) language.Together, these resultssuggest that the effects of language power and argument quality wereintertwined with respect to cognitive responses to the testimonial, sub-

 jective ratings of the speaker and the presented arguments, and atti-tude toward the featured product. This may have compromised the valueof both valenced thought indices as potential mediators.

The difficulty in distinguishing between the effects of argument qual-ity and language power may point to a limitation of previous tests of theELM. In particular, Areni (2002, 2003) has criticized the empirical defi-nition of argument quality within the ELM.Strong (weak) arguments aretypically operationalized as those that generate primarily favorable

(unfavorable) cognitive responses in pretests. However, the actual con-struction of pretested arguments may confound the actual strength of thesupporting evidence with various linguistics markers. The researchreported above improves on previous empirical manipulations of argu-ment quality by removing the effects of power markers via an orthogo-nal manipulation. But there are likely to be several other categories of language markers that also influence persuasion. Future research wouldbenefit by creating orthogonal manipulations for each of these languagemarker categories, hence isolating the effects of argument quality onpersuasion.

It is interesting that argument quality influenced persuasion when

respondents were exposed to the video versions of the testimonial. Underthese conditions it was predicted that message recipients would followthe peripheral route to persuasion, where argument quality has little orno role. However, it is important to remember that the ELM describes acontinuum of elaboration, anchored by the central and peripheral routesto persuasion; it is a matter of degree rather than an absolute dichotomy.Experimental manipulations of “high” versus “low” ability or motiva-tion necessarily create two points on that continuum, but it is not alwayspossible to equate conditions across studies (Areni & Lutz, 1988). Hence,a reasonable interpretation is that the video condition was actually in thelow to moderate range of the elaboration continuum; thus some evidence

of the central route to persuasion was observed. As shown in Figure 3 and 4, there was evidence for the hypotheses

regarding the effect of language power on message recipients’ thoughts.The speech markers in powerless language had the effect of directing thoughts toward the actual speaker. Not only did the number of speaker-related thoughts increase when powerless language was used, but lan-guage power also influenced the corresponding valenced thought index.

LANGUAGE POWER AND PERSUASION 521

Page 16: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 16/20

In this sense, powerless language is similar to the use of rhetorical ques-tions in persuasive communications (Munch & Swasy, 1988; Swasy &Munch, 1985).

Given the relationship between language power and female versus

male “registers” in speech, the gender of the respondent was included ineach of the ANOVAs reported above. In no case did gender produce a sig-nificant main effect or interact with the other variables in the model.However, a more likely scenario is that language power interacts withthe gender of the speaker to influence persuasion and perceptions of thespeaker. Future research would benefit from manipulating speaker gen-der and language power orthogonally to examine possible contingencies.One possible scenario is that female speakers using powerful language may

 ARENI AND SPARKS522

Figure 3. Observed effects of language power on persuasion in the video modalitycondition.

Figure 4. Observed effects of language power on persuasion in the print modalitycondition.

Page 17: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 17/20

be perceived to be more assertive and arrogant than males using the samelanguage, whereas male speakers using powerless language may be seenas more uncertain and lacking in confidence than females using the samelanguage. This would appear to be a fruitful avenue for future research.

Taken together, powerless language appears to act as a negative periph-eral cue and as an attractor of attention, and these combined effectsmake it a rather potent force in persuasion, particularly in contextswhere the source of a message is likely to be salient (i.e., public speak-ing, testimonial advertising, public relations, sales presentations, etc.).Its influence in guiding attention means that it renders otherwise per-suasive elements impotent. Moreover, this may not only weaken theeffects of central arguments, but also the effects of other peripheral cues.Given its effect as a negative cue, this is likely to produce unfavorable atti-tudes, regardless of the merits of the featured product. So an acknowl-edged expert with excellent reasons for advocating a position alreadyaccepted by an audience may, nevertheless, fail to be persuasive if he orshe uses powerless language.

 As with many other laboratory tests of the ELM and other persuasiontheories, the arguments presented in the strong, and especially the weak,argument conditions may seem somewhat artificial and irrelevant toactual advertising claims (Areni, 2002). However, the strong and weakarguments described above were based on a content analysis of severalactual ads for computer scanners. Both the strong and weak conditionswere intended to capture the upper and lower limits of the performancestandards of actual products. For example, the departmental scanner of one of the authors actually does require 5–10 minutes for visually com-plex images, whereas the fastest claim made in an ad was “in seconds.”Nevertheless, in future research more attention could be paid to the real-

ism of the stimuli without compromising the advantages of laboratoryexperiments for developing stimuli that capture construct validity.

REFERENCES

 Areni, C. S. (2002). The proposition-probability model of argument structureand message acceptance. Journal of Consumer Research, 29, 168–187.

 Areni, C. S. (2003). The effects of structural and grammatical variables on per-suasion: An elaboration likelihood model perspective. Psychology & Market-ing, 20, 349–375.

 Areni, C. S., & Lutz, R. J. (1988). The role of argument quality in the elabora-tion likelihood model. In M. J. Houston (Ed.),Advances in consumer research(Vol. 15, pp. 197–202). Provo, UT: Association for Consumer Research.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinc-tion in social psychological research: Conceptual, strategic, and statisticalconsiderations. Journal of Personality and Social Psychology, 51, 1173–1182.

Bradac, J. J., & Mulac, A. (1984a). A molecular view of powerful and powerlessspeech styles: Attributional consequences of specific language features andcommunicator intentions. Communication Monographs, 51, 307–319.

LANGUAGE POWER AND PERSUASION 523

Page 18: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 18/20

Bradac, J. J., & Mulac, A. (1984b). Attributional consequences of powerful andpowerless speech styles in a crisis-intervention context. Journal of Languageand Social Psychology, 3, 1–19.

Bradac, J. J., & Street, R. L., Jr. (1990). Powerful and powerless styles of talk: A theoretical analysis of language and impression formation. Research in Lan-guage and Social Interaction, 23, 195–242.

Chaiken, S., & Eagly, A. H. (1983). Communication modality as a determinant

of persuasion:The role of communicator salience. Journal of Personality andSocial Psychology, 45, 241–256.Erickson, B., Lind, E.A., Johnson,B.C., & O’Barr, W. M. (1978). Speech style and

impression formation in a court setting: The effects of “powerful” and “pow-erless” speech. Journal of Experimental Social Psychology, 14, 266–279.

Gibbons, P., Busch, J., & Bradac, J. J. (1991). Powerful versus powerless lan-guage: Consequences for persuasion impression formation, and cognitiveresponse. Journal of Language and Social Psychology, 10, 115–133.

Greenwald, A. G. (1968). Cognitive learning, cognitive response to persuasion andattitude change. In A.G. Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psy-chological foundations of attitudes (pp. 147–170). New York: Academic Press.

Hosman, L. A., & Wright, J. W. (1987). The effects of hedges and hesitations on

impression formation in a simulated courtroom context. Western Journal of Speech Communication, 51(2), 173-188.

Lakoff, R. (1975). Language and a woman’s place. New York: Harper & Row.MacInnis, D. J., Moorman, C., & Jaworski, B. J. (1991). Enhancing and measur-

ing consumers’ motivation, opportunity, and ability to process brand infor-mation from ads. Journal of Marketing, 55, 32–53.

Munch, J. M., & Swasy, J. L. (1988). Rhetorical questions, summarization fre-quency, and argument strength effects on recall. Journal of ConsumerResearch, 15, 69–76.

O’Barr, W. M. (1982). Linguistic evidence. New York: Academic Press.Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion: Central

and peripheral routes to attitude change. New York: Springer.Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983). Central and peripheralroutes to advertising effectiveness: The moderating role of involvement. Jour-nal of Consumer Research, 10, 134–148.

Petty, R. E., Unnava, H. R., & Strathman, A. J. (1991). Theories of attitudechange. In T. S. Robertson & H. H. Kassarjain (Eds.), Handbook of consumerbehavior (pp. 241–280). Englewood Cliffs, NJ: Prentice Hall.

Petty, R. E.,Wells, G. L., & Brock,T. C. (1976). Distraction can enhance or reduceyielding to propaganda: Thought disruption versus effort justification. Jour-nal of Personality and Social Psychology, 34, 874–884.

Sparks, J. R., & Areni, C. S. (2002). The effects of sales presentation quality andinitial perceptions on persuasion:A multiple role perspective. Journal of Busi-

ness Research, 55, 517–528.Sparks, J. R., Areni, C. S., & Cox, K. C. (1998). An investigation of the effects of 

language style and communication modality on persuasion. CommunicationMonographs, 65, 108–125.

Swasy, J. L., & Munch, J. M. (1985). Examining the target of receiver elabora-tions: Rhetorical question effects on source processing and persuasion. Jour-nal of Consumer Research, 11, 877–886.

 ARENI AND SPARKS524

Page 19: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 19/20

Correspondence regarding this article should be sent to: Charles Areni, Schoolof Business, Economics & Business Building (H69), University of Sydney, Syd-ney, NSW 2006, Australia ([email protected]).

LANGUAGE POWER AND PERSUASION 525

Page 20: Language Power and Persuasion

7/27/2019 Language Power and Persuasion

http://slidepdf.com/reader/full/language-power-and-persuasion 20/20

Reproducedwithpermissionof thecopyrightowner. Further reproductionprohibitedwithoutpermission.