The Semantics of Social Influence: T hreat vs. Persuasion

Edward L. Fink, Deborah A. Cai, Stan A. Kaplowitz, Sungeun Chung,

Mark A. Van Dyke, and Jeong-Naill Killl

7his sludy illue.!tigale:. how lallgullge is used to lIlake moe if ilylllmct altemplJ. Alore specifically, Ih, percewed poteruy and tl}o/uatiofl qf both ir!flul!nct agenl.l and /argt'is and fhe h!lluence tactic llsed OTt examined for their cJficl 011 Ihe aclors' /Jerceived power and ill/luence success. "ntt two influence fachcs examined Off threliis alld altempts to persllad,. II pliol stlld), iN = 1·/5) allti main txperilllelli iN = 189) IVtre cOlltillelet! 10 creale 84 simple jflllnlCfJ, lilt 1I11;ls if alla!;'sis for Ihis inutsligalion. ligtllts are perceivtt! as 1Il0re pOZl"ifiti than largets if illfillence. 77Ie laclic llsed 10 secure compliana (limaltll VersIL! a/tempt 10 persuade) does 1101 offiCI Ihe pC/aivtd POWtT if tilher lilt agml or the largel. A bad agent 0 Ut!1 as mOTe powerjiJllJWII a good agel/I, (Illtl a had lll(gel is comwered mort powerJullhall 1I good la~l!,t'l. rimhmnore, good agents limit more JlICCtH in gaiflillg rompliollCf ~v llJing persuasioll ,hall Ity usiJlg thrmo, (wd Ill')! have more JlICUJS whm UY/llfflCWg a good largelillall a bad largtl. Imp/icfl/ions qf Ihm jllding! are diSCI/net!.

Research on social influence generaU), has treated the topic from one of two perspectives, message production (Brcl\\ n & Levinson, 1987; Dillard, 1990;

]\[c),er, 1996; Wilson , 1997) or message rfkns (e.g., Boster, ~ I itchell , Lapinski, Cooper, Orrego ct aI., 1999; Levine & Boster, 200 I; O ' Keefe, 2002). This article takes a third approach, one that examines how sense is made of social influence messages. I n other ,""ords, when people hear about attempts by one person to influence another, Lhey use their understanding of the agent, larget, and influence tactic to judge the agent and target, and 10 assess the likelihood that the allempt will be successful. T'his study examines this attribution process, focusing on two types of influenc(': LhrealS and attempts to pcr:)uadc.

The perceived power of a communicaLor aflccts the success that the communi­cator may have in task-related communication~, such as in negotiation , as weU as in sociocmOlional communication, such as in establishing a romantic relationship. 1 'T'he influence ta tics that a comllllll1icator cmploys c.g., Ihreats vs. persuasion attempts) affect the perceived power of both the agent and the target of the influence attempt. Previous research has shown that when an agent is perceived by the target to be powerful , the agent is morc likely to succeed in gaining compliance (see, c.g., ~ l ichener, I.twler, & Bacharach 's, 1973, role-pla),ing experiment). Additional research has shown that agenlS who gain compliance with threats are perceived hy participants to be more powerful lhan thos(' who gain compliance with persuasion (Kaplowitz et al.. 1998). What is not known is how the influcnce tactic and lhe

I~(h('(lrd I .. Fillk iJ Pro/wor.(Uu/ Chair alld f)thomll .1. Cni i.1 Auocintt Pr,y,JJ(J( ill 'ht Depart1Tl11lL of CommullIra/loll 01 lilt ( nil'rTJlty oj A/mY/llIu/. S~fln . 1. it a/llml'l!<' iJ Prtifrllor m lilt Sociology DeparlmnJl al .If!(lligoll Stalt l '1III.'tTS1!l'. SUlIgellll G'lluilg, Jjarh A. 1(11/ /~'M. and J(ollg ·.~am Aim art dac/Qra! Jludroll in lil, Dtpart",,,,/ rif COm1ll1l1liral;01l alillt L 'niutTJil) oj.\/lU)'UUuJ. 'fllU (lr/ide ll'flJ "men/td III tIlt l1l(tlmg (J{ Illr . \ itllOnal Commttmcotinn AHocialion, .. \ ol'n'llbn. 20()3. al ,\/tmw. 77" author, wlJh t() IhtmA ,diror frolil. HoJtrr . .,7am'l Priu f)illard, ami lilt hL'O ano'9'm01LI rttttll'fTI .for (ommftl/J Ott (UI rarfi" drqft oj Iht mallUJcnpl.

OmlnwllI((Jltott ,\If1no.l(TaphJ. "01. JO, No. 'I , DC'crrnhn 2003, Pi> 295 ~:Jl6 Ctlp)right :l002, :'\.ltic.n.11 Communication .\.'.:;oci.ltiun 001 , 10. 1 O!\O/IJ:l637i.>O:l20001 791 15

296 COM~IUNIC.\TI()N ~IONO(;RAPHS

perceptions of lhe actors independent of the influence auernpt combine to affect the subsequelll perceptions of the actors and expectations for the influence auempt's success.

A great deal of research has found that aClOrs and obsel"\'ers diffcr in the auributions that they make for the aClOrs' behavior (sec Ajzen, 1996; Fiske & Taylor, 1991 ). People often arc obsel"\'er$. For example, the president may be seen threatening a leader of another nation; a senator may be observed trying to persuade a colleague; in a faculty meeting obsel"\'ations lead to conclusions regarding who has power and who docs not. Obse rvers' future b havior is conditioned upon these awibullons. In short, obselvers are themselves innucnced by these obscrvations. 2

This study builds on the work of Heise (1970) alld Kapluwit< and his colleagues (1998). The sludies by Kaplowitz et al. examine perceptions of agenL' and targets of influence and the tactic used, whereas the work by Heise and his colleagues (e.g., Averett & Heise, 1987; Briu & Heise, 1992; II cise, 1965, 1970; see also Gibbons, Bradac, & Busch, 1992; Alarwell & chmitt, 1967b; Murdock, Bradac, & Bowcrs, 1984; Rolofl~ Paulson, & Vollbrecht, 1998) arc concerned with how language constructs meaning regarding power. In this slUdy, the power of a person (agent or target) will be identified by the potency value of the person 's status (here, simply a noun such as fanner or poet), either by itself (noncolllext) or in a sentence involving an innuencc aucmpl (in-context. This conccplUalizalion is consistent with the approach 10 the semantic analysis of power established by Gollob and Rossman (1973 and Heise (1970). The focus of the present study is not on the ba,es of power (French & Raven, 1959; Zelditch , 2000), but rathrr in how language is used 10 create meaning from information about agents, targets, and inAucnce attempts. j

The first part of this article reviews previous res,'arch on the effect of innuence tactics on perceptions of power. A review of work by Heise (1969a, 1970; Averett & Heise, 1987; Britt & Heise, 1992) and Gollob and Rossman (1973) on how power is judged from language is presented. Next, this article reports the methods used for studying the perception of power within simple senlences, followed by the results of this investigation. Finally, the implications of the research along with directions for future research on influence and perception of power arc discussed.

Power and Compliance Gaining

The perception or attribution of power can have important consequences (Kaplowitz, 1978). For example, those who are perceived as more powerful arc less often challenged (sec, e.g., Bachrach & Bara,,-, 1962). Those with more power "are more able to behave in ways consistelll with their identity" (Cast, 2003, p. 197). Indeed, the consequences of alllibuting power to another affects all illleractions with that olher, and is a criticaJ detenninant of task and relational interactions.

Kaplowitz (1978) prupuscu lhat if an agenl induces compliance, the degree to which observers think lhat the target is powerful depends on the agent's influence tanic. Following Gamson's (1968) reasoning, Kaplowitz proposed that abSOlvers would judge the relationship between an agent and a target to be cordial and to reflect a common interest when persuasion is used, and would be antagon.istic and to

reflect strongly opposed illlerests when lhrcaL' or punishments are used. Thus, when persuasion is used, the target is perceived 10 be responding to a reasonabll' request coming rrom a reasonable person. In contrast, when lhrcaLS arc used, targct compliance is more likely to be perceived as caused by the age ill'S power. This logic

SE~IANTICS OF SOC IAL INFLUENCE 297

demonstrates that the influence tactic used affects the power attIibuted to the target and the a!(cnt as well as perceptions of their relationship.

771reals II ersIIs Persuasion . f tlempts

There are a number of diHcrclll tactics by which an agent may wish to influence or affect a target's behavior. The two tactics compared here are threats and persuasion attempts. Agents threattll a target when they indicate that they will make the target's situation worse if the ta rget does not comply with thei r req uest. By comrast, agents persuade a target when they change the target 's behavior, attitudes, or heliefs, but do not propose to help or hurt the target direcuy as a consequence of that change or lack thereof.

These two tactics are contrasted because u,cy imply diametrically different relationships between actors. Persuasion works well if the agent is LIusted by the target (see, e.g., Hovland & Weiss, 195 1). In contrast, threats arc likely to be both a consequence and a cause of a negalive, uncooperative relationship bet""een two actors (see Gamson, 1968; Gaski , 1986; Shapiro & Bics, 1994) 'L~ one actor attempts to establish dominance ovcr another (Pruitt, 19B I ; Putnam & Poole, 1987). The effectiveness of threats is associated with the size of the threatened punishment, the perceived likelihood that the threat will be enforced , and u,e magnitude of the resources that enable the threat to be enforced (sec Lewicki, Barry, Saunders, & Minton , 2003; Pruitt & Camcvale, 1993). Finally, as Kelm",' (1958, 1961 ) has noted, the process of change associated with the presence of sanctions, which he calls "compliance," differs from the process associated with the incorporation of new information with one's prior beliefs (UimCJ11alizaoon") and that associated \vilh fulfilling a relationship with anou,er person (,'identification"). Threats fall into the first category; ucotllplianrc" requires the agent's sUJvciliance of the target, whereas identification and inlCI11alization do nolo

Among a variety of power strategies, threats and persuasion attempts have been found to he opposites. In Falbo's (1977) study, for example, they were polar opposites on a two-dimensional power-strategy space defined by direct- indirect and rational irrational dimensions (sec also Cody, McLaughlin , & J ordan, 1980; Man"ell & Schmitt, I 967b). The use of these two tactics makes it possible to determine how much the perceived goodness and power of agents and targets depend on the tactic employed.

To suggest that threats and persuasion attempts diOe r for actors docs not indicate how observers make sense of them. To understand and model this process, Heise's (1969a, 1970) approach to the semantics of simple sentences is examined.

Agent, Act, and Target in a Simple Sentence

One approach to measuring U'C perception of power and the vaIiables U,at conll;butc to it is Heise's (1969a., 1970) technique of having responcicllL,) give ,oa tings to components of a simple sentence (subject, verb, and object), and then examin ing how these ratings may be algebraically combined. These component ratings, made outside of a sentence lhal involves an influence altempl, will be lermed non-colltt_\t asstJsmenls.

According to Osgood, Suci, and Tannenbaum (1957), the meaning of a word may be caplun:d by the ward's locaLion on lhree dimensions: eva luation (i.e. ) good vs. bad), potency (i.e., s!rong vs. weak), and activity (i.e ., active vs. passive). Heise (196.'» determined me semanLic-differential scores on each of the three dimensions for

298 CO~I~leNICA'IION ~IO"O(;RAPIIS

I ,000 of the most frequently used English words. Extending Osgood Cl al. 's work, I-Jeist' (1969a, 1970) proposed that if the non-context location of a sentence's subject, verh, and object on the three scmalllic-difTerenlial dimensions is known, the words' f('viseci assessments on the semantic-differential dimensions due to their appearance in a panicular sentence may be dt..'tclmincd. In olher words, the non-context assessments determine the in-context assessments. A1lhough Heise (1969a) was successful in creating regression models that prcdirted the in-context subject, verb, and ol~ject evaluation and activity ralings. his models that predicted the in-context potC'nry ratings Weft' unsuccessful. He concluded lhm "poLency dynamics may he exceptionall} complicated" (I" 212). In a follow-up study Heise (1970) successfully created a predictive model for pOlency dynamics) but this study had methodological difficultit's tha t will be addressed subsequently.

Heise's approach to studying the interpretation of words in a sentence was used by GolJob and Rossman (1973) to determine the perceived power of an agent. Gollob and Rossman predicted participants' perceptions of the power of an agelll from lhe nOll-context cvaluauon and pOlene), orthe agent's bt..'hm·;or (i.e., the sentence's verb) and the non-context evaluation and potency of the target. They did not use the agcllt's non-context potency ratings in their stati"ical model. This study also had some methodological problems, discussed subsequently.

Extensive research indicates that the perceplion of the influence tactic an agent uses tells a great deal aboul the agent, the larget, and their relationship (Gibbons et aI. , 1992; !-l inkin & Schriesheim, 1990; Schlenker & Schlenker, 1974). The influence lactic that an agent uscs (e.g., threat vs. persuasion allcmpls) should have an impact on the perceived potency of the target or influence. In addition, the potency of the influence tactic can vary depending on who (agcllt) is using thc tactic on whom (target). Thi, study blulcis on previous research by examining the potency and evaluation or the agent, the act, and the targeL, and investigat..ing how they arreCl perceived power of an agent and target. 111 addition, the sllIdy examines how the influence tactic and the characteristics or the agellt and target affect the perceived probabilily of compliance 10 an agent.

The probability of compliance with all influenee lactic is a primary interest in sociaJ influence situations, and communication scholars have examined how Stich compliance is brought about (e.g., sec Dillard, 1990). This study's goal is to prQ\;de parsimonious models that can predict ho\\ language creates perceived power and perceived compliance-gaining success for observers or influence attempts.

Methodological Isslier

A major contribution of Heise's (e.g., I 969a) research is the usc of equations with the sentence as the unit of analysis. In Il rise's research thc variable, independent and dependent, are the semantic-differential ratings (evaluation, potency, and activity) or the subject, verb, and objt~ct ill simple sentences. Heise's stl"aLCgy is cmployed here. The main cxpcrimrnt 31lalyzrs how meaning is created in scntrnccs, using data for the terms in each sentence as vatiablcs.

Heise (J 969a, 1970) developed separate equations to predict the e,"aluation, potency, and activity of a senlclln"s subject li'olll the corresponding non-fOlllext ratings of the sentente's subject, verb, and object. Only in later research (Averell & Heise, 1987; Brill & Heise, 1992; Gollob & Rossman, 1973) were cross-ratings used (c.g., using the evaluation of a subject to predict the potency of the subject).

To reduce or avoid multicollinearity among predictors and to csumatc the

SE~IANIlCS OF SOCIAl. I'IFI.L'I·::-'CE 299

parameters for the hypothetical lexical universe, \\ orels must be sampled systemati­cally. Specifically, presuming that there arc three meaningful semamic-differential dimensions. tht· sample of words should reflect each of the eight octants of meaning (i.e., as in a 2 X 2 X 2 design or high vs. low c\'aluation, high vs. low potency, and high vs. 10\\ activily), and orthogonally comhine such words into subject- verb ob­ject sentences. Failing to do so resulls in ro('ffici{"nL~ whose slandard ('rrors reAecl the accidental sampling (sec Kahneman, 1963). This problem is round in Heise (1969a) when he reports that:

the desired sampling d('si~l for potency 0;('111('11('('); fililt-d to matt'rializ(' ... . Thus, the potency elma .,. 'In,o' nOI I'('prcscntati\'c of 1.11(' lOla I possibl{' rdnge or M'Il\l'Il(,('S. (p. 210)

Heise (1970) encountered similar problems. \ 'erbs that were low in potency tended to be ncgalivcly evaluated. In addition, although "a \IIord's connotation was the pl;mc considcraLion for lIsing it ... when a choice was possible, words were selected that also seemed neutral on the evaluation and activity dimensions" (p. 49).

The same issue appeared in Gollob and Rossman (1973). They reponed that "nearly all or the bad verbs and objects were rated on the weak end or the potency scail', and all of'the good verbs and objects were rated on th,' powerrul end or the scale" (I'. 396). I n addition to this problem, Gollob and Rossman used a depcndent vadablc that onflatcd power and influence. asking rrspondcllls to rate the agent' "power and ability to influence others" (p. 395) b) a single scale item. The ability to influrnce others, ho\\'ever, may connote a persuasive strategy, whereas power is more likely LO connOle innucncc aCliviLit's lh~ll could involve threat. These t\VO problems make Gollob and Rossman's result.s problemat ic. This investigation attempts Lo overcome these difficulLics.

Research O1".llio/l alld HY/JolheseJ

AIMllillg qf Ihrent and /JtrJUasio!1 altem/J(j, Threats and pc:rslIa'iion aU(,lllpLS differ in several ways. For example, threatenillg is perceived as a more ncgmivc communi­cation strmegy than is a persua~jon aucmpl. Tht'l"(' appl'ars, hovv('vcr, to bc no study lhal has assl'sscd lhe range or meaning cliOcrcncl's bCl\\CCIl lhreatcning and atlempling to pt.,.,uade. Thererore, the rollowing research question is posed: How do threats and atlempt~ to persuade differ in Illeaning as assessed by the semanuc-differ­ential dimensions?

I'olel/ry '!! Ihe agel/lalld largel. There are se\-eral ways that allributions may be made about the 1""\C'r firth" agcnt and targcl. First, agency suggests power (Britt & Heise, 1992; I leise, 1970; see also mith-Lovin, 19B7). Thererore, a person engaging in an influence attempt should be perceived as more powerrul than the target or such an atlcmpt.

Second, people may auribulc more power lO one \vho thrcmens, 'T'hreats arc counllTnormativl' (rdanvcll & Schmiu, 19671», so an obsen'l'r may assume that ir threats arc employed, the agent must bl' sufficiently PO\"errui to lise such a Lechnique. Threats also convey "indincrcllec to Ihe target's face needs, which suggests !,'J'l'at power" (Kaplo\vitz et aI., 199B, p. 108).' Finally, threats arc morc 1'0\\ cdid acLs than persuasion attempt!;, and Smith-Lcwin 1987) round that an individual ('n~aging in pm>\'crful acts is perc('i\'('d to be 1110rc powcrruJ because or doing so.

Third, an agclll should atlcmpt LO persuadr powrrful largels rather than LhreatC'n

300 CO~I~IUNICA rlO;-< ~IONOGIt.\PIIS

them. All observer may assume that the agent chooses to persuade rather than threaten because of the possibility of the target's retaliation if the agelll were to threaten. The ability to retaliate suggests that the target is powerful (Michener et a1. , 1973). On the other hand, an observer could also assume that persuasion is used on a target because the target is reasonable and responsive to argument. If some observers make the first anriblllion (i.e. , persuasion is used because of the target's power), and some make the second (i.e., persuasion is used because the target is reasonable), on thc average targets who arc subject to persua~ion will be viewed as more powe,ful than targets subject to threat.

In sum, the following hypotheses on the perceived power of the agent and the target are proposed:

H)'P0thts/.s I. The jlgent of an influence aucmpt is perceived to be morc powerful than tlw target. l-I]polllw.J 2. {a '111c agent is perceived to b(' more powerful when using threats than when using

pCr'5uasion, but b', the target is pcrct'ivcd lO be mort' powerful when the subject of a persuasion anempt than when threatened.

Evallllliion 'If Ih, age,,1 and larget. Gollob (1968) found that agents arc perceived by participallls to be more powerful when they do good things to a good target or do harm to a bad target. Th is consistency efTect was supported by Gollob and Rossman 's (1973) finding of a significant interaction eflect of verb evaluation by object evaluation on perceived agent power.

The just-world hypothesis (Lerner, 1980) provides an explanation for these findings. Lerner and " Iiller (1978) proposed that "individuals have a need to believe that they live in a world where people generally get what they deserve" (p. 1030). In other words, good people arc viewed as deserving of, and therefore likely to receive, just rewards (c.g., power), and bad people are viewed as undesenong of, and likely to be denied, lhese rewards. Because threats arc negative as compared to persuasion attempts (1\larw("11 & chmitt, 1967b), an agent who uses a threat against a bad person is acting justly, and thcrefilre should be ,,;ewed as powerful. Because persuasion attempts arc positive as compared to threats (Marwell & Schmitt, 1967b), an agent who attempts to persuade a good person is also acting justly, and therefore should be viewed as powerful. The other two cases (using threats against a good person or persuasion against a bad person) are unjust, and therefore an agelll who docs either should be viewed as Jess good and therefore less powerful (see Thorndike's research on tl,C halo eflect, 1920; sec also Gollob & Rossman, 1973).

Lord Acton 's statement, in a leller to Bishop Mandell Creighton, that "power tends to corrupt and absolute power corrupts ahsolutely" (1887, quoted in Bartlett, 1992, p. 52 1) contrasts with the just-world hypothesis. It suggests that power and goodness are perceived to be negatively related. Research has suppa ned the idea that, by some measures, power indeed docs cormpt (sec Kipnis, 1972; cf. Overbeck & Park, 200 I), although there does not appear to be any research that has looked specifically at the evaluative consequences of power,

These ideas are tested with the following hypotheses on the efTen of tactic, target (,valuation, and agenl evaluation on perceived power:

It)'potJusis 3 .. \n agcnt is perccin'd as more powerful "hen aHcmpling to per~uadc a morf' p()~iti\'dy c\'aluatcd largct or lIsing mrc31s on it more llt'gati\'c1y nalIJated largft thall ",hell IhreaLCning a positively {'valuated targel or attcmpting to peNuade a Ilcgalhdy (,\'aluated om',

f(),polltfJiiJ 4. ~ l ort' positively ('valuated age-illS are perceived lO bc 1110re powc:rhd than I110re ncg.ui\'cl)' evaluatcd agt'l1lS.

SE~Ii\NTICS OF SOC!.\I. INFI.UENCE 301

f(Y/JotJltsis 5. ~Iorc posi tively cvalu~ltcd targeLS arc perceived as more pt)w('rful lhan more ncgauvcly evaluated rarg('lS.

Perceived probability 'If com/lliance. There is very little research on how observers perceive a target 'S probability of compliance \\~dl an agent's influence allcmpt: and much of the existing research is not helpful for the formulation of this study (sec, c.g., Murdock e l ai., 1984; Riccillo & Trenholm, 1983). Because !.he choice of compli­ance-gaining Lactic affeels compliance (e.g., Kaplowilz el ai., 1998, found Lhal LhrCaLS arc perceived as less likely LO g-ain compliance Lhan persuasion; see also Bosler el ai. , 1999; Levine & BOSLer, 200 I), il should also affecl an obselver's perception Lhal Lhe largel will comply. According to reaClance tI,eory, a message that lhrealt:lIs oue's freedom r("sulls in less compliance man a nont.hreatening message (Brehm, 1966). Funhermore, observers undersland lhreaL dynamics (Kane, Joseph, & T edeschi, 1977), and reporl that threats arc a negative compliance-gain­ing melhod (Marwcll & Schmill, 1967b). Thus, observers of threats should, like lhe largeL, anticipate that these lhreats reduce compliance. Therefore:

fljpolhesiJ 6. Threats arc perceived as less likely to gain compliance than persuasion O'lllCmpLS.

Kaplowitz et al. (1998) and Michener el al. (1973) found !.hal the grealer ti,e perceived power of the agent, !.he more observers assume lhat the target will comply with the agent. This finding was retested.

HY/Jol1ims 7. Gremer perceived power of the age-lit resul ts in grealer perceived probability of compliance.

Similarly, sU'onger largelS should be less likely to be perceived as compliant.

f!J'Pol1itJis 8. Greater perceived power of rhe target results in lower perceived probability of compliance.

The evalumion of the agelll and the Larget should also affecl the perceived probabililY of complying. Targels may be mOLivalcd lo respond positively LO agents who arc good, and are Lherefore assumed lo have good intcnlions and perhaps inclined to do good things LO or for olhers. Therefore, such agents should be perceived as morc successful in gaining compliance.

I J;·pollitsis 9. A more posiuve evaluation of the agent results in grealer perceived probability of compliance.

Similarly, targets who are good arc c"pceLed to be more cooperaLive wilh olhers, allempting lo work with ratl,er Lhan againSl ano!.her·s requesl (Thomas, 1976). Therefore, good largelS may be perceived as more likely to comply.

'bpuf/tlSis 10. A 1110rc positive e"aluation of lhe tOlrg-t'l resuhs in the target's greater pcrccin:d probahility or compliance.

A weak version of Gollob's (1968) ronsistency effect (thal age illS are perceived as marc powerful when Lh y do good things LO a good target or do hann to a bad Larget) is thaL good agelll's should be viewed generally as more powerful when their targel is good rmher Ihan bad. Combining lhis idea witll H ypOlhesis 7 (lhe idea lhal mOre powerf'ul agents should be perceived ,L' more likely to succeed in gaining the compliance of another) should result in morc positively ("valuated agents being perceived as more ~ l1ccess rul at gaining compliance when LilC' target is positively cvaluaLCd.

If)POlheJ1S II. ~lorc positively ('\'aiu<lt('d agen ts art' pc·n.:ei\'cd to bc more ukcly to gain compliance

302 COM~ IUNICATION ~IONOGRAI'IIS

\\ ht"11 altempling to inOuence a more pO'iitivdy ('valuatcd target as compared to a more negatively evaluated targel.

Influcnce tactics should also affect the perceived likelihood or compliance. Persuasion attempts to create or preserve a good relationship between agent and tal·get, whereas threats damage such relationships (eai, Wilson, & Drake, 2000). Highly evaluated agents should be associated with inAuence tactics that create or preserve good relationships rather than with those that damage their relationships. Thcrcrore, good agents should be perceived as more likely to seek compljance by llsing a tactic that sustains the good will between the agent and target] and persuasion is such a tactic whereas threat is nolo Furthermore, slralc!:,ries that arc likely 1.0 be ,electnl arc also likely to be viewed as successfUl (sec Meyer, 1990, pp. 64 66; cf O 'K eere, 1990). H ence, good agents should be perceived as more likely to gain compliance by using persuasion rather than threats.

'(}polhlHS 12. ~Iorl' posi tively c\'aJualt'd agenL'l are prrceivcd a'l more likely to gain ('ompliance when using persuasion than \\ hen llsing threats.

The rationalcs underlying Hypotheses II and 12 combine 10 predict greater expected compliance when a positively ('valuated agent uscs positive tactics with positively evaluated targets. In addition, these rationales also justify the parallel prediction that a negatively evaluated agent will bl" perceived 10 be more successrul when lIsing neg-ative lactics with negatively evaluated largets. 'T'hcrcrorc:

'bPQtlteJlJ 13. PCl"Suasion is pcrccivcd LO be mort" likd )' to gain compliance berw("('n morr positiH'ly evaluated agents and I.argf'ls, wlwn·"L~ threats are perceived to be more likdy to gain compliance h('tw('cn more nrgalivrl) ('\'aluatcd agents and targcts. Persuasion attempts involving agents, targcLS, ~U1d taclics that diner in evaluation arc prrccivcd as less likely to gain compliance.

Method

Pilol Stllt!y

Ovm'iew. The purpose or the pilot swdy was (I) to dctermine which person nouns 10 usc in the final investigation , and (2) to establish the non-context ratings or the terms (the person nouns and U1C verbs tried to persuade and threatened) to be used in the final inv(,stigation. The non-context asscssm('nL" wcre used as independent variables in the main expeliment; they were assessed independently or the in-conlext ratings, which were dClcnnincd in the rnain experirnent.

Participant,. Participants in the pilot study were 1·15 students '48 males, 80 remales, 17 gender unspecified). They \\ere voluntcers rrom a variety or undergraduate rumrnullicaLion courses. Course instructors asked students 10 partic ipate in exchange for a small arnount or extra credit, the particip3llls' inslrllclOrs not being involved with the aClual dala culleClion. All aiternmin' class assignment ,·vas ofi'ered as an option lor earning extra credit rol' those who did not 'want to panicipalc in the frsearch project.

Selectioll qf stiJl/uli. Person nouns were seJected initially based on Heise's (1965) semantic-differential dictionary. I leise's list included 45 person nouns. Tht research goal Wl:L') to ha\c the influence agents and targets cover the uni\'erse or person types. Therefore, person nouns were seJectcd that rcprcsl·nted the right combina­tions or high and low (",·aluation (E), a(·tivity VI,. and potency (/~. Some person nouns that were.: not on Heise's li st wcre added because it did not provide

~E~ IANTI CS OF SOC IAl. INFI.UI·; CE 303

person nouns for all eight ceUs. Colleagues were asked to nominate person nouns f'or the empty cells. Because the role of threat versus attempted persuasion was to be assessed, and because the perceived success of the influence allempt was to be measured rather than manipulated, the verb, Ilmalened and lried 10 persuade were also included in the pilot study question nair .' The final list of terms to be rated (and their placement on E, A, and P from lI eise's reported ratings or from the expectations based on colleagues' judgments) was the following (terms not from Heise's list arc underlined):fimner, menlor, sailor (hi E, hi A, hi P);fellow (hi E, 10 A, hi P); rhild, molher, .rympalhizer (hi E, hi If, 10 P); /)Oel, wifr (hi E, 10 A, 10 P); elletlry,jlldge, racisl (10 E, hi If, hi P); lll/tIlall (10 E, hi A, 10 P); loser (10 E, 10 If, 10 I . (No person nouns were available from J-Icise or determined from colleagues' nominations for the 10 E, 10 A, hi P cell.)

/\ lellsurts. The pilot qUc'.'slionnairc asked partkipams to TaLC the 16 terms on six scales ITIcasUling the evaluation, activity, and potency of each terln . Each seman Lic­dirrcrcnlial dimension was assessed by two scales. 'The quC'stionnaire began as follows:

Researrhcrs in the Dcpanmclll of Communication are studying how people make senSe of the "ords thai thc) usc. Bdow is a list of commonly us{'d words. \ \'(. would like you to rate each \,ord on a scries of M.:al(:s that will help us LO learn hcm you personally lInd('l'~lalid each word. There arc no right or wrong answers; "e arc il11ercsH'd in your vit·ws. Thallk you for your help.

The tern,s were rated on the following scait's (and the relevant semantic-differen­tial facLOr for each): strong weak (potenc) ), lively -sti ll (activity), tough tender (potency), active passive (acti'ity), plt-asant unpleasant \naluation), and good bad (evaluation). The terllls were placed in neutral sentences such as " He is a sympathizcr." A1I sentences were structured to refer to mcn (e.g., "it is his wife"), Pilrticipant" rated Ihe t<TIllS on I 7 Likert-type scales.

To control for the effect or term order, sca\(' order, and the polarity of the scales, two forms of the questionnaire were crrated. Onc version had the terms in the order listed previously, with the poles of the scales in lhe order listed previously, with lhe scales in the order listcd previously, whereas the other had term order, scale order, and scale polarit), revcrsed (e.g., the first questioll in version 1 asked for six ratings for child, the first one on the strong weak scalc, with strong = I; the first question in version 2 iu;kcd for SLX ratings for wife, the first rating on the good bad scalr, \\ ith good = 7).

The qU(''ilionnaire also asked participants to report their gender, course and section number} and date and time of questionnaire completion.

Procedure.1. ParticipanLS were approachcd in class and asked to participate in the r(,search by completing a shon questionnaire in exchange for extra credit. The questiollnaire W(.\I\ completed in class. Th' participants w('re instructed to read and sign the consent form attached to the front of the questionnaire ancl detach il from the qu('stionnain', then to complete th(' qu('stionnair('. The lwO versions of the qucstionnairc wcre distributed randomly. Participant completion of the question­naires LOok approximately 10 minutes.

I l llai)'Jir alld mult,. The scale means for each tcrm on each of the six scales were first subjcctt'd to an unrOlated principal cOmpOTlCllts aJlalysis to examine the dimensional strufturc of the scales. All possible factors were extracled. Only two dimensions were lound to have eigcllvalues l{I"('atrr than 1.00, LOgclher accollnting felf ahollt 91 % of Ihe variance in the- terrns. The first, largest dimension is evaluation,

30·1 CO~t~tUNICi\TtON ~tONOGRAJ'HS

accounting for 52% of the valiance among the 16 non-context lcmlS. The second dimension is a combination of potency and activity, which accoul1ls for 39% of the \Oariancc. This finding of two significant dimensions is contral), to Osgood et al. '5 (1957) general finding of three orthogonal dimensions for the semantic-differential scales.b

Because there were only t\\fO significant dimensions, the eight hypothesized scmanLic-di(fcrclllial octants for the location of each term were now reduced LO four quadrants. The sampling strategy now entailed sampling terms (Le., person nouns) from each quadrant The results of the factor analysis for the pilot study aflect only the sampling strategy for electing the person nouns from the universe or potential person nouns. Tn the analyses ror the main ('xp('limclll, power is measured by a single-item potency scalc (strong-weak), either from this pilot study (non-coIllext measure) or from the subsequent data collection for the main eXperlnlcnt (in-context measure).

The lcnns molher, menlor, janner, sailor, and judge \,,'ere found to be in the hi evaluauon hi potency-activity quadrant of the two-dimensional space; the terms sympathizer, poet, child, hUlIlall, and wifi wcrr found to be in the hi evaluation 10 pOlency-activity quadralll; the terms loser and fellow were found to he in the 10 evaluation 10 potency-activity quadrant; and racist and enemy were found LO be in the 10 evaluation hi potency-activity quadrant.

_ ext, the effect of gender and form on the perception of each term was examined. Terms were sought whose meaning was 110l found LO be affected by participant gender, questionnaire version, or the inlcranion of these two valiables. Therefore, analyses of variance were performed on the two factor scores (evalua tion and potency-activity) generated from the p,incipal-components analysis for each of the 16 terms (the 14 nouns plus threatened and trid /0 persuade). The ANOVA de ign was 2 (Gender: male vs. female) X 2 (Form : the first ordered form vs. the revcrse-ordered fOl1ll). Tahle I repons the results of these ANOVAs. Terms whose evaluation and potency scores were not significan tly affected by the independclll variables (i.e., in terms of statistical significance) were candidates for usC' in the main experiment. 7

In addition to these analyse of variance, the reliability of the scales for each factor (evaluation and potency-activity) was assessed using Cronbach's oc For evaluation, two items (good bad and pleasant unpleasant) composed the index whose reliability was assessed; for potency-activity, fOllr items composed the index (stron weak, tough tender, lively still , active passive). The results of the reliability analyses also appear in Table I.

Based on the ANOVA results, ti,e terms child, illlmall,judge, mother, sailor, and wifi were eliminated for use in the main experiment. The term fellow was eliminated because its evaluation score was approximately zero. 'Thjs elimination meant thal lost'r was the only term remaining in the 10 evaluation 10 pOlency-activit)1 quadranl1 so il was retained even though th ere we're' statisti(,a lly significant C'fff'rH of form and Cencier X Form in the relevant Al"lOVA"

Alaill F;'/Jerimellt

ParliripalllJ. Participants were 189 studenL~ (57 males, 126 females, 6 gcnder unspecified). As with the pilot study, they were volunteers from undergraduate communication courses and were olTered the same incentives as participanlS in lhe pilot study.

SE~IAN'IlCS O F SOC I.\1. 1 j'\ 1· I.L · E~CE

IAUU. I ['III E.I H .C! m GL.'\,iOERv\O FOR.\l 0'. A'\1l nn. RUJABIIJIY CIlO .... U"CIi"'i\ Of, E\'AU',\110\ A. .... U

POIt~'C\ •. ..\(: II\,II" ~()R III n R~I 'i 1~1.(lI SIt -m , '" = 12h)

I·OlenC)".Aclh;ty EValUilt.iOIl

Si""nific-.lnt Significiult Effel.:lS Crlmh.ll'h's E ffi,(tll Cronb.ll"h 's

T t'nn p <.05 • IjI <.OS , X

C h ild Nmlt' m F0011 1jJ < .001 .B89 Eucmy NOlle .767 NolU~ .892 Fannl' r NOll!" 835 None .872 FI"IIO\\ '011(' 7BR "on£' 885 Ilum.1Il Form (p < .002 ] .810 None .871

Judgr None ,611 (;C11Cifr p < .001 .755 Form lP < .00 I I

L{J'IC'r ~onc .fin FornI JJ <.OOI ) .88'1 Gender X Fllml (P < JJ:l8)

M t' O(UT Nunl' ,Ii!)l i\'oll .... 77\ ~ l lHhn Form 1ft < .00 I .736 NOIlt" .958 ('ned to Aon(' ,7Q~ I\'on(' 746

pt-rsUiU.1t-POt" :\'0111' .808 1\0l1e .845 R.lCist ~one 75 1 1\"011t' .868 Sailor ;'\Jollr .852 Grncif'T P < .027 .810 S}1up.ll hi/(·r \;olle .7 61 l\OIlt' .845 1111"('.111'11(' <1 ;'IJonc .m)') l\mlt' .735 \Vili, Foon 1,jJ < .001 .7-1 7 Nom' .883

10.5

i\1ea.\IIres !llId de.ligTI. Based on the resul L' li'om Ihe pilo t study, four questionnaire versions were constructed . The four versions differed only in the focal te rms (i.e., the person nouns and verbs) that were lIsed , The terms used were as (o llows: enC111)"

f armer, /osrr, mentor, poel, racisl, sympalhi<.er, I,ied 10 persuade, and Ihrealelled. The unit of analysis for the final study is Ihe sentence. Each sentence was

constructed by using one of Lhe (seven) pel>on nouns as a sentence subject (the agrnl. of influence). one of the ( ix) remaining person nouns as the object (the ta rget of influence), and using one of the (two) ve rbs (threatened or lried to persuade). Thus, 84 srlll cnccs we re genera ted .

Each of the four questionnaire versions contained 2 1 of the 84 semences. Each participant rcc(' ived one of the four versions. The 84 sentenccs wcre placed in random ord(' r across the four versions of the questionnaire, Therefore, the order of scmcnces w1 thin each version was random, and each sentence appears in only one questionnaire . The versions of the questionnai re \\ ere distIibulrd to pa rticipants in random orell'r. All versions of the questionnaire began w1th the following in s lruc~ tions:

R<.":wa rr hrfS in lh{' 1)\' PiUIIl1CO( or Communicat iun a rc sludying how people mak t' SC Il!it' or lhe wOl'ds tha t Ilwy IIS('. Bdow is II li st of srlltenn's contain ing commonly lIsed words. \ Vc would like you t t) rate' carh und l' rlined word o n each of Li lt, th rn sca lt,s f()lllld above;' lht: list u r scntcnc('s. You sh(')uld raw earh word based 0 11 hm\ you unc!rrslil lld wha t it I11 cans within that particular sentence. Write YOll r raling of each word in the hox next to thc (o rrc~p(Jndi ng leue r found belo\\ (' ilr h word. There a rc no right or wrong aJ1 ~'i\\('I'"s ; '\t' an: illltTt'sted in your \'i(~ws. Thank YOll for youI' hdp,

••• For ('i.\lh of t il{' following selllences, (11' to imagine thc intcraction between A and C . Think of A

306 CO~ I~I Ui\"IC.\TI ON ~IO'lO(;RAPII~

and C as rcal peoplr and thrll assess each or iiI(' lhn'c words A, B , and C according to the sralco;; to the righl of the SClllenccs,

Think of each scnlrnc(" as in\'ol\;n ,~ a new pair or people.

Prior to rating each term, panicipants were asked to rate the probability of success (on a 0 100% scale) or lhe influence attempt. For example:

Tho sympathizer (A) threatened (B) the fanner (e) . Whal is lhe probability Ihal A gOl whal A walllNi from C?

Next, parLicipanu~ werc asked to rate the su l~iect (agent), verb, and object (target) of each sentence on three single-i tcm scales: potency (strong weak), aCtivity (anive passive), and evaluation (good had). The firsl-listed anchor in each scale (e.g., strong) was assigned a core or 7, and its corresponding Opposile anchor (e.g., weak) was assigned a SCOre or I .

Procedures. [n ciasses, participants were asked to participate in the research by completing the questionnaire in exchange ror extra credit. As a group in lheir classrooms, participants were instructed to read and sign the consent foml attached to the front of' the questionnaire and detach it from the queslionnaire, then to compiele the queStionnaire. Each respondent's data colieClion LOok approximately 30 minutes.

Alln!~.ri.\. For each semC'nce lhe means or Ihe following variahles were computed based on the responses to th at sentenCf: evalualion of subject, objcct, and verh; potency of subject, object, and verb; anivily of subject, object, and verb; and perceived prohahility of success or the influence attempt. Each of lhese means is based on between 43 and 50 respondents. The unit of analysis in the main experiment is lhe sClllcnce. Because lhe sC'ntences employ the means for each term over participants, the reliability or the variables is expected to be vel)' high.

In the analyses that follow a non-context assessmelll refers to ratings made in the pilot study, in which each term was rated in a simple sentence not invol"ing an influence attempt (e.g., "He is a farmer"). In the ana lyses in which the noncontexl ratiJlgs arc employed as independent variabl(,s, the relevant single-item scales from lhe pilot study are used because these measures correspond to what is used in the main experiment. Therefore, to assess evalualion, the good bad scale is used, and to assess potency, the slrong--wcak scale is used. Continuous variables arc nOt treated as categorical variables in tht: analyses of covariance and t-tesls.9 All reponed I-tcsls are two tailed.

RcsulLS

A/falling qf 77IYealen alld Allem/Ji to Pen-Ill/de

To examille the meaning of Ihrea/flIed and tried 10 persuade, the data from the pilot siudy were "mployed initially (the nOn-cont,'Xl dala; non-missing .N = 143). The scores or (l) Ihe good bad sca lc alld Ihe pl'"sani unpleasant scale wcre averaged to create an evaluative score as a dependent variable. (2) the slrong weak scale and the tough tender scale were 3\'craged to create;' a potency score as a dependent variable, and (3) the lively-stili scale and the aeti,"c passivc scale were a"craged to create an dcLivity score as a dependent variable, Paired-comparison I-tests w('re employcd. Table 2 shows lhal trying 10 persuade was judged to be beller (i.e., higher on the e\'aluative score), weaker (i.e., lower on the pOlency score), and less active (on

'E~I.\"'TIC:S OF 50(.1,\1. I~FLLE:-':C:E 'jill

I.\IU .E 2 Mu'\ S 1 "IHRJ) 01 \"I HIO' I~\ \I.l ,HIO' . POI L'O-, "'I) ,\(.II\"II'Y &;OR~_" fllR I IlRHI ~S~.I) ,,-Xil TR iLl) 10

J1Utsl \[)I., I'II ,1l1 Sit m , \I = 143

Dinwl1~ion

b.IIII.1Iion Pmenn A( li\il)

Inflllt' llt {' 1.ltli(

I n 11 ,(11

I.7h L!I .;.'iI I 18

1.97 U)H , b,()2 1.21 :'.76 I.:to

\ /l fl'. Rilll.1{1:' Ill" dilllt' lhi"lI ratings: I 7, wherc 7 hi,\h ,-.liLw

the aeti,ory store) than threatening, These three dinlTences were all statistically significant and two W(Tr substantial: evalu ation, ~ 14 2) = 18.25, P < .00 I, 1/ ' = .70; potency, «142) = -9.27, 11<.00 1,//'=. 38; .eti, ·iry, tI42 )= -3.47, /1 .;;.00 1, 1/' ) = .08. I'The resuits wcre essentially the saml' when the in-context ralings from lht:" main experiment were lIsed to in\'esliga lc this research question. )

POII'IIC)' 'if the Agel/t alld Target

Non-context potency sco res ranged li'olll 2.90 (for loser) to 5.84 (for mel/lor; .II = 4.44). H)1)othcsis I stat es that an agent is perceived to he Illore powerful than a larget. Hypothesis 2 Su'llCS that an agent is perceived to be more powerful when using thrcats than usi ng when using perslIa!iion, but a target is perceived to be morc powerful when the su bject of a pcrsUl.tl.; ion allcmpl than when threatened. To test these h)1)otheses, a multivariate a nalysis of \"ariancc was conducted, with potency o f the agent yel""Sus potency of the target as a within-subject (i.e., sentence) va tiablc. and vnb (threllte/led \'5. tried to IlenulIde) as a between-suhjects variable. A slalistically significalll e ffec t, I'l l , 82) = 5.42, /1 .;; .02, partial 1/ ' = .062, was found for aclor agelll vs. ta rget). with agents being more potent ,1/ = 4.56) than ta rgets (AI = 4.23).

Thus, Hypothesis I was supporled. Hypothesis 2 predicts an interaction between actor and influence tactic. This

interaction was no t significant, so Hy p()dlC~ is 2 waloi nOt supported"

Ella/llalion 'if Ihe !lgent alld T!IIgel

Non-context (,valu.tion scores ranged from 1.24 (fo r rtlcilt) to 6.35 (for mentor; ,1/ =3.92). H)l)othesis 3 predicts tha t a n a~cnt is perceived as more powerful when a ttempting to persliude a more positively ('valuated target or using threats on a more ncgaLively cval u,lted ta rget than whell thrcatening a posi tively evaluated larget or a ttempting to persuade a nega tively evaluated one. H)l)othesis 4 predicLIi that morc posiliycly evaluated agents arc pcrrci\·cd to be morc powcrful than morc negatively evaluated ag('nL~. To tel-a Hypolheses 3 and 4, an analysis of co\"ariance was conducted with the agent's in-context pOlC'nc), score as the dependent variable, and with the following independent va liables: the non-context evaluation score of' the agent, the non-context c\'a1uation scorc of the target, the inte raction of verb ('hrealened \'s. fried 10 pernwde) and non-romext agent evaluation scorr , and the interaction of verb and thr non-context target evaluation.

The verb by target imeraction \\'a~ tlot a slatistically significant predictor of agent's in-Colllext potency. Therefore, H)1Jothesis 3 was not supported.

Thl' non-context eva.luation sCOr(' of the agent was a staLiSlicaUy significant

308 CO:-'IJ\IUNICATION ~IO:-lOGRi\PIIS

TABI.Ii 1 .\ltA.'J S ·"·\.,\I).\RO I)~XIJ\ 11IJ:\ P~_R(H\ Ell PKOK,\J\IUT' Of. c.:O'\II'I.J;\;'~n., u, bn.U :"'( F. 1\(:l1C, A C.L'\ I POI t .. 'I;C;, .

,"'n T\R(;.:-r J'OJ rMT '.'f,l.I"'l E,\.N.RI\U.' I = 189; "DI8t.R m SL'\TE., ... n .. \ = 84 )

Influcnc(' Tactic

POlrtlc) of A~('1'Il

POU'IIC) of I'arget L"m High

Tried 10 P("l"iuadr

Low

50.53 (11.22 35.20 (13.62.

High

5:UH 11.Yi l 11.6-111.0·1)

J\ oll. R.lll':<' of p('ru'jvcd prohahility or tOlnptiancc: 0 lOoa o.

rhrealcncd

Lc)\\

42.R9 (6.91 ) 28.96 \5.00,

56.86 \6.98; '16.8'1 (8.% ,

predictor of the in-context pOlCncy score or the agcnt, f( I, 79) = 8.96, P '" .004·, partial I,' = . I 0, but the effcci was opposite to Hypothesis 4's prediction. More negatively evaluated agents were perceived to be more powerful (M = 4.86) than more positively evaluated agents (Al = 4.33, evaluation of agents is here subjected to a median split for comparison, not for the A"\lCOVA).

Hypothesis 5 predicts that more positively evaluated targets are perceived as more powerful than more negati\'e1y evaluated largets. To lesl this hypothesis an analysis of covariance was conducted \\·ilh the in-conlext potency score of the target as the dcpendent " .. iable, and wilh the foUowing independent variables: non-conlexi evalualioll score of Ihe targel, verb (I"realmed vs, Iried 10 persuade), and Ihe interaction of verb and non-context (,valuation of the target.

A statistically significam effeci of non-context targel evaluation on target pOlency was found , /'( I, 80) = 20.23, P < .00 I, p,"~ial ;, ' = ,20. The flllding was opposite to the prediction of Hypothesis 5. Nlore negatively cvaluau.'d targets were perceived to be mo)"{' powerflll (M = +.62) than more positively evaluated targets (JII = 3.93, evaluation of largets subjected 10 a median splil).

Perceived Probability 'If Compliance

To test Hypotheses 6 lOan analysis of covariance was conducted. The dependent variable in this ANCOVA was the observers' perceived probabilit), of compliance wilh Ihe larget. The independent variables \\ere the non-COntexl potency score of the agent, non-comexl potcncy score of the target, verb (tJlrealelIed vs. lriti 10 persuade), non-colllext evaluation score of the agent, non-context evaluation score of the target , and each of the interactions (lhree lwo-way and one threc-way) between and among the last three va.iables.

I IYPolhesis 6, which predicted thai threats would be considered less likely to gain compliance than persuasion attempts, was not statistically supported.

The mcan; for HypOlheses 7 and 8 are found in Table 3. Hypolhesis 7 predicts Ihal grealcr power of Ihe agenl resuits in greater per("eived probability of compli­ann". Consistent with Hypothesis 7, a statistically significant difference was found bctwecn low and high potency agcllls in their perceived probability of inducing compliance, f(1. 74 ) = 66.44, P '" .00 I, partial I," = .47; A/(lower pOlency age illS) = 36.98, and Alaligher potency agents) = 49.80, median split. Consistent with Hypothesis 8, high pOlency largelS were perceived as less likely to comply as compared wilh low potency targels, f(J , 74) =66.27,,, '" .001, partial ,,2= .47; MOow pOlency largets) = 52.47, and ,\lehigh potency targelS) = 38. I 7, median split.

The results lesting Hypotheses 9 13 are found in Table 4, Consistem with

SE~ IANTI CS OF SOCI \1. INFl.l1ENCI:

I.\UIJ.4 ~IL" S I \.'\lHRJ) J)~\'I.\ 110'\ Pt:.KO:./\ EO PROI\\BIIJTY en CO\II'U ... ~( I. 11\ hn ,l'f_'CI. T\cTI(;. ,\(; •. '\ I

E\ .\U \lJO'\,"U I \I{('LT F,\'.\U'..\IIO:\ ~I \1'\ EX ... ~...w.I'II.'\-I_ '\' - 189; '\'l \tRI,R OF St:XTF."'U_'i = R-I

42.01 10.:;:2 :H'H 18.8 1

41.85 rs.1b ;-'.'Ui2 (1.87,

l''hrC'.IWl1cd

1.0"

50.fjl 17.77 10.% 13.81

,9.96 IU:l, ·12.,11> H)..')I

309

lIypothesis 9, Ihe more positively evaluated the agenl, the greater the perceived probabilit) of complianct'. I~ I , 7-1 = 16.00, P ,;; .00 I , partial 1/2 = .18; M morc negatively c\'alualed agents" = 43.33. ,II mol'(' positively evaluated agellts = l5.03, median split. Consistent ,,;t!, Hypothesis 10, more positi\e1y ('valuated targets "crt' perceived to be more likely to comply, PI, 74 = 3.68, P ,;; .06, partial 1/2 = .05; ,l/mor,- negativel) e\·a1uated targets· = 11.72, ,II more positively evaluated targets = 45.49. mediall split.

Il ypothl'sis II predicts that morc positi\'ely cvaluated agents arc perceived as more likely to gain compliance whcll alttmpting to influence a more posili,'c1y {'\'alutHcd targct as compared to a morc l1('gaLi\'ciy c\'aluatt'd wrgcl. Th(' cornparison of the perceived probability of compliancc for more positively evaluated agents intrractillg with morc negatively versus more positively ('valuated targets wa"'i statislically significant, 446) = 2.36, /' ,;; .022, Ii' = .11; AI(morc posiu\'ely evaluated targets) = 49.11, M(morc ncgatively evaluated targels) = ·10.92, median split. Thus, II )1)othc.is II was supportcd.

II )l'othesis 12 predicts that more positively e\'aluated agents are perceived as more likrly to gain compliance when using persuasion than using threats, 'T'he comparison of the perceived probability of compliance for the more positively evaluated agents using threat versus pcrsua~ion alLtmpLI) was statislically significant. ~46: = 2.17, P '" .035, 1/2 = .09; "'(persuasion = 48.84, ,1/,threat, = 41.22 . Hypoth­esis 12 II as supported.

Finall), Ilypothesis 13 predict that persuasion is pl'l'l'ei\'ed as more likely to g-ain compliance between more positively evalualed agcl1LS and targcL"'i, whereas thrcaL~ arc perceived as more likely 10 gain compliance between more negatively evaluated agents and targets; persuasion attcmpts involving agents, targets, and tacucs that difTcr in ('valuation are perceived as less likely 10 gain compliance. As seen in Table 4, the two cells with the highest means cOll'l'spond exactly 10 the twO cells predicted to have the highest means; these two ('ells werc significantly and subswlltia lly differelll from the remaining six cells. ~B2) = - 3.79. P ,;; .001, 1/' = .IS. Therefore, H ypolhesis I 3 was supported.

The following is the leasl-squares equation predicting the pcrccivcd probability of compliance based on the analysis of covariance:

P('rProb .Ol11p = 57.37 + 6A6'Agcnt Pot(,ncy 6,./5{Targcl Potency) - 8.68 1r P('rsuade , - 4. I IlAgclll E\"i.tlu.llion l - 0,79 r rarget Evaluation , - 1.99 Persuade X Target Evaluation + .79 Pl'Nuad(' X ,\~('nt Evaluation ) + .'10 Agent Evaluation X Target E,"aluation + .94(pcr­sliadc X Agent Evaluation X Targct EvaJuation I

1 10 CO~I~IUNICATION ~IONOGR.\I'HS

TAKLE 5 S,\..\ I PU . S.~ '\ I k'\ct..') 1\ '\1) TH l:.IR LJl:..SCkWnH. S I i\ I'I~TICS

Agcfl( r<l ~t'l Agent Scnlencr' E\'a1uauon Eva1u.uion POIcnC}

~8 . T he mentor uied to 6.:15 5.27 5.8-1 !X'rsu,tdc lilt' pO("1.

52. The loser tried to 2.2 1 1.61 2.90 pc'fSuad{' lhr en('n1Y. I 'j 111{' I)(W( th n':w'lIl'd 5.27 1.6 1 J.65 lhe ent'm) . W. T he CI1C nl) tl lI'{';tlrned 1.6 1 2.2 1 5. 17 lhe loser.

:\ Olt', Eval uatiQIl and polCllq' ~'Ort'S ,Ire deri \t:d from nOI1-(:ontcXI roltings . .. Predicted pcreci\'cd prol}'1.hilil }, or compliancc is hast-d on "~uation I.

Pl'rcci\'cd Probability

" <lrgt' l or POlcnry Complianc('

:Hi ') 72.29

.~I . I i 18.83

5. 17 25.27

2.90 67. 14

Predicted PCI"Cci\cd

Probahilil)' or

Compllann:·

72.01

25.06

:lb, 12

66.9.">

where PerProbComp is the predicted perceived probability of compliance, and where aU the indcpendcnt variables are non-context (i.e., from the pilot study).

oefficients statistieaUy significant at p .;;; .05 (two tailed) are indicalcd in italics. The fi rst two effects after the intercept show that it was the difference in the potcncy of the agent and ta rget that helped determine the perceived probability of compliance:

6. ~6(Agt' ll t POll-ncy) - 6. · ~5rrargt' 1 POlellCY == 6A6(Agcnt Potency - T arget Potency). (2

The adjusted R for Equation I is 85"/". Thus, th" model predicts the perceived probability of compliance qu.ite well.

T ahle 5 contains examples of the sentences perceived to be most and least effective for both threats and persuasion attempts. Sentence 48 was predicted to generate the highest level of compliance. It depicts a good , powerful agent who attempLed to persuade a good but not very powerful targel. Sentence 40 was also percei\'ed to be very effective in gaj ning compliance; iL depicls a bad but powelful agent who threatened a bad but not velY powerful targe l. Sentence 52 was expected LO yield the least compliance; it depins a bad, not very powerful agent who tried to persuade a bad but powerful larget. S{'ntence 15 was also expected La resuiL in a ve lY 10"'.' level of compliance; it depicts a good, n01 very powerful agel'll who threatened a bad but powerful target.

Discussio n

i nterpretation oj Findillgs

Results of this slUdy showed that agents of innuence allempls are perceived as more powerful than targets of slldl a((('mpts (l-I ypolJwsis I). This finding suggests lhat, if negotiators wish 10 be pcrcei\·cd as powerful , the)' sho uld attcmplLO influence the other, by, for example, making proacti\'e offers. Beca use the percei\'ed probabil­ity of success at gaining compliance is, in part, a Il lllClion of the relatin- perceived power of the agent versus the ta rget, this strategy is also likely LO make ti,e inOuencing individual pereci\'cd to be more likely to succeed at negotiation.

SU'l)Jisingly, no support wa, found ror the notion Lhat agents arc perceived as more powctful when they threatt'n than when they altempi to persuade, nor did

SE~ IANTICS OF SOCIAL INFLUENCE 111

valiability in tactics alter perceptions of the targets' power. The combination of tactic and the target's goodness ,~so had little effect on the percepti n of an agent's power. As may be expected {j'om Lord AClOn's maxim, a bad agelll was seen as more powerful than a good agenl and a bad target was considered morc powelful Ihan a good target. Perceptions of power were influenced more by emluation of the agent or target than by influence tactic employed.

No support \vas found for the notion that ducats are \ricwcd as morc clTccuvc than persuasion attempts. Agents perceived to he more powerful and more positively evaluated agents were perceived as more likdy LO induce compliance than lower potency or more negatively evaluated agents (Hypotheses 7 and 9). Thus, in order to prolTIOlc the appearance that a person might succeed at social inHucnce, the person could choo e to usc the indirect route of increasing one's perceived goodness or power.

Targets perceived 10 be mon' potent and more negatively evaluated were perceived to be less likely 10 comply than those targets perceived 10 have less potency or more positive evaluation (Hypotheses 8 and 10). Therefore, in order to increase one's image as being able to resist influence 3ucmpts, a person needs La be viewed as relatively strong or relatively bad. FlII~ll('rmore, organizers of influencc campaigns may be less inclined to communicate with audiences composed or the powerful or of the stigmatized because the campaign's effecls may be discollllu.'d as not being able to make a difference; observers arc likely to regard attempts to influence such audiences as futile.

~ I ore positively evalU<lled agents "cre more likely to be perceived as gaining ('omp l.i ancc from more posiLivcly evaluated targets than from more negatively evaluated targ ts (Hypothesis 11). Applying this idea, an audience will expect that mcntors, who arc positively Cyalu3tcd, will be effective in gl' tting studen ts' compli­ance if the students arc also positively evaluated. Jf good agents share the audiences perception of their CffcClivcness, thcy rna) wam LO bclic\'c 1hal their targets of influence arc good so that they (the agenl5) can beueve themselves to be effective.

~ lore positively evaluated agents were more likely to be perceived as gaining compliance when using persuasion than when using threats (Hypothesis 12). Thus, morc positively cvaluaLcd agcl1ls in the presence of an audience may \\~sh to frame threats as merely persuasion attempts or avoid threats entirely because they wish to be perceivcd as effcctive influence agents. Persuasion wal) perceived as more likely LO

g"din compliance than threats when the agent and target were both relatively positivcir evaluated. Threats, how("vcr l \H'rc pcrceivcd as more likely to gain compliance when the agent and target \\('re both relatively negati\'e1y evaluated (H)vothesis 13). This finding is similar to that or Collob and Rossman (1973), that participants "rorm a favorable impression of an actor who behaves ~usuy)1 that is, one who benefits a good person [via persll~uiion] or harms a bad person I.via a threatl; and to form an unravorable impression of an aCLQr who behaves 'unjustly,' that is, one who benefits lvia persuasion 1 a bad person or harms a good person lvia a threat]" (I'. +00). If parties know how an audience evaluates them, they can cletermine the influence strategy to employ in order 10 be viewed as effective.

The statistical model (Equation I) ,howcclthat it is the differcnce in the perceived power or the agent and target that was a significant predictor of perceived success of the influcncc auempt. Note that in Equation I the coefficient for the agent's power predicting perceived probability of compliance was not constrained by the corrc-

312 COMMUNICATIO,\ ~ION()GRAI'HS

sponding eSLimaLe ror Lhe LargeL, yeL Lhe coefficients wCre essemjally equal and or OpposiLe sign (see Equation 2).

Implicaliol1f

These findings hm"C broad implications ror interpersonal communication as well as ror negotiation, diplomacy, and organi/.aLionaJ communication. As ~[iller 1987) slaled, persuasion is "both social and symbolic"p. 151 ). These findings broaden understanding or ho\\ the meaning of and relaLionship bel ween a largel and an agent combillc with choice of tactic to affect an obsc.::Tver's perceptions of an agent's power. This understanding advances knowledge abouL how people inler])rel com­munication stralegies to influence oLl1l'J'"s, an issue nut fully addressed in the complian('e-~ainillg literalure (sec Berger, 1985; pcrion; 1993).

The resuits or Ihis study have important implications specificalJ) ror research on compliance gaining. Bosler (1990) noted Ihat more descriptive work is needed to beller understand the "kinds or people [\\ hoJ usc compliance-gaining messages, in "hat kinds or situaLions, and \\;Lh what outcomes" p. 13 '. The fillelings presented here dcsclibc hov,,' words create Ihe meaning of social influence messages, and ho\\ different influence siLUauons arc judged through various combinaLions or good or bad and sLrang or weak agents, good or bad and sLrong or weak Largel>, and p('rslIa"iion attempts or threats.

or course. this study investigated lhe creation of mcaning in a social vacuum. This approach is parsimonious in that context was nOI investigated. I'\c\"(Tthdcss, the social context is not irrelevant to the creation of meaning; rather, a constructed or eV(,1l a conll;\,ed sentence system is used 11('l"r as a staning poinl. This approach is similar to what scienlisLs in oLher domains do: Simplify conlCXI La clarify process and SlrUClure. If this approach is sliccessful, vaIialion in context can surely be invcsti­gawd.

This sorial vacuum also applies to thc person nOUIlS and the relationships bctwcen agems and largets in the semcnCes employed. The cXLensive research by Heise (AverclI & Heise, 1987; Brill & Heise, 1992; I kisc, 1965, 1969a, 1970) and Goilob (1968; Gollob & Rossman, 1973) is consistent wid, and thus validaLes Lhe procedures lIsed here. As these studies and the presclHed rrsuits drmonSlratc, panicipams arc able to make sense of thc sentences c\'cn though they are extremely limited in context and relational information.

One limilalion or Lhis sLudy is that meelialing auribulional variables e.g., reasonableness of agent, percci\'cd ability and motivation of agenL to retaliate' were nOL measured. Future research should explicate the altribuLional process used by observers by measuring Lhese variables.

The evaluative and power relationship betw('en agent and target wcre found to be determining ractors in Lhe percepLion or power and innuence success. ZcidiLCh (2000) has suggestcd that an agent's power exercised o\'er a larget implies morc of a relationship than an altlibutc. Thus, perceptions of power and expectations about innucnce success arc embedded in perceplions or relationships (Dillard, Palmer, & Kinney, 1995). Persuasion presumes Lhat an agent offers a target a choice to comply or nOL to comply. These choices, in tum, help preserve social relationships. On Lhe other hand, thrCaL"i are perceivcd to hann relationships.

i\ universal semantics or social innuene(' assumes Lhat Lhe members or all language communities make meaning about influence attempts in the same way. Yet cultures differ along several dimensions, such as power distance, indi\;dualism-collec-

SE~IANTICS OF SOCI.\!. INFLUENCE 313

lI\"sm, and masculinity' femininity (Hofstede , 1980,. Perhaps the usc of threats is morc nonnative in high power-distance cultures. This study should be replicated in other cultures to determine if the model found here is the same even in the presence of language and other culUlral differences that may innucnec perceptions and meaning of power and compliance.

Finally, in this investigation only tWO tanics of social innuence were considered, thrcaL' and persuasion attempts. The complianec-/,,,,ining literalllre has suggested '('\'cral tactics and categories of tactics (e.g., Edgar & FitLpatrick, 1990; ~ [arwell & Schmitt, 1967a, 1967b; Wheeless, Ban'adough, & StewaJ"t, 1983). Punishment, rewards, moral appeals, aitnlism, and a variety of other Lactics also can be incorpormed into the model of social innllCIlCC gcnci-ated by lhis investigation. Each lactic may affect perceptions of power and outcomes of innucilce by interacting wlth the jud!,>'JTlenl of the agent and targel. Understanding how social influence tactics differ in meaning and incorporating this rncaning inlO lhe model arc fruitful arcas for fUlllre sllIdy.

What is unusual about this study is tha t a small and fixed sct of'variable within simple sentences were examined. These variables were the act and the evaluation and potency of the actors. With this limited set of'variables a !,'feat deal about ho\\ sense is made of influence attempts could he explained, or course) morc research is needed to cxtend the research beyond the limitations of this study.

Footno tes

lK;lplowitL, Fink. alld Lin 1998 use N<lgd 's 19;5 definition;:, of !)O\\cr. "An agent r:ctrcists jJQu'(!T" o\e r ;111

outtome to the extl'rn that the agent's prdc-renc('s help('d 1,." IlIS1' Ihal oli lcornt'\Nagcl, 1975), The agl'nt h<l~ pOltlltlill pfl¥l'tf t)vrr that otltmnw to the extent th.1t Iw/slH' h.IS the ilhili~ 10 cxt'rcise power over it" p, lOt), "N;II.~t:I (1975: lI t' lint,S I)(J\\(.'" ;IS c;lusat ion IhroU.L;h prt'fen'lltc's, rat h('1' than lhrough overt anions, to lakC' into a('COIII)1 til(' nOlion thOl powerful :lctors olll' n net'c! not ,ICt bt'cause o thers fear to chaU(,lIgc lhe known pn·ftrt'nct"S of til(' I){)"'tlful" p, 117 , This rramework is l'lnplo)'('ci hC'rt"

.'Tll(' oh~'l"\r r's Jl("l"iIlt'cliVl' provides insight into thl? annblilion of I'e"p(ln~ibilil)' rel::ued to om influencc allcmpt O-it-idcr, 1~)58J, When the agent ('awit's ,I tMget 10 do x, is the targrt Of' agrnt rr'q>Ollsibll'? ,\uributio ns ahoul huth iI till''1t'l and all agt'nt ilre based on the target's response to d threat rrom the agent. These auributiOIl!i III tum depend on whether a person thinks that allY nonnal person would give in to the thn'al or if any nomml ptrvlIl would IX' abll' to res.i"t iI, Ir ('\ct)' n()rmall){'r~on would giw in, then responsibility is plau'd 011 the Olll' thrc.'~'\(:ning, but ir e\'cl)' nomlal person would he ahle tn resist the threat , then r('''pon"ibility is plact'd 011 lht largt't. In other \'vOrds, "as is usllal with .ulribution, in the ,mribulion or an induced action, Ihc \\hu1t· {,llI',al slnJuurC' is taken into account <lI1d not m('r("I)' the prm;irnal conditions" Heider, 1958, p,2-17 '

lin OIh('r words, Ihe focus is nn how pcoplt' make 'it'IIM' of innucnn' ,IUt'mpts through language, Stud) pMtidpant~, like people in gel1ernJ , assess influence <1tIt'mp l~ b) ima~illin~ sc('ndnos ror th(' situaLion~ pro\idC'd, An imat;int'd 'Kt'nano thai is idiosyncratic 10 a giV<'n paninp'IIH 't',g .. '" saw the threatening ag('IH tarr; .. ing .t gun" \\'ilJ 110t be:: the dl'lel111irlltlH of sllld~ OIiICOIl1(,~, Ralhn, thl' r(,~IIIt~ will reAt'ct tht, social l-onSt'n'lUS <It'rived rrom Ihe ~impk S('nlt.'nets employed. I'llis method usc~ tht' SClHl'OCC <IS the unit of <lnal),,,;s, ~o p.mi{ip.ltll difil'fcn('cs ill imagill('d <;('cnano$ arr, in l'ffCCl. ;l\"(:,ra~cd Oll t (see, c'~" Woelfel & I'ink, 1980, p. 12 11r·.

IAlthou~h Kaplowitl. c t al. , 1998) sho\\rd th.lt ir t\ L"ompli,II1(,(' attempl was successful , Ihe LIS(' of Ihn'al!i incr(,ll-"('cl till' p('rn'ption ()f the agt·Ilt's po\\('r, Ihn';u .. \\'t'rr also shown tu reduce thl' cxpetlittion or ("ulllplianlc,

'1"1'11(' VC'rhs llirralmrd and Irlrd to fn'rslladran' as.<;unwriw nwa n Ih,lI,m influl·nn' altNTlpl was made, and that the OUIl'omc or Ihe aHempt was not nccessa.il)' succes')flll. I '0 determine ir this understanding of thest' \'('rbs is a 1'('<l5Ol1.1hlr Oil{' ror ultdergr<tduatcs al the university rmm \\hidl pilrlicipillllS \\crc drawn, Ihis point was iIlV('Slig'llCd, Data ,\cre ~dthered rrom a (oll\'cn iclllc <;;llllpk of30 unclcrgrnduates willing to complete a ~hon qUC<;lioll lla in' :onl\ 1\\0 people contacted I'cfu"ed 10 compir te the qucstionnail'e;, Thc qucstlonnain' 'L'iked wh(' tllt'r th(' Itnm thrtalnltd, jxrsuadr(/, trUd to jln'IUadt, <llld Intd /() tJlfratm rcpresentt'd an illnucnce attcmpt, and whelher the teml mdicatcd that the attempt was succc~srui. In .lddilioll, r('spondcllu w("rt' askC'd Iheir gt'lldt' l', a'l"t', <lnd nalivr Ianguagt-, Tht' questionnaire took less t.han 5 millllt.('S to complete. Although all four lcnns \\cre illlerprctcd 10 be innuence attempts, IltrtatDlld diWercd .. ignificantiy p < .()() 1, r..lc~('mar Test, binomial di"ttihutinn uS('d rrmll pmuadrd in IX'ing \1e\\t,d as .. un ('",lui, but did not "ib'llificanti) differ Irom tnLtI /()

314 CO~L\'lUN ICKrJON MO;-';OGRAPHS

P(T'.fIl(lf/f in lilis regard I J 0°'0 thought thl' lcnn IJlrto/(!flrJ mc;ull thatlbe aHem!>, was sll('c:cs~ful. 93° 0 thought the len" ptr.fundrd mealll lIl<tl the <lUt!mpl was successful, and J 0% Ihoughllhc ternl lritd /Q lln-.Hlndr meant Ihal the inlluenc(' auempl W~L'! successful}. Thus, the chojce of lenm uscn-,hr((lil'llrd and Incd if) ,,"slIodl results in a comparison o rinnuf'llce aHt'rnpls that do not differ in their implj('d success.

nAlthough factor anaJrsis of sermullit'·difTeremial data bas geller.-llly i(Hlnd tllrt'{' substantial diIllCn'iions. there are many ('xC'cplions to ,hi::; paHl'nl. Indc'cd, Osgood ('(.11. (1957) reported slUdies thaI found four, fivc, and eight slIch faclors. I-leiS(' 1969b) mentioned n;~carch ill which the pOlcnC} and activity dimensions \\,1'')'('

r('prcsclUC'd by a singk dynamism dimension, which is what was found here, 11n olher words, nouns were soughl Ihtlt \.\cre nOI illlerprctcd diff('rC'ntl) by males <Iud females, or b)' the

fonn order tll:t t happened to Ix' USI 'd. NcnUls Ihal w('rc a/reetcd by Lhese varil.lbks have a mea.ning thill is ull'itabk', and this in~tahilit) detracts !i'om the abilil~ to ('f("tIt' paf'!!imoniOllS models of pcrcci\'('d ~()("ial influence. On liw ollwr hand, this stnlu::g) limits Lhis ~tudy\ gCI1('ral izabilitv to nouns Ihat are not dirfcrcntiilUy <lffcCICd by ge'nd('r <llld fonn.

"One migln qucstiu.11 ,\h) (OStr \\dS induded in the l<ilucly giv(,1l Ihat the effect of fonn W;L'i'iI(lti'ilicall)' sipnifieamp < .00 1, 1/1 = .(9), and th(' interaction ofCcnckr X !-'om, was also statistically Si~'11ificallt jJ < .03, '1 = .04). Results for the c\!aluation of" uw:r wen' as foIIO\\s: males, form I: ,\f = ~ 0.17, Sf) = 2.47, n = 29; mal('s, form 2: M = 0.25, SD = 1.37, 11 = 18; femal<.'s, fonn J: M = - 1.2'~, Sf) = 1.38,11 = 3~; f('mah'~, form 2: /ll = 0.78, Sf) = 2.0 I, 11= 45. CollcagJlt's W(,I"(' asked to nOmil1.ll(' per.;on noun$ that fit within ('ells lltat. based 011 l-It'ist's list, \"('rt' mherv .. ist empty. The failure In find noun~ for a givcn cell may rdleellhe fact that su("h nouns ar(' raJ"C. or the 15 pen-on []QUI];:. found in J 165(' (19iiS), nu persOl1 nOUll \\<lS located in the 10 E, 10 A, lu P leU or lilt' 10 E, 10 A, hi P ('cll. Colleagues were unable to SUggl'S( nouns othl'r than lostr .tndftllou: that would he located in Ihe 10 C\aJlIalioll, 10 pOlency-afti"il), quadrant. FiJ(O/t' was eliminated I}('('aus{' of ils evaluative neutrality. Ciwll the statistic-at l"l'!'iulu rl'garding wJrr, the r1l1'lhodological fllll'stion lX'comes whether inclusion of IOJirdctraclS from the validity oftbis s tudy'" resuhs. If ::'Hl\'lhin~, this unstable noun ~h()lI ld make :,ignificiUlt re$ults more di11irull to obtain. In any case. it is hard to '.II'gul' that it invalidates the stud) or hia.ws the r('sults in .111) particular direction.

"Now ,holl lilt'" hypothesis-H'sting .um.lyse's used ;:1 ~implc uumm) \'<It;ablt" IU rt'prtSt'nt all the rclc\ ,Ill{ (artie infonnaliOIl. Thl' usc of this dichulomy I'l'prcscnts lhl' lise of l"O \erbs, rhrroJOItd ,md tntd to pn.fUadt. Hci!'it' 1970) and Golloh a.nd RI1'iSman j 1971· usC'd scmantic-dine rt'ntial ~or('s or the verhs the') cmplo)ed,

References Ajl.c·n,.1. 19961. Tht· social psychology orclcdsion making. III E. T. I lig-gins & A. W. Krugl.lIlski jEd!!.!, Social

p.{)·dlo{ogJ': Handbook of Im.fli pnllnplts (pp. 297 32.1. ~ev.' YOI'k: GuiJii.>rd Prt'ss. A ... ·crell. C .. & ilcise, D. R. (1987. ~ Ioclifitd social iclemili{'.\: Amalgamations, atuibwiom, and ('motions.

]oumal ojMatJlmwtical Sociology, 1.1, 103 132. 8;]("hra('h, P., & Baralz, M . I 1962', Two f."1("('s of PO\\('f". , I",trica" Pulitical &;£11(( Rtflltu', 56, 917952. Bal't1('II , J. 1992). Fmmllflr quotations: tI (()lil'cllOlI f!IIHLHIlJ!,tS. phra.I(.f, alld prfJUrhl ITOud 10 flu';r SOllrUJ ill o1ldm/ olld

I1Wflml IIlmllll" 16lh ed.: .J . Kaplan. Ed.). Boslon: Litlh" Bro\\II, & Compan). Bergrr. C. R. (1905). Soci:ll power ;lIld intcI"}XrsonaJ (·ommunitation. In j'd . I.. Kni.lPP & C. R. ~liUer (&b.),

limit/book ojlllttrpl'I:llJllaf (ommunv(J/iall (Pi>. l39---499). Bc\"(' rly I-lill:o., CA: Sagt'o Boqcr, F. J. (19QO . An eX:llllinalion or the state of compliallC"c-gaining message' behavior I"est':'lreh, In.J. P.

Dillard (Ed.). &tlc-i"!. (omplionu.: nit ProductlO1I l!f intupnro/Ull ;/ylu(IIu mtssOf:ts (pp.7 17\. Scottsdale, AZ: Gorsuch Scarisbrick.

BOSlCl', F . .J., Mitchell , J\t. J\I., Lapinski, M. K. , Cooper, II .. On-cgo, V. 0., & RC' inke, R. 1999). The impacl of guilt and type of C'ompliance-gaining: messagt: on compliance. (,'onullllllicarlOl/ Airmograpkr, 66, 168 177.

Brehm, J W. (1966). Atheery ofN;dlowgictlf r,t1.ctlllIU. Nc\\ York: :\ caclcmic Pres.'), Britl, L., & lI eise', n. R. 11992). Impr('ssiol1s uf' ~c1f-direcl('d 'lClion. Sorial Psydlolog)l QjIflT!u(J', 55. 335 350. Brown, P., & LeviTlson, S. C. (1987). Poilfnlm: Sow 111Iil'maLr ill Imlf:llagi UStl,i!t. Cambridge. U K: Cambridge'

Uni\'crsil) Press. Gli, D. A, Wilson, S. R., & Drake, L. E.. ( 2000~. Culture in thc COl1lcxt or intercultural rwgotiation :

Individua.lism-collectl\;sm and paths to il1tCgrdti\c dgrecmCnts. //umall Commullicatloll Rf!frarrll, 26, 591 617. Cast, A. D. (2003). Power and the abilit), lO dcfin<.' lhl' situ.ltion. Soria{ P",),r/IO/o,{'I OJ.wrlu[l .. 66, 185 201. Cody, 1\1. J., 1\1cL.··\Ughlin, ~ l. I.., & Jordan , lV. J. 1980). A multidimens.ional sC"..tling of three' S<""t!) of

wmpliann'.gainillg sll"at('giCS. t'omlllllllllatloll Qyl1rltr~~, 28, 3 1 16. Oillard, J . P. (Ed.). (1990\. Stf}mlg (omphali,,: 7lli "m(llIefiorJ f.!fllllcr/lt7"JMo/ mtJsagfS. Scoll'>Cla1e, AZ: Gorsuch

Scansbrirk. Dill.lrd ,J. P., Palnwr, ~1. T ., & Kinlll'), T. A (1995!. Rdmjulhujudtpnems in an inl1llcnce ('ontcxt, I1lUl/all

Cl)llllllllJli(lIlion RU(Qrrh, 2 l, 331 353. E.dg'd.r, T.. & Fitzpatrick, ~ I. A. ( I 990}. Communicating :SCXlJ;U desire: t-. lcssage tartics for h:wing and avoiding

inteI"Coul1ie. 1n.J. P. Dillard r&l..l. SuA1J1i cOff/pHtlllc,: Th, productioll if illlrrprrsollal ;'ylll(lIU mwogtJ (pp. 107· 121 ). Scottsdale, AZ: Gorsuch S("arisbrick.

Falbo, T. 1977). Muhidilllcnsiollul scaling of pm\-C'r stralegies. ]ollmall!f Pmollali!y and Social Pv'dm/'W', 35, .1:17.147.

SE~ [ANT[ CS 0[' SOC[ ~L [~FLL' E:-;CE 31.\

I-iske, S. T.. & -t.Jylor. S. E. 1991,. Social {/Jxni1Ion. '\Je\\ Ynrk: r.,lc(;r;u\· II1I1 . Frt'ufh,j. R. P .• Jr., & R,\\i('II, H. 1459,_ The h.l ... t·~ or \()('ial pm\!'r. In C. (:;lmvrigln F,d .I, Slm/itl mwltnl /KIlt'"

J>p. ISO 167,. Ann ,\ rhor: University of r..lid ligan Pn's," Callison, \\', A 1968" POll'" tII,,1 di.l(on/trll. IIOlllt\\ (x)d, II .: Dor\l'Y. C;aski, ,1. F. (1986). Il1u'fT('iali(lIH among .1 dl,llll1('1 (,Illily 's po\.\cr sOllrn's: imp.'tt of the l~x('niS(' of r('\\.lrd

and C'oerCif)ll (111 ('xpl' rl , n'ft'l"cnt, ,mel h·gitinl.ll(' power sUllrn:s.]ounwl rf. l/nrM/mg Rf\(lrcJl. 23, 62 77. Giblx)IIs, P., HradM, .J.J. & Bu~dl . J. D. 1992. TIl(' roll' of"1angua~(' ill lIt'goli.lIion: Thn'ab and prorni\t·s.

In I .. Ih PUllloll1l & ~I . E. Roloff Eds ..• (;nmmummtlon lind IlfJ.:fIllfllion pp. 1.}(; 175,. ~c"bur}' Park, (:.\. S.Il;"t:, Gullob, II. F. I q68 . Imp rt'S$ion fnnnalioll ilOd \\()rd rnmbillatiull ill S<'nt('ll(C's. ]llIImal tif Pnronnli9' Qnd ,wlol

1\)'rholtJgI, 10. :H I li:i. Gollob, II. r ., & Russman, 8 . B. Iq73I.JlIdgmenl.~ oj all al'tor's "po\\c r <\nel ahillty 10 influcllu' nlh('t"i."

]ollmllllJj E • ..:/wnmOllill ,WI(II P~)'d/OIOiff. 9, 191 -lOb, I-ft'id('r. F. (I%8,. n,~ pJ)dwlog), 1!/1II1"/ltnOIlOI ulol;ofll. '\" '\\ York: .John W ik)' & Sumi. 1II'i,/', n. R. Iqh:) !. S('mantiC' dini'rl'l1li<i1 pr'ofil,'" Ii II' 1.000 mO~1 frt'lJ1If'nt En~li,h v"ur,k P,:)'(hlJ/ugiml

,\/ollo,l{TtJ/,hJ: (""noloml. lfJp/ml. 7.9(8), 1 31. I ki ~t', D. R. 1%9<1 .. \fkuu.d dynamics in sill1plt' ~l'nt(·n('('s. }urmJrl/ ilj PmOlra"~) and Sotwl P~ld/Olo(). I I,

20 1 2[3. lI('isC', I). R. 1969h ,. !oxmu" mcthodological i~"iUl'''i in 'iCllhlntic diflen'lIlial n .... l·ardl, P,)·dmhl/(iral Rill/rIm, 72.

106 In II t"iw, D. R. 1970 J. Po tl'nq drn;lInics in ~ il11plf: ~l1tl'lIn"~, .lor/mal tif Ih'(JI/(ditJ! aJ/d .':Mlal PfldllJlogl', /fi, IR :; 1 Il inkin. I ' R., & Slhrit·"hcim. C, A. 1990;. RrlatiOilShipl! Ix'theen suhordin.tlt' pc,"ru"»tiolls (Ir .suIX·I'\<l"M

influcnce laCO('<; iJlld aurihut('d l);ls(''i of SUpl"I'\<lSOI)' jlO\\et'. I/lllf1(lll R,lafIOIB" ., '1, 221 237. Ilufst"d{', C. 1980 .. (;ulluft\ (()I/I''1Unlm; 1"lm/(/lmf/al d!lJnmw m 1I-'01~ rtlolld mllll't. B("\'(' rI~ Hilb , C,\ : S,lgr. 11m land, C. I., & \\'ds\, \\ . ( 195 I I. TIl(' infltl('IKI' of "ioun '" ( rrdihili lY Ull ('()fl IllHllli (";lIi() IlI'rr('C,· li\'("Ilt·"i~. PI/hiu

0ll/lIibll (}jmrlrr(v" 15. 615 6.'10. K.l lult'mall , D. 1'l63,i. Tilt' S{'mantic,' difkr('lltiai <lml tilt' "iu'unlll"(" ofinfl'f!'llU'S . lI n()n~ attrihlH("S. , Imr,.irml

Joumal of J-'f)'rholog'J' 76, 55 I 567, Kan('. T . R., j oscph. J ~I. , & T edeschi. ,j. ' I . Iq77 Pa('('i\'('d rreedom, ;'\j.Q.,r'fl'""iflll, and rcsponsibi li t), ,111£1

the' ass.ignmcnt (If pUlli .. hm('nt. JOllmal oj .f)(lflal P~;(hll"~~" 10.1. 257 26:1. Kaplowit;., S. t\. 1<li81" I"(}\\.m!s;\ s):o.tt·m;llic tilt·or) ofp(mer altribUlioll_ Sunal r~,'fllOlo.t:I" -/I, UI itS. K..tp!{I\\;tz, S . .-\ ., Fink. E. L. & Lin, Y. IQ9RI. Sp(',\kin~ ICludl) ,Uld carl)in~ a hi~ ~t ilk: Tht' cfieu ofpO\\('r

l.\('ti<"s and slrun ural pm .. t'r 0 11 pcrc('ptions of Iht' pmq'r UM'r. In P. C. \\'a.'i1lt1nl E.d., R'.II'llTcli In po/illml I(}(W/og, \'01. 8, pp. 1m 11 9. GrccliwiC"h, (:'1 :.J ,\I Pn'\~.

1\.('[1110111 . II. C. 19.1fl:.. C:ompli.l.IKC, idl' llIifi(";uinll, "lIId int(·'·IMli.tation: 1'11 ... ,(, pnwc,''''CS of ,millld{' dlilll~t" ]nllmal oj Ompirt R,wJIII;on. 2 .. 11 60,

Kt'IIll;tn. II. C. 1961. Pmn· ... "'·s oropiniolil hangt·. Pull/II Opmum (hUlfII'Tb" 25, .17 is. Kipnis. D. I qn . Do("s IXI"""r corrupt? .Joumal ~r Pn,,,,mli!., ami .'illliat PfJ,choloJJ'i. 2-1. :n ,11 . 1.("I'IIt' r , l\1. j I <ISO 'nit' btlit{ In a)rLJI u'OTid: .1 /um/ammta/ dduIWfI. :'\,.\\ " ork: 1)!t·llulll. I .('mrr, ~l.j ., & ~Iill('r, D. rl . II17S, . .Just \ .. orlel rC'\t';\H-h alld Ill(.' a llributif)l1 pron· ... s: Looking b.u:k ,md

ahr;ld. Psyrhol~ta/ Bullrlm. 85, 1030 1051 1A.'\;ne. T. R., & UO'itt'r, F.j . (2001,. Th(' "O('(h IIf pm\l'r .lIltl m,·~~.l~t· variahlt's 011 ('()mplial1fc. (;ommunimlui1I

.HmUlgraplIJ. 68, 2A IR. i.cwiC'ki. R. J , B.tH), n., Sallnell'I'S, D. ~I. , & ~Iilltoll, J. W. 120031. \ rgolwllOlI Ilh rd.\. ~t·w Yurk: ~llGraw- H ill.

l\1.II'\\"dl. G., & Sdnniu, D. R. 1<)61i11. C:ulllpli.mlc-~.lIl1ing Iw h;l"i()r: t\ ~)lHhc," S"is ,tile! mod,'!. .4iMI%gJtal QJ.,fITt(1'!l'. 11, 317 328.

~1 .II'\\dl , C .• & Schmit!, D. R. 1%7h). Dinwnsinlls 01 l"nmpl iantt'-.'{'..t ining lx.'h.I\<lOr: An empirical .m..ti)sl". Soc,omdry, 30, 350 ~16·1

~f(')-('r, J. k. 1990J CognitiH' procesSC"s underl)ill~ tht" retric\-al of compliaIKc-gainin~ ~tnltcgics: An implicit rule'li model. 111.1 P. ilil\;ml Ed., SuA7ng mmpliollcr: lhr pUll/uri/on r!I mlrr/,morlalmjlurnrr ml'f't;r:t.I PI). ~) 7 7:3 \'Koll~I'llr, .V'.: GUt"iurh Sc'lri~brick.

~1(·)cr. J . R. 1996). R(· trif"\'in~ kll()\\kd~(' in sodal ~lIu.III1I11~: . \ I('st of lhr impli,:il rub model. Cmnmuniralulf/ Rmarrh" 23. SAl 61\.

l\lu ·hencr, "1 . 1\ .• 1 .. 11\ kr, E . .1 ., & liilcharach, S. t ~17:S I. jJj'n l'JJUOII 01 pO\\t'r In c()nfllct Sllll.IUom , ]oul7lol qf Pmollali~'1 (lml SoOtJ! PJ)'dlfJlfJg}'. 28. ISS 162.

~Iilkr. U. R. (I~1871, 1\·I~uasion. In C. R. B('fgn & S. II . Ch.tllce IEds. l" lIalldbool. ojrommunitalion trimer Pi>. 116 483\, Nt'\\bury Park, CA: S.agt·.

~lurdock.j . .J.. Brad<lt.J.J , & Uo\\el"S.,j. \\'. 1 CIA 1 . Ellcus ofpcl\\e r on the jX"rcl'Illinn of explicit and implicit thrcaLS. promises, dlld lhromises: A rule-governed pcrspccti\,e. 1I '(Jlml }uumal fir S/Jl«h u",",umcohon, 48. 3 11 '161.

~.lgd,J. H. 197, 1111' dt'lmphl"r OIIt1!,·sU oj potL'tT. ;'\J('\\ 11.1\t'lI, CT: Yak· L'nivl"rsit)· Prt'SS. ()'Kt"t:rt', B. j. 199()). Th,· logi(" or rcgul.uiH rommuni(.ltJon: L'nd('rslalldin ~ thl' ratiunalit) or Ull'SS.lgC

designs. In.J. P. Dill,lrd Ed.). S,tklJlg (omIJlillllrl': 'li't prlJdurtJOII rljm ll!rlJn'wno( mflul1ll't mwagtS Ipp. 87 10,1 ). Sf'()nsdalc, AZ: GOI~lI( h S(,llrisbrick.

316 CO~I~ I UNICATION ~ I ONOGRAI'I-lS

O'Kt't{c, O . .J (2002:. Guilt as a m('chanism of persuasion. In.J. P. Dillard & ~1. Pfall Ws. '. 'n,t /In'.flIQ,flOfl

handbr/Ok." Dn'tlopmmlJ 111 thtory and Pf{ullCt pp.329 341 . Thousand Oaks. CJ\ : Sage, O .. goocl, C. E. , Su("i , C . J. , & Tannenbaum, I). II. 0951' 17t(m(a.surtrTWltofmtanu~. Urbana: Cni\'ersil), of

Illinois Prtss. O\'erbcrk,J R., & Park, B. 2001 . When power doc" not corrupt: Superior individuation procl:sses amon~

pow('rful p(.' rcr"i\"~·rs. J ournal of I'mona"!y aJl(i Soonl Pv'choln;,'. H/. 5 19 565. PerloW, R. M. 1993. TIle ~"'(Jmi(j flj /J"suasio1l. IlilIsdalc, i'U: I"HHcncc Erlbaum Associate ... Pruitl, D. C . 198 1 . ."fgotialloll ht/wL'ior. Nc\\ York: r\c-adcmic PIT''! . Pruitt , D. G., & Camnak. P. j. (I993i . . \ lgolinllon /1/ ronal (oltjhrt. Pacific Crove. CA: Brooks/Colt", Plimalll , I.. 1.. , & Poole , ~1. S. 1987). Confliu and negotialion. In F. ~1..Jablin. L. l .. Putnam, K. H. Robens,

& L W. I>orler Eels. }, Handbook tfforgam.~atiJmal (o"unummt;on pp .. 1-19 599,. Nt"whury Park, Cr\ : Sa~e.

Rinillo, S. C., & Trtnholm, S. '1983. Predicling man"ger's choic(> of influt'nce mode': ' nIt' l·o-eels of illlcqX'rsonal trust ilnd workrr aUlibuuons on milllilgerial laflin in a simulated orgarti~aliollal s<,tting:. II tflml J ouma! if Spttdl Communicatloll. "7, 123 339.

Rolon: J\t . 1:'. .• IJaul!lOIl , G . U., & Vollbrecht , J. (199B .. The illlcq)rClalion of cocrcive communitauon: The dlcu.s or modI' or influence, powerful spe(,I'h, and speaker authority. IlIlmlOlwltal Journal qf COl/jlut j\/(llIaytllnll,9. 139 161 .

Schl,' n.kcr, B. R., & Schlenker, P. A. (19H) Prestige of an infiuenrcr and 1X'ITI'ptions of power. Bul/l'lln oj thl' PJ}c1l1'1noflltc Sorltt'i. 3! 1 A ,3 1 33.

hapiro, D. L., &. 'Bics, R.J. 1994 .. "n1frats. bluff 'i •• mel disd.limCN in n{'~oli'llion. Orgoni:ullonaI IJthlu'lor alld 1-(lImall /Jtri.llfIIl Proa.1Sa. 60, I I 35.

Smith-Lovin. L. 19R7 i. impn:ssions rrom C\'C:IlL'I. J Olimal qf,\[aIJIt111r1{/m/ S4l("wlog). 13. 35 70. Thomas. K . W . 1976'. Connicl and conflict m'Ulagt'n1t'nt. In ~1. n. Dunnettt· Ed." IIr.ndbool. ojmdu.Jlrial and

qrgam~alu.mal /Jf)'rhoiDgJ' PI). 889 935.:. Chicago: Rand ~k:-':all~ Thnmdike, E. 1.. 1920 . A COllstanl error in ps),cholol,rlcal r"dtings. Joumal C!f. 1pplud P9'(holhfJ·. 4. 25-29. \\,heele~ ... , 1 .. R., Barraclough, R ., & Stewart, R. 1983,. C:ompli'Ulu·.gaining and po\\c:r in persuasion. In R.

X Bostrom F.d.,. (;nmmtlllua/IOn]l'arbooJ.. 7 pp. 105 Il5. & \'crl)" lI i1ls. CA: agI". \\"ilsOIl , S. R. I Q97). l)ewloping tlwories or pt'rswl'i .... r m('"<;<;a't~· production: The next gt·ncration. In.J. O.

Greene Ed .l. \[tungt protludlon: AJr:DnuS In (omnmmcali{)1l thtmy pp. 15 13 ). l\ lahwah. ~: L..'lwren('e Erlbaum , \ ssocialc·s.

Wodfel , J. . & J·ink. i'::. I.. 1980. The mt4511Tttllnll qfrom1flumratum /"0(tJIlJ.' (lotiko I"to,)" and mtrJ/Od. :,\cw Yor"-: Ac"ildl'l11ic Prrss.

Zt:ldil("h , ~I.,Jr. ;~lOOO·. Im('rper.;onaJ power. In E. l- . Borh<aHil & R . .J. V. l\ tontgomery (Eels.), EnC)'(wptdia of loc10101:)' (2nd cd .• Vol. 2. pp. 1156 116·1), New York: J\I ,Kmillnn Reference CSA.

Received: March 5, 2003 Revised: October 10, 2003 i\ccqJlcd: October 31, 2003

Copyright of Communication Monographs is the property of National Communication Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.