Journal of Abnormal Psychology1975, Vol. 84, No. 5, 508-518

Effects of Alcohol on Aggression in Male Social DrinkersAlan R. Lang

University of Wisconsin—MadisonDaniel J. Goeckner and Vincent J. Adesso

University of Wisconsin—MilwaukeeG. Alan Marlatt

University of Washington

The purpose of this study was to determine the effects of alcohol on aggressivebehavior in male social drinkers. Ninety-six subjects were randomly assignedto one of eight groups in a 2 X 2 X 2 factorial design. To fully control forexpectation effects, half of the subjects were led .to believe that they would bedrinking alcohol (vodka and tonic), and half believed they would be drinkingonly tonic water. Within each of these two groups, half of the subjects actuallyreceived alcohol, but half were given only tonic. Following the beverage ad-ministration, half of the subjects were provoked to aggress by exposing themto an insulting confederate, whereas control subjects experienced a neutralinteraction. Aggression was assessed by the intensity and duration of shocksadministered to the confederate on a modified Buss aggression apparatus. Theonly significant determinant of aggression was the expectation factor: Subjectswho believed they had consumed alcohol were more aggressive than subjectswho believed they had consumed a nonalcoholic beverage, regardless of theactual alcohol content of the drinks administered. Subjects receiving alcohol,however, showed a significant increase in a reaction time measure, regardlessof the expectation condition. Provocation to aggress was also a significant de-terminant of aggression, but it did not interact with the beverage conditions.

There can be little doubt about the exis-tence of a strong relation between alcoholconsumption and aggressive or destructive be-havior. Research on violent crimes, such asstabbings (Shupe, 19S4) and homicides(Wolfgang & Strohm, 1956), has shown thatalcohol was a factor involved in 64% to 88%of the cases studied. McKay (1963) hasalso reported that the antisocial acts of manyjuvenile delinquents were often associatedwith problem drinking. In addition, the re-sults of other studies (Selzer, 1969; Selzer,Payne, Westervelt, & Quinn, 1970) suggestthat the high proportion of intoxicated driversinvolved in fatal traffic accidents may reflect

This research was supported in part by NationalInstitute of Mental Health Predoctoral ResearchFellowship 1FOL MH58240-01 to Alan R. Lang andby intramural funds from the University of Wis-consin. The authors wish to express special thanksto Richard Lauerman and Michael Pritchard fortheir assistance in the study. Daniel J. Goeckner,who conducted the study as part of his Master ofScience degree under the supervision of Vincent J.Adesso, is now at the University of Illinois.

Requests for reprints should be sent to G. AlanMarlatt, Department of Psychology, University ofWashington, Seattle, Washington 98195.

an enhancing effect of alcohol on aggression,which could interact with the well-docu-mented deleterious effect of alcohol on psy-chomotor coordination. These observationshave contributed to the position of numeroustheorists (e.g., Cameron, 1963; McClelland,Davis, Kalin, & Wanner, 1972; Tucker,1970) that alcohol serves as a stimulator orreleaser of aggressive behavior.

At least two competing theoretical posi-tions can be proposed to explain the correla-tion between drinking and aggression. Thefirst of these is predicated on the assumptionthat alcohol affects aggression-related behav-iors through some physiologically based mech-anism. Some proponents of this approach havestressed the "energizing" effects of alcoholabsorption on general activity level (Barry,Koepfer, & Lutch, 1965), aggressive fantasies(Kalin, McClelland, & Kahn, 1965), andneeds for power and dominance over others(McClelland et al., 1972). For the most part,however, research so far has produced onlyindirect evidence of any directly stimulatingeffect of alcohol on aggression. Others adher-ing to a physiological explanation of ..thedrinking-aggression relation have argued that

508

EFFECTS OF ALCOHOL ON AGGRESSION 509

alcohol does not elicit aggression directly butinstead "disinhibits" and facilitates its ex-pression by acting to reduce fear and anxiety,thereby diminishing the impact of the physi-cal and social consequences of aggression.Studies of the effects of alcohol on anxietyand mood changes (Kastl, 1969; Mendelson,LaDou, & Solomon, 1964; Tamerin & Mendel-son, 1969; Williams, 1966) have yieldedequivocal results, however, with such vari-ables as drinking experience and blood-alco-hol concentration mediating the effects ob-served. Similarly, research testing the effectsof alcohol absorption on approach-avoidanceconflicts that may be associated with the ex-pression of aggression has shown inconsistentfindings, particularly with human subjects(see review by Cappell & Herman, 1972). Atthis time, no strong conclusions can be drawnabout alcohol's disinhibiting effects on aggres-sive behavior.

A second, and often neglected, approachto explanations of the drinking-aggressionrelation calls attention to the distinct possi-bility that it is mediated by a psychologicalexpectancy set regarding the behavioral effectsof alcohol consumption and/or by an abilityand tendency on the part of many people toattribute their antisocial acts to their intoxi-cated state instead of to themselves. In thisregard, Sobell and Sobell (1973) have arguedfurther that one of the rewarding conse-quences of heavy drinking is that it providesa socially acceptable excuse for engaging inotherwise inappropriate behaviors, includingaggression, with a minimum of social dis-approval.

A recent experiment by Boyatzis (1974),which found that male subjects in a partysituation were more aggressive if they drankhard liquor or beer than if they were givensoft drinks, illustrates how simply controllingfor beverage content may lead to discrimin-able differences in behavior, but it does notrule out the role of expectancy factors. Inanother study, Bennett, Buss, and Carpenter(1969) attempted to control for expectationby employing a placebo condition along withthree alcohol dosage conditions and obtainedno significant treatment differences in aggres-sion as measured by the intensity of electricshocks that subjects were willing to administer

to another person. These null results, however,may have been due to the fact that all sub-jects in the experiment had the same beverageexpectation. Finally, in a similar study,Shuntich and Taylor (1972) reported thatundergraduate subjects who received alcoholwere willing to administer higher intensityshocks to another individual than subjects ineither the placebo or control conditions, whichdid not differ significantly from each other.Unfortunately, this experiment did not in-clude a control condition in which subjectswere given the expectation that they werenot receiving alcohol when, in fact, alcoholwas actually given.

To fully explore the role of expectancy asit may interact with the physiological effectsof alcohol, it is necessary to employ additionalplacebo control groups. An experimental de-sign of this sort was described in a recentinvestigation of expectancy effects as deter-minants of loss of control drinking in alco-holics (Marlatt, Demming, & Reid, 1973). Ina 2 X 2 factorial design, half of the subjectswere led to believe that they would be drink-ing alcohol (vodka and tonic) and half be-lieved that they would be drinking only tonicwater. Within each of these two groups, halfof the subjects received alcohol and half weregiven tonic water only. The advantage of thisdesign is that it controls for both the expecta-tion of the effects of alcohol and for the expec-tations associated with drinking a nonalco-holic beverage. The present investigationemployed this design in order to assess theeffects of expectations associated with drink-ing and aggressive behavior.

Another important feature of the presentstudy is that it permits investigation of therole that arousal or prior instigation to ag-gress plays in the drinking-aggression rela-tion. Perhaps the association between drinkingand aggression is the result of the action ofalcohol in lowering tolerance for stress orsensitizing an individual to aggressive cues.To test for this possibility, subjects wereeither insulted or not insulted following anunsuccessful effort to complete a task givenprior to the assessment of aggression.

In view of the equivocal nature of the re-search reviewed above, it was difficult toformulate a priori hypotheses concerning the

510 A. LANG, D. GOECKNER, V. ADESSO, AND G. MAELATT

effects of alcohol on aggressive behavior. Itwas possible, however, to predict that provo-cation or instigation to aggress would increaseaggression relative to nonprovoked subjects.Concerning the effects of alcohol on aggres-sion, the design of the present study permitsan evaluation of two competing theoreticalassumptions: (a) If the physiological effectsof alcohol are primarily involved in the facili-tation of aggression, then subjects who actu-ally receive alcohol (regardless of their ex-pectations) will behave more aggressivelythan subjects who do not receive alcohol; (b)if expectation that alcohol will lead to moreaggression is the major determinant of sub-sequent aggression, then subjects who expectthat their drinks contain alcohol (regardlessof the actual alcoholic content of theirdrinks) will act more aggressively than sub-jects who do not expect to receive alcohol.

METHODSubjects

The subjects were 96 male undergraduates, aged18 to 26, who were enrolled in introductory psychol-ogy courses at the University of Wisconsin—Mil-waukee. Each participant was classified as a heavysocial drinker on the basis of his score on Cahalan'sDrinking Practices Questionnaire (Cahalan, Cisin, &Crossley, 1969), which was administered to under-graduate psychology classes prior to the beginning ofthe experiment. The subjects used in the presentstudy were drawn from the upper half of the heavy-drinker category and represented about 10% of thetotal sample completing the questionnaire. Heavysocial drinkers were selected as subjects for the fol-lowing reasons: (a) Use of a homogeneous sampleof drinkers permits some control over individualdifferences in tolerance and sensitivity to the effectsof alcohol, (b) heavy social drinkers have had con-siderable experience with alcohol and are likely toshare expectancies about its effects, and (c) by study-ing a group of high-risk individuals, some of whommay later develop serious, drinking problems, evidenceconcerning their behavior under the effects of alcoholmay suggest new avenues to explore in the preven-tion and treatment of alcoholism.

Subjects were recruited by telephoning those in-dividuals who met the heavy-drinking criterion andasking them to participate in "a study of the effectsof alcohol on personality, learning, and various otherbehaviors." They were told that they would receiveextra credit in their psychology course for theircooperation. Those who agreed to participate wereasked to abstain from food and drugs of any kindfor a period of 4 hr prior to their scheduled ap-pointment.

Experimental DesignThe independent variables manipulated in this ex-

periment were (a) expectancy—subjects were led tobelieve that their drinks contained either alcoholic ornonalcoholic beverages, (b) alcohol content—subjects'drinks either contained alcohol or did not, and (c)provocation—subjects were either provoked to angerby insult or were treated in a neutral manner. Acompletely randomized 2 X 2 X 2 factorial designwith 12 subjects per cell was employed.

ProcedureA male confederate (posing as another subject)

joined the actual subject in a waiting room wherethey were met by a male experimenter. The experi-menter explained that the experiment was designed toinvestigate the effects of alcohol on personality,learning, and various other behaviors. It was alsopointed out that, as part of the procedure, somesubjects could become intoxicated and some mightreceive moderate electric shocks; acknowledgement oftheir voluntary participation and of their acceptanceof responsibility for any possible consequences of theexperiment was thus required. Before asking forsignatures on the consent forms, the experimentermade it clear that subjects were absolutely free todiscontinue the experiment at any time they wished.Three subjects declined to participate at this point.Then, following a brief outline of the procedures,the subject and confederate were escorted to separaterooms. After seating the subject in an easy chair,the experimenter left momentarily to give the subjecttime to survey the equipment in the room, whichincluded a bathroom scale, an intoxilyzer,1 a barwith the usual accessories, and some shock electrodesadjacent to a large official-looking panel of buttons,lights, dials, and meters.

Shortly after the experimenter left the subject'sroom, another individual, a male wearing a whitelaboratory coat, entered and introduced himself asthe experimenter's assistant. He -stated that, in orderto keep the experimenter blind as to whether or notparticular pairs of subjects had been given alcohol,he would be administering the drinks to all subjects.He stated that, regardless of whether or not thesubject was a member of a pair receiving alcohol, hewould be following the same administration proce-dure for all subjects. It was further explained thatthis precise uniformity of treatment, including theconsumption of comparable volumes of beverage,was necessary to ensure that it was the alcohol, andnot some other difference in the experimental proce-dure, that accounted for any observed variations inbehavior and performance. After this introduction,each subject was weighed in order to determine thevolume of beverage to be administered. An intoxilyzertest was then given to detect any residual blood-alcohol content.

intoxilyzer (Omicrom Model 4011) is adevice for ascertaining blood-alcohol levels throughbreath analysis.

EFFECTS OF ALCOHOL ON AGGRESSION 511

Expectancy and Alcohol-Content Manipulation.The assistant first manipulated the subject's expecta-tion for alcohol in accordance with the conditionassignment. Each subject in the expect-akohol con-dition was told that he and his partner would bereceiving several vodka and tonic drinks; in theexpect-tonic condition each subject was told that heand his partner were part of a control group andwould be drinking only tonic water.

At this point, the assistant consulted a previouslyprepared table to determine what volume of 1 part100-proof vodka (Petrushka) to 5 parts tonic(Schweppes) would be required for an individualof the subject's weight to reach a .10% blood-alcohol concentration within approximately 1 hr. The1: S ratio of vodka to tonic, used in the drinks con-taining alcohol, was chosen on the basis of the pre-viously reported finding (Marlatt et al., 1973) thatsubjects tasting this mixture could not detect thepresence or absence of vodka at a better than chancerate. A .10% blood-alcohol concentration was se-lected for the following reasons: (a) It representsthe lower limit of the range of blood-alcohol levelsthat most authorities define as constituting a stateof legal intoxication (American Medical Association,1959), and (b) it roughly approximates the blood-alcohol level reported to result in a maximum ofaggressive fantasies (McClelland et al., 1972).Through pretesting, it was discovered that about 1.3ml of absolute alcohol/kg of body weight were re-quired for the average subject to reach the desired.10% level within 1 hr. In order to reach this blood-alcohol level, for example, a ISO-pound man wasgiven 6 fluid ounces of 100-proof vodka in a total of36 ounces of vodka and tonic mix.

Once the assistant had determined the volume ofliquid appropriate for a particular subject, he pro-ceeded to prepare the drinks at the bar, while thesubject observed. In the expect-alcohol conditions,the assistant poured about one sixth of the totalvolume required from the vodka battle into a gradu-ated beaker, appearing to measure it carefully, andthen distributed the liquid equally into three largetumblers. This procedure wa-s then repeated usingfive times the volume from the chilled tonic bottles.A squirt of lime juice was added to each tumbler.In the expect-tonic conditions, the procedure was thesame except that the total volume of liquid was takenfrom the tonic bottles only. In order to facilitateaccurate preparation of the beverages and to enhancethe credibility of the deception conditions, all of thebeverages used were premixed and placed into the"legitimate" bottles prior to the subject's arrival.Thus, if a subject was actually to receive alcohol,all the bottles he saw contained a mixture of vodkaand tonic, whereas for subjects in the tonic condi-tion, both the vodka and tonic bottles containedonly tonic. Tonic used in the vodka bottles had beenpreviously decarbonated to make it appear more likevodka.

After the drinks were prepared and placed beforethe subject, the assistant instructed him that a maxi-mum of 30 min was allowed for consumption of allof the drinks. Then, after the subject began drinking,

the assistant removed the bottles from the roomunder the guise that he was going to prepare drinksfor the "other subject." Within several minutes theassistant returned to administer a battery of psycho-logical tests to the subject while he continued todrink. The tests given were selected because of thepotential mediating effects of certain personality fac-tors on the alcohol-aggression relation. The testsadministered included the California F scale, Rotter'sInternal-External Scale (1966), a social desirabilityscale (Crowne & Marlowe, 1960), Eysenck's Person-ality Inventory (Eysenck & Eysenck, 1968), and theShipley Institute of Living Scale Vocabulary andAbstraction Test. Each subject was randomly as-signed to 1 of 12 different orderings of these 5 testsso as to balance any possible effects due to the se-quence in which the tests were presented. The assist-ant was present when the subject completed eachtest so that he could give the appropriate instructionsand could be sure that the drinks were being con-sumed at a reasonable pace. After completion of thetest battery (about 30 min) the assistant left theroom for the last time.

Provocation manipulation. At the end of the30-min period, the experimenter escorted the con-federate and the subject to another room. They wereseated beside each other at a table containing twomirror-tracing apparatuses. Although this task wasdescribed to the subjects as a reliable indicator ofmotor and mental coordination, the actual purpose ofthe task was to provide a context for the provocationmanipulation to be described below. The experi-menter told the subjects that the task consisted ofusing a pencil to trace the shape of a star in thenarrow lane formed by a double-contoured starprinted on a sheet of paper placed in the apparatus.The difficulty of the task was enhanced by the factthat the apparatus screened the target figure fromdirect view, so that the star was visible only in amirror placed vertically beyond the figure. Any pencilmark left on or outside the lines of the double-contoured star was described to the subjects as anerror (see Snoddy, 1920, for details of the apparatusand task). It was then explained that a low numberof errors on the task showed good motor coordina-tion and superior intelligence and a poor performancesuggested the opposite. An ample supply of stardiagrams was placed in each apparatus. Following asignal, 90 sec were allowed for work on the task.After this difficult task was attempted, the subjectand confederate were instructed to place their ownstar diagrams in the space between them on thetable, where they were in full view of both parties.The experimenter then left the room for severalminutes, explaining that he was going to help hisassistant prepare for the next part of the experiment.

While the experimenter was absent, there was timefor a brief conversation between the confederate andthe subject. The entire verbal interaction was taperecorded for subsequent analysis. In the provocationconditions, the star diagram which the confederateproduced was a standard excellent one which he hadprepared beforehand. The subject's relatively poorertracing served as the basis for the confederate's in-

512 A. LANG, D. GOECKNER, V. ADESSO, AND G. MARLATT

suiting and belittling remarks, which constituted theprovocation manipulation. In a sarcastic and con-descending manner, the confederate asked if the sub-ject's attempt had been serious, if he had everbefore looked in a mirror or drawn anything, if hehad to cheat to stay in school, and generally ques-tioned the subject's intelligence. In the no-provoca-tion conditions the confederate's star diagram wasalso a standard one, but one which he had actuallyproduced on his first attempt with the apparatus. Inthis case, his comments to the subject were friendlyand suggested that, although neither of them haddone very well on the task, it was probably nothingto worry about.

Assessment of aggression. Following the interactionbetween the subject and the confederate, the experi-menter returned to the room to introduce the finalphase of the procedure. The participants were in-formed that for this part of the experiment theywould each be performing separate and differentfunctions. One was to serve only as a "teacher" andthe other assumed the role of "learner" for the dura-tion of the session. It was further explained that,because the learner would be receiving rather un-pleasant electric shocks from the teacher, a lotterywas the only fair way to determine who was assignedthe role of teacher. The lottery (a number-guessinggame) was rigged in such a way that the real sub-ject always was assigned the teacher role. The proce-dure that followed was described as a test of "prob-lem solving under stress" in which the learner wouldtry to solve a code while under the threat of electricshock for errors committed. The task was a revisedversion of that described by Buss (1961) and madeuse of a modified "aggression machine" (Geen &Stonner, 1971). As the learner, the confederate wastaken to an adjoining room where ostensibly hewould have the shock electrodes attached and receivefurther instructions.

Upon returning to the subject's room, the experi-menter explained the teacher's role. The subject wasseated before a control panel and given a sheet list-ing 40 three-consonant "code items." His first jobwas to administer the code items to the learner bypressing the three labeled buttons on his controlpanel which corresponded to the three consonants inthe code. He was then to place his hands on the tableand await his partner's response, represented on thecontrol panel by the flashing of either a green lightfor a correct response or a red light for a wrongone. As soon as either light came on, indicating anattempt to decode the item, the subject was to pressa separate button "as quickly as possible" to ac-knowledge receipt of the response. This rapid anduniform acknowledgement of responses was said tofacilitate accurate testing. Then, if the response hadbeen a correct one (indicated by the green light),the teacher was to go on to the next code item andrepeat the procedure. If, however, the learner's re-sponse had been wrong, it was the subject's role toadminister a shock to his partner before going on.The subject was told that the 10 numbered blackbuttons along the bottom of the control panel wereto be used for giving shocks. It was explained that

each of the buttons delivered a continuous shock ofa specific intensity for as long as it was depressed.The low-numbered buttons were said to deliver verymild shocks, the middle ones moderately intenseshocks, and the high-numbered buttons were allegedto deliver fairly strong, painful shocks, The subjectwas instructed that following every flash of the redlight he should choose an intensity button and pressit for as long as he wished in order to give negativefeedback to the learner. This procedure was continueduntil a specified criterion of three consecutive cor-rect responses had been reached.

During the task period the experimenter sat work-ing quietly in a section of the room well isolatedfrom the subject to minimize his contributions toevaluation apprehension. The arbitrary criterion ofthree consecutive correct responses was reached aftera prearranged schedule of 30 trials (10 correct and20 incorrect responses) had been presented to eachsubject. The presentation of the trials was actuallyconducted by an assistant who sat in the next roomat an apparatus that was wired to the subject'spanel and from which he controlled the flashing ofthe right/wrong lights. The assistant also recordedthe subject's reaction times, acknowledging the learn-er's response to each decoding trial. Finally, on eachshock trial, the subject's intended shock intensitiesand the duration of his button presses were moni-tored by the assistant. Thus, although neither theconfederate nor anyone else actually received anyshocks, the subject's behavioral intentions regardingthe shocks could be recorded.

Immediately after this task, the subjects weregiven another intoxilyzer test. Each subject was alsoasked to estimate the number of ounces of alcoholhe had consumed, if any. A thorough debriefing, in-cluding graded questions to assess the effectivenessof the deceptions and cover story, was then given.A complete explanation of the experiment's true pur-poses and the subject's real treatment conditions wasalso given. It was not until this point in the proce-dure that the experimenter and confederate becameaware of the subject's actual beverage or expectancyassignment and the assistant learned of the provoca-tion condition for that particular subject. All sub-jects were offered coffee after the debriefing, andthose who had actually received alcohol were de-tained until their intoxilyzer reading showed them tobe well below the legal limits of intoxication (< .05%blood-alcohol concentration). Before they departed,subjects were asked not to discuss the experimentwith anyone; they were then given their experi-mental credit and thanked for their cooperation.

RESULTSManipulation Checks

Expectancy. The effectiveness of the ex-pectation manipulation was tested by assess-ing the relative contributions of the expec-tancy and alcohol content conditions tosubjects' postexperimental estimates of the

EFFECTS OF ALCOHOL ON AGGRESSION 513

number of ounces of alcohol they had con-sumed. Because these data were collectedafter the provocation manipulation, this factorwas also included in the 2 X 2 X 2 fixed ef-fects analysis of variance. The analysis metthe assumption of homogeneity of variance,^rafl3c(8, 11 = 5.56, p > .05, and revealed ahighly significant effect for the expectancyfactor, F(l, 88) = 113.98, p< .001. Thus,regardless of the other treatment conditions,subjects who expected alcohol later estimatedthat they had consumed considerably morealcohol (X = 4.58 ounces) than those who hadexpected only tonic (X = 1.2S ounces). Itwas also found that the actual beverage ad-ministered had a significant effect on esti-mates, F(l, 88) = 19.39, p < .001, with thosewho had actually received alcohol estimatinga higher alcoholic content (X = 3.60 ounces)than those who had consumed only tonic(X = 2.23 ounces). No effects due to provo-vation or any interaction were significant.

The proportion of variance accounted forby each of the significant main effects wascomputed by Hays' (1963) omega-squaredstatistic. Results showed that expectancy wasresponsible for nearly half of the variance(49.9%), whereas alcohol content accountedfor only 8.1%.

Alcohol. The efficacy of the alcohol manipu-lation was examined by means of the in-toxilyzer test administered immediately afterthe final dependent measure of aggression.Results showed that among those who hadreceived alcohol, the average blood-alcoholcontent was .097%; none of the subjects inthe tonic conditions received a reading otherthan 0. A 2 X 2 fixed effects analysis of vari-ance was performed for only those subjectsreceiving alcohol in order to explore possibledifferences in blood-alcohol levels across theexpectancy and provocation conditions. Thisanalysis" met the assumption of homogeneityof variance, Fmax(4, 11) = 1.97, p > .OS, andshowed no significant effects for these vari-ables. The average levels across the four cellsvaried only from .096% to .098%.

Provocation. The provocation manipulationwas assessed by analyzing two independentjudges' ratings of the confederate's aggressive-ness toward subjects during their taped con-versation. These ratings, which were highly

reliable across judges (r = .92) showed thatwhen subjects were in the provocation condi-tion, the mean score for the confederate'saggression was 2.26 on a scale from 0 to 5.In contrast, every aggression rating exceptone in the no provocation condition was 0.The possible confounding of the confederate'sprovoking behavior with subjects' exposure tothe expectancy and alcohol treatments wasexamined by means of a 2 X 2 fixed analysisof variance for those variables with the taperatings as the dependent measure. This analy-sis met the assumptions of homogeneity ofvariance, Fmax(4, 11) = 1.82, p > .05, andshowed no significant differences.

Psychological TestsOf all the psychological measures employed,

only the Neuroticism scale of the EysenckPersonality Inventory was significantly af-fected by the expectancy and alcohol manipu-lations. The valid, FmRX(4, 23) = 1.88, p >.05, analysis of variance for these two factorsrevealed that alcohol affected Neuroticismscores, F(l, 92) = 14.12, p < .001, such thatsubjects who had consumed alcohol showedhigher scores (X = 12.88) than those whohad not (X = 9.SO).

Psychomotor Reaction TimeSubjects' latency to respond following the

confederate's signal on each trial of the de-coding task was used to determine the effectsof the experimental treatments on reactiontime. Means and standard deviations for thisvariable are presented in Table 1. The analy-sis of variance employed to assess possibleeffects of the three treatment conditions wasfound to be valid, Fmax(8, 11) = 6.S9, p >.05. Results indicated that the actual contentof the drinks consumed was a highly signifi-cant determinant, F(l, 88) = 15.43, p <.001, of response latency. Those subjects whohad actually consumed alcohol showed aslower reaction time (X = 1.23 sec) thanthose who had not (X = 1.07 sec). It wasalso found that provocation had an effect onresponse latency, F(l, 88) = 6.22, p < .05,such that provoked subjects reacted moreslowly (X = 1.20 sec) than those who werenot provoked (X = 1.10). Finally, the within-subjects variable of average latency on the 20

514 A. LANG, D. GOECKNER, V. ADESSO, AND G. MARLATT

shock trials versus average latency on the10 no shock trials also produced a significanteffect, F(l, 88) = 81.74, p < .001, with la-tencies being longer (X = 1.24 sec) on themore novel no shock trials than on the shocktrials (X = 1.10 sec). Neither the expectancyvariable nor any of the interactions in thisanalysis achieved significance.

Verbal AggressionRatings on each subject's verbal aggression

during his taped conversation with the con-federate were made by two independentjudges. A test of reliability showed high inter-rater agreement (r=.96). Using this mea-sure, it was found that subjects in the pro-voked conditions had an average verbal ag-gression score of 1.15 on a scale of 0 to 5.Nonprovoked subjects, on the other hand, allreceived 0 ratings, with the exception of oneindividual whose average score was .50. Toexplore further the effects of alcohol and ex-pectancy on verbal aggression in this situ-ation, a 2 X 2 analysis of variance was em-ployed, using only the data from provokedsubjects. This analysis met the assumptionsof homogeneity of variance, JFmax(4, 11) =2.50, p > .05, but failed to yield any signifi-cant results. Thus, it appeared that thisdependent measure was either insensitive toany effects other than provocation, or that nosuch effects existed.

Direct AggressionShock intensity. The average shock inten-

sity that subjects selected for delivery to the

TABLE 1MEANS AND STANDARD DEVIATIONS TOE

RESPONSE LATENCIES (IN SEC)

Beveragecondition'

Provoked Not provokedCondi-

Expect Expect Expect Expect tionalcohol tonic alcohol tonic mean

Condi-tion

Receive alcoholM 1.270 1.318 1.186 1.161 1.234SD .210 .267 .230 .251 .225

Receive tonicM 1.165 1.054 1.017 1.025 1.065SD .195 .176 .095 .134 .160

Condition meanM 1.217 1.186 1.101 1.093 1.149SD .198 .224 .152 .190 .202

• N = 96.

TABLE 2MEANS AND STANDARD DEVIATIONS FOR

SHOCK INTENSITY

Provoked Not provokedBeverageconditions

Expect Expect Expect Expectalcohol tonic alcohol tonic

Condi-tion

mean

Receive alcoholMSD

Receive tonicMSD

Condition meanMSD

5.06 3.80 4.59 2.90 4.092.65 2.53 1.95 1.29 2.26

5.00 3.59 4.80 2.70 4.022.32 1.59 1.90 1.29 1.99

5.03 3.70 4.70 2.80 4.062.43 2.08 1.88 1.26 2.06

•N =96.

confederate on shock trials of the decodingtask was the main measure of aggression. Con-dition means and standard deviations forshock intensities over all 20 shock trials arepresented in Table 2. The assumptions ofhomogeneity of variance were met by theanalysis of variance, Fmnx(8, 11) = 4.23, p >.05. Results showed a highly significant effectfor the expectancy factor for this measure,F(l, 88) = 15.43 p < .001. Thus, subjectswho thought that they had consumed alcoholgave considerably more intense shocks (X =4.87) than those who thought they had con-sumed only tonic (X = 3.25), regardless ofthe actual content of their drinks. No signifi-cant results were obtained for the actual al-cohol content, provocation, or any interactionof factors. The provocation manipulation did,however, achieve trend-level significance, F(l,88) = 3.35, p < .10, with provoked subjectsdelivering more intense shocks (X = 4.37)than nonprovoked subjects (X = 3.75).

Figure 1 presents the average shock inten-sities delivered by all 96 subjects for eachblock of five shock trials. Examination of thefigure reveals that intensity increased overblocks of trials. Such an effect is common inresearch with this apparatus (cf. Bennett, etal., 1969) and is perhaps attributable to sub-jects' growing frustration across trials. In thepresent study this effect was found to behighly significant, F(3, 264) = 23.20, p <.001, but did not interact significantly withany other factors.

Shock duration. In conjunction with everyshock intensity selection, subjects also deter-

EFFECTS OF ALCOHOL ON AGGRESSION 515

• • Expect Alcohol / Receive Alcoholo——o Expect Alcohol/Receive Tonic•—• Expect Tonic / Receive Alcoholo o Expect Tonic / Receive Tonic

o-

PROVOKED

10 15 20 5SHOCK TRIAL

10 15 20

FIGURE 1. Mean shock intensity delivered by provoked and nonprovokedsubjects (based on intensities for each subject averaged over blocks of 5shock trials).

mined their intended shock duration by thelength of time they held the shock buttondown. Unfortunately, the raw data for thisresponse duration measure of aggression didnot conform to the analysis of variance as-sumption of homogeneity of variance. It waspossible, however, to achieve test validity byusing a logarithmic to base ten transformationon the reciprocals of duration (Kirk, 1968).This resulted in Fmax(8, 11) = S.44, p > .05.The analysis revealed a significant effect forexpectancy F(l, 88) = 4.13, p < .05, con-sistent with that of the shock intensity de-pendent measure. Subjects who thought theyhad received alcohol gave longer shocks (X =1.22 sec) than those who thought they hadconsumed only tonic (X = 1.11 sec), regard-less of the actual content of the drinks. Alco-hol content, provocation, and all interactionsfailed to attain conventional significancelevels, although, as with the intensity mea-sure, provocation showed a borderline signifi-cance level, F(l, 88) = 3.52, p < .10. Asexpected, provoked subjects' shock durationswere longer (X = 1.46 sec) than those of un-provoked subjects (X = .87 sec) with a highlevel of variability being evident.

Correlation Between Dependent MeasuresAside from the strong overall correlation

between shock intensity and shock duration

(r — .53, p < .001), there were very few sig-nificant results among the many overall, con-dition, and cell correlations performed on thedependent measures. One rather strong andconsistent relation did occur, however, whenconsidering only the 48 provoked subjects.For this group it was found that F-scalescores were positively correlated with aggres-sion for both the shock intensity (r = .38,p < .01) and shock duration (r = .40, p <.01) measures. This relation did not appearin the unprovoked subjects. No other strong,consistent correlations occurred in these data.

DebriefingDebriefing interviews revealed that four

subjects were sufficiently aware of the decep-tions of the experiment that they could ver-balize the hypothesis and/or were sure thattheir partners had not received shocks. Thesewere distributed across conditions as follows:two in the expect tonic-receive tonic/pro-voked condition and one each in the expecttonic-receive tonic/not provoked and the ex-pect alcohol-receive alcohol/provoked condi-tions. These subjects, all of whom had eitherparticipated in similar experiments before orhad read about them, were eliminated fromthe study and replaced without regard fortheir data. A post hoc examination of theirdata, however, revealed that with the excep-

516 A. LANG, D. GOECKNER, V. ADESSO, AND G. MARLATT

tion of the individual in the expect tonic-receive tonic/not provoked condition, all ofthese subjects were substantially slower inreacting and less aggressive than the averagesubject in the same condition.

DISCUSSIONThe most important finding of the present

study was that differences in the level ofaggression observed were determined largelyby subjects' expectations or belief about thecontent of the beverage they had consumed.Those who thought their drinks containedalcohol, regardless of the actual alcoholiccontent, gave more intense and longer dura-tion shocks to their partners than those whobelieved they had consumed only nonalco-holic drinks. Another important outcome ofthis study was the failure to uncover anyinteractions between the expectancy, alcohol,and provocation factors. This result suggestedthat the first two factors did not influencesensitivity to the third in such a way thataggressiveness was affected.

Taken together, the above findings stronglychallenge the energizing and the disinhibitiontheories, both of which hold that it is somephysiological effect of alcohol that accountsfor a major portion of the positive correlationbetween drinking and aggression. Based onthe consistent and powerful results from thedirect, objective measures of aggression in thepresent study, the most reasonable conclusionis that expectancy factors play a major rolein the drinking-aggression relation. Only theverbal aggression measure, which confirmedthe provocation hypothesis but yielded nosignificant findings for the expectancy oralcohol content factors, did not support sucha conclusion. This apparent inconsistency,however, may be a result of the possibilitythat the subjectively rated audio tapes ofsubject-confederate conversations representeda relatively insensitive measure of aggression.

One potentially important point that shouldbe made about the present study is that itsmeasures of aggression were essentially mea-sures of "relative strength of aggression,"rather than "decision to aggress," becausethere was considerable experimental pressureon the subjects to administer at least low-intensity, short-duration shocks. There is,

however, no strong reason to believe that evena more personal, "decision to aggress" de-pendent measure would have resulted in sub-stantially different findings.

It might be argued that the results of thisstudy, especially as they relate to expectancy,were due primarily to experimental demandcharacteristics rather than to the manipu-lated variables. Several factors argue againstthis possibility. First, if it is maintained thatsubjects acted more aggressively as a co-operative response to the expect-alcohol in-structional set, this only serves to confirm thehypothesis that there is a widely held beliefand set of behavioral norms that dictate thatdrinking should lead to greater tendenciestoward aggressive behavior. Furthermore, thisargument does not explain the absence of anactual physiological effect of alcohol on ag-gression. A second counterargument to thedemand characteristics position can be de-veloped from the results of the reaction timemeasure. On this measure it was found thatthe actual alcohol content of the beveragesconsumed and not subjects' expectationsabout them accounted for the major finding(significantly slower responding occurred withsubjects who had received alcohol). It isdoubtful that any subject acting in accor-dance with demand characteristics would fol-low expectations about drinking and aggres-sion while behaving contrary to expectationsabout how drinking affects reaction time.Finally, the data of the experimentally"aware" subjects who were replaced in thestudy did not support demand arguments.

Three other results are deserving of somediscussion. First, it was found that subjects'reaction time was affected by the provocationmanipulation in such a way that insultedsubjects responded more slowly than non-insulted subjects. One possible way to accountfor this difference is to assert that provokedsubjects had a decreased interest or reducedmotivation to cooperate in the experiment. Itmight also be that the insulted individualswere preoccupied with the aversive experi-ence they had just been through and did notattend as carefully to the reaction time task.Another finding of some interest in this studywas the positive correlation between F-scalescores and aggressiveness in provoked sub-

EFFECTS OF ALCOHOL ON AGGRESSION 517

jects. This result seems to suggest that highlyauthoritarian subjects were especially likelyto act out aggressively when provoked—afinding worthy of further experimental in-vestigation.-The third finding of some interestwas that subjects who received alcohol scoredsignificantly higher on the Neuroticism scaleof the Eysenck Personality Inventory com-pared with subjects who did not receive alco-hol. This result may be accounted for by thefact that this scale contains items that tapemotional arousal (or autonomic reactivity),and the effects of alcohol may have promptedthese subjects to endorse more of these items.

Several reasons can be given for the dis-crepancy between the findings of the presentstudy and earlier research on drinking andaggression (Bennett et al., 1969; Boyatzis,1974; Shuntich & Taylor, 1972). The in-adequacy of the expectation and placebomanipulations of these former studies is onemajor explanatory factor alluded to earlier.In addition, a methodological difference be-tween the Shuntich and Taylor (1972) ex-periment and the present study concerns thecircumstances under which aggression wasmeasured. The former study employed a pro-cedure in which all subjects expected andreceived retaliatory shocks from their part-ners for each shock they delivered. This pro-cedure obviously resulted in a physically aver-sive and provoking situation for all subjects,whereas the present study tested for provoca-tion effects by the independent manipulationof a verbal insult. Perhaps these differencesin provocation alone account for the differingresults obtained, but a more specific explana-tion is that the subjects who received alco-holic drinks in the Shuntich and Taylorstudy had their threshold for pain increasedby the physiological effects of alcohol. A con-sequence of this effect, then, might have beenthat these subjects were relatively less de-terred by the repeated retaliations of theirpartner and, because of this reduced sensitiv-ity to shock, tended to deliver more intenseshocks themselves.

Another difference between the presentstudy and previous research concerns the typeand quantity of alcohol used. This experimentutilized intoxicating quantities of vodka,whereas most previous research has generally

employed lesser quantities of bourbon orother liquors with higher congener content.Further studies are needed to determine thesignificance of such variations. Finally, thepresent study was the only one to sampleexclusively from a population of heavy drink-ers; other studies used a random sample ofsocial drinkers.

The unique characteristics of the popula-tion studied in this experiment are importantto keep in mind when attempting to generalizefrom the results. It may be, for example,that only heavy social drinkers are particu-larly dependent on drinking as an excuse orpermit for engaging in such socially offensivebehaviors as aggression. Moderate or lightsocial drinkers, on the other hand, may usealcohol primarily as a "social lubricant"which enables them to behave more positivelyin relation to other people. It would be ofinterest to know whether the latter group re-acts differently to the situation described inthe present experiment.

The findings of the present study suggestthat increases in aggression following the con-sumption of alcohol may be the result of thedrinker's expectations concerning the effectsof alcohol. Expressions of aggression may beattributed to the effects of alcohol, thus re-ducing the individual's own responsibilitiesfor his actions. A related question concerningthe nature of the relation between drinkingand aggression may also be asked: What arethe effects of an instigation to aggression(provocation) on subsequent drinking be-havior? In a companion study (Marlatt,Kosturn, & Lang, in press), it was foundthat subjects who were insulted by a con-federate, in a procedure similar to the onedescribed in this paper, consumed significantlymore alcohol than provoked subjects whowere given the opportunity to counteraggressagainst the confederate. From the results ofboth studies, we conclude that (a) subjectswho believe they have consumed alcohol willbehave in a more aggressive manner, and (b)subjects who have been provoked to angerand who have been deprived of the oppor-tunity to behave aggressively will consumemore alcohol. One implication for treatmentintervention that arises from these findings isthe need to investigate the effects of training

518 A. LANG, D. GOECKNER, V. ADESSO, AND G. MAELATT

problem drinkers in socially acceptable alter-native methods of expressing anger (such asassertive training). The result might be adecrease in both aggression and drinking.

REFERENCESAmerican Medical Association, Committee on Medi-

colegal Problems. Chemical tests for intoxication,manual. Chicago: American Medical Association,19S9.

Barry, H., Ill, Koepfer, E., & Lutch, J. Learningto discriminate between alcohol and nondrug con-ditions. Psychological Reports, 1965, 16, 1072.

Bennett, R. M., Buss, A. H., & Carpenter, J. A.Alcohol and human physical aggression. QuarterlyJournal of Studies on Alcohol, 1969, 30, 870-876.

Boyatzis, R. E. The effect of alcohol consumption onthe aggressive behavior of men. Quarterly Journalof Studies on Alcohol, 1974, 35, 959-972.

Buss, A. H. The psychology of aggression. New York:Wiley, 1961.

Cahalan, D., Cisin, I. H., and Crossley, H. M.American drinking practices: A national study ofdrinking behavior and attitudes (Monograph No.6). New Brunswick, N.J.: Rutgers Center of Alco-hol Studies, 1969.

Cameron, N. Personality development and psycho-pathology: A dynamic approach. Boston:Houghton Mifflin, 1963.

Cappell, H., & Herman, C. P. Alcohol and tensionreduction: A review. Quarterly Journal of Studieson Alcohol, 1972, 33, 33-64.

Crowne, D. P., & Marlowe, D. A. new scale of socialdesirability independent of psychopathology. Jour-nal of Consulting Psychology, 1960, 24, 349-354.

Eysenck, H. J., & Eysenck, B. G. Manual for theBysenck Personality Inventory. San Diego, Calif.:Educational and Industrial Testing Service, 1968.

Geen, R., & Stonner, D. An extended apparatus formeasuring aggression in humans. Behavior Re-search Methods and Instrumentation, 1971, 3,197-198.

Hays, W. L. Statistics for psychologists. New York:Holt, Rinehart & Winston, 1963.

Kalin, R., McClelland, D. C., & Kahn, M. Theeffects of male social drinking upon fantasy. Jour-nal of Personality and Social Psychology, 1965, 1,441-452.

Kastl, A. Changes in ego-functioning under alcohol.Quarterly Journal of Studies on Alcohol, 1969, 30,371-382.

Kirk, R. E. Experimental design: Procedures forthe behavioral sciences. Belmont, Calif.: Brooks-Cole, 1968.

Marlatt, G. A., Demming, B., & Reid, J. B. Loss ofcontrol drinking in alcoholics: An experimental

analogue. Journal of Abnormal Psychology, 1973,81, 233-241.

Marlatt, G. A., Kosturn, C. F., & Lang, A. R. Pro-vocation to anger and opportunity for retaliationas determinants of alcohol consumption in socialdrinkers. Journal of Abnormal Psychology, inpress.

McClelland, D. C., Davis, W., Kalin, R., & Wanner,E. The drinking man. New York: Free Press,1972.

McKay, J. Problem drinking among juvenile de-linquents. Crime and Delinquency, 1963, 9, 29-38.

Mendelson, J., LaDou, J., & Solomon, P. Experi-mentally induced chronic intoxication and with-drawal in alcoholics. III. Quarterly Journal ofStudies on Alcohol, Supplement No. 2, 1964, 40-52.

Rotter, J. B. Generalized expectancies for internalversus external control of reinforcement. Psycho-logical Monographs, 1966, 80 (1, Whole No. 609).

Selzer, M. L. Alcoholism, mental illness and stressin 96 drivers causing fatal accidents. BehavioralSciences, 1969, 14, 1-10.

Selzer, M. L., Payne, C., Westervelt, F., & Quinn, J.Automobile accidents as an expression of psycho-pathology in an alcoholic population. QuarterlyJournal of Studies on Alcohol, 1970, 28, 508.

Shuntich, R. J., & Taylor, S. P. The effects of alcoholon human physical aggression. Journal of Experi-mental Research in Personality, 1972, 6, 34-38.

Shupe, L. M. Alcohol and crime: A study of theurine alcohol concentration found in 882 personsarrested during or immediately after the commis-sion of a felony. Journal of Criminal Law, Crimi-nology and Police, 1954, 44, 661-664.

Sncddy, G. S. An experimental analysis of a case oftrial and error learning in the human subject.Psychological Monographs, 1920, 28, 124.

Sobell, M. B., & Sobell, L. C. Individualized behaviortherapy for alcoholics. Behavior Therapy, 1973, 4,49-72.

Tamerin, J., & Mendelson, J. The psychodynamicsof chronic inebriation: Observations of alcoholicsduring the process of drinking in an experimentalsetting. American Journal of Psychiatry, 1969, 125,886-889.

Tucker, I. F. Adjustment, models and mechanisms.New York: Academic Press, 1970.

Williams, A. Social drinking, anxiety, and depression.Journal of Personality and Social Psychology, 1966,3, 689-693.

Wolfgang, M. E., & Strohm, R. B. The relationshipbetween alcohol and criminal homicide. QuarterlyJournal of Studies on Alcohol, 1956, 17, 411-425.

(Received December 6, 1974; revision receivedMarch 13, 1975)