Journal of Abnormal Psychology1984, Vol 93, No 4, 355-372

Copyright 1984 by theAmerican Psychological Association. Inc

Observational Conditioning of Snake Fear in Rhesus MonkeysSusan Mineka, Mark Davidson, Michael Cook, and Richard Keir

University of Wisconsin-Madison

For years theorists have hypothesized on the basis of meagre evidence thatobservational conditioning is involved in the origins of many human andnonhuman primates* fears and phobias The present experiments provide strongsupport for this hypothesis by demonstrating observational conditioning of snakefear in rhesus monkeys Experiment l demonstrated the usefulness of a newindex of snake fear in rhesus monkeys and, using this new measure, alsodemonstrated that young monkeys raised by parents who have a fear of snakesdo not acquire this fear in the absence of any specific experience with snakes InExperiment 2, however, five out of six adolescent/young-adult rhesus monkeysdid acquire an intense and persistent fear of snakes as a result of observing theirwild-reared parents behave fearfully in the presence of real, toy, and model snakesfor a short period of time. The fear was not context specific and showed nosignificant signs of diminution at 3-month follow-up Implications of the presentresults for current theories of the origins of human fears and phobias arediscussed.

In recent years many theorists have pro-posed that vicarious classical conditioningprobably accounts for the origins of a greaterproportion of human fears and phobias thandoes direct classical conditioning (e.g., Ban-dura, 1969; Marks, 1969, 1977; Rachman,1978). Thus they have argued that simplyobserving someone behave fearfully in thepresence of a neutral stimulus may often besufficient for the observer to acquire an in-tense fear of that stimulus. Given the impor-tance that has been placed on the role ofobservational conditioning in the origins offears and phobias, it is unfortunate that "theevidence in support of vicarious acquisitionof fear in humans is indirect and largelyanecdotal" (Rachman, 1978, p. 190). This isperhaps in part because of ethical constraintsthat limit study of the conditions that resultin the induction of long-lasting and intensefears in human subjects.

Of the studies that have been conductedon vicarious fear conditioning in human sub-

This research was supported by grants from the Uni-versity of Wisconsin Graduate School and by GrantsBNS-7823612 and BNS-8216141 from the National Sci-ence Foundation.

Requests for reprints should be sent to Susan Mineka,Department of ftychology, 1202 West Johnson Street,University of Wisconsin, Madison, Wisconsin 53706

jects, all appear to have been single-sessionlaboratory experiments involving the condi-tioning of autonomic responses such as gal-vanic skin response and heart rate to neutralstimuli as a result of observational experiences(e.g., Bandura & Rosenthal, 1966; Berger,1962; Brown, 1974; Craig and Lowery, 1969;Hygge, 1976; Kravetz, 1974). These studiesclearly have extended our knowledge of factorsthat influence vicarious classical conditioningof autonomic responses. However, there areat least four major reasons why these studiesare of limited usefulness m furthering under-standing of the role that vicarious condition-ing actually plays in the origins of real humanfears and phobias. The first and most obviouslimitation stems from the observation madeabove: They have all involved conditioningof autonomic components of fear, which areknown to not always correlate highly withother components of fear (e.g., Hodgson &Rachman, 1974; Lang, 1968, 1971; Mineka,1979). Therefore we do not know how wellthe other components of fear, which areprominent aspects of real human fears andphobias, can be conditioned observationally.

The second and third limitations stemfrom design problems of the experimentsthemselves. In none of these studies were thesubjects ever tested for their fears in situationsother than the one in which their fear had

355

356 MINEKA, DAVIDSON, COOK, AND KEIR

been acquired. Thus the context specificityof these fears acquired vicariously in thelaboratory is unknown, although the naturallyoccurring fears that such studies are attempt-ing to model and understand are clearly notcontext specific. (See Bouton & Bolles, 1979;Konorski, 1967; Rescorla & Wagner, 1972,for discussions of context effects.) The thirdlimitation is that most of these studies havetaken place in a single laboratory session andhave not involved any follow-up to determinethe persistence of the fears. In the few studiesthat have examined extinction of vicariouslyconditioned fears, extinction has been quiterapid even though many naturally occurringhuman fears, and certainly phobias, are no-toriously resistant to extinction.

The fourth limitation of the studies con-ducted to date stems from the fact that allexcept one (Hygge & Ohman, 1978) of thestudies employed highly arbitrary neutralstimuli as the conditioned stimulus (CS). YetMarks (1969) and Sehgman (1971), amongothers, have pointed out that the objects ofpeople's fears and phobias usually do notcome from a random arbitrary group ofobjects, such as electric outlets, stoves, ham-mers, bicycles, and so forth, even thoughsuch stimuli may frequently be associatedwith trauma and/or with parents' verbal in-structions to avoid these objects. These ob-servations, in combination with the work ofOhman and his colleagues in recent years(Ohman, Ericksson, & Olofsson, 1975; Oh-man, Fredrikson, Hugdahl, & Rimino, 1976;Ohman, Fredrikson, & Hugdahl, 1978a,1978b), suggest that there are qualitative and/or quantitative differences in the characteris-tics of conditioning to potentially phobic asopposed to nonphobic objects. Thereforemore studies of vicarious conditioning shouldbe conducted that use potentially phobicstimuli as CSs. (The study that has been doneshares the first three limitations discussedabove; Hygge & Ohman, 1978.)

Over the years a fruitful interchange hasdeveloped between animal researchers andhuman researchers interested in the originsof, and the cures for, fears and phobias. Inparticular, animal research has had a largeimpact on both theory and research on thetopic of human fears and phobias (e.g., Ban-

dura, 1969; Baum, 1970; Eysenck & Rach-man, 1965; Marks, 1969; Mineka, 1979, inpress; Rachman, 1978; Riccio & Silvestri,1973; Wolpe, 1958). It is therefore unfortu-nate that the paucity of evidence on vicariousconditioning of fear in humans is paralleledin the animal literature, where there are alsoonly a handful of studies. Probably the twobest studies are those of Stephenson (1967)and Crooks (1967, cited in Bandura, 1969),who both demonstrated vicarious condition-ing of fear in rhesus monkeys. However, thesestudies share the four limitations discussedabove for the human literature in that (a)they used only one measure of fear (behavioralavoidance), (b) they had no test for contextspecificity or for retention of the fear, and (c)they used arbitrary stimuli as CSs. The focusof the present research was to find moreclear-cut and convincing evidence for thevicarious acquisition of an intense and long-lasting fear in rhesus monkey subjects andthereby to provide a more convincing animalmodel for the vicarious acquisition of phobicfears in humans.

Snakes were chosen as the potentially pho-bic stimuli for the present research becauseit has long been known that many adultprimate species exhibit an intense fear ofsnakes. There has been a long-standing con-troversy over whether this fear is innate orspontaneous (e.g., Hebb, 1946; Masserman& Pechtel, 1953; Morris & Morris, 1965), asopposed to being based on some form oflearning (e.g., Haselrud, 1938; Schiller, 1952;Yerkes & Yerkes, 1936). Probably the mostconvincing evidence to date that the fearmust indeed be learned stems from the ob-servations of Joslin, Fletcher, and Emlen(1964), who made the first systematic com-parisons of the response of wild-reared andlaboratory-reared rhesus monkeys to a livesnake and other snakelike objects. They foundthat only the wild-reared monkeys consistentlyexhibited a strong fear of snakes, and theysuggested that these monkeys had probablylearned the fear when they were in the wilda number of years prior to the study. Theyalso argued that it was unlikely that all ofthese monkeys had had a direct traumaticconditioning experience with snakes andtherefore that observational conditioning

OBSERVATIONAL CONDITIONING OF SNAKE FEAR IN MONKEYS 357

rather than direct conditioning was the morelikely source of the fear. Some years laterthese results were replicated by Murray andKing (1973) with squirrel monkeys, and byMineka, Keir, and Price (1980) again withrhesus monkeys.

Mmeka et al. (1980) also demonstratedthat certain components of the fear are ex-traordinarily persistent even in the face ofseven repeated flooding sessions designed toreduce the fear. In particular, although therewere significant changes in behavioral avoid-ance of the snake (as measured by the latencyto reach for food on the far side of the snake),there were no significant changes in the be-havioral disturbance component of the fear,which is probably more akin to the autonomicand/or subjective components seen in hu-mans. The same pattern of results occurred6 months later when the monkeys were givena follow-up test and further flooding sessions(Mmeka & Keir, 1983). This pattern of resultsis analogous to findings in the literature onflooding of human phobias (e.g., Hodgson &Rachman, 1974; Rachman, 1978) and there-fore further substantiates the contention thatthese monkeys' fear of snakes has a phobic-like quality.

There is then good reason to suspect thatwild-reared monkeys acquired their fear ofsnakes vicariously and that the fear shares anumber of the characteristics of human pho-bias. However, as seen above, the evidencesupporting the proposition that a fear thisintense and persistentin a monkey or in ahumancould in fact be learned observa-tionally is meagre at best. The present exper-iments were designed to test this propositionas directly as possible in a laboratory setting.In the first experiment, the usefulness of anew index for measuring fear of snakes thathas not previously been used in this line ofresearch was explored. It was important tohave this new index both because the proce-dures most commonly used in the past arenot always well suited for young monkeyslike those being used here, and because itprovided a different context in which toassess the context specificity of the fear. Inthe second experiment, the facility with whichyoung monkeys acquire a fear of snakes byobserving their wild-reared parents' fearful

behavior in the presence of both real andmodel snakes was examined. Several indicesof fear were used, and the context specificityand retention of the fear were also examined.

Experiment 1The primary purpose of this experiment

was to devise a new test of snake fear and toexamine its relation to other indices of snakefear used previously. This was necessary be-cause of difficulties in adapting some of theyoung monkeys to be used in Experiment 2to the Wisconsin General Test Apparatus(WGTA) that has been used in most studiesin this area (e.g., Joshn et al., 1964; Minekaet al., 1980; Murray & King, 1973). A specialproblem in adapting young monkeys thatwere still living with parents and peers to thisapparatus is that each test session constitutesa brief separation, which results in agitatedbehavior (cf. Mineka & Suomi, 1978, for areview) not conducive to rapid responding(reaching for food) in the WGTA. Becauserapid responding is one primary index oflack of fear, it was necessary to devise analternative way to test for avoidance of snakesin these monkeys. The Sackett Self-SelectionCircus (Sackett, 1970), which is normallyused to test social preferences, was adaptedfor this purpose by using fear stimuli ratherthan social stimuli. As will be seen, in usingthis new procedure the results with wild-reared monkeys essentially duplicate the re-sults of fear testing in the WGTA, thusvalidating this as an index of fear.

A secondary purpose of the present exper-iment was to explore one alternative expla-nation suggested by Joslin et al. (1964) as towhy laboratory-reared monkeys do not showa fear of snakes. They suggested that therewas a slight possibility that the social envi-ronment of laboratory-reared monkeys hasbeen sufficiently aberrant that such a feardoes not develop, although it might developspontaneously in a normal social environ-ment. For example, perhaps some subtle de-velopmental experience that can only be fos-tered by wild-born parents is necessary toproduce a fear of snakes. To test this possi-bility a group of laboratory-reared monkeysliving with their wild-born parents need to

358 MINEKA, DAVIDSON, COOK, AND KEIR

be tested for their fear of snakes. Althoughthe ideal test would be one in which all ofthe monkeys were living in a naturalisticoutdoor environment that did not containsnakes, the closest approximation that canbe made in a laboratory setting is to testmonkeys that have been reared and are livingin a Nuclear Family Unit (M. Harlow, 1971).In these living units, 4 wild-born adult male-female pairs live in adjacent cages with theirlaboratory-born offspring. Small openingsfrom the cages allow the infant and adoles-cent/young-adult offspring (but not the full-grown adults) to enter a central cage wherethey have access to a variety of toys and caninteract directly with the offspring of theother parents. This environment has consis-tently been shown to produce the most "nor-mal" and socially sophisticated monkeys thatexist at the Wisconsin Primate Laboratory(Suomi, 1974). Thus it seems to provide agood first test of the hypothesis that thelaboratory-reared monkeys of the Joskn etal. (1964) and Mineka et al. (1980) studiesdo not show snake fear because of aberrationsin their social development.

MethodSubjects

Subjects were 7 wild-reared and 9 laboratory-rearedrhesus monkeys (Macaca mulatto.) living in one NuclearFamily Unit at the University of Wisconsin PrimateLaboratory The wild-reared adults were living m pairs,each confined to a 1.3 m X 1 3 m X 2 m living cage(One of the wild-reared females had a young infant attesting time, so she could not be used as a subject Thusthere were 7 wild-reared subjects instead of 8.) However,their laboratory-reared off-spring had access not only tothe cage of their parents, but also to a central 2.6 m X1.3 m X 2 m play area and to the living cages of theother 3 adult male-female pairs in the unit Thus theoffspring had access to their biological parents and siblings,to peers, and to other adults of both sexes (See M.Harlow, 1971, for a more complete description )

The 4 male and 3 female wild-reared monkeys rangedin estimated age from 19 to 28 years, with a mean of22.7 years. They had all been imported from India atleast 13 years prior to the study and had had at least 2years experience in the wild prior to capture Of the 9laboratory-reared monkeys, there were 4 sexually matureyoung adults (two males and two females, all approxi-mately 5-6 years old), 2 adolescent females (approximately3-3'A years of age), and 3 infants (two males and onefemale, 8-10 months of age). At the time of the experimenteach family consisted of one adult male-female pairliving with 1-3 offspring, and one sexually mature 5-year-old young adult

Apparatus

The following objects were used to determine eachsubject's reactions, (a) A young 2Vb-ft-long (75 cm) hveboa constrictor (Constrictor constrictor), approximately1.2S in (3 25 cm) in diameter, (b) a nonmoving, 20-inlong (56 cm), sinuous lifelike model of a snake, approx-imately '/i-in. (1.3 cm) in diameter and colored to imitatethe appearance of a grass snake, (c) a 24-in -long (61cm), sinuous, rubber, toy snake, approximately 1 in (2.6cm) in diameter, which jiggled slightly when moved bythe experimenter, (d) a 3-ft-long (91.5 cm), sinuous,black, rubber, electrical cord, approximately % in. (1 cm)in diameter, (e) a 3%-ft-long (114 cm), sinuous, yellowplastic-covered electrical cord, and (f) four neutral ob-jectsan unpainted square wood block, a red paintedtriangular block, a blue painted cross-shaped object, anda red painted round block The first two neutral objectswere used alternatively in the WGTA, and all four wereused in the Circus pretest.

Some of the testing was done in a Wisconsin GeneralTest Apparatus The WGTA consists of a cage for thesubject, a movable cage blind, a movable gray tray usedto present stimulus objects, and a one-way mirror behindwhich the experimenter sits (see H Harlow, 1949, for adetailed description) The objects were all presented inone of two uncovered 53 4 cm X 21.3 cm X 15 2 cmPlexiglas boxes, each with a small Plexiglas ledge at thetop of the back side (side away from the monkey) wherea food reward could be placed The Plexiglas box wasplaced on the movable gray tray

The second testing apparatus was the Sackett Self-Selection Circus (Sackett, 1970), which consists of ahexagonal center cage, surrounded by six rhomboidalouter cages The walls separating the center compartmentfrom the six outer cages were formed by guillotine-typedoors that could be hydraulically raised and lowered bycable Only four of the six outer cages were used in thepresent experiment as stimulus chambers, and for eachof these, both the outer walls and the guillotme-typedoors adjacent to the center cage were Plexiglas The sidewalls separating the six different outer cages from eachother were made of '/s in plywood A fifth outer cagewas used as a start compartment, and its guillotine-typedoor was made of plywood, preventing visual access tothe central compartment (The sixth outer cage was notused, its central wall was wooden as well.) Light wasprovided by a horizontal fluorescent fixture attached justoutside the outer Plexiglas wall of each outer cage Avideo camera was mounted directly above the apparatusand was positioned m such a way that the entire apparatuswas seen from above on a monitor in an adjacent room.An observer recorded the number of entries into andamount of time spent in each compartment.

Each of the four stimulus objects used in Circus testing(real, toy, and model snakes, and neutral unpainted woodobject) were placed inside a transport cage made of wireand Plexiglas The Plexiglas side of each transport cagefaced the outer Plexiglas wall of one of the four outercages, and so the stimuli were seen through two layers ofPlexiglas. (The real snake was also placed inside aPlexiglas box before being placed in the transport and sowas seen through three layers of Plexiglas.)

OBSERVATIONAL CONDITIONING OF SNAKE FEAR IN MONKEYS 359

ProcedureWisconsin General Test Apparatus Prior to the test

day, each subject was adapted to the apparatus andprocedure. Although some of the subjects (especiallyamong the wild-reared adults) had had previous WGTAexperience, they were all readapted to ensure that theirperformances would be comparable The adaptationtechnique was similar to that used in the general adap-tation procedure for the WGTA A food reward (eithera fruit loop, raisin, chocolate chip, or marshmaliow,depending on individual preference) was placed on theledge of the Plexiglas box An adaptation object (one ofthree different, colored, wood blocks) was placed in thebox The moving tray with the Plexiglas box on it wasadvanced toward the subject, and then the blind wasraised by a hydraulic mechanism, at which point anelectric timer was started The timer stopped when thesubject touched the food reward The subject was consid-ered to be adapted when it reached for and touched thefood reward within 10 s on 18 out of 20 trials Mostsubjects required at least several days of adaptation beforereaching criterion and some required many days, especiallythe subjects with no prior WGTA experience.

In the fear-testing session, the same procedure wasfollowed except that different test stimuli (including thesnake and snakelike objects) were used on each trial Inaddition, although the tuner was stopped when the subjecttouched the food reward, the trial continued until 60 shad elapsed It should be noted that with this procedurethe monkey was required to reach across the open boxcontaining the test object m order to obtain the rewardThe trial was terminated after 60 s even if the subjecthad not yet reached for the food

There were 22 trials, which included 2 presentationsof each of the five snakelike objects and 12 presentationsof the neutral objects. The first two trials were alwayswith two of the neutral wood objects (unpamted and redtriangular blocks) The remaining five snakelike objectswere presented in one of two randomly predeterminedorders that differed in whether the real or the toy snakewas presented first, (a) toy snake, yellow cord, modelsnake, black cord, and real snake, or (b) real snake,yellow cord, model snake, black cord, and toy snakeOne of the two wood objects was always presentedbetween each of the other objects, the two wood objectsbeing used alternately The stimub were then repeated inthe same order as fiist presented, starting with the twowood objects and then proceeding with a wood objectbetween each of the other test stimuli

The latency of the food-reaching response and thedisturbance behaviors that occurred during the 60-s tnalwere observed and recorded The disturbance behaviorswere scored every 20 s using a 1-0 modified frequencysystem (Sackett, 1978) The occurence of a particulardisturbance behavior was given a score of 1 with nofurther recording of that behavior during that 20-s intervalThere were 12 disturbance behaviors that had previouslybeen shown to occur in this situation (cf Mineka et al.,1980)' cage clutch (holding onto the side or the back ofthe cage), cage shake (obvious moving or shaking of thecage), eye aversion (rapidly looking away from the stim-ulus), fear grimace (stretching the lips over the gums,thus exposing the teeth), spasm/tic (a vigorous shakingor jerking of the hands or upper body), threat flips thrust

forward, ears retracted and flattened against the head),vocalization, fear withdraw (sudden retreat to the backof the cage), ear flap (ears flattened along the head butwithout the lips thrust forward as with the threat), lipsmack (lips repeatedly moving up and down, chatteringof teeth), piloerection (fur raised up on shoulders andtorso), and stare (prolonged fixed gaze into the box fromthe back half of the cage). Because each behavior couldbe scored from 0 to 3 times in a given 60-s trial, themaximum composite emotionality score for any subjecton any tnal was 36

An evaluation of the real snake's movement was madeupon each presentation because it was not feasible tocontrol for this variable The rating system was based ona scale of increasing degree of movement from 0 to 3,with 0 indicating no movement at all and 3 indicatingthat the snake was on the verge of coming out of thePlexiglas box The highest scored received on any trialwas 1 (slight movement), although the snake was com-pletely free to move about

The WGTA test was administered to all 7 wild-rearedadults and to the four 5-year-olds as well Adaptationproved to be extremely difficult for the 2 adolescents(they did not adapt in 20-30 sessions) and so was noteven attempted for the 3 infants

Sackett Self-Selection Circus On each Circus test allof the subjects first entered the center compartment ofthe apparatus from a transport cage After the opaqueouter wall of the side cage used as a start compartmenthad been put in place, they were allowed to enter thestart compartment, with its four opaque walls, from thecenter compartment and were detained there for 5 mm.During this time the four stimulus objects were all putin their appropriate positions outside the stimulus cham-bers Five minutes after entry into the start compartment,the door to the center compartment was opened andsimultaneously the four doors to the four outer cagestimulus compartments surrounding the center compart-ment were raised as well When the monkey exited fromthe start compartment (almost always instantaneously),the wooden wall to it was lowered to prevent reentry.The monkey was then free to enter and reenter any ofthe four side compartments or to stay m the centercompartment for the next 5 mm, at which point the trialterminated

Each subject was first adapted to the apparatus andprocedure by conducting neutral object tests duringwhich all four stimuli were the wooden blocks describedin the Apparatus section The purpose of the adaptationtrials was to familiarize the animal with the apparatusand to check for compartment preference Each subjectreceived from 3 to 6 adaptation tests until they met acriterion on three consecutive sessions of no preference,which was defined as spending from 10% to 35% of thetotal entry time in each compartment All of the subjectsmet this criterion, and in fact only 2 of the wild-rearedsubjects took more than three sessions to do so

Each subject was given four S-mm fear tests with thethree snake stimuli and the unpainted neutral objectfollowing the adaptation sessions The four stimuli wereeach placed behind each compartment once Number ofentries and total entry time for each compartment wererecorded. An animal was defined as having entered acompartment when its shoulders and head had passedthe line dividing the central compartment and one of the

360 MINEKA, DAVIDSON, COOK, AND KEIR

side compartments All of the subjects were tested in theCircus within 4 weeks after their WGTA test. The threeCircus adaptation tests occurred approximately 1 weekprior to the four fear tests, which were conducted over a10-day period with each test being separated by 2-3days.

In all statistical analyses, the rejection level was set atp < .05; all significant effects are reported.

Results

Wisconsin General Test Apparatus

The WGTA tests of the 7 Nuclear Familywild-reared adults produced the typical pat-tern of results that was previously reportedby Joslin et al. (1964) and Mineka et al.(1980). All 7 adults exhibited avoidance ofthe real and toy snakes by not responding atall within 60 s on either of the presentationsof either stimulus. In addition, 4 failed torespond to the model snake, and the meanlatency of the other 3 subjects was 43 s. Bycontrast, the 4 laboratory-reared subjects thatwere WGTA-trained, showed rapid respond-ing in the presence of all of the stimuli. Theseresults are illustrated in Figure 1.

To affirm that the differences in respondingof the two groups to the different objectswere statistically significant, a 2 X 6(Groups X Stimulus Objects) analysis of vari-ance (ANOVA) was performed. The analysisrevealed a highly significant main effect bothfor groups, f\l,9)= 12.48, p < .01, and forstimulus objects, F(5, 45) = 50.75, p < .001.There was also a highly significant Groups XStimulus Objects interaction, ^(5, 45) =19.91, p < .001. Further analysis revealedthat there was a significant simple main effectof stimulus object for the wild-reared group,F(5, 30) = 247.83, p < .001, but not for thelaboratory-reared group, F\5, 15) = 1.29.Duncan's post hoc comparisons (a = .01)revealed that the wild-reared group had longerlatencies to the real, toy, and model snakesthan to the yellow and black cords or theneutral object. In addition, the wild-rearedsubjects showed longer latencies than thelaboratory-reared subjects with the real, toy,and model snakes, but not with the yellowand black cords or the neutral object (Dun-can's, a = .01). (For the remainder of thearticle the yellow and black cords are consid-ered to be two of the three neutral objects

BEHAVIORAL AVOIDANCE OF SNAKES IN WGTA60

g g l WILD-REAREDI I LABORATORY-REARED

j L

/ * * / * / / /

Figure 1 Mean latency to reach for the food in theWisconsin General Test Apparatus in the presence of thesix different objects for the 7 wild-reared and 4 laboratory-reared monkeys of Experiment 1

because their results are always identical tothose with the other neutral object.)

To determine whether there were significantdifferences in emotionality or disturbancebehaviors exhibited by the two groups to thedifferent stimulus objects, another 2X6(Groups X Stimulus Objects) ANOVA wasperformed on disturbance behavior scores.The analysis revealed highly significant maineffects both for groups, F(l, 9) = 39.31, p ILUCD

366 MINEKA, DAVIDSON, COOK, AND KEIR

Table 1Individual Sackett Circus Data

Observer

AA32AA38AA82AC30AG39AG47

R

61.355 828 057 060.858 0

Pretest

T

28 542.545.355.356 0640

M

68.365 360.057.564 559 3

N

55 867 293.861346 852 3

R

0.045.013.311.00 8

23 3

Posttest

T

2.048.020 7166513440

M

9.340.829.736.054 453 7

N

193.355.8

136 7149.3137 4115.0

Three-month follow-up

R

0

031110

T

0No

0203325

M

12follow-up

1530530

N

208

19714115220

Note Table entries are time spent (in seconds) in each stimulus compartment Pretest and posttest values are averagesbecause each subject received more than one of each R = real snake, T = toy snake, M = model snake, N = neutralobject

test.4 A 3 X 4 (Tests X Stimulus Objects) AN-OVA revealed a significant main effect forstimulus object, F(3, 9) = 22.90, p < .001,and a significant Tests X Stimulus Objectsinteraction, F\6, 18) = 5.90, p < .01. Dun-can's post hoc comparisons revealed thatresponding to the real and toy snakes did notchange over sessions, but that responding tothe other two stimuli did change from thefirst test session to the last test session. Theobservers spent more time with model snakesand less time with the neutral stimulus duringthe first test session than during the last testsession. The amount of time spent with thesestimuli during the second test session wasintermediate and did not differ significantlyfrom the amount of time spent during eitherthe first or the last sessions.

For the 5 observers who learned, a test wasalso performed to determine whether therewere significant changes from baseline to3-month follow-up, or from acquisition to3-month follow-up. The average pretest per-formance (average of four sessions), the av-erage acquisition performance (average ofthree sessions), and the single follow-up ses-sion were included in a 3 X 4 (RepeatedMeasures X Stimulus Objects) ANOVA. Theanalysis revealed a significant main effect forstimulus object, FO, 12) = 19.32, p< .001,and a significant Stimulus Objects X RepeatedMeasures interaction, F(6, 24) = 13.28, p

B

0010

001100

0

0000

Note These scores reflect the sums of two trials with each stimulus R = real snake, T = toy snake, M = modelsnake; N = neutral object, Y = yellow cord, B = black corda Maximum score = 120 (60 s for each of two trials)

b Maximum score = 72 (36 s for each of two trials)

368 MINEKA, DAVIDSON, COOK, AND KEIR

Table 3Three Measures of Fear Retention at Three-month Follow-up for Experiment 2

MeasureCircus test (in s)

Real snakeToy snakeModel snakeNeutral objects

WGTA latency (in s)Real snakeToy snakeModel snakeNeutral objects

WGTA disturbancebehaviors(mean no)

Real snakeToy snakeModel snakeNeutral objects

Posttest

9.726.936 6

146 3

60 057.04494.6

10 2942.20.3

Three-monthfollow-up

8.415 622.0

143.6

60.052 445.8

59

8.66.62400

Note These data are only for the 5 subjects that learnedWGTA = Wisconsin General Test Apparatus

the posttest, although none had done so inthe pretest. The same general pattern ofresults existed at 3-month follow-up, with all5 observers showing disturbance behaviors tothe real and toy snakes, and 4 of the 5 to themodel snake as well. By contrast, disturbancebehaviors to the neutral stimuli occurred veryinfrequently during the posttest, and not atall during the 3-month follow-up.

Comparison of the Models' PretestPerformance With the Observers'Posttest Performance

One further question of interest is whetherthe fear acquired by the 5 adolescent/young-adult observers was as intense as that of theirparental models. Strictly speaking, these testscannot be performed because in two casesthe same animal served as model for twodifferent observers, and so the data are notstrictly independent. Nevertheless, as can beseen from Figures 3, 4, and 5, the resultsfrom the posttest of the observers were highlysimilar to those of the pretest of the models.As can be seen in Figure 5, the parentalmodels tended to show more total disturbancebehaviors with the real snake than did theobservers. However, as can be seen in Figure6, there were no apparent differences in thenumber of different fear behaviors shown by

the parental models and the observers wholearned in the presence of the real, toy, ormodel snakes.

Correlation of Model's Disturbance BehaviorDuring Conditioning With Observer's Fear

To assess whether there was any relationbetween the level of disturbance behaviorsshown by the parental models and the ob-server's level of acquired fear, Spearman cor-relations were computed. The total numberof disturbance behaviors shown by the modelsacross all the observational conditioning ses-sions was first correlated with the 6 observer'sdisturbance behavior on the WGTA posttest.When the correlations were computed sepa-rately for the real, toy, and model snakes,they were all highly positive, but only thecorrelation for the real snake was statisticallysignificant (real snake, rs = .90; toy snake,rs = .757; model snake, rs = .643). A furthercorrelation between the total number of dis-turbance behaviors exhibited by the models

FEAR BEHAVIORS IN WGTACO _cc 6O