GENERALIZED EXPECTANCIES FOR INTERPERSONAL TRUST

JULIAN B. ROTTKR 2

University of Connecticut

A long last it seems that many thoughtfulpeople have become aware that the strong-value conflicts in our society can no longer

be ignored and that radical change is necessary forsurvival. Many have also recognized that changein itself is not necessarily good and that changethat will produce a better society must be controlledor planned. The atom bomb, the Asian war, thecollege student, the hippie, the black revolutions,and the problems of pollution and overpopulationhave helped even politicians see that our problemshave become deeper and that slogans and mythswill no longer suffice. Where many psychologistshave individually understood the serious and debili-tating conflicts in society's values, most have pre-ferred to follow more traditional, easier, or saferlines of study rather than to probe deeply into theseculture conflicts and their consequences for person-ality development and social behavior. In recentyears, however, a continually increasing number ofsocial and personality psychologists have directedtheir energies toward investigations of our socialills. Tt is gratifying to observe the multiplica-tion of studies of such problems as alienation,trust, altruism, bystander intervention, vandalism,Machiavellianism, and moral development.

Tf our society is to be improved, it will not comeabout because one group or another has seized orobtained power, but rather because social plannersand people in power will have access to knowledgeabout how socially desirable traits or characteristicsare developed and maintained, and will make use

of that knowledge.

1 Presidential address to the Division of Personality andSocial Psychology, read at the meeting of the AmericanPsychological Association, Miami Beach, Florida, September1970. The author is indebted to Donald Stein for hisvaluable assistance in preparing this article.

The research reported in this article was supported bythe University of Connecticut Research Foundation (1965-1966 and 1967-1968) and the National Institute of MentalHealth (Grants No. 11455 and 17741).

2 Requests for reprints should be sent to Julian B.Rotter, Department of Psychology, University of Con-necticut, Storrs, Connecticut. 06268.

I f it ever held sway, the morality of the tencommandments now has little effect on public orprivate behavior. Religion as a mass method forthe development of moral behavior seems to havefailed. Political creeds have lost their appeal, and,for better or worse, it now seems to be up tosocial scientists to provide the basis for a bettersociety. Our society seems to be falling apart, andthough it may be true that something better willnaturally arise from the shambles, the chances thatsomething better will develop may be materiallyenhanced by the knowledge that the social scientistmay be able to provide. How successful we willbe in producing the knowledge and in guiding itsapplication is yet to be determined. The attemptis surely worth the effort.

Jn accord with these remarks, I would like tospeak this afternoon on a series of investigationsin the field of interpersonal trust conducted bymyself and a group of students. This work hasbeen carried out over the past seven years but stillrepresents only a beginning. Most of the researchhas concentrated on individual differences, and Jwould like also to discuss in more theoretical termsthe nature of such differences and how their studyhelps to clarify problems of broad social issues.

It is belaboring the obvious to discuss the signifi-cance of interpersonal trust in our society. Itseems clear that disarmament will not proceedwithout an increase in trust on one or both sidesof the iron curtain, as Osgood (1960) pointed out10 years ago. The entire fabric of our day-to-dayliving, of our social order, rests on trust—buyinggasoline, paying taxes, going to the dentist, flyingto a convention—almost all of our decisions involvetrusting someone else. The more complex the soci-ety, the greater the dependence on others. I f trustweakens, the social order collapses. The absence oftrust between government and the forces of lawand order and the groups seeking change in oursociety has reached alarming proportions, and theresults are likewise alarming.

I am aware that it makes no sense to try torebuild trust in the absence of some social change.

443

444 AM ICKICAN PSYCHOLOGIST

However, the attempt to decelerate what appearsto lie increasing distrust and to built a society inwhich people trust each other more may in itselfdemand changes in the behavior of individuals andgroups that constitute positive social change.

Interpersonal trust is defined here as an expect-ancy held by an individual or a group that theword, promise, verbal, or written statement ofanother individual or group can be relied on.I f such expectancies are generalized and constitutea relatively stable personality characteristic, theyshould be readily amenable to investigation. Thisdefinition clearly departs significantly from otheruses of the concept of basic trust which is used ina more general way, often synonymous with thehealthy personality or belief in the goodness ofothers.

I will not attempt here to review the extensivel i terature on trust and trustworthiness. Much ofthe recent research falls into three groups: studiesof the prisoner's dilemma or the two-person nonsumzero game recently reviewed by Vinacke (1969);studies of source credibility recently reviewed byGiffin (1967) ; and studies involving the choice ofn smaller immediate reward versus a promisedlarger delayed reward, reviewed by Mischel (1966).Most of these investigations have dealt with condi-tion differences or experimental manipulations andnot wi th individual differences. With some excep-tions, when individual differences were studied, theyusually failed to find relationships between person-al i ty variables and trusting behavior, particularlyin the prisoner's dilemma paradigm. One excep-tion, evidenced from such studies as those ofDeutsch (I960) and Wrightsman (1966), incli-catos that trust and trustworthiness are positivelyassociated.

The failure of individual differences measures ofstable personal characteristics in these studies canbe attributed either to the fact that trust is aspecific rather than a generalized characteristic,that the personality measures were themselves in-val id , or that the measures used were poorlyselected and perhaps irrelevant. However, a fourthpossibility is that the prisoner's dilemma and simi-lar games produce a relatively specific reaction,characteristic of competitive games but not of otherkinds of interpersonal interactions. In fact, if theresults of these studies were characteristic of every-day behavior, the normal adult is so competitive,

uncooperative, and untrusting that he could hardlyget through a normal day's activities.

Studies in which the experimental conditions aremanipulated as a sole focus of the investigationhave their limitations as well as their advantages.One limitation is that there are a large number ofconditions that are worth studying but that cannoteasily be manipulated experimentally. Perhaps amore serious limitation is that the parameters ofgroups that are used in such studies are not them-selves varied so that it is not known whether theresults could be generalized to other groups or,indeed, what variables in group selection are rele-vant in generalizing results. The advantages ofconditions manipulations studies or so-called ex-perimental studies are the ease of design and thefact that when condition differences appear, theymay provide important leads not only for im-mediate application, but also for the origin or thenature of the conditions that produce individualdifferences.

Alternatively, the study of individual differencesmay produce important hypotheses for studies ofconditions effects. Individual differences, particu-larly in social behavior, are largely the result ofearlier condition differences, and individual differ-ence measures are of primary significance for in-vestigations involving the development, origin,maintenance, and stability of the behavior beingstudied. Most efficacious is the interaction designwhere conditions differences are combined withrelevant personal measures. The effect of variouskinds of experimental manipulations will vary as aresult of using different groups of subjects regard-less of whether we are studying trust, aggression,anxiety, cautiousness, dissonance reduction, or so-cial attitudes. Only when we know whether or notinteraction effects are present, do we begin to ap-proach a meaningful understanding of the behaviorwe arc studying.

Our strategy, therefore, was to develop a theo-retically based measure of interpersonal trust fo-cused on college students, not only as our mostavailable subject population, but also because theyrepresent the future policy makers of our society.Before going on to determine whether or not thereis generality to trusting behavior, how differencesin trust develop, and how individual differences intrust interact with a variety of experimental condi-tions, it is necessary to discuss the meaning of a

RXPKCTANCIKS I'OK INTKKPICKSONAL TlUIST 445

generalized expectancy within the framework ofsocial learning theory.

In social learning theory, the potential of theoccurrence of a behavior is considered to be afunction of the expectancy that that behavior willlead to a particular reinforcement or reinforce-ments, and the value of these reinforcements in agiven situation. Tn other words, there are fourclasses of variables with which we deal—behaviorpotentials, expectancies, reinforcement values, andsituations.

The relationship we have postulated among thesefour variables allows for the prediction of a specificbehavior in a given situation. The process of gen-eralization accounts for the consistency and stabil-ity of behavior across situations. In social learningtheory, an expectancy is a function of a specificexpectancy, and a generalized expectancy resultingfrom the generalization from related experience.The relative importance of the specific expectancyis a function of the degree of experience in thatspecific situation, or conversely, the importance ofgeneralized expectancy is a function of the degreeof novelty, ambiguity, or unstructuredness of aparticular situation. The more novel the situation,the greater weight generalized expectancies have.Because expectancies generalize along lines of per-ceived similarity, relatively stable modes of re-sponding develop, and a learned basis for a theoryof personality is provided. A fu l l statement ofthis theory is provided elsewhere (Rotter, 1954;Rotter, Chance, & Phares, 1971). What is impor-tant here, however, is that the theory itself providesfor general characteristics and also for specificity.The situation partially determines the response, andthe theory predicts that situations of considerablefamiliarity are less predictable from a generalizedtendency than those involving more novelty.Whether or not one trusts his wife would give usless information about whether the same personwould trust the butcher than if we knew whetheror not he trusted his automobile mechanic. Tf onewanted to relate trust of marital partners to maritaladjustment, a test specifically designed to measuresuch trust would be better than one that sampledtrust of butchers and automobile mechanics. Thedifference between specific and generalized expect-ancy is relative and arbitrary, but the distinctionis of considerable significance for the develop-ment of certain kinds of personality measurementinstruments.

Early research in social learning was concernedwith psychological needs, that is, with the predic-tion of classes of behavior, functionally related inthat they led to the same or similar reinforcements.In recent years, we have turned our attention toanother class of content variables or personalityvariables which are referred to as generalized ex-pectancies for problem solving. Man is a cate-gorizing animal. He continuously forms concepts,changes concepts, and discovers new dimensions ofsimilarity. Where similarity is perceived, the basisfor generalization and a functional personality con-cept is present. Psychological situations, both socialand nonsocial, may also be perceived as similar inthat they present similar problems. For example,all of us are continuously faced with the problemof deciding whether what happens to us is contin-gent on our own behavior and can be controlledby our own actions or whether it depends uponluck, the intervention of powerful others, or influ-ences which we cannot understand. Some of youwill recognize that the general problem just de-scribed has been conceptualized as internal versusexternal control of reinforcement (Rotter, 1966).We have also done work with generalized expect-ancies for cautiousness, for guilt, and a generalizedexpectancy that problems can be solved by thetechnique of looking for alternatives (Rotter et al.,1971). It is not the reinforcement which is similarin the various situations in which the individualfinds himself, but it is some aspect of the situationswhich presents a similar problem.

Another common human experience is that ofbeing provided with information from other people,either promises of reinforcements to come, ormerely statements of presumed fact. Implicit inall these situations is the problem of whether ornot to believe the other person. On this basis, wehave hypothesized a generalized expectancy of trustor distrust.

Following these theoretical lines, we have con-structed what I refer to as an additive test ofinterpersonal trust. An additive test is one whosepurpose is to sample a broad range of situationsof more or less equivalent strength with the ef-fectiveness of the instrument dependent upon theadequacy of the sampling of situations. In con-trast, a power test varies items in strength andsamples from a narrow or restricted range of events.Achievement tests are typical power tests. Additivetests have some characteristics that should be noted,

446 AMKKICAN PSYCHOLOGIST

In general, they may be expected to provide lowerprediction in any specific situation than a powertest devised to measure in that situation, hut beable to predict to a greater range of situations.They may also be expected not to predict at all insome situations in which the subjects have had agreat deal of consistent experience. Internal con-sistency of additive tests would be expected to belower than that of power tests.

The test construction procedure followed was todevise a large number of items presented in aLikert format which appeared to deal with beliefin the communications of others. The others dealtwith were not specific persons with whom the indi-vidual had had a long time close association, butrather classes of significant others whose behaviorcould affect the subject's l ife. Consequently, itemsdealt with such social agents as teachers, otherstudents, judges, newspaper writers, politicians, etc.Following administration of these items to a largenumber of male and female college students, itemswere retained or rejected on the basis of threecriteria. The first of these was a significant cor-relation with the total of all the other items withthat item removed, the second was that the itemproduced a spread of scores providing maximumdifferentiation of subjects, and the last was thatthe item did not correlate too highly with theMarlowc-Crowne Social Desirability scale (Crowne& Marlowe, 1964).

We did not attempt to eliminate all relationshipto social desirability because we felt that theoreti-cally the motive for social approval was not entirelyunrelated to trust of others.

The final form of the test included 25 trust itemsand 1 5 filler items. The scale has an internal con-sistency of .76, and retest reliabilities for five weeks,three months, and seven months were, respectively,.69, .68, and .56. For males and females combined,the test correlates .29 with the Marlowe-CrowncSocial Desirability scale which suggests that trustis regarded as a socially desirable trait, but thatthe total amount of variance in the test scale ac-counted for by the social approval motive is rela-tively small. If is clear that the expression ofdistrust in many situations is not followed bysocial disapproval. There is no significant cor-relation with college entrance scores. Contrary tothe belief of many, high intelligence does not neces-sarily lead to distrust. Alienation, in the sense ofpowerlessness, as measured by the .Internal-External

Control scale is significantly related to scores onthe trust scale. Since both scales arc question-naires, method variance could account for the cor-relation. But the relationship remains significantand in the thirties in four large samples when socialdesirability is partialed out. In any case, it islogical that those who distrust would feel lessability to control and that those who felt that theycould not control what happened to them would besomewhat lower in trust.

Some demographic data are available on thescale. Two studies indicate that youngest childrenare less trusting than only, oldest, or middle chil-dren (Geller, 1966; Rotter, 1967). College stu-dents reporting themselves as agnostics and atheistsare significantly less trusting than others. Studentswho perceive their parents as believing in twodifferent religious are less trusting than those whoperceive both parents as believing in the samereligion or lack of religion. Differences amongreligious affiliations arc present, but they appearto reflect socioeconomic differences in our popula-tion. Tn any case, at the University of Connecticut,Jewish students were found to be more trusting,followed by Protestants, and Catholics being leasttrusting. Socioeconomic level of parents shows asmall but consistent decrease in trust with decreas-ing socioeconomic levels. Since large numbers ofsubjects were used in the study of demographiccharacteristics, relatively small differences weresignificant, and overlap among groups was verygreat.

Tn order to assess the construct validity of thisscale, a first study was conducted (Rotter, 1967)using a sociomctric method in college fraternitiesand sororities. Two sororities and two fraternitieswere used, and all members of these organizationswho had lived together for a period of at leastsix months were included in this study. In addi-tion to asking subjects to nominate members ofthe group who were highest and lowest in inter-personal trust, three other variables were includedthat might be related to trust—gullibility, depen-dency, and trustworthiness. As control variables,humor, popularity, and friendship were included.

The results of this study indicate that the scalecould significantly predict sociometric ratings oftrust and that these predictions were significantlyhigher than the control variables of popularity,friendship, and humor, hi other words, discrimi-nant as well as construct validity was demonstrated.

EXPECTANCIES TOR TNTERVKRSONAL TRUST 447

Interestingly enough, gullibility and dependencywere negatively related to trust, both as measuredby the. trust scale and sociometric trust. The hightruster does not trust out of a need to have some-one else take care of him, and apparently he is notregarded as someone who is easily fooled, tricked,or is naive. On the other hand, as demonstratedin previous studies, trustworthiness is closely relatedboth to sociometric trust and to the interpersonaltrust scale (see Rotter, 1967).

An unpublished dissertation by Geller (1966)assessed the validity of the scale under experimentalconditions with a behavioral criterion. The studyby Geller and some of the others referred to in-volved deception. .It seems paradoxical that aninterest in interpersonal trust would lead to decep-tion experiments, and we have avoided themwherever possible. However, for some problems itis extremely difficult to study trust in the labora-tory without some mild deception. In Geller'sstudy, trust scale scores were collected in theelementary psychology course on the first day ofclasses, and the experiment that took place fromtwo to four months later was not connected in anyway by the subjects with the test administrationearlier. A similar procedure was used in the otherstudies referred to later in this article.

Geller (1966) had three groups of subjects. Thefirst group volunteered for a study of auditory dis-crimination. They were brought into the labora-tory and went through the Crutchfield variationof the Asch conformity experiment. They werethen asked to fill out a brief form, and while theywere doing so, a student confederate told themthat they had just been tricked and that therewere, in fact, no other subjects in other boothsand that the whole thing was done by the use ofa tape recorder. Following this, they were takento another room where the second part of theexperiment began. The second group was similarto the first except that the confederate only madesome small talk and did not reveal that the Aschsituation was a hoax. The third group was notplaced in the Asch situation at all, but broughtimmediately to the second part of the experiment.

In the second phase of this experiment, subjectswere asked to discriminate two chords played ona tape recorder as to whether they were the sameor different. The chords had been pretested andwere easily identified by 100% of the pretest sub-jects. Subjects were told that we were interested

in how quickly they could make the discriminationand were shown a simple apparatus with two brassknobs, one of them labeled "Same," and one,"Different." The experimenter demonstrated,"When the two chords were the same, you pull outone knob; when they were different, you pull outthe other." As he pulled out the knob for"Different," a big blue spark leaped out from theapparatus making a loud noise. The experimenterjumped up and down, swore, and finally said, "1thought they had fixed that." He then retreatedbehind the apparatus, stayed there 10 seconds,came out and assured the subject that it was nowfixed and that he would not get a shock.

Following the behavioral criterion situation, sub-jects were taken to another room and interviewedabout their feelings about the experiment using astructured open-ended interview. Judges ratedresponses to determine whether or not they trustedthe experimenter when he said that the apparatuswas fixed. Blind ratings by judges who did notknow which group the subject was in were used.

In the behavioral criterion situation, the firstfour chords on the tape were all different, and itwas possible to measure how long it took the sub-ject before he pulled out the knob, whether hepulled it out at all, and other spontaneous com-ments and behavioral indications that he did nottrust the experimenter's reassurance. Judges be-hind a one-way screen rated behavior withoutknowing a subject's group membership. Geller(1966) hypothesized that the group that had beentold they had been tricked earlier would show thegreatest distrust in this situation, the control groupthe least, and those who had gone through the Aschsituation and who had been told nothing would bein the middle. In general, these results were ob-tained, and they are best illustrated by the cor-relations of the trust scale with trusting behaviorin the behavioral situation and in the interviewthat followed.

In the behavioral criterion situation, the trustscale correlated with trusting behavior significantlyonly in the control group. The correlation of .38was about the same as that obtained in the socio-metric study. In the deception-concealed situation,the correlation fell to .14, and in the deceptionrevealed, to —.04. Apparently subjects who havebeen told that they have been tricked, even thoughtrusting to begin with, will not trust further. Thesefindings support the sociometric study which indi-

448 AMERICAN PSYCHOLOGIST

catcs that, gullibility and trust are two differentvariables. I t also suggests (hat the Asch situationin itself produces more suspicion than manyexperimenters seem to be aware of.

One of the most interesting findings from thisstudy is that in the interview situation a signifi-cant correlation of .43 appeared between the trustscale and admission that the subject was suspiciousof the experimenter only in the group that hadhad the deception revealed. That is, in this group,high trtisters said that they were suspicious andlow Irustcrs said they were not. Analysis of thebehavior of this group indicated that on the behav-ioral criterion, most subjects, both high and lowtruslers, were suspicious. The high trustcrs toldthe truth and admitted it, the low trusters deniedit. The point is of considerable significance fordeception studies that rely on interviews to dem-onstrate the success of a deception. What theseresults indicate is that low trusters, the peoplewho might be expected to be most suspicious, arethe ones who will deny suspicion. While theseresults may not be generalizable to all decep-tion studies, they are true for this study underconditions where the subject has good reasonto feel that he has already been tricked by theexperimenter.

The study by Roberts (1967) deals with thesame topic. In Roberts' research, the subjects weregiven some experimenter tips on how to use theirtime most effectively in preparing for a readingcomprehension test. Both high and low trusters,who were given correct tips, made increaseduse of the suggestions over three trials. For sub-jects, however, who were given incorrect tips, thelow trusters quickly abandoned the experimenter'sadvice, but the high trusters persisted in followingit over the three trials of the experiment. Thehigh trusters showed a gradual decline, the lowtrusters, a precipitous one. The results appearcontradictory to the previous study which indi-cated high trusters were not gullible. The differ-ence between the two studies, however, was thatin Roberts' study, after having given the wrongadvice, the experimenter admitted it was the wrongadvice, apologized, and said the next time he woulddo better. High trusters, in other words, willpermit a mistake or two and still trust providingthe mistake is admitted and an apology made. Ifthey are shown clear-cut evidence that they havebeen tricked, they will continue to extend trust.

The point may be of significance for governmentand other institutions that have lost credibilitywith the public and hope to regain it.

Other studies illustrate the breadth or rangeand limitations of behavioral prediction possiblefrom this scale. Gilbert (1967) studied subjects'willingness to disclose personal and uncomplimen-tary information about themselves and found suchwillingness significantly related to their trust scores.

MacDonald, Kessel, and Fuller ( 1 9 7 0 ) , however,failed to obtain any relationship between willing-ness for self-disclosure using the Jourard Scale andinterpersonal trust scores; nor did they find a cor-relation between interpersonal trust scores and amodified prisoner's dilemma game. As they pointout, however, their game did not require believingor not believing someone else, but rather dealtwith belief in the benevolence or malevolence ofthe other person. At least in this game situation,the two arc not related. Similarly, in two studiesinvolving sensitivity groups (Aronson, 1970; Stein,1970) , the interpersonal trust scale did not relateto judges' ratings of subjects' openness in thissituation. The kind of trust we are measuring, itappears, is what we set out to measure, an expect-ancy that communications can be believed, not awillingness or desire to speak about private mattersto strangers or belief in the benevolence of othersin competitive situations. Hamsher (1968) alsofailed to find a statistically significant relationship,although a trend was present, between interpersonaltrust scores and trusting behavior in a competitivestock market game where communications wereexchanged and the subjects had the choice of lyingor telling the truth. He did, however, find a sta-tistically significant relationship between inter-personal trust scores and trustworthiness in thegame. In other words, although high trusters inthis game showed only a trend toward believingcommunications of competitors more than lowtrusters, they found it more difficult to lie than

did low trusters.Hamsher, Geller, and Rotter (1968) related trust

scale scores given at the beginning of the semester,as in the other studies reported, to a survey onbelief in the Warren Commission Report givenone month later. A strong relationship betweentrust scale scores and the belief that the WarrenCommission knew there was a conspiracy and wascovering it up was present. .Eighty-five percent

EXPECTANCIES FOR INTEKPJJKSONAL TRUST 449

of the subjects said that they had followed thecontroversy closely.

An interesting sidelight of this study was thatthe responses to the question "T believe that therewas, in fact, a conspiracy and information aboutit has been kept from the public" was respondedto positively by 28% of the subjects; 19% said,"I don't know"; and only 52% said, "No."These figures are not much different from thoseof public opinion polls reported in the newspaperswhich asked a similar question. That 28% ofthe public could unequivocally believe that a com-mittee which had been formed with the purposeof placing on it those people who were mostbelievable and respected in our country could feelthat these people were deliberately lying to themis indeed shocking.

In a study whose primary focus was investigationof sex guilt, Boroto (1970) obtained an interest-ing incidental finding regarding the trust scale.Boroto's subjects were brought into a room wherehe presumably had some personal information aboutthem in a folder and where he also accidentallyknocked over a folder containing "filthy pictures"on his way out of the office to answer a telephonecall. Scooping up the pictures and replacing themin the folder, he left the office where the subjectswere being observed surreptitiously. When Borotodivided his subjects into those who looked into thefolder involving their personal information, or thedirty pictures, or both, and compared them withthose subjects who did not invade the experi-menter's privacy, he found considerable differencein the mean trust scores. Those who looked atneither folder had a mean interpersonal trust scoreof 80, and those who looked at either of the foldersaveraged a score of 66. The difference is not onlyhighly statistically significant, but is approximatelyone and a half standard deviations different. Theaverage score for those who looked at either folderwas close to the mean of the larger population. Itwas those who did not look at either folder whowere deviant in being unusually high on trust.Interestingly enough, in a later interview, abouthalf of the subjects who had looked at the foldersdenied looking at them.

Two studies have investigated the antecedents ofinterpersonal trust in college students. Katz andRotter (1969) sent the parents of 100 students,who were one-half of a standard deviation aboveand 100 students who were one one-half of a stan-

dard deviation below the mean of a larger group,trust scales using a strong appeal and received67% returns. The students themselves had takenthe scale approximately one year earlier. Theletters were sent so that they would arrive onMonday, and it was requested that they be re-turned by Friday. They were sent only to parentsof students who lived some distance from campusso that the potential of contamination from thestudents themselves was minimal. The results indi-cated no differential return rate from the parentsof high trusters or low trusters. The fathers ofhigh-trusting sons were significantly higher ontrust than the fathers of low-trusting sons. Thefathers did not appear to have a significant effecton daughters, and mothers' scores showed a weakstatistical trend toward a small and equal effect onsons and (laughters. The important role the fatherappears to have in influencing the trust of sons wehave interpreted as due to the fact that it is thefather who interprets the trustworthiness of out-side agents to the family. His major concern is ineducating his son who must, like him eventuallymake his way in a world where he is continuouslydealing with strangers.

A study by Into (1969) investigated a largenumber of child-raising behaviors by interviewingcollege students and obtaining their statementsabout their parents' behavior during childhood.While such studies have their obvious limitations,they can be the source of valuable hypotheses, andthey do provide some suggestive evidence. A largenumber of variables were investigated, and the re-sults are too extensive to be reviewed here. Whatis clear is that the strongest effects were modelingeffects, both of the father and the mother, anddirect teaching effects. The parents of high trust-ing subjects were more trusting to their children,were more trustworthy, trusted outsiders more, anddirectly taught trust and trustworthiness. Althoughthese effects were significant for both fathers andmothers, they were again a little stronger forfathers. One finding of interest in relationship tothe earlier reported differentiation of trust as anexpectancy for truthfulness versus trust in thebelief in a benign world, was that low trusters werelikely to report that their parents made no threatsor made them and did not keep them, and hightrusters were likely to report that their parentsboth made and kept threats.

450 AMERICAN PSYCHOLOGIST

Hochreich (1066) has developed a cartoon typetest to study trust in children. The test different i-ates children who select, an immediate lesser rewardfrom a larger promised delayed reward at less thanthe .10 level of significance in the expected direc-tion. Too little data on this instrument, however,are available to assess its validity.

While discussing developmental antecedents oflow trust, mention should be made of the relation-ship of trust to adjustment. Four independentsamples of subjects from various experiments pro-vided scores on the Incomplete Sentences Blank(Rotter & Rafferty, 1950), a measure of collegestudent adjustment; and the Interpersonal TrustScale (Gcllcr, 1966; Hochreich, 1968; Mulry,1966; Rotter & Stein, 1970). It seems to beclearly established that low trust correlates withmaladjustment as measured by the IncompleteSentences Blank. The correlation is low but sig-nificant in all samples when sexes arc combined.Most samples show a higher relationship for fe-males, but whether or not the difference betweenmales and females is characteristic of the totalstudent population is still unclear. In some sampleswhere it is possible to partial out the effects ofsocial desirability, the significant relationship be-tween adjustment and trust remains. Of course,it cannot be determined from this data whether lowtrust leads to adjustment problems or adjustmentproblems lead to low trust, or whether, in fact,both are the effects of developmental experiences.

With Donald Stein, T have also been studyingpublic attitudes of trust toward various occupa-tional groups (Rotter & Stein, 1970). We have setup separate rating scales for trust, competence,and altruism. We have obtained four samples ofsubjects—200 elementary psychology students fromthe University of Connecticut, 100 from the Uni-versity of Maryland, 50 secretaries in a publicschool system, and 50 high school teachers. Mostof the secretaries had only a high school educationplus some business training and had an averageage of 38. The 50 teachers all had bachelor's ormaster's degrees and had an average age of 35.The results of this study are still being analyzed,but. some preliminary results are of interest.

Surprisingly, the mean ratings given to the 20occupations correlated, from one sample to another,in the nineties for all variables and groups. Thatis, the attitudes of the teachers, male and femalecollege students of two different universities, and

secretaries—populations which varied in age, sex,education, and geography-—were all surprisingly andhighly similar. The average rating, by the subjects,of the truthfulness of the 20 occupations, alsocorrelated significantly with their trust scales. Thissuggests that the studies done about attitudes ofcollege students would not be too dissimilar fromthose of the more general population which mightbe called white collar or middle class. The secondfinding of interest is that although these scalesintcrcorrelated, when all the samples arc combined,the correlation between trust and competence is .60;between trust and altruism, .63; and betweencompetence and altruism, .55. There was still aconsiderable independent variance. If the correla-tions between the means of the different samplescan be taken as an estimate of reliability, it isclear that the three scales are measuring at leastpartially independent variables.

Time permits only a brief description of someof the interesting findings regarding specific occupa-tions. Of particular interest is the discrepancybetween rankings on the different scales for someof the occupations. For example, clergymen whoranked first on altruism, ranked only eighth oncompetence. Dentists, on the other hand, rankedsecond on competence but sixth on altruism.Psychiatrists and psychologists ranked higher onaltruism than they did on competence and truth-fulness. Surprisingly, law enforcement officials wererated as ranking ninth on altruism, tenth on truth-fulness, but sixteenth on competence. Executivesof large corporations ranked ninth on competence,fourteenth on truthfulness, and eighteenth on altru-ism. The anchor occupation which ranked loweston all three scales was used car salesman, andphysicians are seen as highest in competence andtruthfulness.

The psychologists' rating on truthfulness wasseventh, better than the average but not reallyanything to brag about. It should be a matter ofconcern to psychologists who desire to make con-tributions to our society that this rating does notbecome any lower and, hopefully, will becomehigher. Continuation of large numbers of decep-tion experiments and the public announcement ofopinions as substitutes Cor replicable data obviouslywill not help it.

Finally, the question of whether or not trust insignificant others in this society is indeed dropping

EXPECTANCIES FOR INTERPERSONAL TUTJST 451

during recent years should be considered. Since1964, my assistants and I have tested the elemen-tary psychology students under comparable testingconditions every fall semester, and in some yearsalso in the spring semester. Analyses have indi-cated that the samples have not changed signifi-cantly in class rank, in college entrance examina-tion scores, or ethnic composition. In all, 4,605students were included in the six-year study, andeach of the individual samples is large. The resultsare somewhat startling in that each year there wasa significant drop in mean trust scores comparingthe September testing with the prior Septembertesting. These results are in agreement with arecent report of Wrightsman and Baker (1969)who gave a Philosophies of Human Nature Scale toincoming freshmen at Peabody University on sev-eral occasions over the period 1954-1968. A sub-scale of this test dealing with trust showed a sig-nificant decline over the period. Trust was definedby them as "the extent to which people are seenas moral, honest or reliable [p. 2991."

In the 1964 study, the average score was 72.4;in 1969, it was 66.6 (Hochreich & Rotter, 1970).The difference is not only highly statistically sig-nificant, but it indicates that a student who was atthe mean of the distribution in 1969 would havebeen in the lower one-third of the distribution in1964. That such a drop is not the result of itbecoming less socially desirable to express trust,is attested to by the fact that correlations withthe Marlowe-Crowne Social Desirability scale andthe trust scale did not differ significantly over theperiod studied.

An item analysis was made to discover how gen-eral was the drop in trust and in what items wasit greatest. In all, 14 items dropped significantly,and 1 increased significantly from 1964 to 1969.The items that showed the greatest decrease in-cluded those in the areas of politics, peace keeping,and communications. The items that showed littleor no change included mainly those dealing withsocial agents with whom the subjects were likelyto have had some direct contact, such as parents,repairmen, salesmen, ordinary people. The declinein trust over the last six years may not be sur-prising to many, but it appears to the author toindeed be precipitous, and should it continue, oursociety would be in serious trouble, if it is notso already.

REFERENCES

AEONSON, S. R. A comparison of cognitive vs. focused-activitics techniques in sensitivity group training. Un-published doctoral dissertation, University of Connecticut,1970.

BOROTO, D. R. The Mosher Forced Choice Inventory as apredictor of resistance to temptation. Unpublishedmaster's thesis, University of Connecticut, 1970.

CROWNE, D. P., & MAKI.OWK, D. The approval motive:Studies in evaluative dependence. New York: Wiley,1964.

DEUTSCH, M. Trust, trustworthiness, and the F scale.Journal oj Abnormal and Social Psychology, 1960, 61138-140.

GKIXEK, J. D. Some personal and situalional determinantsof interpersonal trust. Unpublished doctoral dissertation,University of Connecticut, 1966.

GIFFIN, K. The contribution of studies of source credibilityto a theory of interpersonal trust in the communicationprocess. Psychological Bulletin, 1967, 68, 104-120.

GILBERT, J. Interpersonal trust: Implications for psycho-therapeutic technique. Unpublished master's thesis, Uni-versity of Connecticut, 1967.

HAMSIIEK, J. H., JR. Validity of personality inventoriesas a function of disguise of purpose. Unpublished doc-toral dissertation, University of Connecticut, 1968.

HAMSJIER, J. H., GEIJ.ER, J. D., & ROTTER, J. B. Inter-personal trust, internal-external control, and the. WarrenCommission report. Journal oj Personality and SocialPsychology, 1968, 9, 210-215.

HociiREicir, D. J. A children's scale for measuring inter-personal trust. Unpublished master's thesis, Universityof Connecticut, 1966.

HOCIIKEICH, D. J. Refined analysis of internal-externalcontrol and behavior in a laboratory situation. Unpub-lished doctoral dissertation, University of Connecticut,1968.

HociiREien, D. J., & ROTTER, J. B. Have college studentsbecome less trusting? Journal of Personality and SocialPsychology, 1970, 15, 211-214.

INTO, E. C. Some possible childrcaring antecedents ofinterpersonal trust. Unpublished master's thesis, Uni-versity of Connecticut, 1969.

KATZ, H. A., & ROTTER, J. 13. Interpersonal trust scores ofcollege students and their parents. Child Development,1969, 40, 6S7-661.

MAcDoNAi.n, A. P., KESSKL, V. S., & FULLER, J. B. Self-disclosure and two kinds of trust. Unpublished manu-script, Rehabilitation Research & Training Center, WestVirginia University, 1970.

MTSCTIEI,, W. Theory and research on the antecedents ofself-imposed delay of reward. In B. A. Maher (Ed.),Progress in experimental personality research. Vol. 3.New York: Academic Press, 1966.

MULUY, R. C. Personality and test-taking behavior. Un-published doctoral dissertation, University of Connecticut,1966.

OSGOOD, C. E. A case for graduated unilateral disarmament.Bulletin of Atomic Scientists, 1960, 16, 127-139.

452 AMKKTCAN PSYCHOLOGIST

ROBERTS, M. D. The persistence of interpersonal trusLUnpublished inaslcr's thesis, University of Connecticut,1967.

KOTTKK, J. B. Social learning and clinical psychology.Englcwood Cliffs, N. J.: Prentice-Hall, 19S4.

ROTTER, J. B. Generalized expectancies for internal versusexternal control of reinforcement. Psychological Mono-graphs, 1966, 80(1, Whole No. 609).

ROTTKR, J. B. A new scale for the measurement of inter-personal trust. Journal of Personality, 1967, 35, 651-665.

HOTTER, J. B., CHANCE, J'., & PJIARES, K. J. Applicationsof a social learning theory of personality. New York:Molt, Rinehart & Winston, 1971, in press.

ROTTER, J. B., & RAI.-TERTY, J. E. Manual for the RotterIncomplete Sentences Blank. New York: PsychologicalCorporation, 1950.

ROTTER, J. B., & STEFN, D. K. Ratings of the trustworthi-ness, competence and altruism of twenty representativeoccupations. Unpublished manuscript, University of Con-necticut, 1970.

STEIN, D. K. Kxpectalion and modeling in sensitivitygroups. Unpublished doctoral dissertation, University ofConnecticut, 1970.

VINACKE, W. E. Variables in experimental games: Towarda field theory. Psychological Bulletin, 1969, 71, 29.5-318.

WRIGIITSMAN, L. S. Pcrsoaalily and at t i tudinal correlatesof trusting and trustworthy behaviors in a two-persongame. Journal of Personality and Social Psychology,1966, 4, 328-332.

WRIOIITSMAJST, L. S., & BAKER, N. J. Where have all theidealistic, imperturbable freshmen gone? Proceedings ofthe 77th Annual Convention of the American Psycho-logical Association, 1960, 4, 299 -300. (Summary)