Anticipating informed consent: An empirical approach

Anticipating Informed Consent

An Empirical Approach

ELLEN BERSCHEID University of Minnesota±

ROBERT STEVEN BARON University of Iowa

MARSHALL DERMER University of Minnesota 2

MARK LIBMAN University of Minnesota 3

Section 3-4, "Obtaining Informed Consent to Par-ticipate," of the new Ethical Standards for Re-search with Human Subjects, recently adopted bythe American Psychological Association (Cook etal., 1972) begins:

The psychologist's ethical obligation to use people as re-search participants only if they give their informed consentrests on well-established traditions of research ethics andon strong rational grounds. The individual's human rightof free choice requires that his decision to participate bemade in the light of adequate and accurate information[p. VI.

The principle of informed consent has longguided much psychological research. A prerequi-site for obtaining funds from the U.S. PublicHealth Service (1969), for example, has been aprior review of proposed research by an appropriatecommittee of the applicant institution. Amongother considerations, the review is to assure that"the methods used to obtain informed consent areadequate and appropriate [p. 1]."

In conflict with the principle of informed con-sent is the practice of misleading prospective re-search participants as to the true purpose of theexperiment and withholding from them informationconcerning the exact procedures to be followed dur-ing the course of their participation. "Deception"

1 This research was funded in part by National Instituteof Mental Health Grant MH 16729 to Ellen Berscheid.The authors wish to express their appreciation of KinleyLarantz' statistical advice.

Requests for reprints should be sent to Ellen Berscheid,Department of Psychology, University of Minnesota, Min-neapolis, Minnesota 55455.

2 Now at University of Wisconsin, Milwaukee.3 Now at York University, Toronto, Ontario, Canada.

methodology, however, often serves to facilitate themanipulation of independent variables that other-wise would not be amenable to rigorous investiga-tion and to minimize the extent to which "demandcharacteristics" (Orne, 1964) are operative in ex-perimental settings.

The new APA Ethical Standards (1972) docu-ment recognizes that methodological requirementsfrequently present the investigator with ethical di-lemmas. As a guideline for resolution in the caseof deception methodology, the document outlinesa "number of considerations that may make the useof deception more acceptable [p. VII]": The re-search problem is of great importance; it may bedemonstrated that the research objectives cannotbe realized without deception; the subject is al-lowed to withdraw from the study at any time andto withdraw his data if he so wishes when the con-cealment or misrepresentation is revealed to him;the investigator takes full responsibility for detect-ing and removing stressful aftereffects; and finally,

there is sufficient reason for the concealment or misrepre-sentation that when the research participant is later in-formed . . . , he can be expected to find it reasonable, andso suffer no loss of confidence in the integrity of the in-vestigator or others involved . . . [p. VIII.

The USPHS (1969) directive also appears to makeallowance for research in which "informed consent. . . can be anticipated [p. 10]."

How can one determine in advance of experimen-tation whether a potential subject would consentto participation if he were informed of the aimsand procedures involved in the experimentation?

Under what circumstances is he likely to later findthe concealment and deception .reasonable? A

AMERICAN PSYCHOLOGIST • OCTOBER 1973 • 913

number of strategies may be followed to answerthese questions.

Perhaps the most usual procedure for determin-ing whether consent may be anticipated is for theinvestigator himself to estimate whether his poten-tial subjects would consent to the experimentationif given full procedural information. The potentialpitfalls of this strategy need not be outlined.

Another current practice is to present the pro-posed experimental procedure, along with a descrip-tion of the subject population, to an "ethics ad-visory group" in order to obtain their collectiveopinion as to whether participants are likely tofind the concealment reasonable. While such judg-ments may not be as vulnerable to self-interest asthose of the principal investigator, a number ofquestions about the predictive accuracy of anygiven "advisory" group could be raised.

We propose that a simple, but more accurate,alternative to the strategies outlined above mightbe to draw a sample from the proposed subjectpopulation, present it with the full procedure tobe followed in the experiment along with the pur-pose of the experimentation, and determine the ex-tent to which these subjects would be willing toparticipate in the experiment described. Baumrind(1971, p. 889) has recently suggested a similarstrategy. From this "role-play-sampling" proce-dure, consent rates could be projected for the sub-ject population. In addition, the degree to whichalternative procedures (perhaps equally desirableon methodological grounds) would be acceptable topotential subjects could be easily and inexpensivelydetermined prior to actual participation in experi-mentation by any human subject.

The proposed strategy is admittedly a "role-playing" procedure. As such, it may be liable toseveral of the criticisms that have been lodgedagainst role-play methodology with regard to itssuitability as a substitute to the nonvicarious ex-periencing of an experimental situation (Freedman,1969). It has been charged, for example, thatwhen a person is asked to imagine himself in asituation, his behavior may diverge considerablyfrom his behavior when actually placed in thatsituation. With respect to the role-play-samplingprocedure, it is clear that a prospective participantmay not always accurately foretell the degree ofdiscomfort or stress that will be aroused by theexperiment even when it is carefully described indetail. It is possible, in addition, that he may

entertain self-enhancing projections of the behaviorhe is likely to exhibit in the situation. Unfortu-nately, these problems are also inherent in theprinciple of informed consent. Adoption of thisprinciple alone does not guarantee that an indi-vidual who has given fully informed consent to par-ticipate in an experimental procedure will not laterregret his decision due to the inaccuracy of hissubjective judgments made prior to consent. Forthis reason it should be emphasized that the pro-posed procedure is addressed only to the questionof consent which plays a central, but by no meansexclusive, role in the attempt to protect the sub-ject's welfare.

In the present study we wished to investigatethe degree to which "informed consent" to a num-ber of published social-psychological experimentscould have been anticipated had a sampling proce-dure for determining likelihood of consent beenused prior to experimentation. To meet the criti-cism that role-play subjects, unlike potential par-ticipants, may fail to become "involved" since theirdecisions are without consequence, two variants ofthe role-play consent procedure were investigated.A high-consequence condition was constructed inan attempt to approximate an actual consent situa-tion in which subjects believe their consent re-sponses will determine actual experimental partici-pation. In a low-consequence condition, responseswere obtained under the more usual role-playingcircumstance in which subjects anticipate thattheir decisions are largely academic and will mostprobably never have any real consequence for them.

The subject sample was drawn from the popula-tion most typically used in social-psychological ex-perimentation, the lower-division college student.Included for investigation of prospective subjects'reactions were both those experiments that mightbe considered innocuous in terms of subject dis-comfort (nonstressful) and those that we suspectedto have been stressful to subjects during the courseof experimentation (stressful). It was hypothe-sized that subjects in the high-consequence condi-tion would respond more cautiously to stressfulthan to nonstressful experiments, while subjects inthe low-consequence condition would not respondas differently to the two types of experiments.

Use of the proposed strategy for determiningwhether consent may be anticipated necessitates ajudgment of whether the consent that follows thereceipt of a particular body of inforrhation should

914 • OCTOBER 1973 • AMERICAN PSYCHOLOGIST

be considered "informed" or "uninformed." Whatare the relevant pieces of information that shouldbe made available to prospective participants inorder for their consent to be deemed informed?Principle 3 of the APA Standards (Cook et al.,1972) document states, in part, "The investigatorshould inform the participant of all features ofthe research that reasonably might be expected toinfluence willingness to participate [p. VI]." Theinfluence of various types of information upon will-ingness to consent was investigated in the presentstudy. Stressful and nonstressful experiments weredescribed to potential subjects at one of seven levelsof information:

(a) Rationale (cover story information only).A glance at any recent journal documents the factthat potential volunteers tend to consent to partici-pation in studies even when they are given onlya cursory rationale for their participation. It wasexpected that this high willingness to participatein experimentation would be reproduced when onlythe "cover stories" which were used in each of theexperiments were presented to the subjects.

(b) Procedure (procedural information pre-sented after cover story). In the case of stressfulexperiments, it was expected that additional knowl-edge of procedure should cause a significant dropin level of consent. A significant drop was alsopredicted for the nonstressful experiments, al-though the decrease was expected to be less thanin the case of stressful experiments. In making thelatter prediction it was reasoned that the effect ofrendering consent on the basis of limited informa-tion is probably due not only to an implicit trustin the experimenter, but also to the surrender ofresponsibility for the decision as a result of theunequal power inherent in the prospective volun-teer's lack of information. If more information isprovided, even if that additional procedural infor-mation no more than reinforces the initial rationale,the potential volunteer might take more responsi-bility for his decision to consent, and, hence, con-sent level should be lowered even in the case ofnonstressful experiments.

(c) Desirable behavioral expectation versus (d)undesirable behavioral expectation (expectation in-formation presented after cover story and proce-dure). The potential subject's subjective predic-tion of whether he would be likely to behave ina "normal," justifiable, or socially desirable mannerduring the course of the experiment might affect

likelihood of consent. Most usually, of course,this information is not accessible to the investigatorto give to subjects prior to experimentation since thequestion of how subjects will behave is usually thevery purpose of the experimentation. Nevertheless,we thought it interesting to investigate the impactof this variable on consent, since it appears that itis partially on these grounds that the ethicality ofseveral experiments has been called into question.For example, one can question the extent to whichthe Milgram (1963) experiment would have re-ceived the adverse publicity and criticism it re-ceived if most of his subjects had defied the ex-perimenter early in the procedure. Within thepresent study, rather than leave it to the individ-ual subject's imagination as to how he would be-have in each experimental context, his expectationthat he would be likely to engage in either sociallydesirable or undesirable behavior was directly ma-nipulated.

(e) Neutral debrief (debriefing information pre-sented after cover story, and procedure), (/) de-sirable debriej (debriefing information presentedafter cover story, procedure, and desirable behav-ioral expectation information), and (g) undesirabledebrief (debriefing information presented aftercover story, procedure, and undesirable behavioralexpectation information). Since some of the in-formation that is typically contained within a de-briefing session can be hypothesized to affect aprospective volunteer's decision to consent, threedebriefing conditions were included for low-conse-quence subjects. Debriefing information was with-held from high-consequence subjects because itwas feared that such information might jeopardizethe credibility of the high-consequence manipula-tion.

Debriefing sessions usually provide the partici-pant with some insight into the problem which hasbeen studied and the purposes of the research inwhich he has participated. In experiments employ-ing deception methodology, a participant alsolearns that he was deliberately misled as to thetrue purpose of the experiment and that his priorconsent to participate was not fully informed.While it was hypothesized here that the receipt ofthis information would increase willingness to con-sent, the effects of debriefing have not always beenassumed to be ameliorative (e.g., Baumrind, 1964).

To test the hypotheses outlined above and toexplore the feasibility of the proposed "role-play-


sampling" procedure, subjects were randomly as-signed to 1 of 11 conditions. Eight conditions weredefined by crossing consequence level (high versuslow) with nature of information (rationale, proce-dure, desirable behavioral expectation, undesirablebehavioral expectation). In addition, the threedebrief conditions outlined above were run at thelow-consequence level. Crossed with the abovebetween-subject factors was a within-subject stressfactor such that subjects read descriptions of bothstressful and nonstressful experiments within eachof the 11 conditions.

Method

SUBJECTS

The subjects were 106 undergraduates, 38 males and 68females, enrolled in an introductory psychology course atthe University of Minnesota, who participated for coursecredit. In order to qualify for participation, it was nec-essary that each subject commit himself to attend a "sec-ond session" the following week for which he would re-ceive $2. Commitment to a second session was necessaryto establish the credibility of the high-consequence ma-nipulation.

PROCEDURE

The subjects arrived at the assigned room at various timesduring the day. There were typically 5-10 subjects work-ing on test booklets at long desks at any one time. Thesubjects were randomly assigned to booklets; the contentsof the booklet determined the experimental condition. Noverbal instructions were given by the experimenter. Thesubjects were not aware of their participation in a multi-dimensional study since a filler task was included at theend of the booklet so that the task required the sameamount of time to complete in all conditions. When thesubject had worked for approximately SO minutes, the ex-perimenter interrupted him and asked him to self-addressan envelope so that he could be contacted for the secondsession. The subject was then free to leave. The envelopewas later used to mail a description of the aims of thepresent study to the subjects and to relieve them of thenecessity of a second session.

The first page of each booklet contained an introductionto the study and instructions. All of the subjects wereinformed that the study had two purposes, one of whichwas "to investigate how people feel about participating invarious types of psychological experiments."

CONSEQUENCE MANIPULATION

High-consequence-condition subjects were told that thesecond purpose of the study was "to allow you to exercisesome choice as to the experiment in which you will be asubject." In addition, these subjects were told, "We wouldlike to know how you would personally feel if you hadparticipated as a subject in each experiment. We will usesome of your answers (e.g., the question 'How willingwould you be ... ?') to determine in which of these ex-periments you will later participate as a subject."

Low-consequence-condition subjects were told that thesecond purpose of the study was to use their responses tothe current booklet to help interpret additional "judg-ments you will make on a series of experiments nextweek."

BACKGROUND INFORMATION

Before reading the experimental descriptions, the subjectswere asked to answer a series of items that provided back-ground information. The sex and major of the respondent,as well as the number of previous psychological studies inwhich he had participated, were assessed. In addition, 16items from the Berger (1952) Acceptance of Self Scalewere administered. The items were chosen from the "self"subscale that presumably measures chronic self-esteem.

After responding to these items, the subjects moved onto a series of bipolar adjectives designed to measure theirattitudes toward psychology and psychologists (e.g., trust-worthy-untrustworthy, rewarding-punishing) and experi-mentation (e.g., interesting-dull, pleasant-unpleasant).

ROLE-PLAYING INSTRUCTIONS AND RESPONSES

After completing this section, the subjects read six experi-mental descriptions which were presented in random order.The importance of projection into these experimental situa-tions was stressed. To aid the subjects in this effort , allwere instructed that:

Details will be given in the descriptions to help you toimagine yourself in the settings. For example, referencemight be made to a typical subject—in this case youare to imagine that you are that typical subject, andthat his behaviors and feelings are your own. After read-ing about an experiment please answer the questionsabout it in the separate answer booklet. Remember toanswer these questions as if you had just participatedin the experiment.

Alter reading each description, the subjects answered aseries of items designed to measure their reactions to theexperiment. Among these was the response measure ofmajor interest: "How willing would you have been to par-ticipate in this experiment?" Included also were bipolaradjective scales designed to assess the subject's estimate ofhis well-being following the experiment and the favorabilityof his evaluation of the experiment itself.

EXPERIMENTAL DESCRIPTIONS

Choice of the six experiments included in the present studywas arbitrary, but the following were thought to be fairlyrepresentative of a range of experiments that have appearedin the social-psychological literature in the past 10 years.The first four were considered to be stressful, each for dif-ferent reasons; the last two were thought to be relativelynonstressful.

Milgram (1963). This study of obedience in a harm-doing situation was included since it has probably receivedmore public criticism on ethical grounds than any otherpiece of work in social psychology. The majority of Mil-gram's subjects continued to pull a lever which they be-lieved administered electric shock to a victim to the pointat which they were ostensibly delivering dangerous levelsof punishment.

Berscheid, Boye, and Walster (1968). This study usedharm-doing as an independent rather than a dependentvariable. The subjects tended to derogate "victims" whom


they had ostensibly harmed through administration of elec-tric shock and from whom they expected no retaliation.

Barber and Hahn (1962), The control condition fromthis study of hypnotic pain reduction was used since itinvolved actual physical pain. The subjects were requiredto immerse one of their hands in painfully cold water forthree minutes.

Orne and Scheibe (1964). This procedure was adoptedfrom Orne and Scheibe's "demand" condition in which thesubjects responded to danger cues by interpreting a harm-less situation as a very threatening one.

Festinger and Maccoby (1964). This was a study ofthe effect of distraction on attitude change. A persuasivecommunication was presented while the subjects watchedan amusing film.

Deutsch and Krauss (1960). The subjects were allowedto compete or cooperate in a simulated bargaining situationinvolving the movement of toy trucks.

INFORMATION MANIPULATION

As previously noted, seven information conditions wereemployed. To illustrate each, we present below the com-plete text used to describe the Milgram (1963) experiment:

(a) Rationale. Only a typical cover story was given:

This experiment was designed to study the effects ofpunishment on learning. We know very little about thistopic because almost no truly scientific studies havebeen made of it in human beings. For instance, we don'tknow how much punishment is best for learning—andwe don't know how much difference it makes as to whois giving the punishment, whether an adult learns bestfrom a younger or an older person than himself—ormany things of that sort.

So in this study we are bringing together a numberof students and asking some of them to be teachers andsome of them to be learners. We want to find out justwhat effect different people have on each other as teach-ers and learners, and also what effect punishment willhave on learning in this situation.

(b) Procedure. In this condition, the subjects receivedboth the rationale and information regarding the experi-mental apparati and procedure. That is, in addition to theinformation above, the subjects read the following:

You were randomly assigned the role of teacher. Youmet another student who was randomly assigned therole of learner. The learner was taken into the nextroom and strapped in an "electric chair" apparatus. Thestraps were to prevent excessive movement while thelearner was being shocked. It would be impossible forhim to escape the shock. An electrode was attached tothe learner's wrist and electrode paste was applied toprevent blisters and burns. Although the shock could beextremely painful, it could cause no permanent tissuedamage.

The lesson consisted of a paired-associates learningtask. You read a series of word pairs to the learner andthen read the first word of the pair along with fourterms. The learner indicated which of the four termshad originally been paired with the first word. He com-municated his answer by pressing one of four switches infront of him which in turn lit one of four numberedsquares in an answer box on a shock generator in frontof you.

The shock generator had an on-off switch and 30 leverswitches to set the voltage level -of the shock. Voltage

ranged from IS to 450 volts. This resulted in a rangefrom slight shock to extremely dangerous, severe shock.

When you depressed a switch a pilot light correspond-ing to each switch was illuminated in bright red, andelectric buzzing was heard, an electric blue light labeled"voltage energizer" flashed, the dial on the voltage meterswung to the right and various relay clicks sounded.

You were told to administer a shock to the learnereach time he gave a wrong response. Moreover, youwere told to move the voltage one level higher on theshock generator each time the learner gave a wrong an-swer. Before administering the shock you announced thevoltage level. The voltage level started at 15 volts andyou were 'told to increase it one step (i.e., 15 volts) eachtime the learner gave a wrong answer. The experimentalsession would end when the learner had mastered 10pairs. Before beginning you were given a sample shockat the third voltage level, i.e., 45 volts, to give you anidea of the intensity of the shock. The learning trialswere then begun. Typically your learner had great diffi-culty mastering the task and successively higher shocklevels up to and including 450 volts were demanded.

(c) Desirable behavioral expectation. The subjects inthis condition read all of the informatipn above and, inaddition, learned that:

You as the typical teacher initially raised the voltage andshocked the learner for each mistake. The learner gaveno subjective reaction to the shock up to 285 volts.When the shock reached around 300 volts, however, thelearner displayed signs of great distress, e.g., poundingon the wall of the room in which he was bound to theelectric chair. You were able to hear these reactions. Atthis point, as a typical subject, you refused to continuein spite of the experimenter's urgings that cooperationwas essential.

'(d) Undesirable behavioral expectation. The subjects inthis condition received information identical to that of Con-dition b and, in addition, read the following:

As the typical teacher, you continued to raise the voltageand shock the learner. The learner gave no reaction tothe shock up to 285 volts. When the shock reachedabout 300 volts, however, the learner displayed signs ofgreat distress, e.g., pounding on the wall of the roomin which he was bound to the electric chair. You wereable to hear these reactions. After this outburst nomore responses were given by the learner but this ab-sence of responding in itself constituted an error. Youcontinued to raise the voltage level. At around 315 voltsthe learner was again heard in great agitation but afterthis the learner fell silent. But you as the teacher con-tinued to increase the voltage in steps until 450 voltswere being given. This level was labeled as "extremelydangerous severe shock." You displayed very noticeabledistress as you continued to shock the learner. Youtwitched, trembled, and laughed uncontrollably in a bi-zarre fashion. You were aware that you were givingextremely painful shocks. You appealed to the experi-menter to terminate the procedure but he told you tocontinue since your cooperation was absolutely essential.You, as the typical teacher, continued to increase thevoltage and shock the learner until the end of the scale—450 volts—was reached.

(e) Neutral debrief. The subjects in this condition re-ceived the information contained in Condition b and, inaddition, the following debriefing information:


After the experiment you were informed of the true pur-pose of the study. The experimenter was really measur-ing how long and to what voltage level you would con-tinue to shock a victim. This was part of a larger studyinvestigating some of the conditions under which peoplewould obey orders. Many other subjects had partici-pated and had responded in the same manner as you did.The study was part of a long-overdue investigation ofthe common belief that Americans are not blindly obe-dient. No shock was actually being given since theshock generator was not really connected to the elec-trodes on the learner. The learner was revealed to bean employee and confederate of the experimenter. Youwere introduced to your "victim" and assured that youhad caused no harm whatsoever. Every effort was madeto assure you that you left the situation with a completeinsight into the purpose and nature of the experimentand that you left in a state of well-being.

This condition differed from Conditions / and g below inthat the subject was given no information concerning howhe would have been likely to behave during the course ofthe experiment and presumably reached his own conclusions.

(/) Desirable debrief. These subjects received the in-formation provided in Condition c and, in addition, thesame debriefing information provided in Condition e.

(g) Undesirable debrief. The subjects received the in-formation provided in Condition d and, in addition, thesame debriefing information provided in Condition e.

Resets and Discussion

WILLINGNESS TO PARTICIPATE

The primary dependent measure, willingness to par-ticipate in a particular experiment, was operation-alized as the subject's response to the followingquestion: "Assuming that you would have reactedas described and that it had been possible for youto know everything about this experiment that younow know, would you have chosen to participatein it?" The response categories, "definitely not,""probably not," "maybe," "probably," and "defi-nitely," were respectively scored from 1 through S.Responses were averaged for the two nonstressfuland also for the four stressful experiments.

It will be recalled that it was hypothesized thathigh-consequence as compared to low-consequencesubjects would respond more cautiously to the ex-periments since the former were instructed so thattheir responses would help determine in whichstudy they would subsequently participate. Inorder to assess the extent to which the consequenceand information (between-subject) factors, to-gether with the stress (within-subject) factor, af-fected willingness to participate, a 2 (between) X1 (within) analysis of variance was performed onthe "willingness to participate" data. This analysisrevealed neither a consequence main effect (F =

.004, df - 1/68, p - .95), nor did it indicate thatconsequence interacted with the other variables(Consequence X Information, F — .53, df = 3/68,p = .66; Consequence X Stress, F = .056, df = I/68, p = .81; Consequence X Stress X Information,F= .64, d/ = 3/68, /> = .59).

The inclusion of the consequence factor was anattempt to approximate an actual consent situa-tion in which the prospective subject believes hisresponse will determine participation. In the ab-sence of a consequence main effect and related in-teraction effects, it is especially unfortunate thatwe have no means of assessing the absolute levelof consequence, or "involvement," that our datareflect. It must be assumed that this level issomewhat less than would have been the case hadthe subjects been placed in an actual consent situa-tion. Despite the necessity of adopting a conserva-tive view, evidence, which will be presentedshortly, indicates that the subjects did take thetasks seriously; their responses do not at all appearto be thoughtless, undiscriminating, or particularlyaccepting. In any event, the absence of conse-quence main effects and interactions, on all de-pendent measures examined, was capitalized on bycollapsing across this factor in the analyses whichfollow.

It was predicted that both information, a be-tween-subject factor, and the stress associated withthe experiments, a within-subject factor, would af-fect consent. The relevant means for each condi-tion are presented in Table 1. An unweightedmeans analysis of variance performed on these datayielded both significant information (F = 3.08, df= 6/99, p - .008) and stress (F - 90.40, dj =1/99, p< .00001) main effects. In addition, theinteraction between these two factors was signifi-cant (F = 3.22, df = 6/99, p = .006).

Tests for the simple effects of stress within in-formation conditions revealed that subjects werereliably more willing to participate in nonstressfulas compared to stressful experiments with the ex-ception of the rationale (t — .57, dj = 99, p —.57) and desirable debrief (t = 1.60, df = 99, p =.11) conditions. The absence of a stress effectwithin the rationale condition is explicable in termsof the paucity of stress information within thiscondition. The experments were, however, de-scribed extensively in the desirable behavioral ex-pectation and desirable debrief conditions. Thepresence of a stress effect in the former condition


TABLE 1

"Would You Have Chosen to Participate" Means

Type of experiment

Nonstressful"Stressful1"

Information level

Rationale(n - 19)

3.79d

3.57<,d

Procedure(n = 18)

3.67d2.42ab

Desirablebehavioralexpectation

(« - 17)

4.24d

2.80obo

Undesirablebehavioralexpectation

(n = 22)

3.91d

2.29a

Neutraldebrief(« =8)

4.19d

3.41bo

Desirabledebrief

(» = 11)

3.59d

2.97Bb0d

Undesirabledebrief

(» = 11)

4.05d

3.10.bc

Note.i »* "definitely not," 2 « "probably not," 3 «* "maybe," 4 « "probably," 5 » "definitely." Means with common subscripts do not differ,at the .05 level, as determined by two-tailed I tests between stress conditions and Duncan multiple-range tests within stress conditions.

« Each cell mean In this row is the average (across subjects) of each subject's average response to two nonstressful experiments.' Each cell mean in this row is the average (across subjects) of each subject's average response to four stressful experiments.

and absence of such an effect in the latter condi-tion suggest that debriefing information did exert aslight ameliorative effect. The moderating effectof debriefing information was restricted, however,to the desirable debrief condition (see Table 1).

To examine the simple effects of informationwithin stress conditions, Duncan multiple-rangetests were performed (see Table 1). These anal-yses revealed no significant differences due toinformation within the nonstressful conditions.Within the stressful conditions, however, type ofinformation did affect willingness to participate.Willingness to participate was significantly higherin the rationale condition than it was in either theprocedure or undesirable behavioral expectationconditions. These differences again reflect the pau-city of stress information within the rationale con-dition as compared to the other conditions. Will-ingness to participate was also significantly higherfor the subjects who were never sensitized to unde-sirable behaviors and who knew the purpose behindeach experiment (neutral debrief), as compared tothe subjects who were sensitized to undesirable be-haviors but did not receive the debriefing informa-tion (undesirable behavioral expectation).

The Duncan tests for the simple effects of infor-

mation within the stressful condition also revealed aslight tendency for debriefing information to in-crease consent, such that the neutral, desirable, andundesirable debrief conditions did not differ signifi-cantly from the rationale condition. However,levels of consent within the former conditions didnot differ from those found in the procedure, de-sirable behavioral expectation, and undesirable be-havioral expectation conditions where debriefing in-formation was absent. Thus, debriefing informa-tion was again found not especially effective inincreasing consent levels for stressful experiments.

In the preceding analyses, consent was treated asa continuous variable. Within discussions of in-formed consent, however, it is often conceptualizedas a dichotomous variable: A subject either agreesor refuses to participate. To investigate how eachof the experiments would have fared if the presentrole-play procedure had been used prior to actualexperimentation and a "percentage of refusals" de-cision rule had been formulated and enforced, re-sponses to the willingness-to-participate questionwere rescaled such that "definitely not" responseswere considered instances of consent refusal whilethe remaining responses were considered expressionsof consent. These data are presented in Table 2,

TABLE 2Withholding of Consent Expressed as a Percentage of Decisions

Information level

Type of experiment

NonstressfulStressful

Rationale

33

Procedure

335


028


042

Neutraldebrief

09

Desirabledebrief

07

Undesirabledebrief

016

Note, Observations within any one cell are dependent. Observations reflecting the simple effects of stress within a particular level of informationare dependent.


TABLE 3

"Personal Reaction following Experiment" Means

Type of experiment

Nonstressful0

Stressful6

Information level

Rationale(« = 19)

11.93bo11.20b

Procedure(« - 18)

11.37.8.12ab


(» = 17)

12.410

8.75ab


(n - 22)

9.9807.10.

Neutraldebrief(n = 8)

11.50.10.00.b.

Desirabledebrief

(» - 11)

10.870

9.86.bo

Undesirabledebrief

(n - 11)

11.09.9.62lbo

Note. 2 f= "sadder/less satisfied than before," 10.0 s= scale midpoint, 18 <& "happier/more satisfied than before." Mejans with common subscriptsdo not differ, at the .05 level, as determined by two-tailed (tests between stress conditions and Duncan multiple-range tests within stress conditions.

» Each cell mean in this row is the average (across subjects) of each subject's average response to two nonstressful experiments.•> Each cell mean in this row is the average (across subjects) of each subject's average response to four stressful experiments.

where each figure represents the percentage of con-sent refusal decisions for each condition. For ex-ample, in the nonstressful rationale cell there were19 subjects who judged two experiments, yieldinga total of 38 decisions. Of these 38 decisions, 1was indicative of refusal. Thus, 1/38 X 100 =.03 X 100 = 3%.

Unfortunately, a conceptually meaningful anal-ysis of this measure was difficult due to the re-peated-measures design. Analyses that met statisti-cal canons required further rescaling, with theconsequence that the analyses lost substantive im-port. It will be noted, however, that the resultspresented in Table 2 are similar in pattern to thosepresented in Table I,4 with the exception thatwhen the data are viewed from this perspective,the effects seem substantially more robust withinthe stressful condition. Thus, in the stressful cells,procedural information produced more than 10times the refusals (35% refusals) than rationaleinformation alone (3% refusals). Moreover, withinthe stressful cells a comparison of the refusal ratesin the two behavioral expectation conditions tothose in the procedure cell shows that when thesubjects were given desirable behavioral expecta-tions, their refusal rate dropped by one-fifth (to28%) and that when they had undesirable expec-

*The high-consent refusal rates within the stressful con-ditions, when considered together with the presence of re-liable information and interaction effects in previous anal-yses, underscore and provide a conservative demonstrationof the general caution with which subjects approached therole-playing task. We could have made subjects appeareven more cautious by dichotomizing the original measuresuch that both "definitely not" and "probably not" re-sponses were scaled as instances of consent refusal. Thisscaling problem can, of course, be avoided in future stud-ies by providing only yes versus no as response alternatives.

tations, refusal rates increased by one-fifth (to42%).

Finally, the effects of debriefing seem noticeablymore powerful when one focuses on definite refusals.The pooled rate of refusal for the procedure, de-sirable behavioral expectation, and undesirable be-havioral expectation conditions is 35%. Thepooled rate of refusal in the three debrief conditionsis 11%. This represents a two-thirds reduction inrefusals as a result of debriefing information.While debriefing did not reduce refusals to thelevel observed in the rationale cells (3%), thesedata suggest that the effects of the informationtreatments might be more substantial than exami-nation of the continuous "willingness" data indi-cates.

Baumrind (1971) has suggested that if morethan 5 out of 100 pretest subjects question an ex-perimental procedure the study should not be run.This type of rule considers only subject risk, ofcourse, and does not take into consideration the po-tential merit of the experimentation. It is inter-esting nonetheless that application of such a ruleto our anticipated informed consent rates wouldresult in the stressful experiments not being run;when procedural information about the stressfulexperiments was provided, more than 5% of thesubjects refused consent. This remained true evenin instances in which debriefing information wasadditionally included. The nonstressful experi-ments, however, passed the 5% criterion.

PERSONAL REACTIONS

Baumrind (1964) has speculated that debriefinginformation may embarrass the subject and resultin a loss of self-esteem. To tap this dimension, sub-jects were asked after the description of each ex-


periment: "If you had participated as a subjectin this particular experiment, how do you thinkyou would have felt after it was over?" Responseswere collected on two 9-point scales (r = .75)which ranged from "Happier than before" to "Sad-der than before," and "More satisfied with yourselfthan before" to "Less satisfied with yourself thanbefore." Responses to these two scales weresummed and averaged for the two nonstressful andalso for the four stressful experiments. A score of2 corresponds to "sadder/less satisfied," while ascore of 18 corresponds to "happier/more satis-fied." The relevant means for each condition arepresented in Table 3.

An unweighted-means analysis of varianceyielded both significant information (F = 3.64, df= 6/99, p= .003) and stress effects (F = 74.34,df- 1/99, p < .00001), as well as a significantinteraction effect (F = 3.39, df = 6/99, p = .004).

Tests for the simple effects of stress indicatedthat subjects reliably expressed more favorable re-actions to the nonstressful as compared to thestressful experiments, except in the rationale (t =.81, df = 99, p- .41) and the three debrief condi-tions (neutral debrief: t = 1.67, df = 99,p = .10;desirable debrief: t = 1.12, df = 99, p = .27; un-desirable debrief: t = 1.64, df = 99,p = .10). Thepresence of a stress effect within the procedure,desirable behavioral expectation, and undesirablebehavioral expectation conditions, together with theabsence of such an effect when the debriefing in-formation was included, suggests that debriefinginformation increased the favorability of personalreaction. The debriefing response means are at ap-proximately the scale midpoint, which indicatesthat the subjects believed they would feel aboutthe same after the experiments as they did beforethe experiments.

TABLE 4

"Evaluation of Experiments" Means

Looking at the differences between informationlevels within stress conditions, Duncan multiple-range tests again revealed no significant differencesdue to information within the nonstressful condi-tions (see Table 3). Within the stressful condi-tions, however, one significant difference due to in-formation level emerged. As before, the subjectswithin the rationale condition responded more fa-vorably than the subjects in the undesirable be-havioral expectation condition.

EVALUATION OF EXPERIMENTS

The subjects evaluated each experiment by re-sponding to the following question: "If you hadparticipated as a subject in this particular experi-ment, how do you think you would rate it withrespect to the following dimensions?" Responseswere collected on two 9-point scales (r = .85)which ranged from "worthwhile" to "not worth-while," and "valuable to science" to "not valuableto science." These responses were combined, suchthat the final composite ranged from 2, "not worth-while/not valuable," to 18, "worthwhile/valuable."The relevant means are presented in Table 4.

An analysis of variance again revealed significantinformation (F = 4.82, df = 6/99, p = .0002),stress (F = 7.98, df = 1/99, p = .006), and inter-action (f = 3.85, df=l/99, # = .002) effects.The latter effect, however, was due to the presenceof a stress effect within the undesirable behavioralexpectation condition, and the absence of this effectwithin the remaining information conditions. Nosignificant differences in evaluation between infor-mation levels within the nonstressful conditionswere found ,(see Table 4). Within the stressfulconditions, however, level of information did affectevaluation. Examination of Table 4 will reveal

Type of experiment

Nonstressful*Stressful11

Infor mation level

Rationale(n - 19)

lS.OS0dlS.73bd

Procedure(« - 18)

12.470

10.01.0


(« - 17)

12.590

11.21..


(n - 22)

13.890

9.85.

Neutraldebrief(n - 8)

13.570

13.60lbo

Desirabledebrief

(n - 11)

11.27c12.12.to

Undesirabledebrief

(n - 11)

12.0S0

12.66.bo

Note. 2 » "not worthwhile/not valuable," 10 — scale midpoint, 18 *= "worthwhile/valuable." Means with common subscripts do not differ, atthe .05 level, as determined by two-tailed t tests between stress conditions and Duncan multiple-range tests within stress conditions.

* Each cell mean In this row is the average (across subjects) of each subject's average response to two nonstressful experiments.b Each cell mean In this row Is the average (across subjects) of each subject's average response to four stressful experiments.


that evaluations were reliably more favorable inthe rationale Condition than in either the procedure,desirable behavioral expectation, or undesirable be-haviorarexpectation conditions.

To summarize the findings across dependent mea-sures, responses to nonstressful experiments wereunaffected by variations in the amount and natureof experimental information, while stressful experi-ments were substantially affected by this manipula-tion. Focusing on stressful experiments, the mostfavorable responses were observed within the ra-tionale condition, while the least favorable re-sponses occurred within the undesirable behavioralexpectation condition, with the differences betweenthese conditions being reliable on each dependentmeasure. The favorability of responses within theformer condition suggests that students, in the ab-sence of additional information, hold a generallypositive view of psychological experimentation. Thefavorability of this view often diminished, how-ever, as additional procedural information was pre-sented. The ameliorative effect of debriefing in-formation as measured by "willingness to partici-pate" was weak, being contingent on the presenceof desirable behavioral expectation information.In contrast, the ameliorative effect of debriefing asmeasured both by the percentage of definite refusalsand "personal reaction" was substantially stronger.Indeed, regardless of the presence or nature ofbehavioral expectation information, debriefingraised "personal reactions" to stressful experimentsto about the level associated with nonstressful ex-periments. Finally, on the "evaluation of experi-ments" measure, debriefing information eliminatedthe single stress effect detected; further, regardlessof the presence or nature of behavioral expectationinformation, debriefing raised favorability of evalu-ation to that found in the rationale condition.

The differential effectiveness of debriefing asmeasured by "personal reaction" in contrast to"willingness to participate" is of interest. Whiledebriefing information increased the favorability ofthe subjects' personal reactions to stressful experi-ments to about the level associated with nonstress-ful experiments, such information did not concomi-tantly increase willingness to participate. In termsof the respective scales, debriefed subjects indi-cated on the average that they would feel about thesame after stressful experiments as they did beforethem, while also indicating that "maybe" theywould be willing to participate in such studies.

Thus, within the current study, the fact that thesubjects on the average believed they would notleave stressful experiments with greater anxiety orlower self-esteem than when they first arrived wasnot a sufficient condition for eliciting a high levelof consent. The ethical requirement that the ex-perimenter must not only remove stressful after-effects, but must additionally provide the subjectwith positive gain, would therefore appear judicious.

THE MILGEAM OBEDIENCE EXPERIMENT

While responses across either stressful or nonstress-ful experiments were assessed in the preceding anal-yses, individual reactions to the Milgram experi-ment are of particular interest. These may becompared with consent-related responses obtainedin an experiment conducted by Ring, Wallston, andCorey (1970) and, thus, the predictive validity ofthe current role-play-derived consent estimates canbe roughly assessed.

Ring et al. (1970) essentially replicated Mil-gram's (1963) procedure with the intent of empiri-cally resolving some of the ethical questions sur-rounding the paradigm. After actually participat-ing in the replication, the subjects completed aquestionnaire in which their candid reactions to theexperiment were solicited. Some of the subjectswere given debriefing information before filling outthe questionnaire; others were not. The ques-tionnaire was presented to the subject as an at-tempt to determine "whether any experiments inwhich you've participated in any way violate therights of subjects as stated in the Ethical StandardsCode of the American Psychological Association[Ringetal., 1970, p. 74]."

On a dichotomous consent-related response, 4%of the Ring et al. subjects who had received de-briefing information indicated that they regrettedthey had participated in the experiment; on a re-lated dependent measure, 4% of the debriefed sub-jects indicated the experiment should not be per-mitted to continue. The corresponding percentagesfor subjects who had not received debriefing infor-mation were 43% and 57%, or, on the average,50%. Debriefing, thus, had a substantial ameliora-tive effect on subjects who actually participated inthis replication of the Milgram paradigm.

When the current role-play responses for theMilgram experiment alone are examined via one-way analyses of variance, a highly significant in-


TABLE 5

Role-Playing Responses Obtained for Milgram Study

Variable

Would you have chosen to par-ticipate?"

Percentage checking "definitelynot"

Personal reaction following ex-periment11

Evaluation of experiment"

Information level

Rationale(n = 19)

3.69b

5

11.16b

16.260

Procedure(n = 18)

1.83.

61

6.12a

8.6U


(n - 17)

2.00.

53

6.06a

lO.SSab


(n - 22)

1.68.

77

4.68a

6.96.

Neutraldebrief(n = 8)

3.38b

12

9.38b

13.13b.

Desirabledebrief

(» - 11)

3.00b

9

9.64b

12.64bo

Undesirabledebrief

(n - 11)

3.19b

9

9.5Sb

12.91bo

formation effect on each of the continuous responsemeasures is revealed. The relevant data are illus-trated in Table S. Furthermore, the effect was con-sistent across these measures. Favorability of re-sponse was highest in the rationale condition, wassignificantly reduced with the addition of procedu-ral information (regardless of behavioral expecta-tion information), and was significantly increasedwith the addition of debriefing information (againregardless of behavioral expectation information)such that the debrief cells did not differ signifi-cantly from the rationale condition. Thus, debrief-ing information, as in the Ring et al. replication,had a reliably ameliorative effect on consent-relatedresponses.

A comparison of role-play-derived consent rateswith rates derived upon the Ring et al. replicationadds further support to the predictive validity ofthe role-play procedure. As before, our subjects'responses to the participation question were dichot-omized (see Table S).5 Of the subjects exposedto the cover story, procedure, and debriefing infor-mation in the present study (neutral debrief), 12%

6 The critical reader may wonder why we persisted indichotomizing the original consent measure between "defi-nitely not" and "probably not." On a priori grounds, bothresponses would appear indicative of consent refusal andmight be so scored. Such scoring yields consent refusalrates of about 16% for the nonstressful experiments in therationale and/or procedure conditions. There would, how-ever, seem little for subjects to be genuinely upset aboutsince the experiments appear innocuous in these conditions.Thus, such scoring was judged to be empirically inappro-priate (see Footnote 4).

multiple-range test.

indicated they would "definitely not" wish to par-ticipate; the corresponding percentage for the sub-jects who did not receive debriefing informationwas 61%.e A comparison of these role-playingconsent estimates with consent-related rates ob-tained from actual participants in the Ring et al.study indicates that our estimates are not badly offthe mark. These estimates (12% for debriefedsubjects and 61% for nondebriefed subjects) tendto be on the conservative side such that they areoverestimates of the percentage of subjects whoactually reported participation undesirable in theRing et al. experiment (4% for debriefed subjectsand 50% for nondebriefed subjects). This con-servative bias may be due to the possibility thatthe Milgram (19.63) procedure was more severethan the Ring et al. (1970, p. 77) procedure.

Thus, not only does the proposed role-playingprocedure possess face validity, but, more impor-tantly, the correspondence between the Ring et al.findings and our role-playing estimates suggests theprocedure possesses predictive validity as well. Ifexperiments routinely included an ethics-relatedquestionnaire which could be anonymously com-pleted by the subjects, as in the Ring et al. study,the validity and generality of role-playing-derivedconsent estimates could be better established. Ofequal importance, such questionnaire feedback

8 These role-playing-derived estimates were based onlyon the procedure and neutral debrief conditions since be-havioral expectation information is typically unavailable tothe experimenter.


would provide a stimulus for procedural alteration,should this be deemed necessary.

BACKGROUND MEASURES

To explore the relationship between the disposi-tional variables collected (self-esteem, need for so-cial approval, attitudes toward psychological ex-perimentation, etc.) and consent (scaled as a con-tinuous variable), a number of correlational anal-yses were conducted.

To correct for the effects of the manipulatedstress variable, correlations between the disposi-tional variables and the consent variable were runwithin each stress condition. That is, one set ofcorrelations was run so as to predict the averageconsent response for the four stressful experiments,and a second set of correlations was run in an at-tempt to predict the average consent response forthe two nonstressful experiments. In each of theseanalyses, the average consent response was ex-pressed as a deviation about its associated infor-mation cell mean so as to correct for the simpleeffects of information. In analyzing the consentresponses to the stressful experiments, for example,a rationale condition subject's average response tothe four stressful experiments was computed andwas then expressed as a deviation from the meanresponse in the rationale condition.

Five dispositional variables accounted for 20%of the consent response variance (F = 6.31, df —4/101, p < .0005) for stressful experiments and2% of the variance (F = .52, df = 4/101, ns) fornonstressful experiments:

(a) Attitude toward psychological experimenta-tion was measured as the sum of responses to five9-point scales which were, respectively, bounded bythe descriptors: interesting-dull, worthwhile-notworthwhile, pleasant-unpleasant, valuable to sci-ence-not valuable to science, and soothing-threat-ening. The more favorable a subject's attitude to-ward experimentation, the greater his willingness toparticipate in stressful experiments (r — .38, df —104, p = .0001; nonstressful r = .05, ns).

(b) The more favorable a subject's anticipatedpersonal reaction to experiments as measured bythe sum of responses to four 9-point scales (happierthan before-less happy than before, more satisfiedwith yourself than before-less satisfied with your-self than before, more tense than before-more calm

than before, and more favorable toward psychologythan before-less favorable toward psychology thanbefore), the greater his willingness to participatein stressful experiments (r = .29, df = 104, p =.003; nonstressful r — .07, ns).

(c) and (d) Similarly, the more favorable asubject's attitude toward psychology as measuredby a 9-point scale (extremely favorable-extremelyunfavorable) and the more favorable his attitudetoward psychologists as defined as the sum ofresponses to nine evaluative dimensions (e.g.,trustworthy-untrustworthy, interesting-dull), thegreater his willingness to participate in stressful ex-periments (r = .29, df = 104, p - .003; r = .14,df = 104, p = . 1 5 ) ; for nonstressful experimentsthese variables respectively correlated .11 and .04with consent.

(e) In addition, the higher a subject's self-es-teem, as measured by the Berger scale, the less hetended to be willing to participate in stressful ex-periments (r = —.12, df — 104, p = .22; nonstress-ful r= .04, ns).

Summary

A procedure in which a sample of subjects from apotential population are asked to role play a pro-posed experimental procedure and to judge theirwillingness to participate in that experiment hadthey been fully informed was investigated as analternative to current practices for determiningwhether consent can be anticipated. On the basisof the current study, the procedure appears to pos-sess predictive validity. Validity could be moregenerally ascertained, however, if experimentersregularly administered a postexperimental, ethics-related, questionnaire to the subjects. Responsesto such a questionnaire could additionally be sub-mitted as partial fulfillment of an ethics require-ment, as well as submitted to individuals challeng-ing the ethicality of a study.

In determining whether informed consent wouldbe likely, the procedure would appear to have thevirtue of being less susceptible to error than wouldjudgments made by the principal investigator orby advisory board members who usually differ fromthe subject population on a wide variety of char-acteristics which may be critical for determiningconsent. The procedure has the further advantageof being an inexpensive and quick device for screen-ing a variety of procedures, all of which may be


equally acceptable on methodological grounds, with-out extensive pretesting and actual participation bysubjects. It might be noted that the role-play pro-cedure also may be used to evaluate the adequacyof various "payment" plans for experimental par-ticipation and to determine whether the proposedplan assures "freedom from coercion to participate"(cf. Section 5 of APA Ethical Standards, Cook etal., 1972, p. VIII) for a particular subject popula-tion as opposed to another subject population.

Adequacy of representation of the informationprovided sample subjects may be judged by anadvisory board. Information concerning likelysubject behavior along the socially desirable-un-desirable dimension probably ought to be providedwhere it is known, as well as information concern-ing the hypotheses under test and the purpose ofthe experiment, since these factors affected level ofconsent within the current study.

One problem with the role-play consent proce-dure is that, paradoxically, while deception method-ology is usually used in an attempt to induce"naive" reaction to the independent variables underinvestigation, the procedure may "contaminate" thesubject population for this type of experiment.One possible way to avoid this effect may be toscreen cross-university or cross-city, etc., choosingthe sample from a population deemed to be assimilar as possible to that proposed for actual use(e.g., two midwestern universities).

While use of the proposed procedure would nottotally guarantee that subjects' rights and welfarewould be protected, it might provide a data basethat would reduce the necessity for scientists todeal with subjective, conflicting, and emotionaljudgments on this important problem, as is toofrequently the case.

REFERENCES

BARBER, T. X., & HAHN, K. W. Physiological and subjec-tive responses to pain producing stimulation under hyp-notically suggested and waking-imagined "analgesia."Journal of Abnormal and Social Psychology, 1962, 65,411-418.,

BAUMRIND, D. Some thoughts on ethics of research: Afterreading Milgram's "Behavioral study of obedience."American Psychologist, 1964, 19, 421-423.

BAUMRIND, D. Principles of ethical conduct in the treat-ment of subjects: Reaclion to the draft of the Commit-tee on Ethical Standards in Psychological Research.American Psychologist, 1971, 26, 887-896.

BEROER, E. M. The relation between expressed acceptanceof self and expressed acceptance of others. Journal ofAbnormal and Social Psychology, 1952, 47, 778-782.

BERSCHEID, E., BOYE, D., & WALSTER, E. Retaliation as ameans of restoring equity. Journal of Personality andSocial Psychology, 1968, 10, 370-376.

COOK, S. W., HICKS, L. H., KIMBLE, G. A., McGuiRE,W. J., SCHOCGEN, P. H., & SMITH, M. B. Ethical stan-dards for research with human subjects. (Draft) APAMonitor, 1972, 3(5), I-XIX.

DEUTSCH, M., & KRAUSS, R. M. The effect of threat oninterpersonal bargaining. Journal of Abnormal andSocial Psychology, 1960, 61, 181-189.

FESTINGER, L., & MACCOBY, N. On resistance to persuasivecommunications. Journal of Abnormal and Social Psy-chology, 1964, 68, 359-366.

FHEEDMAN, J. L. Role-playing: Psychology by consensus.Journal of Personality and Social Psychology, 1969, 13,107-114.

MILGRAM, S. Behavioral study of obedience. Journal ofAbnormal and Social Psychology, 1963, 67, 371-378.

ORNE, M. T. On the social psychology of the psychologicalexperiment: With particular reference to demand charac-teristics and their implications, American Psychologist,1964, 17, 776-783.

ORNE, M. T., & SCHEIBE, K. E. The contribution of non-deprivation factors in the production of sensory depriva-tion effects: The psychology of the "panic button."Journal of Abnormal and Social Psychology, 1964, 68,3-12.

RING, K., WALLSTON, K., & COREY, M. Mode of debriefingas a factor affecting subjective reaction to a Milgram-type obedience experiment: An ethical inquiry. Repre-sentative Research in Social Psychology, 1970, 1, 67-88.

U.S. PUBLIC HEALTH SERVICE. Protection of the individ-ual as a research subject. Washington, B.C.: U.S. Gov-ernment Printing Office, 1969.

AMERICAN PSYCHOLOGIST • OCTOBER 1973 • 92S

Documents

Anticipating informed consent: An empirical approach