verbal cues in the mediation of experimenter bias

9
~sycholo~ical Reporis, 1968, 22, 1045-1053. @ Southern Universities Press 1968 VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS' JOHN G. ADAIR AND JOYCE S. EPSTEIN University of ManifoLa Nm York State Division for Youth Summary.-In a replication of Rosenthal's experimenter bias study, 6 male Es obtained significanrly higher ratings when expecting high ratings from their 5 female Ss than when expecting low ratings (n = 5 each) in a person-percep- tion task. The instructions read by these Es were tape recorded and played back in the absence of any E (mon-visual condition) to a second set of Ss. It was con- cluded from the significent bias effect in the non-visual condition that verbal cues are sufficient mediators of E bias. Suggestions were made as to the nature of the bias transmission process in this smdy. A number of studies have demonstrated the existence in psychological re- search of E bias, i.e., Es obtained from Ss the data they expected to obtain in- dependent of any treatment effect. E bias has been extensively studied in the context of person-perception where Es with pre-established biases obtain from their Ss' ratings of photographs of neutral stimulus value. A statistically signifi- cant difference between the ratings obtained by Es expecting positive ratings and those expecting negative ratings has been consistently found in a number of studies (Rosenthal, 1966). Recent investigations have attempted to specify the mode by which the bias effect is uansrnitted. Rosenthal, Fode, Friedrnan and Vikan-Kline ( 1960) exam- ined Ss' perceptions of biased Es and found that extent of bias correlated with such E attributes as friendly, interested, expressive-voiced, and use of hand, head, and arm gestures. More recently Rosenthal and his colleagues (Friedman, 1967; Rosenthal, Friedman, & Kurland, 1966) have taken sound motion piccures of E-S interactions. Analysis cf these pict~~res has revealed a number of significant behavioral variables associated with Es who are successful in obtaining results in line with their expectations. In general, Ss' ratings of biased Es' behavior and observations of Es' behavior via motion pictures have suggested that both kinesic (visual) and paralinguistic (verbal) aspects of the E-S interaction serve to com- municate the bias. In only one previous study, however, has there been an attempt to con- trol Ss' exposure to visual or verbal cues during the experiment and thus to exam- ine the relative contribution of each to the bias effect. Fode (1960; Rosenthal 8: Fode, 1963) resrricted E-S visual contact in one group and Es' verbal commun- ication in a second group and compared Ss' performance in these groups with conditions in which there were no restrictions. Verbal communication was re- 'The study is based in part on a thesis by the second auchor. The s ~dy and the prepara- tion of this paper were supported in pact by Grant APA-166 from the National Research Council of Canada and by a gcmt from the Canada Council to the first author. An ab- breviated version of this paper was presented at the meeting of the Midwestern Psychologi- cal Association, Chicago, May, 1967.

Upload: joyce-s

Post on 16-Apr-2017

233 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

~ s y c h o l o ~ i c a l Reporis, 1968, 22, 1045-1053. @ Southern Universities Press 1968

VERBAL CUES I N THE MEDIATION OF EXPERIMENTER BIAS'

J O H N G. ADAIR AND JOYCE S. EPSTEIN

University of ManifoLa N m York State Division for Youth

Summary.-In a replication of Rosenthal's experimenter bias study, 6 male Es obtained significanrly higher ratings when expecting high ratings from their 5 female Ss than when expecting low ratings ( n = 5 each) in a person-percep- tion task. The instructions read by these Es were tape recorded and played back in the absence of any E (mon-visual condition) to a second set of Ss. It was con- cluded from the significent bias effect in the non-visual condition that verbal cues are sufficient mediators of E bias. Suggestions were made as to the nature of the bias transmission process in this smdy.

A number of studies have demonstrated the existence in psychological re- search of E bias, i.e., Es obtained from Ss the data they expected to obtain in- dependent of any treatment effect. E bias has been extensively studied in the context of person-perception where Es with pre-established biases obtain from their Ss' ratings of photographs of neutral stimulus value. A statistically signifi- cant difference between the ratings obtained by Es expecting positive ratings and those expecting negative ratings has been consistently found in a number of studies (Rosenthal, 1966).

Recent investigations have attempted to specify the mode by which the bias effect is uansrnitted. Rosenthal, Fode, Friedrnan and Vikan-Kline ( 1960) exam- ined Ss' perceptions of biased Es and found that extent of bias correlated with such E attributes as friendly, interested, expressive-voiced, and use of hand, head, and arm gestures. More recently Rosenthal and his colleagues (Friedman, 1967; Rosenthal, Friedman, & Kurland, 1966) have taken sound motion piccures of E-S interactions. Analysis cf these p ic t~~res has revealed a number of significant behavioral variables associated with Es who are successful in obtaining results in line with their expectations. In general, Ss' ratings of biased Es' behavior and observations of Es' behavior via motion pictures have suggested that both kinesic (visual) and paralinguistic (verbal) aspects of the E-S interaction serve to com- municate the bias.

In only one previous study, however, has there been an attempt to con- trol Ss' exposure to visual or verbal cues during the experiment and thus to exam- ine the relative contribution of each to the bias effect. Fode (1960; Rosenthal 8: Fode, 1963) resrricted E-S visual contact in one group and Es' verbal commun- ication in a second group and compared Ss' performance in these groups with conditions in which there were no restrictions. Verbal communication was re- 'The study is based in part on a thesis by the second auchor. The s ~ d y and the prepara- tion of this paper were supported in pact by Grant APA-166 from the National Research Council o f Canada and by a gcmt from the Canada Council to the first author. An ab- breviated version of this paper was presented at the meeting of the Midwestern Psychologi- cal Association, Chicago, May, 1967.

Page 2: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

1046 J. G. ADAIR & J. S. EPSTEIN

moved by requiring S to read the instructions handed to him by the silent E. Vis- ual contact was restricted by a screen which separated E from S while the former read the instructions aloud and recorded S's ratings. Since Es in the non-visual and non-verbal groups had been given high ( f 5 ) biases, their data were com- pared with the ratings obtained by Ss in a -5 bias group in which there were no communication restrictions. The bias effect was obtained in the comparison with the non-visual group and led Fode to conclude that "verbal ales are suf- ficient to mediate E bias."

The extent to which Fode's study conclusively demonstrated the role of verbal cues in transmitting E bias, however, is questioned on two points. First, the procedure employed in the non-visual condition did not effectively eliminate visual factors since E was visible during the pre-instructional phase of the experi- ment. Before stepping behind the screen E "greeted and seated" his S. Thus, the non-visual group was not completely without some visual interaction. The significance of this initial interaction is suggested by a study (Rosenthal, Fode, Vikan-Kline, & Persinger, 1964) which indicates that at least part of the bias transmission may occur before the instructional phase of the experiment actually begins.

In addition, the appropriateness of the -5 bias control group for testing the significance of the bias effect for the non-visual group is questioned. A -5 bias non-visual group would have been a more comparable treatment condition for this analysis.

The present study, therefore, is an attempt to improve upon the design to examine the role played by verbal cues in the transmission of bias by the re- moval of visual cues from all phases of the experiment, and by providing for both high and low bias groups in the non-visual condition. It is hypothesized that under these conditions, support for the mediation of bias by verbal cues alone would not be obtained.

Experimenters and Silbjects

Six male college scudents from an undergraduate social psychology class vol- unteered to serve as Es. At the time of recruitment Es were told by their instruc- tor (JGA) that their performance would reflect on their abiliry to conduct re- search and that participation in this study would provide valuable experience. Each E tested 10 female introductory psychology students on a person-perception task. An additional 60 female Ss were tested in the same task by means of tape recordings of Es' voices. Both Es and Ss were students at the University of Mani- toba.

The Photograph-judging T a ~ k

The person-perception paradigm previously employed by Rosenthal was used. This task required S to race the degree of success or failure experienced

Page 3: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

VERBAL CUES IN EXPERIMENTER BIAS 1047

by each photographed face in a set of ten pictures. The ratings were made on a scale ranging from +lo, indicating the person has experienced extreme suc- cess, to -10, indicating the person pictured has experienced extreme failure. N o neutral responses were permitted.

The stimulus materials were photographs standardized after the procedure suggested by Rosenthal (Rosenthal & Fode, 1963). From an original collection of 46 photographs, 10 of neutral stimulus value were selected (Epstein, 1966). Each of the selected photographs had an average rating of less than f 1.50 and the sum of the ratings of the 10 selected photographs was zero.' Since E was not able to handle the pictures in the non-visual condition, the 10 photographs were mounted in a row on a poster-board strip. Each of the pictures was identi- fiable by a number placed beneath it on the poster-board.

Vi~zlal Conditions Each E volunteered for two separate hours of participation. At the first

hour, one-half of the Es were biased to expect average ratings of + 5 from their Ss while the remaining Es were biased to expect -5 ratings. The bias for each E was reversed for the second 1-hr. session which was held approximately 1 wk. later. Biases were induced by instructions to E that previous research of this nature indicated chat the type of Ss they would test had previously averaged a

+5 (-5) rating and that they should therefore expect their Ss to average at that level. The instructions ro E were similar to those used previously (Rosenthal, 1964) and were read by the junior investigator who also assigned Es to the treat- ment conditions.

At the first session, each E tested five Ss individually. Ss were assigned to a given E and condition in the order they signed up for the experiment. After seating S, E read the standard instructions for the task which were modified only slightly to accommodate their subsequent presentation in the non-visual condi- tion.

I am going to read you same instructions. No questions can be answered about this experiment. OK?

W e are in the process of developing a test of empathy. This test is designed to show how well a person is able to put himself into someone else's place. Before you are a series of photographs. For each one I want you to judge whether the person pictured has been experiencing success or failure. To help you make more exact judgments, you ate to use the rating scale on your left. As you can see, the scale runs from - 10 to + 10. A rating of -10 means that you judge the person to have experienced extreme failure. A rating of + 10 means chat you judge the person to have experienced extreme success. A rating of -1 means that you judge the person to have experienced mild failure while a rating of + l means that you judge the person to have experienced mild success. You are to rate each photo as accurately as you can. Just mark down the rating you assign to each photo on the form provided. Indicate the number of the rating and the sign. All ready? You may begin.

"A test-retest reliability coefficient of .80 (0 < ,001) was recently obtained with the task administered at approximately 2-day intervals by Es without induced bias.

Page 4: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

1048 J. G. ADAlR & J. S. EPSTEIN

In contrast to earlier studies, S was provided a copy of the rating scale and a sheet on which he recorded his responses. Although the 10 photographs were simultaneously presented on a poster-board strip, Ss were encouraged to make their judgments in the numerical order of the pictures. At the second session E was given the opposite bias and tested five more Ss. Thus, 30 Ss were tested under each bias condition.

Non-visual (Taped) Conditions The instructions read by Es in the visual conditions were monitored by a

concealed tape recorder. Concealment of the tape recorder was necessitated by the observation (Rosenthal, Persinger, Mulry, Grothe, & Vikan-Kline, 1964) that monitoring E's behavior may remove the usual bias effect. There was no evi- dence to suggest that either Es or Ss were at any time aware of the recorder's presence.

These tape-recorded instr~lctions were then played to another 60 Ss who comprised the non-visual conditions. Ss were assigned, in groups of five, to each E in the manner employed in the visual condition with 30 Ss per bias condition. Ss were brought into the experimental room by a prominently placed sign ask- ing them to be seated and informing them that a tape-recorded message would give them their instructions. The junior investigator operated the tape recorder from behind a one-way vision screen and presented the taped instructions when S was seated. S was required to write her responses on the record sheet since E was not present,

S's Ratings of E's Behavior At the conclusion of each session, S was asked to go to another room to com-

plete a modified form of the rating scale for recording S's perceptions of biased E's behavior (Rosenthal & Fode, 1963). I t was expected that S's perception of her E's behavior would be related to the extent to which S was influenced by E. Unfortunately, since Ss were given credit for participating in the experiment be- fore this request was made and since the second experimental room was some distance from the first, only a little more than one-half of the Ss completed the questionnaire.

RESULTS The mean rating of the photographs obtained by each E under the four

treatment conditions and the mean for each condition are presented in Table 1. Since the same E was used in all four treatment combinations, it seemed best to conceptualize the experiment as one in which a two-factor study, bias by mode of presentation, was replicated six times. The summary of this analysis is presented in Table 2.

The hypothesis that the bias effect would be present only in the visual con- dition was rejected on the basis of a non-significant Bias X Mode of Presentation interaction. Rejection of the hypothesis was further supported by a significant

Page 5: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

VERBAL CUES IN EXPERIMENTER BIAS 1049

TABLE 1 MEAN RATINGS OF PHOTOGRAPHS OBTAINED BY BIASED ES IN

VISUAL AND NON-VISUAL CONDITIONS

E Visual Non-visual +5 -5 +5 - 5

El 3.68" 1.44 5.36 3.08 Ez 3.42 1.46: 5.54 4.18 E 3 3.58 1.04" 6.44 4.92 E4 1.28 1.40* 4.52 -0.08 EL 1.94* 2.16 2.08 2.98 Eo 2.28" 1.40 1.78 1.52 M 2.70 1.48 4.29 2.77 S D 1.74 1.44 2.47 2.27

"Bias first induced in E.

main effect for bias ( p < .05), indicating Es obtained higher ratings from their Ss when a high expectancy was induced than with a low expectancy for both the visual and non-visual (taped) conditions.

Differences between Es were obtained as a function of the mode of pre- senting the instructions ( 9 < .05). This interaction, presented in Fig. 1, was the result of three Es obtaining much higher ratings in the non-visual (taped) condition than in the visuzl condition. These Es also obtained greater bias than

TABLE 2 ANALYSIS OF VAR~ANCE OF MEAN RATINGS OBTAINED BY ES

Source df MS F P Mode of Presentation (M) 1 61.94 7.18 <.05 Bias (B) 1 56.05 8.72 <.05 Experimenters (E) 5 19.65 M X B 1 0.69 .12 M X E 5 8.62 3.05 <.05 B X E 5 6.43 2.28 M X B X E 5 5.99 2.12 Within 96 2.82 Total 119

the other three Es under both conditions, the mean difference between high and low bias mean ratings being 3.97 for good biasers and 1.48 for poor biasers. These differences were not attributable to differences in the order of bias induce- ment.

Fig. 1 also shows two other significant differences in the ratings obtained from Ss with the taped instructions. Under the non-visual conditions mean rat- ings were generally higher (F = 3.05, p < .05) and showed a greater variabil- ity between Es' mean ratings than in the visual condition. While the greater var-

Page 6: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

1050 J. G. ADAIR & J. S. EPSTEIN

0 I VISUAL WN-VISUAL

FIG. 1. Mean ratings of photographs obtained by each E in the visual and non- visual (taped) conditions

iability between Es may be accounted for by variations in the quality of the tapes, speaking ability of Es , and background noise, intuitively, however, it would be expected that the ratings obtained in the taped condition would be more con- sistent within a given S than in the visual condition. This assumption is based on the fact that the only cues transmitted from E co S in the taped condition come from E's voice at the beginning of the session. S's responses in the visual condition, however, should be more variable as S continues to receive cues from her interaction with E during the picture presentacion. This hypothesis was sup- ported by a greater within& variability in the visual ( S D = 4.80) than in the non-visual ( S D = 3.94) condition ( t = 3.185), df = 118, p < .001). In addition, a comparison of the intercorrelations among the 10 pictures indicated a greater mean intercorrelation for the non-visual condition (.209) over the visual condition (.047) which was significant beyond the ,001 level ( t = 3.76, df = 8 8 ) . This indicated a greater consistency of responding within Ss in the taped condition and supported the hypothesis presented above.

Page 7: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

VERBAL CUES IN EXPERIMENTER BIAS 105 1

Correlations of the extent of bias with S s perception of biased E's behavior yielded the significant correlations indicated in Table 3. The extent of bias was defined as the difference between 5"s mean rating of the photographs and the grand mean rating of the photographs obtained from Ss under the opposite bias by the same E (Friedman, Kurland, & Rosenthal, 1965). In other words, if E had a +5 bias, the grand mean rating of Ss run by that E under a -5 bias were subtracted from S s mean rating; if E had a -5 bias, S's rating was subtracted from the grand mean for rhat E obtained under a +5 bias.

TABLE 3 PEARSONIAN CORRELATIONS OF EXTENT OF BIAS WITH S'S PERCEPTIONS

OF BIASED E's BEHAVIOR

E is Visuals Non-visual$

Satisfied with the experiment Personal Relaxed Interested Courteous Pleasant-voiced Expressive-voiced Pleasant Use of head gestures Expressive-faced

$Correlations differ significantly from zero at the .05 level *, .02 level t, . O 1 level $ by t rest, two-tailed; df = 38 for visual and 30 for non-visual conditions.

The significant correlztions in Table 3 indicated that E who was perceived by his Ss in the visual condition to be more pleasant, relaxed, interested, cour- teous, satisfied with the experiment, expressive-faced, and pleasant- and expres- sive-voiced tended to bias his Ss more. The use of head gestures was also associ- ated with greater bias in the visual conditions. Only two variables correlated significantly with extent oP bias in the non-visual condition. The more per- sonal and expressive-voiced E, the more biased data he obtained from his Ss. While rhese latter correlations are consistent with common sense expectations, the lack of consistency between the correlations for the visual and non-visual con- ditions is noticeable.

These significant relationships should be interpreted with care, however, as the correlations are based on responses from only 40 Ss in the visual condi- tion and from 32 Ss in the non-visual condition. The need for this caution is reinforced by a significantly greater extent-of-bias score for Ss who failed to com- plete the rating scale than for Ss who completed the scale under both modes of presentation ( F = 5.07, df = 1/116, p < .05). Failure to find a greater num- ber of completions of the rating scale associated with Es who more readily ob- tained bias, however, suggested that this latter result was probably due to the

Page 8: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

1052 J. G . ADAIR & J. S. EPSTEIN

procedure for obtaining the data and not directly related to E's ability to influ- ence S.

DISCUSSION The results clearly support Fode's (1960) conclusion that verbal cues are

sufficient to mediate E bias. Even with all visual cues eliminated by physically removing E, the biasing effect was transmitted by the auditory stimuli from the instructions read by E. Although the paralinguistic cues which were appar- ently used to communicate bias were not specified in the present study, some s~~ggestions as to the nature of the verbal bias transmission process may be made from examination of the effect of the taped instructions on S s ratings.

In general, under the non-visual conditions Ss' ratings were higher and less variable than under the visual condition. The significantly less variable re- sponses elicited by the pictures and the greater intercorrelation among the pic- tures in the non-visual conditions indicated that S's responses in these conditions were consistently restricted to a fairly narrow range of ratings. This suggested that the bias was verbally transmitted by a single cue or a single set of ales in the initial instructions to S. These cues, discerned by S at the very beginning of the experiment, served as a general impression or frame of reference for guid- ing her responses to all of the pictures. The extent to which these verbal cues were attended to by S was apparent from the significant correlations of extent-of- bias scores with ratings of the expressiveness of E's voice which were found to be significant in both the live and taped conditions.

These suggestions as to how the bias transmission process operates are con- sistent with the recent observation (Duncan & Rosenthal, 1966), based on para- linguistic analysis of E's instruction-reading behavior, that ratings are of more successful or failing people if the corresponding response alternative is given greater vocal emphasis by E during reading of these portions of the instructions. Barber and Calverley ( 1964) similarly found manipulations of E's tone of voice produced significant differences in S s level of hypnotic-like suggestibility.

The taped instructions were also uniform in producing higher ratings in both positive and negative bias conditions. Possibly the absence of E increased S's anxiety which in some undetermined manner led to higher ratings. Simi- larly, a "Hawthorne effect," i.e., the perception by S of special treatment being accorded her by experimentation without the E being present, could have affected S's ratings. In any event, the elevated ratings appeared to be an artifact of the tape procedure. It should be noted, however, that even automated procedures may have effects on Ss which need to be considered in data interpretation.

In both the visual and non-visual conditions, the correlations between extent- of-bias scores and S's ratings of E's behavior were consistent with previous re- sults obtained by Rosenthal and his colleagues and with common sense expecta- tions. Even the discrepancies between visual and taped conditions in the direc- tion and magnitude of the correlations are consistent with the previous finding

Page 9: VERBAL CUES IN THE MEDIATION OF EXPERIMENTER BIAS

VERBAL CUES IN EXPERLMENTER BIAS 1053

(Rosenthal, 1766) of "channel discrepancies" when judgments are made on the basis of only single modes of communication.

The noticeable differences among Es in the ratings of the photographs ob- tained from their Ss were sufficient to suggest chat Es in the present sn~dy came from two distinct populations. Although the lack of an adequate sample pre- cluded a special analysis of Ss' ratings of Es' behavior, future research should con- sider the extent to which individual differences in biasing ability are determined by certain E attributes, attitudes, and behavior, and more basically, the extent to which this ability is general across experimental sessions and varieties of casks.

In conclusion it shodd be noted that the use of taped instructions is a val- uable method for gaining better experimental control in studies of the E-bias phe- nomenon. Controlled presentation of instructions enables better examination of the important variables involved in the visual transmission of the bias effect and more complex designs involving controlled visual and verbal interaction will lead to understanding of the complex contribution of these variables.

REFERENCES BARBER, T. X., & CALVERLEY, D. S. Effect of E's tone of voice on "hypnotic-like" sug-

gestibility. Psychological Reporlr, 1964, 15, 139-144. DUNCAN, S., & ROSENTHAL, R. Vocal emphasis in experimenters' instruction reading as

unintended determinants of subjects' responses. Unpublished manuscript, Univer. of Chicago, 1966.

EPSTEIN, J. S. The role of verbal cues in the mediation of experimenter bias. Unpub- lished master's thesis, Univer. of Manitoba, 1966.

FODE, K. L. The effect of non-visual and non-verbal interaction on experimenter bias. Unpublished master's thesis, Univer. of North Dakota, Grand Forks, 1960.

FRIEDMAN, N. The social nature o f psychological research: the psychological experiment as a socMl interaction. New York: Basic Books, 1967.

FRIEDMAN, N., KURLAND, D., & R O S E N ~ A L , R. Experimenter behavior as an un- intended determinant of experimental results. Jorrrnal of Projeclive Techniques and Personality Assessnzent, 1965, 29, 479-490.

ROSENTHAL, R. The effect of the experimenter on che results of psychological research. In B. Maher (Ed.), Progress in experimental personality research. Vol. 1. New York: Academic Press, 1964. Pp. 80-1 14.

ROSENTHAL, R. Experjmentcr effec;s in behavioral research. New York: Appleton-Cen- mry-Crofts, 1966.

ROSENTHAL, R., & FODE, K. Psychology of the scientist: V. Three experiments in ex- perimenter bias. Psychological Reports, 1963, 12, 491-511.

ROSENTHAL, R., FODE, K., FRIEDMAN, C., & VIKAN-KLINE, L. Subjects' perception of their experimenter undrr conditions of experimenter bias. Perceptrral and Motor Skills, 1960, 11, 325-33:.

ROSENTHAL, R., FODE, K., VIKAN-KLME, L., 8: PERSINGER, G. Verbal conditioning: mediator of experimenter expectancy effects. Psychological Reports, 1964, 14, 71- 74.

ROSENTHAL, R., FRIEDMAN, N., & KURLAND, D. Instruction-reading behavior of the experimenter as an uninrended determinant of experirnenral results. Journal of Ex- perimental Personaliiy Research, 1966, 1, 221-226.

ROSENTHAL, R., PERSINGER, G., MULRY, R., GROTHE, M., & VIKAN-KLME, L. Em- phasis on experimental procedure, sex of subjects, and the biasing effects of experi- mental hypotheses. Iou~nal of Projective Techniques and Personality Assessment, 1964, 28,470-473.

Accepted April 15, 1968.