AN INVESTEGMWN 65 EXIMINER
INFLUENQE 0R WEEKSEER ENTELLEGENCE

SCALE 50R CHILD-BER SCORES,

Ykosts foo Hm Degree cg: D“. D.
MECHEGAR SM’EE UNEVERSII‘Y

WiEkiam Harvey Gikiingham
i970

 

THES‘S

 

This is to certify that the

thesis entitled

An Investigation of Examiner
Influence on Wechsler Intelligence
Scale for Children Scores

presented by

William Harvey Gillingham

has been accepted towards fulfillment
of the requirements for

m degree in_C.Qu.D.§_¢_Ll_Dg , Personnel

Services and Educational PsychOIOgy

we z. 0W;

Major professo
Date l/l 9/7 0

 

 

M W?" Ell-9W
RWV

We”

.a‘ I) tb'rsfli‘l

00/5;
1

A /.,

.7!
I

I» _-

 

 

ABSTRACT

AN INVESTIGATION OF EXAMINER INFLUENCE 0N WECHSLER
INTELLIGENCE SCALE FOR CHILDREN SCORES

BY

William Harvey Gillingham

A. The Problem

Discussion of the possibility of observer influence has
recently become controversial largely as a result of the work of
Robert Rosenthal (1966; 1967a). Observer influence has been
examined by studying different types of observers such as inter-
viewers, research experimenters, teachers, and projective and intelli-
gence test examiners. This thesis was a study of examiner influence
in that it looked at the person who administered the Wechsler Intelli-
gence Scale for Children (WISC) as a possible source of undesired

variation of intelligence test scores.
Be “Sign

The sample consisted of four male and four female "180 exami-
ners who had just completed VISC training. Each examiner tested
eight junior high school students (four boys and four girls) who were
randomly selected from a population of students who were "average” in
intelligence (that is, they scored from ninety to one hundred-ten on

the California 'rest of Mental Maturity-Short Form, 1963 Revision).

William Harvey Gillingham

Eight examinees were randomly assigned to each examiner and then
they were randomly designated 'above average“ or “below average“ so
that each examiner tested two "above average'I boys, two "below average“
boys, two Iabove average“ girls, and two 'below average' girls. The
order in which an examiner tested his examinees was also randomized.
Parental approval for testing was secured and confidentiality was
assured. Neither parents nor examiners were told that they were per-
ticipating in a study or an experiment. All WISC's were administered
in private offices during a one week period.

A rating sheet was used by the grand experimenter to transmit
an expectancy condition to examiners. On each examinee's rating sheet
a discrepancy between California Test of Mental Maturity-Short Form
score and school achievement was fictitiously indicated, and a 'pre-
dicted WISC score" was advanced.

Analysis of the data was by a mixed model, four-way analysis of
variance having three fixed variables (expectancy, sex of examines,
and sex of examiner), and two random variables (examiners nested in
sex of examiner, and replications nested in all other variables). The
five per cent level of confidence was arbitrarily chosen for signifi-

cance tests.

0. Analysis of Results

Both the sex of examiner and sex of examines effects were sig-
nificant (p(.05). Female examiners obtained higher mean WISC scores
and male examinees achieved higher mean WISC scores.

The expectancy effect was not statistically significant. It

was feared that "130 examiners did not retain the expectancy condition

William Harvey Gillingham
given to them by the grand experimenter and, therefore, examiners'
expectancy was neither transmitted to nor received by examinees. It
was also felt that the W180 was a structured and factual experimental
task and that the W150 examiners were relatively well-trained and
experienced. Neither the examiner effect nor any of the interaction
effects was significant.

The results of this study pointed to the conclusion that exami-
ner influence was not a great problem in intelligence testing and that
observer influence was difficult to demonstrate when experienced

observers administered structured tasks.

AN INVESTIGATION OF EXAMINER INFLUENCE ON WECHSLER

INTELLIGENCE SCALE FOR CHILDREN SCORES

By
William.Harvey Gillingham

A THESIS
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Counseling, Personnel Services
and Educational Psychology

1970

G é « 7 37,2.»
7- /.,7 0

ACKNOWLEDGEMENTS

The writer is especially indebted to the late Buford Stefflre
under whose direction this thesis originated. Dr. Stefflre's impact
on the writer's life is continually felt. Special gratitude is also
felt for Dr. Robert Craig who kindly and patiently assumed an advisory
role, and to other committee members, Dr. Donald Grumman and Dr. Fred
Vescolani.

The writer would also like to acknowledge his wife, Judy, for
her support, understanding, and unselfishness throughout his doctoral

program.

ii

TABLE OF CONTENTS

CHAPTER

I 0 THE PROBLEM O O O O O O O O O O O 0

Need for the Study .

Purpose of the Study . . . . . .
Hypotheses . . . . . . . . . . .
Definition of Terms . . . . . . .
Overview of the Study . . . . . .
II. REVIEW OF RELATED RESEARCH . . . .
Examiner Influence . . . . . . .
Experimenter Influence . . . . .
Discussion of PreVious Research .
III. DESIGN . . . . . . . . . . . . . .
The Sample . . . . . . . . . . .
Instrumentation . . . . . . . . .
Procedure . . . . . . . . . . . .
Design and Analysis . . . . . . .
Summary . . . . . . . . . . . . .
IV. ANALYSIS OF RESULTS . . . . . . . .
Hypotheses and Results . . . . .
Discussion of Results . . . . . .
Summary . . . . . . . . . . . . .

iii

PAGE

(DODKJOka

11
23
26
26

27

51
55
37

57
#2
#9

CHAPTER
V. SUMMARY, CONCLUSIONS, AND IMPLICATIONS
FURTHER RESEARCH . . . . . . . . . .
Summary ..............
Conclusions ............
Implications for Further Research .

BI BLI OGRAPHY O O O O O O O O O O O O O O I 0

FOR

APPENDIX A. TABLE 6. CTMM MEANS AND STANDARD

DEVIATIONS FOR ALL SUBGROUPS . . . .
APPENDIX Be Rating Sheet e e e e e e e e e

APPENDIX C. TABLE 7. WISC TOTAL SCORES . .

iv

PAGE

52
52
59
62
6h

71
75
74

LIST OF TABLES

TABLE

1. SOURCES OF VARIATION AND DEGREES OF FREEDOM . . . . . . .

2. SUMMARY OF MEAN TOTAL NISC SCORES, STANDARD DEVIATIONS,
AND F-RATIOS FOR MAIN EFFECTS . . . . . . . . . . . . .

3. SUMMARY OF MEAN TOTAL VISC SCORES, STANDARD DEVIATIONS,
AND F-RATIOS FOR TWO-WAY INTERACTION EFFECTS . . . . .

4. SUMMARY OF MEAN TOTAL NISC SCORES, STANDARD DEVIATIONS,
AND F-RATIO FOR THREE-WAY INTERACTION EFFECT . . . . .

5. SUMMARY or mus, STANDARD DEVIATIONS, AND r-m'rms
FOR SIGNIFICANT EFFECTS O O O O O O O 0' O O I O O 0 O O

6. APPENDIX A. CTMM MEANS AND STANDARD DEVIATIONS
PORALLSUBGROI’PSOOOOOOeOOOOOOeeeee0

7. APPENDIX C. WISC TOTAL SCORES . . . . . . . . . . . . .

PAGE

54

58

42

71
7h

LI ST OF FIGURES

FI GU RE PAGE

1 O R...‘r°h D0318!) O O O O O O O O O O O O O O O O I O O O O 52

vi

CHAPTER I

THE PROBLEM

In this chapter the need for the study is developed and then
the purpose of the study is specifically stated. The research hypo-
theses are stated and certain terms are defined. The chapter is con-

cluded with an overview of the study.

A. Need for the Study

The possibility that observers may influence what they observe
or may be influenced in how they record what they observe has been
discussed for years (Kintz, Delprato, Mettee, Persons, &.Schapps,
1965). Discussion of the possibility of observer influence has
recently become controversial largely as a result of the work of
Robert Rosenthal (1966, 1967a). Observer influence has been examined
by studying different types of observers. Thus, Rosenthal (1966, 1967a)
studied the research experimenter to determine the existence of
I'experimsnter influence.” Rosenthal and Jacobson (1966, 1968)
examined the possibility of "teacher influence'I by attempting to see
if teachers influenced pupils' scores on intelligence tests. Socio-
logists (Suchman, 1962; and Wilkie, 1965) looked at the sociological
interviewer as an observer to determine the possibility of an "inter-
viewer influence” effect. The possibility of "examiner influence“ has
been checked by studying the examiners of individual intelligence

1

2

tests (Cieutat, 1965; Cohen; and Larrabes and Kleinsasser, 1967),
and by studying the examiners of projective tests (Simmons and Christy,
1962; Turner and Coleman, 1962; and Rosenthal, 1965b).

Rosenthal (1966; 1967a) is convinced that observer influence
(particularly experimenter influence and teacher influence) exists.
He claims to have repeatedly demonstrated experimenter influence in
experimental studies. His data and their interpretation have been
sharply criticized (Ingraham and Harrington, 1966, 1967; Barber and
Silver, 1968a, 1968b; Thorndike, 1968; Barber, Forgione, Chaves,
Calvsrlsy, McPeake, & Bowen, 1969; and Claiborn, 1969). Some critics
concede the possibility of experimenter influence but question the
pervasiveness of observer influence in more structured situations
such as individual intelligence testing. Other criticisms center
around shortcomings in Rossnthal's research designs and his statisti-
cal treatment of data.

The existence of observer influence needs to be examined in
other than experimenter influence studies. An area of concern for
the writer has been the use of individual intelligence tests to make
important and sometimes irreversible decisions about people. If
this practice continues it seems imperative that the possibility of
examiner influence be investigated. Furthermore, future studies of
observer influence of all kinds need to be statistically well-designed
so that they do not meet with the same criticisms leveled at Rosenthal.
In short, there seems to be a need for a well-designed study to
investigate the existence of examiner influence on individual intelli-

gence test scores.

B. Purpose of the Study

The purpose of this study was to provide a well-designed field
testing of experimental research results. Specifically, this study
looked at the person who administered the Wechsler Intelligence Scale
for Children as a possible source of undesired variation of intelli-
gence test scores. The research design also provided for an exami-
nation of variables that had been rarely isolated in previous research.
In this design, test examiners were given both expectancy conditions
in random order; that is, examiners were led to believe that half of
their examinees were “above average" in intelligence and half were
"below average" in intelligence. The sex of examiner and sex of
examines variables were crossed so that male examiners tested male and
female examinees and female examiners tested male and female examinees.
The examinees were junior high school students and therefore, were
much younger than the examiners who were graduate students at Michi-

gan State University.

C. Hypotheses

Eight hypotheses were generated from the design of this study.

hypothesis 1

 

Female examiners will obtain a higher mean intelligence test
score than male examiners. Some research (Cieutat, 1965) suggests
that female examiners of individual intelligence tests obtain higher

scores from examinees.

Hypothesis 2

 

There will be no difference between mean intelligence test
scores achieved by boy and girl examinees. No research was found that
would suggest that male examinees score higher or lower than female

examinees on the Hechsler Intelligence Scale for Children.

Hypothesis 3

 

Examinees who are expected by examiners to be above average
will achieve a higher mean intelligence test score than examinees
who are sXpected to be below average. This hypothesis is a test of
the expectancy effect which has been experimentally demonstrated
(Larrabeo and Kleinsasser, 1967; Rosenthal, Pods, Friedman, and
Vikan-Kline, 1960; Rosenthal, Persinger, Vikan-Kline, and Mulry,
1963; Rosenthal, Mulry, Parsinger, Vikan-Kline, and Grothe, 1964a,
1964b; and Friedman, Kurland, and Rosenthal, 1965). This hypothesis

will be directionally tested.

Hypothesis k

 

There will be no reliable difference between mean intelligence
test scores obtained by the examiners. Since examinees were randomly
assigned, there is no reason to expect that examiners will obtain
different mean intelligence test scores. If this test is statisti-
cally significant, it probably indicates that some examiners regu-
larly assign scores that are higher than scores assigned by other

examiners.

Hypothesis 5

 

Opposite sex combinations of examiners and examinees will

5

obtain higher mean intelligence test scores than same sex combina-
tions. There is sufficient research (Stevenson, 1961; Stevenson and
Allen, 1964; Cieutat, 1965; Hill and Stevenson, 1965; and Rosenthal,
1967a) to support a directional hypothesis concerning the sex of

examiner interacting with the sex of examines variable.

Hypothesis 6

 

There will be no interaction effect between the sex of the
examiner and the extent to which the obtained scores are in the
direction suggested by the expectancy set. It cannot be determined
whether male or female experimenters are more susceptible to an
expectancy effect. Rosenthal, Mulry, Persinger, Vikan-Kline, and
Grothe (196ha) found that male experimenters are 'better biasers,"

but a directional hypothesis is not advanced.

Hypothesis‘?

 

There will be no interaction effect between the sex of the
examinee and the extent to which the obtained scores are in the direc-
tion suggested by the expectancy set. Again, a directional hypothesis
is not advanced because of insufficient research to support the con-
ceptualization that one sex of examinees is more influenced by the
direction of the expectancy than the other sex of examinees. In two
studies (Rosenthal, Mulry, Persinger, Vikan-Kline, and Grothe, 1964a;
and Stevenson and Allen, 1964), it was found that female subjects
were easier to influence than male subjects. As in Hypothesis 6, a
directional hypothesis is not formulated because it is felt that the

subjects' age may have accounted for the results found in the studies

6

cited. That is, a young male examiner may be better able to influence

a college coed than an eleven year old junior high school girl.

Hypothesis 8

 

There will be no interaction effect among the sex of examiner,
sex of examines, and direction of expectancy variables. This hypo-
thesis is advanced to determine the existence of a three-way inter-

action effect.

D. Definition of Terms

The following terms need to be defined to assure common meaning.

Expectancy

 

Experimentsrs may have a preconceived idea concerning the
results of their research. A test examiner may have a preconceived
idea concerning the test score of the subject he is testing. That is,
he has an "expectation” of what the outcome will be. For the pur-
pose of the study, an "expectancy" was believed to exist when the
grand experimenter (the writer) gave test examiners an esthmate of

the ability of each subject tested.

Experimenter of Examiner Influence

 

"Experimenter or examiner influence" involves a conscious or

unconscious effect on the outcome of research or test scores.

E. Overview of the Study

In the next chapter, a review of recent (1960 to present)

research on experimenter and examiner influence is presented. The

7

reader will be introduced to the research that has been done on the
experimenter and examiner variables. He will also be shown that the
present study represents a necessary extension of the previous research.
The research design (sample, instruments, procedure, and
analysis) is discussed in Chapter III and results are presented in
Chapter IV. Chapter V'includes a summary, conclusions, and implica-

tions for further research.

CHAPTER II

REVIEW OF RELATED RESEARCH

Chapter II contains a review of recent (1960 to present)
literature concerning examiner influence on individual intelligence
test scores and on projective test scores, research on experimenter
influence, and a discussion of previous research. Studies of exami-
ner influence on individual intelligence and projective test scores
were included because of their relevance to this study's design in
which an individual intelligence test was used. These studies were
conducted to determine whether physical or personality characteris-
tics of test examiners influenced the test scores they obtained.

The rest of the studies in Chapter II were concerned with
Robert Rosenthal's (1967a) contention that experimenters influence

the results of their research.

A. Examiner Influence

Examiner Influence on Individual Intelligence Test Scores

 

Very little research was found concerning examiner influence
on intelligence test scores. Cieutat (1965) also noted this lack of
research. In his study, seven male and six female examiners tested
24} boys and girls with the Stanford-Einst. Cieutat found that
female examiners obtained significantly higher (pg .001) intelli-
gence test scores from subjects. Analysis of variance revealed a

8

9
significant interaction effect (p5 .05) between sex of examiners and
sex of subjects. Highest performances were obtained when examiners
tested epposite sex subjects. Cohen (1965) also found a significant
(pg .005) examiner effect using the Vechsler-Bellevue. Cohen believed
that examiner influence reduced the subtest validity of the Wechsler-
Bellevue.

Exner (1966) studied the effect of "rapport building" on
Stanford-Binet intelligence test scores. Twenty-five subjects took
the test after substantial attempts were made by examiners to "build
rapport! The other twenty-five examinees took the test without pre-
liminary "rapport building." The experimental group scored signifi-
cantly higher (p-‘. .001) than the control group.

Wartenbsrg-Ekren (1962) found no significant difference on
Block Design scores of the wschslsr Adult Intelligence Scale when
eight examiners gave that subtsst to two examinees who were allegedly
earning higher grades than the other two examinees. However, Larra-
bee and Kleinsasser (1967) found a significant difference when five
examiners administered the Wechsler Intelligence Scale for Children
to twelve sixth graders. Each examines was tested by two examiners:
one examiner administered the even items; the other examiner admini-
stered the odd items. One examiner was told that the examinee was
"above average"; the other examiner was told that the examinee was
"below average." On the Verbal part of the test the difference was
over ten IQ points (p5 .05).

Friedman (1967) reviewed literature concerning examiner dif-
ferences in testing and found that the examiner's and subject's race

made a difference in testing. Same race combinations obtained best

10
scores. He noted, however, the lack of research on the examiner
variable in testing. He pointed out that most subjects were still
tested by only one examiner. He also suspected an examiner expectancy
effect in testing because most examiners review the examinee's records

before testing.

Examiner Influence on Projective Test Scores

Turner and Coleman (1962) and Simmons and Christy (1962) studied
examiner influence on Thematic Apperception Test (TAT) responses.
Turner and Coleman designed their study to maximize the probability
of significant differences but failed to obtain many significant
results. They did find that examiners who exhibited warmth elicited
significantly more hostile responses from subjects. They did not
find examiner competency, experience, or preference for the TAT signi-
ficantly related to subjects' responses.

Rosenthal (1965b) reported two studies conducted in the 1950's
in which subjects' scores on Rorschach cards correlated significantly
with examiners' scores. Furthermore, analysis of subjects' pretest
and posttost scores revealed that their score became more like the
examiners' scores as a result of the examiner-subject interaction.

Examiner influence on individual intelligence test scores or
on projective test scores was investigated in a limited number of
research studies. In these studies, the examiner's personality
(warmth, rapport) and physical characteristics (sex, race) had an
effect on the dependent variable. Specifically, opposite sex and
same race combinations obtained the best scores; “warm? examiners

elicited more hostile responses; and examiners who built rapport

11
obtained significantly higher Stanford-Binet intelligence test scores.
Examiner influence on intelligence and projective test scores prevents
accurate assessment of examinees and therefore, needs to be more fully

understood and controlled.

B. EXperimentsr Influence

There were many studies of experimenter characteristics and
their influence en experimental research. These included studies of
the sex of experimenter, prestige of experimenter, experience of experi-
menter, personality of experimenter, and race of the experimenter.
Several studies were reviewed in which the cues which mediate experi-

menter influence were examined.

Sex of Experimenter

 

None of the twelve studies reviewed in this section demon-
strated a significant sex of experimenter effect. Two of seven
studies showed a significant sex of subject effectp-in one study male
subjects performed better and in the other study female subjects per-
formed better. Hhen the sex of experimenter variable was examined in
interaction with another variable, significant results were usually
obtained. The interaction of the sex of experimenter and sex of sub-
ject variables was sometimes significant--ospecia11y when experi-
menters worked with opposite sex subjects.

Stevenson and Allen (1964) found significant sex ef subject
and interaction effects when adult subjects were verbally reinforced
in performing a marble-sorting task. Female subjects performed sig-

nificantly better than males and opposite sex interactions resulted

12
in the best performances. The authors (Stevenson and Allen, 1967)
repeated this study later, however, and found no significant inter-
action between ths experimenter and sex of subject variables.

Miller and Solkoff (1965) did not find a significant sex of
experimenter effect in a verbal conditioning experiment. Stevenson
(1961) studied children with the marble-sorting task and found sex of
experimenter related to the age and sex of subjects. Opposite sex
combinations of experimenters and subjects resulted in the best per-
formances. In another study (Hill and Stevenson, 1965) using marble-
sorting as the dependent variable, no significant sex effects were
found until the interaction of sex and verbal reinforcement was
examined. Best performances were obtained by male experimenters in
the reinforcement condition.

A verbal conditioning study (Sarason and Minard, 1965) revealed
that male subjects were conditioned easier than female subjects, but
sex of experimenter was not a significant variable. The interaction
effect of the sex of experimenter and hostility of experimenter
variables was significant. Low hostility male experimenters and
high hostility female experimenters conditioned their subjects best.
In another verbal conditioning experiment (Ogawa and Cakes, 1965)
it was found that female experimenters conditioned low anxiety male
subjects significantly better than high anxiety male subjects. The
authors reasoned that high anxious males became more anxious which
reduced the quality of their performances.

A summary of these verbal conditioning studies indicated that
the sex of an experimenter alone did not influence the results of the

studies. Only when another independent variable interacted with the

15

sex of experimenter variable were significant results obtained.

 

Prestige of Experimenter

Pour studies were reviewed which dealt specifically with the
prestige of experimenters as a variable. All of these studies
involved verbal conditioning. Two of the studies revealed signifi-
cant prostige effects; the other two did not. Of the former two
studies, one showed that low prestige experimenters obtained better
performances; the other study reported that high prestige experimenters
obtained better results. These studies were discussed below.

Prince (1962) found that more prestigious experimenters con-
ditioned the verbal behavior of children significantly better than
less prestigious experimenters. In another verbal conditioning
experiment (Katkin, Risk, and Spielberger, 1966) an undergraduate
experimenter (low prestige) obtained a significantly greater per-
formancs increment than a professor (high prestige). Vith only two
experimenters it was possible that other variables may have influenced
the subjects' responses in this experiment.

Sarason and Minard (1965) manipulated the experimenters'
prestige level by the way they dressed and contacted subjects. They
found no significant prestige of experimenter effect using sixteen
experimenters in.a verbal conditioning experiment. Blaufarb (1960)
found no significant prestige of experimenter effect using ten experi-
menters.

In a social science survey research study, Jones (1965) dis-
cussed a “courtesy biasI found in Southeast Asia. She referred to the

experimenter effect of prestige on people who traditionally treated

11+
visitors courteously, and stressed the importance of constructing

questionnaire items which were not susceptible to influence.

Experience of Experimentsr

 

The effect of an experimenter's experience on research results
was studied by several authors. Rosenthal (196A) once suggested that
use of unsophisticated and less ego-involved experimenters might reduce
an experimenter's influence. But he found that naive experimenters
influenced results also and retracted his former suggestion. In an
animal conditioning experiment, Brogdsn (1962) found that naive experi-
menters did not condition rabbits as well as experienced experimenters.
Hewever, the difference diminished with practice. Cordare and Ison
(1965) reported an interesting study of planaria. Experimentsrs who
expected high planarian activity reported significantly (p5 .001) more
activity than those experimenters expecting little planarian activity.
They attributed this extremely significant difference to the naivetd'of
the experimenters. Another study (Ingraham and Harrington, 1966)
appeared to support this notion. Some of the experimenters worked with
rats which they were led to believe were ”bright"; others worked with
rats which were supposedly 'dull.‘ They found no significant differ-
ence and even the insignificant difference disappeared by the fifth
day of the experiment. The authors conceptualized that naive experi-
menters must rely on cues given by the chief experimenter because of
the ambiguity of the situation. As experimenters gained experience
in the experimental situation, they increasingly responded to factual
cues emitted by the rats. The authors concluded that experimenter

influence was not a problem if experimenters were experienced in

15
working with subjects. In a sociological study, Kish (1962) also
found that experienced interviewers influenced their data signifi-
cantly less than naive interviewers. He suggested the use of struc-
tured questionnaire items to reduce interviewer influence.
In three of four studies, naive experimenters influenced their
results more than experienced experimenters and this influence was

reduced as experimenters practiced or became more experienced.

Personality Characteristics of Experimenter

 

Many studies were reviewed which examined the effect of experi-
mentsrs' personality in experiments. variables included hostility,

anxiety, need for social approval, and warmth of the experimenter.

Hostility. Two studies (Sarason, 1962; and Sarason and Minard,
1965) demonstrated a significant experimenter hostility effect on the
results of verbal conditioning experiments. Hostile experimenters of
both sexes elicited more hostile verbs in one study (Sarason, 1962).
In the other study, low hostility male experimenters and high hostility

female experimenters were able to condition subjects best.

Anxiety. In discussing the effect of sxperimsnters' anxiety on
results, researchers (Rosenthal, Persinger, and Feds, 1962; linkel and
Sarason, 1964) felt that a curvilinear relationship existed. That is,
experimenters exhibiting medium anxiety influenced subjects more than
experimenters showing high or low anxiety. The results of their
studies and others' did not clearly confirm this hypothesis. Winksl
and Sarason found that female subjects performed better for low

anxiety experimenters. Rosenthal's study demonstrated that medium

l6

anxiety experimenters were the least effective influencers. In another
study (Rosenthal, Persinger, Vikan-Kline, and Mulry, 1963) it was found
that high anxiety experimenters were most influential on subjects'
responses and that high anxiety subjects were most susceptible to being
influenced. Finally, Rosenthal, Kohn, Greenfield, and Carota (1965)
reported a study in which "less anxious' experimenters influenced
subjects significantly better than ”more anxious' experimenters. Thus,
while anxiety of the experimenter appeared to be an important variable,
its direction was unpredictable. The different results may reflect
some difficulty in correctly identifying "high," "medium,“ and "low"
anxiety experimenters. It is possible that experimenters affected

their data in some unexamined way rather than by their anxiety level.

Need for social approval. Rosenthal conceptualized that experi-

 

menters who had a high need for social approval would be better
influencers in order to be approved by the grand experimenter. This
conceptualization was supported in his review of the literature (1967a).
Need for social approval was not found to be related to the degree of
experimenter influence, however, in a study by Rosenthal, Persinger,

Vikan-Kline, and Mulry (1965).

[Eggmth. The warmth of experimenters seemed to be a significant
variable. Reece and Whitman (1962) found that 'warm' experimenters
conditioned subjects significantly better (piE.OOl) than “cold” experi-
menters. Rosenthal, Kohn, Greenfield, and Carota (1965) reported that
more friendly experimenters influenced subjects better. Authors of

both studies felt that subjects were more likely to want to please

warm, friend 1y experimenter.

17

Race of Experimenter

The influence of an sxpsrimentsr's race has not been exten-
sively studied. Two studies were reviewed which indicated that an
experimentor's race alone does not affect data. Williams (1964)
found that experimentsr's race was significant only in interaction
with a social distance (between interviewer and subject) variable,
and when interview questions were potentially threatening. Katz,
Robinson, Epps, and Waly (1964) administered a disguised verbal test
of aggression to Negro high school males. When the test was neutrally
described, Negro and white experimenters obtained the same results.
When described as an intelligence test, Negro experimenters obtained
significantly higher aggression scores than those elicited by white

experimenters.

Cues Which Mediate Experimenter Influence

If experimenters do influence their research, the question
must be asked, "How?" One factor analytic study (Rosenthal, Fods,
Friedman, and Vikan-Kline, 1960) revealed that experimenters who read
directions more slowly and used more hand gestures influenced their
data the most. Thus, these are examples of I'cues" which mediate some
kind of experimenter influence. In more recent studies (Rosenthal
and Fode, 1965a, 1965b; Rosenthal, Feds, Vikan—Kline, and Persinger,
19611; Friedman, Kurland, and Rosenthal, 1965; Rosenthal, Friedman,
and Kurland, 1966; Friedman, 1967; and Rosenthal, 1967a) the following
cuss have been found significant in the transmission of experimenter
influence: frequency of glances, rate and accuracy of reading direc-

tions, body activity, and touch. Rosenthal concluded that verbal

18
communication was the most significant mediator of experimenter
influence.

That verbal cues were important mediators was confirmed by
Sarason (1962). He found verbal reinforcement significantly more
influential on subjects than a visual reinforcer (flashing light).
Contrarily, Reece and Whitman (1962) reported a verbal conditioning
study in which visual cues (body movement, smiling, glancing by
experimenter) reinforced subjects more than verbal cues. Similarly,
Garlsmith and Aronson (1965) found visual cues to be effective medi—
ators of experimenter influence.

A majority of the authors of the studies reviewed in this sec-
tion suggested that experimenters do influence their data. The results
of these studies were too conflicting and inconsistent to warrant many
definite conclusions regarding how experimenters influence their data.

The sex and race of experimenters seemed to be significant
variables, especially when studied in interaction with other variables.
The warmth of an experimenter appeared to have a consistent effect on
data; that is, "warm" experimenters influenced their research more
than eXperimenters perceived as being less warm.

The research.on other experimenter personality characteristics
and also on the prestige of the experimenter revealed little consis-
tent evidence to support an assessment of their importance. Similarly,
the effect of the experience of experimenter was discussed in an unre-
solved debate by Rosenthal, Harrington, and Ingraham. While visual
and verbal cues were found to mediate experimenter influence, more

research needs to be done to discover how experimenters influence

19

their data.

In several studies the grand experimenter deliberately gave
experimenters an expectancy concerning the results of the study.
There seemed to be two phenomena that created experimenter influence
in these studies (Rosenthal, 1964); "expectancy effects," and “effects

of early data returns.“

ExPectancy Effects

Rosenthal (196k) described "expectancy effects” when he rea-
soned that researchers usually studied variables in which they were
interested and, therefore, often had certain 'expectations' regarding
that variable. If one's expectations in any way distorted the research
data, experimenter influence existed.

To test the expectancy effect on data, Rosenthal, Mulry, Per-
singer, Vikan-Kline, and Grothe (1964b) presented a sequence of
twenty photographs of people to a large non: group and asked them to
rate each picture from -10 to +10 according to perceived failure
(~10) or success (+10) exhibited by the faces. The mean rating was
zero. In subsequent studies, experimenters were asked to administer
the photographs to subjects and were told that the purpose was to
develop an empathy test. Then experimenters were given differential
expectancies concerning how their subjects would rate the photographs.
That is, some experimenters were told to eXpect mean ratings of #5
(moderate success), and the other experimenters were led to expect
-5 (moderate failure) mean ratings. Of course, subjects were assigned
randomly and if significantly different mean ratings were obtained

by +5 and -5 experimenters, a significant expectancy effect was

20
demonstrated. Using this basic design, Rosenthal and his students
reported five studies (Rosenthal, Fode, Friedman, and Vikan-Kline,
1960; Rosenthal, Persinger, Vikan-Kline, and Mulry, 1965; Rosenthal,
Mulry, Persinger, Vikan-Kline, and Grothe, 1964a, 1964b; and Fried-
man, Kurland, and Rosenthal, 1965) in which a significant experi-
menter expectancy effect was found. One study (Resenthal, Fode,
Vikan-Kline, and Persinger, 1964) revealed no significant effect.

Rosenthal and Halas (1962) and Cordaro and Ison (1963) found
significant expectancy effects using planaria as subjects. Ingraham
and Harrington (1966) reported no significant eXpectancy effect when
experimenters were led to believe they had either I'm.aze--bright" or
”maze-dull" rats for subjects. With the same design, Rosenthal and
Fode (1965a) had earlier found a significant difference; the sup-
posedly "maze-bright" rats learned a discrimination task signifi-
cantly faster.

Cooper, Eisenberg, Robert, and Dohrenwend (1967) gave opposite
expectancy conditions to ten experimenters, each of whom had ten sub-
jects. A significant expectancy effect was demonstrated. Using the
photo-rating task described above, Friedman (1967) also found a sig-
nificant expectancy effect.

Rosenthal and Jacobson (1966; 1968) examined the expectancy
effect hypothesis in the classroom. They described their study at

great length in Eygmalion in the Classroom (1968). The authors

 

explained what they meant by an "interpersonal self-fulfilling
prophecy“ (1968, p. vii): ' . . . how one person's expectation for
another person's behavior can quite unwittingly become a more

accurate prediction simply for its having been made.” Specifically,

21

they Wondered if a teacher's expectation of a pupil's ability some-
how actually helped determine the pupil's ability. Working in an
elementary school (Grades 1-6), Rosenthal and Jacobson administered
a pretest (Flanagan Tests of General Ability) to all pupils in May
of 196k, and told the teachers that it was the 'Harvard Test of
Inflected Acquisition.“ Then they randomly selected twenty percent
of the pupils and told the teachers in September of 1964 that the
test indicated that these pupils were about to take an “intellectual
spurt” or were about to "bloom." Retesting occurred in January,
1965, the basic posttest was administered in May, 1965, and a followb
up posttest was administered in May, 1966. Rosenthal and Jacobson
hypothesized that younger I'bloomers" would shew greater gains when
compared to control pupils than older "bloomers.” This hypothesis
was confirmed in that significant results were found only at the
first and second grade levels. Significant results were not obtained
for an ability effect (fast, medium, slow tracks) or for a minority
group status effect (Mexican versus American). Analysis did indicate
a significant sex of pupil effect: boy "bloomers" spurted more than
girl “bloomers." Rosenthal and Jacobson concluded that teachers
apparently did communicate an expectation of performance to the
"bloomers“ which accounted for their gains on the intelligence test
posttest. They conceptualized that the "quality of interaction"
between teachers and “bloomers" probably made the difference.

In view of recent criticism (Thorndike, 1968; and Claiborn,
1969) of Rosenthal and Jacobson's study, one must continue to question
the existence of teacher influence. Thorndike (1968, p. 711) con-

cluded that the basic data were ' . . . so untrustworthy that any

22

conclusions based upon them must be suspect." Claiborn (1969) repli-
cated parts of the Rosenthal and Jacobson study and found no signifi-
cant differences on the hypotheses tested in both studies. He dis-
cussed the "failure to replicate'' and concluded that the question of
teacher influence remained “equivocal.” Taking the strength of
Rosenthal's convictions and assuredness of his critics together, the
writer was left rather confused. It seemed, however, that Rosenthal

may be guilty of that which he has warned us about—-observer influence.

Effects of Early Data Returns

 

"Early data return effect" referred to the tendency for experi—
menters to be influenced by the hypotheses suggested by data collected
early in the life of an experiment. Rosenthal directed two ingenious
studies of this effect. In the first study (Rosenthal, Persinger,
Vikan-Kline, and Redo, 1965a) all experimenters were led to expect
-§5 mean photograph ratings. The researchers arranged for four of the
experimenters to obtain 'good' scores from their first two subjects
by the use of coached accomplices and arranged for four experimenters
to obtain "bad" scores by the same procedure. Four experimenters
experienced only naive subjects who were not accomplices. The dif-
ference between the two eXperimental groups was significant and, as
hypothesized, the control experimenters obtained a mean rating between
the two eXperimental groups. Furthermore, there was some evidence
of a sequence effect; that is, "good" data got better and 'bad” data
got worse. In a similar but more complex eXperiment (Rosenthal,

Kohn, Greenfield, and Carota, 1965), the authors found a significant

(p $.05) early date return effect. In this study, early data return

25
effect was strong enough to change an opposite, initial experimenter
expectancy. The effect was strongest, however, when it confirmed
an initial expectancy. The authors concluded that when experimenters
experienced 'good' data they became warmer and more friendly and

exercised greater influence on their subjects.
0. Discussion of Previous Research

In this chapter studies of examiner influence on test scores
were reviewed. There was only one study (Larrabee and Kleinsasser,
1967) reviewed in which the Wechsler Intelligence Scale for Children
was used as an instrument.

Several studies concerning experimenter influence on experi-
mental research were reviewed. The sex of the experimenter sometimes
had an influence on experimental results when combined with other
variables. Four studies were discussed in which the prestige of the
experimenter was examined and that variable did not appear to have
a consistent influence on research data. In a review of literature
concerning the influence on research results of an experimenter's
experience, it was concluded by this writer that experienced experi-
menters were less likely to influence their data. It was also found
that certain personality characteristics (hostility, anxiety, need
for social approval, and warmth) of eXperimenters sometimes influ-
enced results of experiments.

If experimenter influence exists, it is important to understand
how it is mediated. Studies were reported in which it was demonstrated
that both verbal and non-verbal cues mediated experimenter influence.

Several studies were reviewed in which the grand experimenter

24
deliberately gave experimenters an expectancy regarding the experi-
ment. In these studies an experimenter influence effect was consis-
tently demonstrated.

This review of the literature revealed that experimenter
influence has been extensively examined in experimental research.
Rosenthal and Jacobson (1968) applied experimental research findings
to a practical setting. They are vulnerable to criticism in regards
to their research design and the interpretation of their data. The
review of literature revealed weaknesses in research design of other
studies as well. In many studies experimenters were given only one
expectancy condition. This procedure was seen as a weakness because
an expectancy "set" could easily deve10p which might account for the
influencing effect. Much of the research involved male experimenters
and female subjects of similar age. The sex of experimenter and sex
of subject variables need to be crossed and younger subjects need to
be tested. Most of the studies employed meaningless rating instru-
ments which required little or no training to administer. This pre-
vented an examination of the experimenter's experience or training
as a variable in research on experimenter influence.

There seemed to be confusion concerning just how experimenter
influence operated. Rosenthal (1967a) believed that experienced
experimenters were more likely to influence their research because
they were more ego-involved with their research and were ”better
biasers." That is, they were better at communicating their expectancy
to subjects and better at reinforcing ”correct" responses. Thus, for

Rosenthal, experimenter influence increased with the experimenter's

25
experience and "snowballed" as an experiment progressed. Friedman
(1967), one of Rosenthal's associates, agreed with this hypothesis.
This hypothesis has been continually criticized by Ingraham

and Harrington (1966, 1967) in a running debate in Psychological

 

Reports. They found (Ingraham and Harrington, 1966) that experi-
menters influenced less as an experiment progressed and as experimen-
ters gained experience in the experimental task. They also found that
training of experimenters reduced the experimenter influence effect
and that experimenters who were given both expectancy conditions
influenced their results less. They concluded (Harrington and Ingra-
ham, 1967) that experimenter influenoe existed when an inexperienced
and untrained experimenter began an experiment in which he was given
an expectancy condition. As the eXperiment progressed, the experi-
menter responded more and more to factual cues presented by the sub-
jects and.less and less to the experimentally induced expectancy effect.
Rosenthal (1967b, 1967c) retorted by manipulating Ingraham and Harring-
ton's statistical methods in such a way that their research confirmed
his hypothesis.

From this discussion it can be seen that the operation of
experimenter influence is difficult to understand. For example,
does the subject in an experimenter influence study actually perform
better or does the experimenter just think the subject does better?

In Chapter III, the design of this study is presented. The
sample, instrumentation, procedure, and the design and analysis are
described. The research design was chosen to correct some of the

weaknesses mentioned in the discussion of previous research.

CHAPTER III

DESIGN

Chapter III includes an explanation of the samples, instru-

mentation, procedure, and the design and analysis.
A. The Sample

Examiners

Eight trained Wechsler Intelligence Scale for Children (WISC)
examiners participated in this study. They were volunteers from two
Michigan State University individual testing classes (Education 866A)
totalling approximately thirty-five people. Random selection of
examiners would have been statistically desirable; however, due to
insufficient funds, the writer could not make the research attractive
enough to insure a large pool of examiners from which to randomly
draw eight. Four men and five women volunteered to test eight exami-
nees each for $2.50 per test. A scheduling conflict eliminated one

woman examiner which left four men and four women examiners.

Examinees

There were sixty-four examinees--thirty-two boys and thirty-
two girls. Examinees were sixth, seventh, and eighth grade students
at DeWitt Junior High School. DeWitt, Michigan is a middle-class

suburb of Lansing, Michigan. The writer identified fifty-eight boys
26

27

and sixty-nine girls who had previously scored from ninety to one
hundred-ten on the California Test of Mental Maturity-Short Form.
From this population, thirty-two boys and thirty-two girls were ran-
domly selected for the sample. The means and standard deviations
obtained by examinees on the California Test of Mental Maturity are
presented in Appendix A.

The examinees ranged in age from ten years, ten months to
fourteen years, six months, with a median age of twelve years, seven

and one-half months.

B. Instrumentation

Wechsler Intelligence Scale for Children

The Wechsler Intelligence Scale for Children (WISC) was used in
this study for two reasons: first, a pool of trained WISC examiners
was more accessible than Stanford-Binet or other individual intelli-
gence test examiners; and second, although the WISC had high reliability,
scoring was somewhat subjective. The nature of this study demanded
that some degree of subjectivity in the scoring of protocols be perk
mitted.

The WISC is an individual intelligence test for children between
age five and age fifteen (Wechsler, l9k9). There are five Verbal
Scale subtests: Information, Comprehension, Arithmetic, Similari-
ties, and vocabulary. There are five Performance Scale subtests;
Picture Completion, Picture Arrangement, Block Design, Object Assembly,
and Coding. In addition, there are two optional subtests; Digit
Span and Mazes. Verbal, Performance, and Total I.Q.'s are computed

by a standard score formula with mean set at 100 and the standard

28
deviation at fifteen. The WISC was standardized on a stratified
random sampling of 2200 White American boys and girls. Published in
l9h9, the sampling was based on 1940 Census Bureau data.

Scoring of the W180 was made as objective as possible. How-
ever, Cronbach (1960, p. 19#) pointed out that,

”The skill of the examiner may influence the score greatly.

In some of the verbal tests, the examiner must make rather

sensitive judgments as to the correctness of an answer since

it may be necessary to request the subject to elaborate his
meaning. Answers that seem.wrong may be correct when the
subject explains himself. Subjectivity in scoring border-
line answers is also a potential problem."

Littell (1960) noted that the predictive validity of the W180
had not been demonstrated. Though Burstein (Buros, 1965) did not
mention this limitation, no recent studies were reviewed in which an
attempt was made to demonstrate the WISC's predictive validity. Con-
current validity has been successfully demonstrated. In various
studies, the WISC correlates with the Stanford-Binet from .h9 to
.94 with a median correlation of about .80. Correlations between
the WISC and the Wechsler-Bellevue ranged from .72 to .87. With the
California Test of Mental Maturity, correlations ranged from .77 to
.81. When correlated with achievement tests, coefficients ranged
from .14 to .81 with a median correlation of .66.

WISC reliability is very high. Split-half reliability coef-
ficients vary from .86 to .96 depending on the age level of examinees
(Littell, 1960). Cronbach (1960, p. 198) felt that the Wechsler
Performance Scale was the most reliable performance scale ever deve-
loped. Coefficients of internal consistency ranged from .59 to .91
with standard error of measurement ranging from 5.00 to 5.61. Only

one coefficient of stability was reported (Littell, 1960) and it

29
was .77 over a four year period. Reliability coefficients were
greatest from age ten and one-half to age thirteen and one-half.
The subjects in this study were between the ages of ten and fourteen.
Fraser (Buros, 1959) felt that the W130 seemed most valid for
normal range subjects. All the subjects in the present study were

drawn from the "normal” range. In the Sixth Mental Measurement

 

Yearbook (Buros, 1965), Burstein supported the WISC as having good
reliability. He pointed out that much of the W180 research since
1960 concerned "psychopathological applications." That is, the

W180 was used to measure the intelligence of special groups such as
the retarded, the gifted, brain-damaged children, and the disadvan-
taged. A review of WISC literature since 1960 brought the writer

to the same conclusion.made earlier by Littell (1960): Studies need
to be made concerning possible sources of score variation other than

intelligence.

Rating Sheet

 

An expectancy influence was experimentally induced by giving

test examiners a rating sheet for each examinee (see Appendix B).

The rating sheet included the examinee's name and fictitious ratings
of his California Test of Mental Maturity-Shcrt Form 1.0. and school
achievement. The fictitious ratings were marked on continua from
"bottom one-fourth" to "tap one—fourth." In each case a discrepancy
between intelligence test score and school achievement was indicated.
For example, an "above average" examinee was rated at the top one-
fourth on the CTMM, and second one-fourth on school achievement. The

last item on the rating sheet was a "predicted WISC score" which was

50
above average or below average depending on whether the examinee had
previously been designated as "above average" or "below average."
That is, examinees who were supposedly "above average” were given
predicted WISC scores above one hundred. Examinees who were supposedly
”below average" were assigned predicted WISC scores below one hundred.
In this way it was hoped that examiners would have an ”expectancy"

regarding the outcome of an examinee's WISC score.

C. Procedure

After the thirty-two boys and thirty-two girls were selected,
they were randomly assigned to one of the eight examiners so that
each examiner had four boys and four girls. Then “above average“ or
"below average" designations were randomly assigned to examinees so
that each examiner would test two ”above average“ boys, two 'below
average" boys, two "above average" girls, and two 'below average“
girls. Finally, the order in which examinees were tested was ran-
domized for each examiner.

It was necessary to call seventy parents to secure approval
for testing sixty-four subjects. When an originally selected subject
could not participate, a replacement was randomly drawn from the
remaining papulation. Testing was offered as a service to parents
and confidentiality was assured. Neither the subjects nor their
parents were told that the testing was part of a “study" or an ”experi-
ment." Test scores were interpreted to parents at a later date.

Examiners were given the following information before testing

began:

51

"These students are being tested because there is a dis-

crepancy between their I.Q. (CTMM) and school grades. Be

sure to read the information for each student before test-
ing because we are interested in knowing your Opinion with
regard to whether he has been over-achieving or under-
achieving. When turning in test scores please let us know
the student's 1.0. and any special circumstances con-

cerning the Verbal and Performances scores."

Examiners were led to believe that the writer was merely an
agent of the DeWitt Public Schools. In no case did an examiner indi-
cate that he "guessed" that he was taking part in an eXperiment or
was indeed a subject in an experiment.

All testing was completed within a one week period starting
the day after examiners completed their WISC course. Most examiners
tested four examinees in each of two sessions. Two examiners tested
eight examinees in one day. Scheduling conflicts prevented a more
uniform testing schedule which would have been desirable in terms of
research design.

Examinees were scheduled at seventy-five or ninety minute
intervals. Each examiner had a private office in which to administer
the W180. No more than three examiners were testing at one time and
conversation among examiners was infrequent. It was feared that too
much discussion among examiners might accidentally lead examiners to

the realization that they were subjects in the research design. All

tests were scored by the examiners and returned to the writer.

D. Design and Analysis

Analysis ef the data was by a mixed model, four-way analysis
of variance (Hays, pp. 459-hh7). A I'mixed model” was used because

the design had both "fixed'' and "random" variables. A variable is

52

"fixed" when there are qualitatively distinct levels of the variable.

 

A "random'' variable is one in which the levels chosen represent only
a sample of the pOpulation of possible levels. The design of the
study (see Figure 1) consisted of three fixed variables (expectancy,
sex of examinee, and sex of examiner), and two random variables
(examiners nested in sex of examiner, and replications nested in all
other variables). A variable is nested in a second variable when
each level of the nested variable does not appear in all levels of
the second variable. For example, from Figure 1 it can be seen that
Examiner 2 can only be in one level of the sex of examiner variable

because he is a male.

 

Expectancy of Examiner by Examines Sex

 

 

 

 

 

 

 

 

 

 

Male Examines Female Examinee
”Above IffiBelow "Above I'Below

Sex of Average" Average" Average“ Average"
Examiner Examiner Expectancy Expectancy ExJectancy Expectancy

1 R l, R 2 R 1, R 2

2

Male a;

5

n

5

6
Female

7

8

 

 

 

 

 

 

 

 

Fig. 1 Research Design

55

As seen in Figure 1, the sex of examiner variable is crossed
with the sex of examinee and expectancy variables. Variables are
"crossed" when each level of one variable occurs with each level of
the other variable. From the design diagrammed in Figure 1 it can
be seen that four male examiners and four female examiners each tested
two "above average" male examinees, two 'below average“ males, two
”above average" females, and two ”below average" females. The five
per cent level of confidence was arbitrarily chosen for significance
tests.

Analysis of variance with fixed and random variables involves
the usual assumptions of independence of observations, equality of
variance, and normality of distributions. In less complex designs,
at least, the violation of these assumptions in designs with equal
numbers in the subgroups has been shown to have negligible effects
on the significance tests. See, for example, Scheffe (1959, p. 554),
Norton (1952), Young and Veldman (1965), and Boneau (1969).

Analysis of variance with fixed and random variables, as con-
ducted here, involves the additional assumption (Hays, p. 465) that
the degree of relationship, if any, among the different observations
for the ammo fixed variable (expectancy, sex of examiner, and sex of
examinees) are the same for all levels of that variable. There is no
reason to believe this assumption is violated with the present data.

The use of analysis of variance in complex designs requires
the assumption, accepted here and by most investigators, that the
technique continues to be robust and that it is insensitive to vio-

lations of more basic assumptions in these applications.

54
The sources of variation and their degrees of freedom are

presented in Table 1.

TABLE 1

SOURCES OF VARIATION AND
DEGREES OF FREEDOM

 

 

Sources of Variation Deggges of Freedom
Sex of Examiner 1
Sex of Examinee 1
Expectancy 1
Examiner Nested in Sex of Examiner 6
Replications 52
Sex of Examiner 1 Sex of Examines 1
Sex of Examiner x Expectancy 1
Sex of Examinee x Expectancy l

Examiner x Expectancy Nested in
Sex of Examiner 6

Examiner X Sex of Examinee Nested
in Expectancy 6

Examiner x Expectancy x Sex of Exami-
nee Nested in Sex of Examiner 6

Sex of Examiner x Expectancy X Sex of
Examinee 1

Total 65

E. Summary

Chapter III contained an explanation of the samples, instrumen-
tation, and analysis of the research design. The sample consisted of
four male and four female Wechsler Intelligence Scale for Children
(WISC) examiners who had just completed WISC training. Each examiner
tested eight junior high school students (four boys and four girls)
who were randomly selected from a population of students who were
”average“ in intelligence; that is, they scored from ninety to one
hundred-ten on the California Test of Mental Maturity--Short Form,
1965 Revision. Examinees were randomly assigned to examiners.

The WISC is an individual intelligence test for children
between ages five and fifteen. It has good reliability with split-
half reliability coefficients ranging from .86 to .96. A rating
sheet (see Appendix B) was used by the grand experimenter to trans-
mit an expectancy condition to examiners. On each examinee's rating
sheet a discrepancy between California Test of Mental Maturity--
Short Form score and school achievement was fictitiously indicated,
and a ”predicted WISC score" was advanced.

Eight examinees were randomly assigned to each examiner and
then they were randomly designated "above average“ or “below average”
so that each examiner tested two "above average" boys, two ”below
average" boys, two “above average" girls, and two ”below averageI
girls. The order in which an examiner tested his examinees was also
randomized. Parental approval for testing was secured and confi-
dontiality was assured. Neither parents nor examiners were told that

they were participating in a study or an experiment. All testing

55

56

was done in private offices during a one week period. Analysis of
the data was by four-way analysis of variance.
In Chapter IV the hypotheses will be restated and the results

will be presented.

CHAPTER IV

ANALYSIS OF RESULTS

In Chapter IV each hypothesis will be presented with a state-
ment concerning whether the hypothesis in question will be rejected
or not rejected. The significance of the obtained F-ratios will be
determined by using Table IV in Hays (1964, pp. 677-688). Following
this section, the results of the hypotheses will be discussed. Wechs-
ler Intelligence Scale for Children (WISC) Total scores are presented

in Appendix C.
A. Hypotheses and Results

Hypothesis 1

Female examiners will obtain a significantly higher mean
Total WISC score than male examiners.

For a one-tailed F-test with one and six degrees of freedom,
the required F was 5.99. Since the obtained F-ratio of 8.17 exceeded
5.99, the sex of examiner hypothesis was not rejected. The sex of
examiner effect was significant (p 5.05) in that female examiners
obtained a higher mean Total WISC score than male examiners (see

Table 2).

57

58

TABLE 2

311?»an or MEAN TOTAL WISC scones, STANDARD DEVIATIONS,
AND F—RATIOS FOR MAIN EFFECTS

 

 

Standard
Main Effect Means Deviations F-Ratio
Sex of Examiner
Males 97.55 8.75
Females 105.12 8.06 8.17‘
Sex of Examines
Males 101.66 9.05
Female! 99.00 9e 07 9e 50.
Expectancy
"Above Average“ 101.17 9.89
”Below Average" 99.50 8.25 .84
Examiner
1 97.75 7.49
2 101.50 9.52
5 94.00 8.60
4 96.88 11.51
5 104.75 7.85
6 105.50 6.65
7 104.62 9.98
8 99.62 8.79 .92

 

I"Sig. at .05 level of confidence

Hypothesis 2

 

There will be no significant difference between mean Total
WISC scores achieved by boy and girl examinees.

For a two-tailed F-test with one and six degrees of freedom,
the required F was 8.81. Since the obtained F—ratio was 9.50, the
null hypothesis was rejected. The sex of examinee effect was signifi-
cant (p 6.05) in that male examinees achieved a higher mean Total WISC

score than female examinees (see Table 2).

59
Hypothesis_5

 

The mean Total WISC score achieved by ”above average“ exami-
nees will be significantly higher than the mean Total WISC score
achieved by "below average" examinees ("above average" examinees
were those whose predicted WISC score was above one hundred; ”below
average” examinees were those whose predicted WISC score was below
one hundred).

For a one-tailed F-test with one and six degrees of freedom,
the required F was 5.99. Since the obtained F-ratio was .84, the
directional hypothesis was rejected. The expectancy effect was not

significant (see Table 2).

Hypothesis 4

 

There will be no significant difference between.mean Total
WISC scores obtained by the eight examiners.

For a two-tailed F-test with six and thirty-two degrees of
freedom, the required F was 2.87. Since the obtained F-ratio was
.92, the null hypothesis was not rejected. The examiner effect was

not significant (see Table 2).

Hypothesisg5
The mean Total WISC score obtained by opposite sex combinations

of examiners and examinees will be significantly higher than the mean
Total WISC score obtained by same sex combinations.

For a one-tailed F-test with one and thirty-two degrees of
freedom, the required F was 4.17. Since the obtained F—ratio was

.28, the directional hypothesis was rejected. The interaction effect

between the sex of examiner and sex of examinee variables was not

40

significant (see Table 5).

TABLE 5

SUMMARY OF MEAN TOTAL WISC SCORES, STANDARD DEVIATIONS,
AND F-RATIOS FOR TWO-WAY INTERACTION EFFECTS

 

 

Standard
Interaction Effect Means Deviations F
Sex of Examiner x Sex of Examines
Male Examiner-Male Examines 98.51 9.52
Male Examiner-Female Examines 96.75 9.28
Female Examiner-Male Examinee 105.00 7.54
Female Examiner-Female Examinee 100.75 12.76 .28

Sex of Examiner X Expectancy
Male Examiner-"Above Average“ Examinees 99.00 11.25

Male Examiner-“Below Average" Examinees 96.06 6.86
Female Examiner-"Above Average" Examinees 102.81 15.26
Female Examiner-"Below Average" Examinees 102.94 8.27 .59
Sex of Examines X EXpectancy
Male Examines-"Above Average" 102.69 9.95
Male Examines-"Below Average“ 100.62 8.19
Female Examines-"Above Average' 99.62 9.92
Female Examines-"Below Average' 98.58 8.41 .01

 

Hypothesis 6

 

There will be no significant interaction effect on mean Total
WISC scores between the sex of the examiner and the extent to which
the obtained Total WISC scores are in the direction suggested by the
expectancy set.

For a two-tailed F-test with one and thirty-two degrees of
freedom, the required F was 5.57. Since the obtained Fiwas .59, the
null hypothesis was not rejected. The interaction effect between the
sex of examiner and expectancy variables was not significant (see

Table 5).

41

Hypothesis 7

 

There will be no significant interaction effect on.mean Total
WISC scores between the sex of the examinee and the extent to which the
obtained Total WISC scores are in the direction suggested by the expec-
tancy set.

For a two-tailed F-test with one and six degrees of freedom,
the required was 8.81. Since the obtained F-ratio was .01, the null
hypothesis was not rejected. The interaction effect between the sex

of examinee and expectancy variables was not significant (see Table 5).

Hypothesis 8

 

There will be no significant interaction effect among the sex
of examiner, sex of examinee, and direction of expectancy variables on
mean Tetal WISC scores.

For a two-tailed F-test with one and thirty-two degrees of
freedom, the required F was 5.57. Since the obtained F was .01, the
null hypothesis was not rejected. The three-way interaction effect
between the sex of examiner, sex of examinee, and expectancy variables

was not significant (see Table 4).

Il‘lr A I]? 4

 

42

TABLE 4

SUMMARY OF MEAN TOTAL WISC SCORES, STANDARD DEVIATIONS,
AND F-RATIO FOR THREE-WAY INTERACTION EFFECT

 

 

Standard
Interaction Effect Means Deviations F

Sex of Examiner x Sex of Examines x Expectancy
Male Examiner-Male Examines-"Above Average“ 100.12 12.12
Male Examiner-Male Examines-“Below Averags'I 96.50 6.50
Male Examiner-Female Examines-"Above Average” 97.88 11.02
Male Examineeremale Examines-"Below Average” 95.62 7.79
Female Examiner-Male Examines-“Above Average|| 105.25 7.09
Female Examiner-Male Examines-“Below Average" 107.75 8.07

Female Examiner-Female Examines-"Above Average' 100.28 17.19
Female Examiner-Female Examines-”Below Average" 101.12 8.59 .01

 

B. Discussion of Results

Female WISC examiners obtained significantly (pé.05) higher
WISC scores from examinees than male examiners obtained. This result
supported Cieutat's (1965) research with the Stanford-Binst. As
Rosenthal (1966, p. 47) concluded, sex of experimenter seemed to be
an 'aotive' rather than a “passive' variable but not a very predic-
table one. The design of this study did not provide for detsnnining
whether significant effects were brought about by examinees or by
examiners' perceptions of examinees. It was felt, however, that the
WISC testing was a novel and, perhaps, an anxiety-producing event
for most examinees. Junior high school students were asked to go
to the nearly empty high school at a designated time and be tested by
a complete stranger. A possible explanation is that examinees felt
more comfortable in the presence of female examiners since most of
their teachers were females. There were only three male teachers in

Grades K through 8 of the school district. Three of four female

45
examiners obtained.mean WISC scores higher than the highest mean
WISC score obtained by any male examiner.

The significant (pr.05) sex of examinee effect was unexpected.
Cieutat (1965) did not find a significant sex of examinee effect. It
was conjectured that boys were more aggressive and competitive during
the testing situation which accounted for their mean WISC score. No
further attempt was made to explain this effect.

The central hypothesis of this study was that examinees who
were allegedly ”above average" would obtain a mean WISC score signi-
ficantly higher than supposedly ”below average“ examinees. That is,
did the grand experimenter's estimate of an examinee's WISC score
create an expectancy on the part of the examiner which influenced
the obtained WISC score? Though the results were in the desired
directiOn, the eXpectancy effect did not approach significance.
Several explanations for this failure to demonstrate examiner influ-
ence were made.

First, it.must be considered possible that the WISC examiners
did not receive and/or retain an expectancy regarding the outcome of
an examinee's WISC score based on the grand experimenter's estimate
of that WISC score. This possibility was believed very plausible.

In criticizing Rosenthal's research, Barber and Silver (1968a, p. 25)
outlined an “eight-step transmission process" involved in inducing
experimenter influence.

'(a) The student experimenter attended to the expectancy come

munication from the principal investigator. (b) The experi-

menter comprehended the expectancy communication. (c) The
experimenter retained the communication. (d) The experimenter

(intentionally or unintentionally) attempted to transmit the

expectancy to the subject. (e) The subject (consciously or
unconsciously) attended to the expectancy communication from

414

the experimenter. (f) The subject (consciously or uncon-

sciously) comprehended the experimenter's expectancy. (g)

The subject (consciously or unconsciously) retained the

experimenter's expectancy. (h) The subject (wittingly or

unwittingly) acted upon (gave responses in harmony with)

the experimentsr's expectancy.“

It was believed that WISC examiners "attended to the expectancy comp
munication'I and comprehended it. However, a break in the transmis-
sion process may have come when examiners failed to retain the expec-
tancy communication. Therefore, little attempt was made to inten-
tionally or unintentionally transmit the expectancy to examinees.
The grand experimenter gave an expectancy for each examinee to the
examiner at the very beginning of testing. It was believed that
examiners listened and understood that each examinee would be either
'above average" or ”below average.” Furthermore, the grand experi-
menter reinforced the expectancy condition for each.exsminee immedi-
ately preceding each test administration. However, once testing
began, it was felt that examiners may have forgotten the expectancy
regarding the examinee being tested. Examiners did not appear to be
ego involved with the expectancy for each examinee. The possible
failure to control and.measurs the expectancy condition.must be
regarded as a limitation of this study.

In addition to a breakdown in the transmission process, the
failure to show a significant expectancy effect was possibly due to
the nature of the experimental task and the experience of examiners.
Ingraham.and Harrington (1966, 1967), and Barber and Silver (1968a)
suggested that expectancy effect was difficult to demonstrate in
relatively structured tasks. Barber and Silver (1968a, p. 26) con-

cluded, 'Ssveral studies in this area used relatively structured or

45
factual tasks, such as the Wechsler Adult Intelligence Scale, the
Taylor Manifest Anxiety scale, and a number-estimation task; none of
these studies showed an experimenter bias effect.‘I Compared to
Rosenthal's photo-rating task, the W130 was a very structured experi-
mental task. Examiners did not have to respond to the ambiguous cues
given them by the grand experimenter; rather, they very quickly
responded to factual cues presented to them by examinees. Since
these factual cues represented continuing information concerning an
examinee's 2523 intelligence, the “expected” WISC score for an exami-
nes may have become increasingly less important and disregarded or
forgotten.

Ingraham and Harrington (1966, 1967) also believed that experi-
enced and well-trained experimenters did not allow their expectan-
cies to influence their results as much as less experienced experi-
menters. Though the WISC examiners were inexperienced when compared
to experienced WISC examiners, they were well-trained when compared
to the experimenters employed in most of Rosenthal's research. Each
examiner had just completed a five-week course in WISC-WAIS test
administration. Presumably, they were trained to rigorously follow
directions of administering the test and recording examinees'
responses. They were trained not to intentionally influence or dis-
tort an examinee's responses. Examiners did not need to remember a
“predicted” WISC score for an examinee because they would soon deterb
mine an "actual“ WISC score for an examinee.

Barber and Silver (1968a, p. 26) postulated that an expectancy
effect was easier to demonstrate when a subordinate-superordinate

relationship existed between experimenters and the grand experimenter.

46
The WISC examiners in this study were responsible for eight test
administrations but in no other way were they subordinate to the
grand experimenter. Furthermore, the amount of compensation they
received for testing was fixed and not dependent on the test results.
In some studies experimenters received more compensation if their
results confirmed the expectancy condition.

Finally, it was believed that the design of this study mini-
mized the possibility of Type I error (rejecting null hypothesis
when it should not be rejected) in regards to the expectancy effect.
As mentioned above, the WISC was a relatively structured and factual
experimental task, and examiners were relatively well-trained and
experienced. In addition, each examiner tested both male and female
examinees, and worked under both expectancy conditions which were
randomly assigned. It was impossible for an examiner to think that
gll_his examinees were “above average” or that they were all 'below
average.“

Considering the failure to demonstrate a significant expec-
tancy effect, the results of this study seemed to support the con-
clusions reached by Ingraham and Harrington (1966, 1967). and Barber
and Silver (1968a, 1968b): experimenter influence was not as easy to
demonstrate as Rosenthal claimed--especially in structured tasks
using experienced experimenters.

The examiner effect was examined in the study to obtain evi—
dence on Rosenthal's contention that this variable should always be
tested in research, and also because adding this variable strengthened
the research design. Though some variation existed among the mean

Total WISC scores obtained by examiners, the F—ratio (.92) did not

47
approach significance. This insignificant result contraindicated the
possibility that examiners influenced test scores in idiosyncratic
ways. The result.may also be taken as further evidence of the “180's
reliability.

Turning to a discussion of interaction effects, it was inter-
esting to note that none of the four interaction effects tested was
significant. The review of literature for this study abounded with
significant interaction effects using marble sorting ability or
photo ratings as the dependent variable. Again, the failure to
demonstrate significant interaction effects was possibly due to the
structured nature of the eXperimental task and the experience of the
WISC examiners. Since no interaction effects were significant, no
t-tests were computed between means.

There was considerable evidence in related literature to
support the directional hypothesis that opposite sex combinations of
examiners and examinees would obtain higher’mean WISC scores than
same sex combinations. This hypothesis was not supported in this
study (F=.28). Though the female examiner-male examinee combination
resulted in the highest mean Total WISC score (105.00), the male
examiner-female examinee mean Total V130 score (96.75) was the lowest
combination.

The interaction of the sex of examiner and expectancy variables
was investigated to determine whether~males or females influenced
their data more. Rosenthal,‘flu1ry, Persinger, Vikanrxline, and
Grothe (196#) found that males were better "biasers." With an F-ratio
of .39, that conclusion was not supported in the present study.

Examination of mean WISC scores (see Table 5) showed that male

48
examiners did obtain a mean WISC score for ”above average” examinees
that was almost three points above that obtained for I‘below average"
examinees. There was practically no difference in mean WISC scores
obtained for "above average" and 'below average“ examinees by female
examiners. While the results were in the expected direction, the
obtained F-ratio was far from significant and this discussion should
not be interpreted as support for the hypothesis that males are
better 'biasers' than females. In Barber and Silver's (1968a, p. 25)
terms, examiners apparently did not receive and/or retain an expec-
tancy for examinees and therefore did not transmit an.expectancy
communication to examinees.

There was virtually no (F==.Ol) interaction effect between the
sex of examinee and expectancy variables. Though a directional
hypothesis was not advanced, it was postulated that female examinees'
WISC performance would be more influenced by examiners than male
examinees' WISC performance. Non-statistical examination of mean
scores (see Table 3) revealed insignificant differences in the oppo-
site direction. That is, the difference between 'above average" and
“below average“ male examinees' mean WISC score was greater than the
difference between "above average“ and "below average'I mean WISC
scores for female examinees. As suggested earlier (p. 6), this
result possibly contradicted previous research (Rosenthal, Mulry,
Persinger, Vikan-Kline, and Grothe, 1964a; and Stevenson and Allen,
1964) because of the greater age difference between examiners and
examinees in this study. Specifically, male college experimenters
may have more cues by which to influence female college subjects

than a junior high school girl. Of course one could counter with

49
the hypothesis that the greater age difference between examiner and
examines in this study could increase the subordinate-superordinate
relationship which should increase examiner influence. A more
plausible explanation for the insignificant interaction effect was
believed to be that the WISC was a structured task and that the expec-
tancy effect demonstrated in.aypothesis 3 was just too weak. Appar-
ently the expectancy communication was not received by the exami-
nees of this study.

The three-way interaction effect was investigated to determine
if the three variables (sex of examiner, sex of examinee, and expec-
tancy) were working in some complex way. No significant interaction
effect (F::.01) was found. The hypothesis was advanced for explora-
tory reasons and no attempt was made to explain the insignificant

effect.
0. Summary

In Chapter IV the statistical hypotheses were restated and a
decision was made to reject or not reject each hypothesis. Then the
results of each hypothesis were discussed in reference to previous
research and possible explanations for the results.

Both the sex of examiner and sex of examinee effects were
significant (pé.05). Female examiners obtained higher mean WISC
scores and male examinees achieved higher’mean WISC scores. While
the significant sex of examiner effect confirmed previous research
(Cieutat, 1965), the significant sex of examinee effect was unex-
pected. It was conceptualized that male examinees were less anxious

and more competitive when placed in a novel situation. The

50

interaction effect between sex of examiner and sex of examinee was
not significant. The main sex effects seemed to over-shadow the
interaction.effect.

The major finding in this study was that the expectancy effect
was not significant. Two principle explanations were offered. First,
the eight-step transmission process delineated by Barber and Silver
(1968a, p. 25) broke down in this study. That is, it was feared
that 'ISG examiners did not retain the expectancy condition given to
them by the grand experimenter and, therefore, examiners' expectancy
was neither transmitted to nor received by examinees. Secondly, it
was felt that the WISC represented a structured and factual experi-
mental task, and that the W180 examiners were relatively well-
trained and experienced. Thus, the results of the study seemed to
confirm Ingraham and Harrington's (1966, 1967) and Barber and Silver's
(1968a, 1968b) criticism of Rosenthal's research: experimenter
influence has not been consistently demonstrated when experienced
experimenters are administering structured tasks.

The fact that no significant examiner effect was demonstrated
in this study added to the argument that examiner influence was not a
great problem in intelligence testing, at least not in this study.

The expectancy variable was examined in interaction with both
the sex of examiner and sex of examinee variables and neither inter-
action effect was significant. Thus, the results of this study were
not construed as evidence that.ma1e expert-enters were better influ-
encers or that female examinees were more easily influenced. These
two insignificant interaction effects represented further evidence

of a lack of an expectancy effect in situations having this degree of

51
of examination structure and examiner experience. Analysis of the
three-way interaction effect among the sex of examiner, sex of exami-
nee, and expectancy variables revealed no significant complex inter»
action.
Chapter V will include a summary of this study and conclusions

made from it.

CHAPTER V

SUMMARY, CONCLUSIONS, AND IMPLICATION
FOR FURTHER RESEAICH

A. Summary

The present study was made to examine the claim of Robert
Rosenthal (1967a) and others that experimenters frequently influ-
enced (intentionally or unintentiOnally) the results of their
research. Cieutat (1965), Cohen (1965), and Larrabee and Kleinsas-
ser (1967) investigated the possibility that examiners of individual
intelligence tests affected the scores of examinees.

The purpose of this study was to provide a well-designed field
testing of experimental research results. Specifically, the writer
sought to determine if Wechsler Intelligence Scale for Children
(WISC) examiners influenced their exmminees' scores when the exami-
ners were given an expectancy or estimate of each examinee's score.

Eight.hypotheses were advanced:

hypothesis 1

 

Female examiners will obtain a.higher*mean intelligence test

score than male examiners.

gypothesis 2
There will be no difference between mean intelligence test

scores achieved by boy and girl examinees.
52

53

Hypothesis):
Examinees who are expected by examiners to be above average
will achieve a higher*mean intelligence test score than examinees

who are expected to be below average.

Hypothesis #

 

There will be no difference between mean intelligence test

scores obtained by the examiners.

Hypothesis 5

 

Opposite sex combinations of examiners and examinees will

obtain higher mean intelligence test scores than same sex combinations.

Hypothesis 6

 

There will be no interaction effect between the sex of the
examiner and the extent to which the obtained scores are in the

direction suggested by the expectancy set.

gypothesis 7
There will be no interaction effect between the sex of the
examinee and the extent to which the obtained scores are in the

direction suggested by the expectancy set.

hypothesis 8

 

There will be no interaction effect among the sex of exami-

ner, sex of examinee, and direction of expectancy variables.

Chapter II contained a review of recent (1960 to present )

literature concerning examiner influence on individual intelligence

54
test scores and on projective test score; research on experimenter
influence; and a discussion of previous research.

Examiner influence on individual intelligence test scores
and on projective test scores was investigated in a limited number
of research studies. In these studies the examiner's personality
(warmth and rapport) or physical characteristics (sex and race)
had an effect on the dependent variable. Specifically, opposite
sex combinations and same race combinations produced the highest
scores.

A review of studies of experimenter influence revealed that
experimenters frequently appear to influence the results of their
studies. However, the results were too conflicting and inconsistent
to warrant many definite conclusions regarding how experimenters
influence their data. The sex and race of experimenters seemed to
be significant variables, expecially when studied in interaction
with other variables. The warmth of an experimenter appeared to
have a consistent effect on data; that is, "warm? experimenters
influenced their research more than experimenters perceived as being
less warm.

The research on other experimenter personality characteristics
and also on the prestige of the experimenter revealed little consis-
tent evidence to support an assessment of their>importance. Simi-
larly, the effect of the experience of experimenter was discussed in
an unresolved debate by Rosenthal, Harrington, and Ingraham. Ihile
visual and verbal cues were found to mediate experimenter influence,
more research needs to be done to discover how experimenters influ-

ence their data.

55

The review of studies in which a'grand experimenter deliberately
gave experimenters an expectancy concerning the results of the study
consistently revealed a significant expectancy effect.

Rosenthal and Jacobson‘s (1968) Oak School experiment was dis-
cussed as a study in which the authors attempted to apply experimental
results to a practical setting. Rosenthal and Jacobson concluded that
the teachers in the experiment had communicated an expectation of
ability to “bloomers' which accounted for their intellectual 'spurt."

In a discussion of previous research some weaknesses in research
design were mentioned. In most studies experimenters were given only
one expectancy condition. This procedure was seen as a weakness
because an expectancy "set'' could develop which might account for the
influence effect. Much of the research involved male experimenters
and female subjects of similar age. The sex of experimenter and sex
of subject variables need to be crossed and younger subjects need
to be tested. Most of the studies employed meaningless rating instru-
ments which required little or no training to administer. This pre-
vented an examination of the experimenter's experience or training
as a variable in research on experimenter influence.

There seemed to be confusion concerning just how experimenter
influence operated. Rosenthal (1967a) believed that experienced
experimenters were likely to influence their research because they
were more ego-involved with their research and were 'better biasers.“
That is, they were better at communicating their expectancy to sub-
jects and better at reinforcing ”correct' responses. Thus, for
Rosenthal, experimenter influence increased with the experimenter's

experience and 'snowballed" as an experiment progressed. Friedman

56
(1967), one of Rosenthal's associates, agreed with this hypothesis.

This hypothesis was sharply criticized by Ingraham and Bar-
rington (1966, 1967) and by Barber and Silver (1968a, 1968b). These
authors felt that experimenters who were experienced and well-trained
relied more on factual cues given them during an experiment and less
on ambigqu cues given them by grand experimenters.

Chapter III included a discussion of the samples, instrumen-
tation, procedure, and the design and analysis of the study.

The souple consisted of four male and four female Vechsler
Intelligence Scale for Children (WISC) volunteer examiners who had
just cwpleted WISC training. They were paid 82.50 per test. Each
examiner tested eight junior high school students (four boys and four
girls) who were randomly selected from a population of students who
were ”average“ in intelligence (that is, they scored from ninety to
one hundred-ten on the California Test of Mental Maturity-«Short
Form, 1965 Revision). Examinees were randomly assigned to examiners.

The WISC is an individual intelligence test for children between
ages five and fifteen. It was selected for use in this study because
VISC examiners were available and because the IISC had good reliability
(split-half reliability coefficients ranged free: .86 to .96). A
rating sheet (see Appendix B) was used by the grand experimenter to
transmit an expectancy condition to examiners. On each examinee's
rating sheet a discrepancy between California Test of Mental Maturity-
Short Form score and school achievment was fictitiously indicated,
and a “predicted WISC score" was advanced.

After eight examinees were randomly assigned to an examiner,

57
they were randomly designated ”above average“ or “below average” so
that each examiner tested two "above average" boys, two "below
averageI boys, two ”above average” girls, and two “below average“
girls. The order in which an examiner tested his examinees was also
randaaized.

Parental approval for testing was secured and confidentiality
was assured. Neither parents nor examiners were told that the test-
ing was part of a research study or experiment. In no case did an
examiner indicate that he “guessed” that he was taking part in an
experiment. A11 testing was done in private offices at the high
school during a one week period.

Results were analyzed by four-way analysis of variance. The
variables were sex of examiner, sex of examinee, expectancy (“above
average" or “below average“), examiners nested in sex of examiner,
and replications nested in all other variables. The five per cent
level of significance was arbitrarily chosen.

In Chapter IV the eight hypotheses were restated in statisti-
cal form and rejected or not rejected. Both the sex of examiner
and sex of examinee effects were significant (p 9.9.05). As seen in
Table 5 on the next page, female examiners obtained higher’mean
WISC scores and male examinees achieved higher mean.VISC scores.
While the significant sex of examiner effect confirmed previous
research (Cieutat, 1965), the significant sex of examinee effect
was unexpected. It was conceptualized that male examinees were less
anxious and more competitive when placed in a novel situation. The
interaction effect between sex of examiner and sex of examinee was

not significant. The main sex effects seemed to over-shadow the

58

interaction effect.

TABLE 5

SUMMARY OF MEANS, STANDARD DEVIATIONS, AND F-RATIOS
FOR SIGNIFICANT EFFECTS

 

 

Standard
Effect Means DeViations FbRatio
Sex of Examiner
M810 97055 8’75
Female 103.12 8.06 8.17‘
Sex of Examines
Male 101.66 9.0}
Female 99.00 9.07 9.50'

 

I"Sig. at .05 level of confidence

The major finding in this study was that the expectancy effect
was not significant. Two principle explanations were offered. First,
the eight-step transmission process delineated by Barber and Silver
(1968a, p. 25) broke down in this study. That is, it was feared
that WISC examiners did not retain the expectancy condition given to
them by the grand experhmenter and, therefore, examiners' expectancy
was neither transmitted to nor received by examinees. Secondly, it
was felt that the WISC examiners were relatively well-trained and
experienced. Thus, the results of the study seemed to confirm Ingra-
ham and Harrington's (1966, 1967) and Barber and Silver‘s (1968a,
1968b) criticism of Rosenthal's research: experimenter influence
has not been consistently demonstrated when experienced eXperimenters
are administering structured tasks.

The fact that no significant examiner effect was demonstrated

in this study added to the argument that examiner influence was not

59

a great problem in intelligence testing, at least not in this study.
The expectancy variable was examined in interaction with both
the sex of examiner and sex of examinee variables and neither inter-
action was significant. Thus, the results of this study were not
construed as evidence that,ma1e eXperimenters were better “influ-
encers' or that female examinees were more easily influenced. These
two insignificant interaction effects represented further evidence
that the transmission of an expectancy condition did not take place,
and that the NISC was a structured task administered by experienced
examiners. Analysis of the three-way interaction effect among the
sex of examiner, sex of examinee, and eXpectancy variables revealed

no significant complex interaction.
3. Conclusions

The significant (p£E.C5) sex of examiner effect supported the
feeling that the sex of experimenter or examiner was an active,
though usually unpredictable variable. The mean Total WISC score
obtained by the female examiners in this study was 5.59 IQ points
higher than the mean Total WISC score obtained by male examiners.
If important and somewhat irreversible decisions were going to be
made for an individual on the basis of a WISC score, it would seem
wise to have that individual tested by both a male and a female
examiner. The added expense, effort, and time would be justified
in relation to the importance of the decision to be made.

The fact that a significant expectancy effect was not demons
strated in this study confirmed Barber and Silver's (1968a, p. 26)

conclusion that expectancy influence was not present when the experi-

mental task was structured and factual. The insignificant

6O
expectancy effect also supported Ingraham and Harrington's (1966,
1967) contention that experienced experimenters were not as likely
to influence their results as inexperienced experimenters. The
WISC examiners in this study did not retain the expectancy for an
examinee when they started the testing situation. Instead, it was
hypothesized that they increasingly responded to the factual stimuli
presented them by examinees.

It was concluded that a well-designed study would minimize
the probability of a significant expectancy effect. Specifically,
it was considered important to design a study employing both.male
and female examiners, and both male and female examinees. The sex
of experimenter and sex of subject variables were infrequently crossed
in previous research. In agreement with Ingraham.and Harrington
(1966) was the feeling that it was necessary to give examiners both
expectancy conditions in random order to prevent development of an
expectancy 'set.‘

It was also felt that many of the studies which demonstrated
a significant expectancy effect lacked rigorous statistical metho-
dology. Barber and Silver (1968a, p. 24) stated that,"(l) the vari-
ables to be studied and the statistics to be used should be speci-
fied in advance; (2) the level of significance should be stated in
advance; (3) the data should be analyzed by some 'overall' test such
as multivariate analysis of variance; and (h) conclusions should
not be made from the results of post hoc tests performed upon the
data after an overall test has failed to reject the null hypothesis."

In the present study results were analyzed by an overall test

61

(four-way analysis of variance), and no “post hoc tests' were per-
formed when the overall test failed to reject the null hypotheses.

Rosenthal (1967a) and other experimenters apparently assumed
that if experimenter influence could be demonstrated in laboratory
tasks, the results could be generalized or applied to practical and
more meaningful situations such as individual intelligence testing.
On the basis of this study and criticisms leveled by others (Barber
and Silver, 1968a, 1968b: Barber, 331.31., 1969; and Claiborn, 1969),
this assumption appeared unwarranted. Rosenthal (1967a) often maxi-
mized the probability of obtaining a significant expectancy effect
by (1) designing ambiguous experimental tasks; (2) using opposite
sex combinations of experimenters and subjects; (5) making experi-
menters subordinate to the grand experimenter; (h) giving experi-
menters only one expectancy condition; or even by (5) paying experi-
menters more if their results were in the desired direction. Under
a combination of those conditions, it was not surprising that a
significant expectancy effect was found. However, in practical
settings (such as individual intelligence testing) it was believed
that these conditions usually did not exist. The generalizability
of Rosenthal's experimental research was considered questionable and
was not supported in this study. In.fact, in light of the well-
reascned criticisms of Rosenthal's basic experimenter influence research
(Ingraham and Harrington, 1966, 1967: Barber and Silver, 1968a, 1968b;
and Barber, 3_t_.__g_1_., 1969). one must consider the possibility that

observer influence and expectancy effect simply may not exist.

0. Implications for Further Research

Any replication of the present study should include a more
adequate method of communicating the expectancy condition to V130
examiners and checking its presence during actual testing. Examiners
ego-involvement in the expectancy condition given them must be increased
and sustained. Controls should be included which derive from Barber
and Silver's (1968a, p. 25) 'eight—step transmission process." Before
the existence of examiner influence, if am, can be demonstrated,
future researchers must be certain that the eXpectancy communication
was “attended to,“ "comprehended," and ”retained“ by test examiners
21d “transmitted“ to examinees.

More importantly, further research concerning the sex of the
examiner as a variable influencing individual intelligence test
scores should be conducted. A research design employing male and
female examiners testing the same examinees with alternate forms of
the same test might be profitable.

From the results of their study, Rosenthal and Jacobson (1968)
implied that more research should be conducted in teacher training
programs to explore how teachers' expectations of pupils' ability
affects actual pupil performance. The results of this study, however,
did not imply a great need for further research on the problem of an
expectancy effect on individual intelligence testing. It was not
claimed that this study settled the question 'once and for all;'
however, the insignificant expectancy effect did support the conclu-
sions of other researchers (Ingraham and Harrington, 1966, 19673 and

Barber and Silver, 1968a, 1968b). Experimenter influence was not a

62

65
problem in structured, factual tasks administered by eXperienced

and well-trained experimenters.

BI BLI OGRAPHY

BI BLI OGRAPHY

Barber, T.X. Invalid arguments, postmortem analyses, and the experi-
menter bias effect. Journal of Consultingfnd Clinical Psy-

ChOIOEZ, 1969’ 35. 11-1Ke

Barber, '1'.x., Porgione, A., Chaves, .J.l"., Calvsrlsy, D.S., and McPeake,
J.D. dc Bowen, B. Five attempts to replicate the experimenter
bias effect. Journal of Consultingand Clinical Psycholgg,
1969’ 559 1"6e A

Barber, T.x., do Silver, ILJ. Fact, fiction, and the experimenter
bias effect. Psychological Bulletin Monograph, 1968a, 70,
1-29e ’

Barber, T.X., & Silver,‘M.J. Pitfalls in data analysis and interpre-
tation: a reply to Rosenthal. Psmhclogical Bulletin Mono-
m 1" lfiab’ 70, 48-62e ‘ '

Bass, BM. Measures of average influence and change in agreement of
rankings by a group of judges. Socicmetgy, 1960, 25, 195-202.

Blaufarb, H. The relation of experimenter status and achievement
imagery to the conditioning of verbal behavior. Unpublished
Ph.D. dissertation, Univ. of Illinois, 1960.

Boneau, C.A. The effects of violations of assumptions underlying the
t-test. Psychological Bulletin, 1969, 57, h9—69.

Buros, 0.x. (ed. ). The Fifth Mental Meagxgemehts Yearbook. Highland
Park, N.J.: The Gryphon Press, 1959.

Buros, 0.x. (ed. ).,The Sixth Mental Heasurements Yearbook. Highland
Park, N.J.: The Gryphon Press, W -

Brogden, V.J. The experimenter as a factor in animal conditioning.
Psychological Raports, 1962, 11, 259-242.

Carlsmith, J.M., a Aronson, E. Some hedonic consequences of the con-
firmation and disconfirmation of expectancies. Journal of

Abnormal and Social Psycholog, 1965, 66, 151-156.

Cieutat, V.J. Examiner differences with individual intelligence
tests. Perceptual and Motor Skills, 1965. 20, 1517-1318.

Claiborn, LL. hpectancy effects in the classroom: a failure to repli-
cate. Journal of Educational Psycholofl, 1969. 60, 577-585.

64

65

Cohen, E. Examiner differences with individual intelligence tests.
Perceptual and Motor Skills, 1965, 20, 1524.

 

CoOper, J., Eisenberg, L., Robert, J., do Dohrenwend, B. The effect
of experimenter expectancy and preparatory effort on belief in
the probable occurrence of future events. Journal of Social

chholog, 1967, 71, 221-226.

 

Cordaro, L., dc Ison, J.R. Psychology of the scientist: I. Observer
bias in classical conditioning of the planarian. Psychologi-
cal ”Rom, 1963’ 15’ 787‘789.

Cronbach, L.J., Essentials of Psychological Testing. Second edition.
New York: Harper and Brothers, 1960.

Exner, J.E. Jr., Variations in WISC performances as influenced by
differences in pretest rapport. Journal of Genetic Psflhology,

1966.715 299-506

Friedman, N. The Social Nature of P jchological Research. New York:
Basic Books, Inc., 1967.

 

Friedman, N., Kurland, D., a Rosenthal, R. Experimenter behavior as
an unintended determinant of experimental results. Journal
of Projective Technimies and Personality Assessment, 1965,
29’ 1‘79190e

Griffith, R.M. Rorschach water percepts: a study in conflicting
results. American Psychologist, 1961, 16, 507-511.

Harrington, 6.14., a. Ingraham, 18., Psychology of the scientist: XXV.
Experimenter bias and tails of Pascal. Psychological Reports,
1967. 21. 515-516.

Hays, 11.1.. Statistics for Psycholojgibsts. New York: Holt, Rinehart,
& Winston, 1961‘s

Hill, K.T., in Stevenson, HJ. The effects of social reinforcement
vs. non-reinforcement and sex of E on the performance of
adolescent firls. Journal offiPersonality, 1965, 55, 50-56.

Ingraham, L.H., b Harrington, (Mi. Psychology of the scientist:
XVI. EXperience of E as a variable in reducing experimenter
bias. Psycholggipal Reports, 1966, 19, 455-461.

Jones, E.L. The courtesy bias in South-East Asian surveys. Interna-
tional Social Science Journal, 1965, 15. 70-76.

Katkin, 3.8., Risk, R.T., dc Spielberger, C.D. The effects of experi-
menter status and subject awareness on verbal conditioning.
Journal of Echrimental Research in Personality, 1966, 1,
155-160.

 

66

Katz, 1., Robinson, J., Epps, 3., do Naly, P. Race of experimenter
and instructions in the expression of hostility by Negro boys.
Journal of Social Issues, 1964, 20, 54-60.

 

Kintz, B.L., Delprato, D.J., Mettee, D.R., Persons, 0.3., b Schappe,
3.1!. The experimenter effect. ngcholcgical Bulletin, 1965,

 

Kish, L. Studies of interviewer variance for attitudinal variables.
Journal of the American Statistical Association, 1962, 57,
92-115e

Larrabee, L. I», do Kleinsasser, L.D. The effect of experimenter bias
on WISC performance. Unpublished paper, St. Louis, Mo: Psy-
chological Associates, 1967.

Levy, L.H. Reflections on replications and the experimenter bias
effect. Journal of Consulting and Clinical Psychology, 1969
53’ 11-1he '

Littell, U.A. The Wechsler Scale for Children: review of a decade
of research. Psychological Bulletin, 1960, 57, 152-156.

Masling, J. The influence of situational and interpersonal variables
in projective testing. Psychological Bulletin, 1960, 57, 65-85.

McFall, R.M. Unintentional comunication: the effect of congruence
and incongruence between subject and experimenter constructions.
Unpublished Ph.D. dissertation, Ohio State Univ., 1966.

McGuigan, 1".J. The experimenter: a neglected stimulus object.
Psychological Bulletin, 1965, 60, 421-428.

Miller, M.E., & Solkoff, N. Effects of mode of response and sex of
E upon recognition thresholds of taboo words. Perceptual and
Motor Skills, 1965. 20, 575-578.

Mulry, R.C. The effects of the experimenter's perception of his own
performance on subject performance in a pursuit rotor task.
Unpublished master's thesis, Univ. of North Dakota, 1962.

Norton, D.w. An empirical investigation of the effects of non-
normality and heterogeneity upon the F-test of analysis of
variance. Ph.D. thesis, State Univ. of Iowa, 1952.

Ogawa, J ., & Cakes, NJ. Sex of experimenter and manifest anxiety as
related to verbal conditioning. Journal of Personality, 1965,

55. 555-569-

Prince, A.I. Relative prestige and the verbal conditioning of children.
Paper read at the Seventieth Annual Convention of the American
Psychological Association, St. Louis, Missouri, 1962.

67

Reese, M.M., Jo Whitman, R.N. Expressive movements, warmth, and verbal
reinforcement, Journal of Abnormal and Social Psycholog, 1962,

64, 254-256.

Rosenthal, R. Covert communication in the psychological experiment.
Psychological Bulletin, 1967a, 67, 556-567.

Rosenthal, R. Experimenter attributes as determinants of subjects'
responses. Journal of Proaective Techniques and Personality
Assessment, 1 5a, 27, 52 -551.

Rosenthal, R. Experimenter Effects in Behavioral Research. New York:
Appleton-Century-Crofts, 1966.

 

Rosenthal, R. Experimenter expectancy and the reassuring nature of
the null hypothesis decision procedures. Psychological Bulle-

tin Mongfimlh’ 1%8’ 70’ 30-470

Rosenthal, R. Experimenter modeling effects as determinants of sub-
jects' responses. Journal of Prpjective Techniques and Personality
Assessment, 1965b, 27, 467371.

 

Rosenthal, R. Experimenter outcome-orientation and the results of the
psychological experiment. Psychological Bulletin, 1964, 61,

Rosenthal, R. On not so replicated experiments and not so null results.
Journal of ConsultingCand Clinical Psycholpfl, 1969, 55, 7-10.

Rosenthal, R. Psychology of the scientist: XXIII. Experimenter
expectancy, eXperimenter experience, and Pascal's wager.
Psychological Reports, 1967b, 20, 619-622.

Rosenthal, R. Psychology of the scientist: XXVI. Experimenter expec-
tancy, one tail of Pascal, and the distribution of three tails.

Psychological Reports, 1967c, 21, 517-520.

Rosenthal, R., 8c Pods, K.L. The effect of experimenter bias on the
performance of the albino rat. Behavioral Science, 1965a, 8,
185-189.

Rosenthal, R., do Pods, K.L. Psychology of the scientist: V. Three
experiments in experimenter bias. Psychological Reports, 1965b,

Rosenthal, R., Pods, K.L., Friedman, C.J., & Vikan-Kline, L.L. Sub-
jects' perception of their experimenter under conditions of
experimenter bias. Perceptual and Motor Skills, 1960, 11,
325-551-

Rosenthal, R., Pods, K.L., Vikan-Kline, L.L., do Persinger, GJ.
Verbal conditioning: mediator of experimenter expectancy
effects. Psychological Reports, 1964, 14, 71-74.

68

Rosenthal, R., Friedman, C.J., Johnson, C.J., Fode, K.L., Schill,
T., White, R.C., &.Vikan-Kline, L.L. Variables affecting experi-
menter bias in a group situation. Genetic Psychology Meno-
ra hs, 1964, 70, 271-296.

 

Rosenthal, R., Friedman, N., & Kurland, D. Instruction-reading behavior
of the experimenter as an unintended determinant of experi-
mental results. Journal of Experimental Research in Personality,
1966, 1, 221-226.

 

Rosenthal, R., &,Halas, E.S. Experimenter effect in the study of
invertebrate behavior. Psychological Reports, 1962, 11,
251-2%e

Rosenthal, R., &.Jacobson, L. Pygmalion in the Classromm: Teacher
Expectation and Pupils' Intellectual DevelOpment. New York:
Holt, Rinehart, and Winston, 1968.

 

 

Rosenthal, R., &,Jacobson, L. Teachers' expectancies: determinants
of pupils' IQ. gains. Psychological Reports, 1966, 19, 115-118.

 

Rosenthal, R., Kohn, P., Greenfield, P.M., & Carota, N. Data desira-
bility, experimenter expectancy, and the results of psycholo-
gical research. Journal of Personality and Social Psychology,
1966. 5. 20-27.

 

Rosenthal, R., Kohn, P., Greenfield, P.M., aiCarota, N. Psychology
of the scientist: XIV. Experimentsrs' hypothesis-confirmation
and mood as determinants of experimental results. Perceptual
and Motor Skills, 1965, 20, 1257-1252.

Rosenth.1’ Re, ““117, R.C., "nimr, Go's, Vikan-Klim, Lelia, &
Grothe, M. Changes in experimental hypotheses as determinants
of experimental results. Journal of Pro'ective Techniques
and Personality Assessment, 196ha,257'435-h69.

 

Rosenthal, R., Mulry, R.C., Persinger, G.W., Vikan-Kline, L.L., 8r.
Grothe, M. Emphasis on experimental procedure, sex of subjects
and the biasing effects of experimental hypotheses. Journal

of Projective Technigues and Personality Assessment, 1964b,
2 . 70 75.

Rosenthal, R., &.Persinger, G.W. Let's pretend: subjects' perception
of imaginary experimenters. Perceppual and Motor Skills,
1962, 14, 407-409.

 

 

Rosenthal, R., Persinger, G.W., a Rode, K.L. Experimenter bias,
anxiety, and social desirability. Perceptual and Motor Skills,
1962. 15. 75-74-

 

Rosenthal, R., Persinger, G.W., Vikan-Kline, L.L., & Fode, K.L. The
effect of early data returns on data subsequently obtained by
outcome-biased experimenters. Sociometry, 1965a, 26, 487-498.

 

69

Rosenthal, R., Persinger, G.W., Vikan-Kline, L.L., & Fode, K.L. The
effect of experimenter outcome-bias and subject set on aware-
ness in verbal conditioning eXperiments. Journal of Verbal
Learningyand Verbal Behavior, 1965b, 2, 275-285.

 

 

Rosenthal, R., Persinger, G.W., Vikan-Kline, L.L., & Mulry, R.C.
The role of the research assistant in the mediation of experi-
menter bias. Journal of Personality, 1965, 51, 515-555.

 

Sarason, I.G. Individual differences, situational variables, and
personality research. Journal of Abnormal and Social ng-

chology, 1962, 65, 576-580.

Sarason, I.G., & Minard, J. Interrelationships among subject, experi-
menter, and situational variables. Journal of Abnormal and
Social Psycholggy, 1965, 67, 87-91.

 

 

Scheffe, H. The Analysis of Variance. New York: Wiley, 1959.

 

Simmons, W.L., & Christy, E.G. Verbal reinforcement of a TAT theme.
Journal of Projective Techniques and Personality Assessment,

1962. 25. 557-541-

Stevenson, H.W. Social reinforcement with children as a function of
CA, sex of E, and sex of S. Journal of Abnormal and Social
Ps cholo y, 1961, 65, 147-154.

Stevenson, H.W., & Allen, 8. Adult performance as a function of sex
of experimenter and sex of subject. Journal of Abnormal and
Social Psychology, 1964, 68, 214-216.

 

Stevenson, H.W., & Allen, 8. Variables associated with adults' effec-
tiveness as reinforcing agents, Journal of Personality, 1967,
55’ 2146-264.

Suchman, E.A. An analysis of "bias” in survey research. Public Opinion
Quarterly, 1962, 26, 102—111.

Thorndike, R.L. Review of Rosenthal, R. & Jacobson, L. :yggalion in
the Classroom. American Educational Research Journal, 19 8,

5. 708-711.

Turner, G.C., &,Coleman, J.C. Examiner influence on Thematic Appor-
ception Test responses. Journal of Projective Technigues and
Personality Assessment, 1962, 26, 478-486.

 

Wartenberg-Ekren, U. The effect of experimenter knowledge of a sub-
ject's scholastic standing on the performance of a reasoning
task. Unpublished master's thesis, Marquette Univ., 1962.

Wechsler, D. WISC Manual. New York: The Psychological Corporation,
1949.

 

7O

Wilkie, C.H. A study of distortions in recording interviews. Social
work, 1965’ 8’ 51-56.

Williams, J.A. Interviewer—respondent interaction: a study of bias
in the information interview. Sociometry, 1964, 27, 558-552.

 

Winkel, G.H., a Sarason, I.G. Subject, experimenter, and situational
variables in research on anxiety. Journal of Abnormal and

Social Psycholggy, 1964, 68, 601-658:

Young, R.K., & Veldman, D.J. Heterogeneity and skewness in analysis
of variance. Perceptual and Motor Skills, 1965. 16, 588.

APPENDICES

APPENDIX A

APPENDIX A

TABLE 6

CM MEANS AND STANDARD DEVIATIONS
FOR ALL SUBGROUPS

 

 

Standard
Effect Means Deviations
Sex of Examiner
Males 101.78 5.81
FemaIQ. 102s 00 5e61
Sex of Examinee
Males 100.97 6.05
Females 102.81 5.21
EXpectancy
”Above Average' 102.88 5.29
"Below Average" 100.91 5.94
Examiner v
1 105.58 6.19
2 99.62 5.55
5 101.25 5.96
4 102.88 7.22
5 105.75 5.86
6 101.75 5.42
7 101.65 5.50
8 100.88 7.72
Sex of Examiner X Sex of Examinee
Male Examiner-Male Examinee 100.44 6.46
Male Examiner-Female Examinee 105.12 74.92
Female Examiner-Male Examinee 101.50 5.75
Female Examiner-Female Examinee 102.50 5.65

71

TABLE 6 (cont'd.)

 

 

Standard
Effects Means Deviations

Sex of Examiner X Expectancy
Male Examiner-“Above Average“ Examinees 102.94 5.52
Male Examiner-”Below Average" Examinees 100.62 6.22
Female Examiner-"Above Average" Examinees 102.81 6.02
Female Examiner-“Below Average" Examinees 101.19 5.25

Sex of Examinee x Expectancy
Male Examinee-"Above Average” 102.44 5.75
Male Examinee-"Below Average' 99.50 6.07
Female Examinee-'Above AverageII 105.51 4.88
Female Examinee-"Below Average“ 102.51 5.64

Sex of Examiner X Sex of Examinee x Expectancy

Male Examiner-Male Examinee-'Above Average" 102.00 6.48
Male Examiner-Male Examinee-'Below Average' 98.88 6.47
Male Examiner-Female EXaminee-"Above Average" 105.88 4.09
Male Examiner-Female Examinee-“Below Average” 102.58 5.85
Female Examiner-Male Examinee-”Above Average“ 102.88 5.46
Female Examiner-Male Examinee-”Below Average“ 100.12 6.01
Female Examiner-Female Examinee-”Above Average“ 102.75 5.80
Female Examiner-Female Examinee-“Below Average” 102.25 5.90

72

APPENDIX B

APPENDIX 8

Rating Sheet

NAME

Group 1.0. Score

 

 

(CTMM-S Form) 1 _l ‘__ 1
Bottom } [5 2nd «} I 5rd }- II Top 3}
School Aciev-
ment Average | ] 1?
Bottom f l 2nd 3} I 5rd '5 I Top If

Predicted WISC Score

 

75

APPENDIX C

APPENDIX 0

TABLE 7

WISC TOTAL SCORES

Expectancy of Examiner by Examinee Sex

 

 

 

 

 

 

 

 

 

 

 

 

Male Examinee Female Examinee
'Above |'Below ‘Above “Below
Sex of Average" Average' Average” Average”
Examiner Examiner Expectgpcy Expectancy Expectgpcy Expectanc
1 107 106’ 95 101
88 101 497 87
2 116’ 101 101 90
Male 107 99 88 110
' ""5: " 111 88 8'5 101
89 2L 9: 95
4 100 89 109 89
85 94 117 94
T 98 106 107 91
104 118 108 106
6 163‘ 105 975‘ 101?
Female 109 102 111 101
7 112 109 ‘85 107
115, «98 99 114
8 110 91 99
95 199 109 98

 

 

 

74

 

"I1i!1111'1171111111111S