iNCUMS‘SYENT STATUS QHARACTEMSTICS
AND \NFLUENCE PROCESSES: _
A REPUCAT‘OM AND REFORMULMIOM

“We-"sis for the Dams M M A.
meme»: STATE UNWERSETY
PAUL t-L mess '
x972 '

......

c . . ,

  

LI B R A R y V
Michigan State
University .

 
     

ABSTRACT
INCONSISTENT STATUS CHARACTERISTICS

AND INFLUENCE PROCESSES:
A REPLICATION AND REFORMULATION

By

Paul H. Tress

Theoretical and experimental social psychology in recent years
has concentrated on how status variables affect the power and pres-
tige orders of small groups. The chief finding of such investigations
is that individuals who possess the higher state of a characteristic
that differentiates members of a group are less prone to influence
than individuals who possess low states of the characteriStic.

Until recently, work has concentrated on the case of one char-
acteristic, showing that the differences in acceptance of influence
among group members is related to the status differential existing
among the group members. Berger and Fisek (1970) have conducted an
experiment with the intent of generalizing such a finding to the case
of multiple characteristics. More specifically, Berger and Fisek used
two characteristics in a dyad where the characteristics were relevant
to the group's task. Our work was initially concerned with issues
raised by the Berger and Fisek paper and resulted in a reformulation
of the theory.

In order to do this we first show the relationship between status
characteristics and the pattern of interaction a group undergoes in
a two-step decision task. If one member of the group changes his
decision between the first and second step, the member is said to

be influenced.

Paul H. Tress

Influence is related to possession of different states of status
characteristics relevant to the group's task. Individuals can possess
similar or dissimilar states of characteristics, that is, they can
evaluate themselves according to whether or not they are univalent
and possess the same state for each characteristic. We agree with
Berger and Fisek that individuals who are univalent in the high state
will be less prone to influence than individuals who are univalent in
the low state. However, our major concern is the direction and nature
of univalence in the multivalent or non-univalent case. Whereas Berger
and Fisek argue an individual in such a case will combine the two char-
acteristics or cognitively balance them, our opinion differs. Berger
and Fisek fail to realize combining or balancing implies some univalent
process that needs to be specified. Such a specification takes the
form of inquiring into how information individuals have about each
other on their differentiating characteristics is related to or maps
onto some expected level of perfbrmance on the group's task.

An experiment was conducted to force disagreement among subjects
in a problem-solving task. The only information subjects had about
each other were the states of the characteristics they and their
partner in the dyad process. The resolution of such disagreement
provides an indirect measure of influence.

Despite some differences in the design of our experiment and
Berger and Fisek's, the results are almost similar. Statistical
analysis gives little insight into the nature of the univalent
process. However, the process appears to be an independent trials

process and reaches stability quite rapidly.

INCONSISTENT STATUS CHARACTERISTICS
AND INFLUENCE PROCESSES:
A REPLICATION AND REFORMULATION

By
3“" (I. L}.
Paul HVWTress

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS
Department of Sociology

1971

ACKNOWLEDGMENTS

The ideas developed in this thesis have evolved over the past
two years due to continual interaction with the members of the
thesis committee. I am especially indebted to Dr. Thomas L. Conner,
the chairman of the committee. Continual debate and dialogue with
Dr. Conner resulted in an almost periodic re-evaluation of the intent
and meaning of the thesis.

The time spent in discussion with Dr. Conner was matched with
time spent with other members of the committee. I would like to
thank these members for their guidance and advice. These individuals
are Drs. Bo Anderson, Santo F. Camilleri, and Hans B. Lee.

The research reported in this thesis was sponsored by a Social
Science Research Training Grant, N.I.H.M.H. lIUIO—02,administered

by Dr. Camilleri.

ii

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . .
LIST OF FIGURES . . . . . . . . . . . . . . . . . . .
INTRODUCTION . . . . . . . . . . . . . . . . . . . . .
GENERAL THEORETICAL CONSIDERATIONS . . . . .
PROCESSUAL CONSIDERATIONS . . . . . . . . .

THE EXPERIMENT . . . . . . . . . . . . . . . . . . . .
Phase I O O 0 O I O O O O O O O O O O O O O O I 0
Phase II 0 I O O O O O O O O O O O O O O O O O O 0

RESULTS . . . . . . . . . . . . . . . . . . . . . .
Subjects . . . . . . . . . . . . . . . . . . .
Comparison of the Original Experiment and the

Replication . . . . . . . . . . . .
The Nature of Univalence . . . . . . .
Processual Observations . . . . . . . . . . . . .

SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . . .
BIBLIOGRAPHY O O O O O O O O O O O O O O I O O C .

APPENDIX A: INSTRUCTIONS GIVEN TO SUBJECTS . . . . . .
Phase I . . . . . . . . . . . . . . . . . .
Phase II . . . . . . . . . . . .
Boardman: Part II . . . . . . . . . . . . . . . .

APPENDIX B: INTERVIEW SCHEDULE . . . . . . . . . .
Phase I . . . . . . . . . . . . . . . . . . .
Phase II . . . . . . . . . . . . . . . . . .
Debriefing . . . . . . . . . . . . . .

APPENDIX C: REASONS FOR ELIMINATING SUBJECTS FROM
SAMPLE O O O O O O O O 0 O O O O O C I O I 0

APPENDIX D: BERGER—FISEK HOST PROCEDURE: PHASE II . .

iii

iv

19
2O
22

27
27

27
29
3M

ui
as
1+7
1+7
53
58
60
so
61
62
64

66

LIST OF TABLES

P(S), Mean, and Variance for Inconsistent Symmetric
Studies 0 I O O O O O O O O O O O O O O O O .

Tests for Combining of Data . . . . . . . . . . . .
Previous Studies of Specific Characteristics .

Distribution Vector and Observed Values for
Critical Trials until Theoretical Stability . . .

iv

28

29

30

37

10.

11.

LIST OF FIGURES

The Influence Process Arising from Initial

Disagreement . . . . . . .

P's Evaluation of Alternatives

P's Evaluation and Selection of Alternatives
Initial Evaluation and Selection of Alternatives
Subsequent Evaluation of Self and Other . . . . .
Observed Frequency of Stay Re8ponses

Observed and Expected Frequency of Stay Responses,

Given Two Parameters . . .

Processual Matrices . . . . .

Observed and Predicted Values of P(S) for Each

Critical Trial . . . . . .

Plot of Trial Blocks . . .

Detail of Plot of Trial Blocks

10

10

11

ll

31

32

36

38

39

#0

INTRODUCTION

Within the last decade, researchers have shown how experimentally
induced status variables affect the power and prestige order of small
groups. This work has attempted to formalize well-known influence
processes in human interaction. The main finding of these studies
is that individuals who possess the higher of two different states
of a status characteristic will be less prone to influence than indi-
viduals with a lower state of the characteristic.1

This process occurs whether the status characteristic is diffuse
or specific. Diffuse characteristics are those that are not initially
relevant to the group's task. As a diffuse characteristic, income
level is not related to chess-playing ability. However, one would
base his expectations of an individual's chess-playing ability on
income level if no other basis of differentiation were present (Berger,
Cohen, 8 Zelditch, 1966). Specific characteristics are those that are
germane to the group's task throughout the task-focused activity of
the group. High verbal ability would result in a high expectation for
an individual possessing this ability, if the group's task were some

type of word game, and if the relevance between the level of verbal

 

1This interpretation slightly contrasts with the statement that
high status members exercise greater influence on the task than low
status members. Our statement is more amenable to empirical verifi—
cation; (See Berger 8 Fisek, 1970, p. 287, for an example of the
above.

ability and the expected performance on the word game were made
explicit.

The details of such influence processes have been investigated.
However, such studies have concentrated on only one differentiating
dimension, i.e., either one diffuse or one Specific status character-
istic. A recent paper has tried to investigate such influence processes
in the case of more than one characteristic. The work of Berger and
Fisek (1970) was concerned with two characteristics in a dyad.2 Berger
and Fisek were concerned with two specific types of arrays of the char-
acteristics: (a) a consistent state of exact opposites, with one person
having a high state of each characteristic and the other person having
a low state of each characteristic; and (b) an inconsistent state of
polarized mirror images. (One person is high on one characteristic
and low on the other, while the other person is respectively low and
high on these characteristics.)

In the inconsistent case, each individual is faced with dissonant
or incongruent information about himself and the other individual.

Such inconsistency must be resolved. The bulk of the Berger and
Fisek work was a concern with the resolution of such dissonance or
incongruity.

Our study has two concerns, a specification of the resolution
process and an attempt to replicate Berger and Fisek's experiment.

We will first discuss general theoretical formulations and then detail

 

2Most of the work done on the relationship of status character-
istics and influence processes is in the dyad. Theoretical and
methodological considerations with other size groups increase at an
exponential rate. This is a limitation of work done on the formalization
of influence processes.

the resolution of the influence process in the inconsistent case.

This will be followed by a description of our experiment. The experi-
ment was designed to replicate Berger and Pisek's work and to give us
some insight into the nature of the resolution process.3 Following
this, our data analysis is given and conclusions about the nature of

the influence process are drawn.

 

3Our replication is not concerned with other possible arrays.
Included in this is the "status edge case" where one individual has
consistent and the other inconsistent states of the characteristics.
In this case, the characteristics are not in a symmetric array as in
the inconsistent case (Berger 8 Fisek, 1970, p. 301). Berger and Fisek
neglect to investigate two other cases: individuals may be exactly
the same and consistent or inconsistent. Our replication is concerned
only with the inconsistent polarized mirror image case.

GENERAL THEORETICAL CONSIDERATIONS

A brief discussion of theoretical considerations common to the
relationship of status characteristics and influence processes is
needed before we detail the nature of the influence process in the
multiple characteristic case.

In all cases we assume interaction takes place in a dyad. Such
interaction is seen through the eyes of one actor whom we designate
as p. The other actor is designated as o. P and o possess states of
Ci’ a status characteristic. This will be denoted as Cl and C2 in the
case of two characteristics.

We will assume that interaction between p and o is task-focused
and that p and o are oriented to each other. The interaction is task-
focused in that p and o are oriented to successful completion of the
task. The orientation of p and o to each other means that p and o
evaluate each other's task performance and have expectations as to
each other's future performance.

At the initial part of the interaction process p and 0 have no
prior knowledge of which Ci is instrumental to successful completion
of the group's task. However, we assume p and 0 know that possession
of some state of some specific characteristic is necessary to success—
ful completion of the group's task.

In order for some differentiating attribute of individuals to be

considered a specific Ci’ three conditions must hold:

4

(1) Ci must have different states or degrees which are recog-
nized by p and o.

(2) P and 0 must associate each of these differentiated states
with certain levels of expectations as to the future performance of
himself and other on the task if task performance is related and rele—
vant to possession of some state of Ci'

(3) Knowledge of the states of C1 possessed by p and o generates
general expectations to p and 0 about the personal qualities of indi-
viduals who possess such states of Ci'

For simplicity of analysis, we further assume each Ci in the
case of multiple Cis is equally relevant to the group's task and that
the Cis are differentiated in a dichotomized sense of high and low
degrees. These degrees are designated by H and L respectively. Thus,
X:ab designates person X's state of C and C . For example, p:HL,

1 2

o:LL means p is high on C1 and low on C2 respectively. PzHH, o:LL
and p:HL, o:LH are examples of consistent and inconsistent symmetric
cases respectively.

Our major theoretical concern is the relationship between the
Cis and influence processes within the group. It has been found that
the rate at which p and o reject influence from each other in task
performance activities is related to possession of high states of
Ci’ if Ci is relevant to the group's task (Berger, Conner, 8 McKeown,
1969; Berger 8 Conner, 1970; Moore, 1969). To paraphrase Moore, the
argument in its most general form thus asserts that the influence
differential between S and 0 is a direct function of the status dif-

ferential existing between S and 0 (Moore, 1969, p. 1M7; Moore's S

is our p).

In order to understand the nature of this "direct function," we

need to detail the influence process.

PROCESSUAL CONSIDERATIONS

Our analysis of the process will be seen through the eyes of p.
Let us assume that the group's task involves a Choice among a set of
alternatives. More specifically, the task is binary with two alter-
natives, A and B. With this basis we state the following:

Assumption 1: An alternative positively evaluated by p

(or 0) will be selected by p (or o). A negatively evaluated
alternative will be rejected.

 

Let us further assume that p and o are differentiated on some
C18 and that these Cis are relevant to the group's task. The Cis
are specific and the only information p and 0 have about each other.
If p gives a performance output to o in the form of an attempted
solution to some task problem, and if 0 has a positive reaction to
p's performance output (or vice versa), we say that p and 0 have
agreed on a solution to a task problem. If one member of the dyad
has a negative reaction to the performance output of another, the
members of the dyad are said to be in disagreement. If the solution
to a problem involves an initial and final choice of alternatives by
p and 0, initial and final agreement and disagreement are possible.

Obviously, we are interested in the case of initial disagreement.
If one member's selection is in initial disagreement with another
member's selection, a performance output has been given from one
member to the other in the form of a negative reaction to selection
of alternatives. This may result in an influence process, since p

7

and 0 may change their evaluations of each other's initial perform-

ance output. Figure 1 illustrates this process.”

X(+R)X
O‘1

X(PO)X
X(IB)X
31
X(-R)X ”(130)?

 

B2
Y<IB)x

X and X are different actors.

a1, a2, 81, 82 are constants such that a1 + a2 = l and 81 + 82 = 1.

(PO) "gives a performance output to"

(+R) "positively reacts to" or "selects the performance output of"

(-R) "negatively reacts to" or "does not select the perfbrmance
output of"

(IB) "is influenced by"

Fig. l.--The influence process arising from initial disagreement.

Since the branches of Figure l are mutually exclusive and ex—
haustive, we may conceive of the as and 88 as rates or mathematical
probabilities such that the as and the Bs sum to one. To preview our
experimental design, it should seem clear that by fixing the as such
that a2 is high, the 83 will exist. The size of 81 and 82 will reflect
the amount of influence. A comparison of the 83 with the states of
Ci possessed by p and 0 will operationalize our postulated direct

function between the differential statuses of p and o and the influence

 

”Our Figure l is a modification of Berger and Conner's (1970)

figures.

differential between p and o. For example, in Figure l we would expect
82 to be larger than 81 if p, as X, possesses higher states of the

Cis than 0 and the Cis are seen by p and o as being equally relevant

to the group's task.

If p and 0 have negative reactions to each other's performance
output, that is, initial selection of alternatives, the path a2 will
be followed. P and 0's selection of available alternatives is assumed
to occur after p and o differentially evaluate these alternatives.

If p chooses alternative A and o chooses B, or vice versa, p must
decide who is correct, himself or 0.5 Since the Cis that p and o
possess are relevant to the task, the Cis inform p about his and 0's
ability to perform on the task. Since our interest is an initial dis-
agreement of p and 0, we make the following assumption:

Assumption 2: P associates a negative reaction to his

performance output as a difference in evaluation of
alternatives by himself and o.

 

The critical part of the influence process is the relationship
of the Bs and p and 0's states of the Cis. Let's diagram our process
as discussed so far.6 As indicated in Figure 2, p has differentially
evaluated alternatives A and B. In Figure 3 p selects the alter-
native he has positively evaluated and rejects the alternative he
has negatively evaluated.

Let's further assume 0 undergoes the same process as p. Since

we are interested only in influence processes which occur if p and o

 

5We are assuming p and o are collectively oriented towards the
correct solution of the task. Our analysis is seen through the eyes
of p.

6These figures were originally developed in Conner (1965) and
presented in Berger, Cohen, Conner, 8 Zelditch (1966).

10

initially select different alternatives, we assume 0 has selected
alternative B and rejected alternative A. We will assume p is

able to be an object of orientation to himself, that is, p can
reflect about his actions and develop attitudes and cognitions

about himself, p'. Thus p is aware of his and 0's evaluation and
selection of alternatives. This is shown in Figure H. Assumption 1

implies o's selection of his positively evaluated alternative.

+ positively evaluates

- negatively evaluates

P
Fig. 2.——P's evaluation of alternatives.
+ positively evaluates
- negatively evaluates
P

+ selects

- __rejects

 

Fig. 3.——P's evaluation and selection of alternatives.

However, the only bases p has for an evaluation of his and 0's
present and future performances are the C13. This is the only informa-
tion p and 0 have about each other, and this information is relevant
to successful completion of the task. This is shown by Figure 5.

On the basis of his self and other evaluation, p will make a

final choice of alternative A or B. The final choice will be in

ll

balance with the initial choices made by p and o and with p's evalua-

tion of p' and o.

P and o -- actors

A and B -- alternatives

p' -- p as object of orientation
to self

+ positively evaluates

 

- negatively evaluates

\l B Z + selects

:___rejects

Fig. u.--Initia1 evaluation and selection of alternatives.

 

P and o —- actors

A and B —- alternatives

p' -- p as object of orientation
to self

+ positively evaluates

- negatively evaluates

 

_ i __ selects

— _ rejects

Fig. 5.-—Subsequent evaluation of self and other.

Let's look at the left-hand part of Figure 5. This concerns the
relationship between actors p and p' over alternatives or objects A
and B. Suppose p positively evaluates himself, p'. This relation of
p with p' over alternatives or objects A and B is assumed to be bal-
anced or to seek balance. According to formalized arguments about
balance theory, the product of the signs in a relation among people
and objects consisting of evaluations linking the people and objects
must be positive for the relationship to be balanced (Berger, Cohen,

Snell, 8 Zelditch, 1962). For example, if p positively evaluates

l2

himself after an initial selection of alternative A, p will stay with
his initial choice and resist the influence of 0's choice of alter—
native B if p and o initially disagree in their evaluation and selec-
tion of alternatives. If p had negatively evaluated himself and
positively evaluated 0 in the same situation, p's final choice would
be B. Balance theory would predict the same results in our example

if p negatively evaluated 0 in the first case and positively evaluated

0 in the second case, independent of his evaluation of himself.

 

If, after initial disagreement of alternatives, p changes his
choice of alternatives, p is influenced. Thus p's final choice in-
dicates whether he is influenced by 0 if the final and initial choices
are compared. Influence is a function of p and 0's initial choice of
alternatives and p's evaluation of p' and 0 based on possession of
levels of the Cis relevant to the group's task. Therefore:

Proposition 1: At the final stage of a two-step decision

process, p will tend to select and positively evaluate

alternatives A and B such that his selection is in balance

with his initial evaluation of alternatives and his self-
evaluation or evaluation of o.

 

This proposition is Moore's postulated direct relationship between an
actor's possession of states of C18 and his propensity to be influenced.
It should be intuitive to the reader that p's selection of
alternatives in the second part of the decision process is a binary
process consisting of independent trials. If we assign a l to the
event wherein the same alternative is selected at both stages of the

decision process and a O to other selection conditions, we can con-

n X.
ceive of a random variable X and Z -i-, where n is the number of
i=1 n

times p undergoes the two-step decision process and Xi = l or O.

This is the maximum likelihood estimator and the best estimator of

13

P(S), the probability of a self or stay response. Thus if P(S) is
high, p undergoes little influence. P(O), the probability of an other
response or of being influenced, is equal to 1 — P(S). P(S) and P(O)
are the Bs in Figure 1.

Our major concern is the relation of possession of states of Cis
and the influence process. Therefore, we are concerned with a situation
such that in Figure 1 a2 equals unity. This is the situation where
one actor negatively reacts to the initial selection of another, or
Figure u. If we can experimentally induce CiS relevant to the task,
Figure 5, P(S) can be computed. If high states of the Cis are necessary
to completion of the task and p possesses these high states, P(S)
will be high.

The Berger and Fisek study concerned two specific characteris-
tics. As implied in our introduction, these characteristics can be
arranged in one of three types of sets: consistent, e.g., p:HH, o:LL;
inconsistent symmetric, e.g., p:HL, o:LH; and inconsistent non—symmetric,
e.g., p:HH, o:LH.7

Since we are concerned with the nature of the influence process,
we need to explicate the nature of the process in the three types of
sets. Two of the sets have one common factor: the characteristics
of at least one of the members of the set are perceived by p as being
of the same type, high or low. This occurs in the consistent and

inconsistent non-symmetric set types and does not occur in the

 

7Our use of terms is slightly different from Berger and Fisek's
(1970). In general, there are n different types of arrays of the
C. s in the dyad where n is the number of 018. A complete test of the
processes postulated above would involve sixteen studies in the
Berger and Fisek case. Three Cis involve 6H studies and 4 Cis involve
256.

14

inconsistent symmetric case. Thus all possible arrays of specific
characteristics can be categorized into one of two generic types-—
univalent (e.g., our consistent and inconsistent non-symmetric cases)
or multivalent.

In the univalent case of two specific characteristics, one can
easily predict the direction of influence. In the consistent set,
p's evaluation of p' and o are univalent. As argued above, the nature
of p's final choice indirectly measures the degree of influence.
According to formalized balance theory, the three-way relationship
consisting of p, p', and alternative A or B will be balanced and
stable if and only if the product of the signs in the relationship
is positive. Since each specific characteristic has some state that
is preferred to be possessed by individuals over other states of the
characteristic, and since one alternative has been preferred over
another, we can predict the final preference of alternatives, our
indirect measure of influence. For example, if p initially negatively
evaluates alternative A while it is positively evaluated by 0, and if
p univalently evaluates himself since he possesses preferred states
of the specific characteristics, p's final choice will be alternative
B. This is a continued negative evaluation of alternative A and a
rejection of 0's influence. In the inconsistent non-symmetric case,
p may evaluate either himself or o in a univalent manner. If p
evaluates himself in a univalent manner, p will be influenced in a
manner similar to the consistent case. If p evaluates o in a univalent
manner, p will tend to balance the three—way relationship of himself,

0, and alternative A or B.

15

If we refer again to Figure 5, we see there are four three-way
relationships, each involving p with either 0 or p' and alternative
A or B. Since selection of alternatives is a mutually exclusive
event, we can predict the nature of influence by seeing p's final
choice. P's final choice can be predicted by a knowledge of the
univalent evaluation of p' or o by p and the initial alternative
chosen by the individual p univalently evaluates.

In the inconsistent symmetric case p does not evaluate himself
or o in a univalent manner, so we cannot predict the nature of in-
fluence via a balance argument. If p is forced to make a final de-
cision, given initial disagreement with o, and if p sees himself and
o as being non-univalent in their possession of specific character-
istics, p will act in one of two ways depending on how p sees possession
of specific characteristics in relation to his and 0's capacity to
complete the task before them. Specific characteristics determine
p's evaluation of his and 0's ability to complete the task before
them via one of two types of relationship in the multi-specific
characteristic case. The various characteristics may map onto the
dimension of ability to complete the task in a non-isomorphic manner
(similar to a map from n—space to one-space in linear algebra).

Since this results in one bit of information for p about himself

and o, and since the information is the only basis p has for evaluation
of his and 0's future performance of the task, the single bit of
information is evaluated in a univalent manner as a single bit of
information is evaluated in only one way.

Likewise, we can argue the map from information about possession

of specific characteristics to possession of an evaluated state as to

16

future task performance may be isomorphic (a map from n-space to n-
space). This does not result in a univalent evaluation of p' and 0
unless p averages the individual projections of the specific charac-
teristics into the future performance dimension. The average may be
a simple average of the characteristics if they are equally relevant
to the group's task, or a weighted average otherwise. A weighted
average may result if different utilities are assigned to the con-
tributions which possession of different specific characteristics
makes to successful completion of the group's task. However, this
is beyond the scope of this investigation, as we are assuming here
that each of the characteristics is equally valued and subsequently
given equal weight by p and 0. Since such an averaging effect results
in one bit of infbrmation, the result of the process is a univalent
evaluation, an averaging out of the information possessed by p about
himself and o.

The Berger and Fisek paper discussed two types of cognitive
mechanisms that p may undergo in the inconsistent symmetric case,
which are called combining and balancing mechanisms. In the first,
p combines his information about himself and 0 (our map from n-space
to one-space). In the case of two Cis, if both characteristics are
seen as being equally relevant to the group's task, the result may
be a characteristic that is a medium state for p and o. In the bal—
ancing mechanism, Berger and Fisek feel p will behave in a manner to
balance his evaluation of himself and o by evaluating himself and o
in a consistent manner. Unlike the single reflected evaluation in
the combining mechanism, the balancing mechanism may be due to a

reflection of multiple univalent characteristics.

17

In general, the knowledge of the existence of a univalent
evaluation by p of p' or o and p and 0's initial choice of alter-
natives provides us with enough information to predict p's final
choice of alternatives. Final choice is always based on some univa-
lent evaluation of p' or o by p. Berger and Fisek's failure to view
the influence process in such a manner limits the fruitfulness of
their analysis. Their postulated mechanisms are the same type of
univalent mechanism, since they are based on a univalent evaluation
of p' and o by p. In the case of the combining mechanism, the com-
bination of states results in one bit of information about the states
possessed by p' and o; and this must be univalent in the trivial
sense. In the case of the balancing mechanism, p is faced with dis-
crepant information about the characteristics possessed by himself
and o. This is resolved by p seeing himself and o as possessors of
only one type of characteristic, resulting in univalence.

We need to understand the dynamics of the influence process that
result in the way the univalent evaluation is brought about, that is,
we need to specify the type of mapping of the specific characteristic
dimension to the future performance evaluation dimension.

All influence processes involve a univalent evaluation by p of
p' and o. The direction of such influence is of little theoretical
interest to us in the consistent and inconsistent non—symmetric cases,
since it can readily be predicted. Since the means by which influence
occurs cannot be predicted in the inconsistent symmetric case, the
direction of influence similarly cannot be predicted in such a case.

The Berger and Fisek experiment is the first reported work of

influence processes in the case of multiple status characteristics.

18

We decided to conduct a replication of their experiment in order to
examine two questions:

(1) Do similar results occur if the experiment is replicated
in a modified manner; that is, can we reproduce Berger and Fisek's
results?8

(2) Does further data analysis shed light on the theoretical
questions raised in our discussion of the nature of influence in the
inconsistent symmetric case? More specifically, can we specify the

means by which the influence mechanism occurs?

 

8There were some changes made in the experimental procedure
which are described in the following section.

THE EXPERIMENT

The last section of this thesis has argued that a situation
like that depicted in Figure 5 would provide us with the information
needed to measure the influence process between p and o. This would
involve three types of manipulations:

(1) We would need to force p to differentially evaluate p'
and 0 only on the basis of Cis relevant to the group's task. That
is, we need to induce states of Ci to p and o.

(2) It must appear that p and 0 select different alternatives
for their initial choice in the two-step decision process. That is,
we need to force subjects to follow the path of a2 in Figure 1.

(3) P and 0 must be oriented towards each other and must be
individually oriented towards successful completion of the group's
task.

The study was conducted in two parts or phases.9 Phase I con—
sisted of the creation of two Cis. In Phase II, the Cis were made
relevant to the task, the subjects were fOrced to be oriented towards
each other and towards successful completion of the task, and subjects
were forced to make evaluations of their own and their partner's

performance outputs.

 

9The actual instructions given to subjects in the two phases are
found in Appendix A.

19

20

Phase I

Subjects were run in pairs to set up the dyad. After a brief
introduction in a waiting room, the subjects were led to a small-
groups laboratory. Each of the subjects was randomly assigned to
sit at one of the two tables in the room. The tables and subjects
were separated from each other by a screen.

Two tests were administered to subjects to establish degrees of
fictitious abilities. The tests consisted of subjects making a binary
choice about a series of stimuli. The stimuli were series of slides
presented to subjects on a screen in front of them. All of our sub-
jects saw these slides in the same order.

Subjects were told that these tests would measure their levels
on each of the two abilities. They were also told that these abilities
were unrelated to other skills or abilities an individual may possess
and were unique attributes of individuals. The first test measured
"Meaning Insight Ability," the ability of an individual to "intuitively
understand or grasp the overall meaning or significance" of unfamiliar
objects or events. The second test measured "Relational Insight
Ability," a "unique attribute of individuals" which, subjects were
told, refers to the ability of an individual to "understand the re-
lationship between unfamiliar objects or events."

In the first test, subjects were given five seconds to choose
which of two non-English words or phrases, which were phonetically
spelled from a "primitive language," meant the same as an English

word or phrase. All the words were presented on slides. Subjects

21

recorded their responses on panels in front of them.10 The ”Meaning

Insight Ability" test consisted of twenty decisions or trials. The
test for "Relational Insight Ability" was conducted in much the same
manner. Subjects were asked to make a binary choice after five seconds
on each of twenty slides. The stimuli presented in the "Relational
Insight Ability" test were different from that presented by Berger
and Fisek. However, we presented the same stimuli in the "Meaning
Insight Ability" test.

The Berger and Fisek stimuli were unavailable. Their stimuli
consisted of a series of slides asking subjects to match the phonetic
spelling of a Japanese word with one of two "ancient Japanese ideo—
graphs" which had the same pronunciation as the phonetic spelling.
Our task involved asking subjects whether or not they thought a
standard figure drawn on a card that was held in front of them could
be contained in the slide shown on a screen. Subjects were told that
each of the slides was quite complex and might contain more than one
standard. The same set of five slides, in the same order, were shown
four times, with a different standard each time, making a total of
twenty trials.

Having completed the two tests in Phase I, subjects were told
the scores were being standardized by an experimenter assistant, and

the host experimenter proceeded to describe Phase II.

 

10In both tests in Phase I, there was no communication between
subjects. In Phase 11, subjects only communicated their initial
choice. This will be described below.

22

Phase II

Subjects were told they would be working in a "critical choice
situation," a situation where the most important thing was for each
of the subjects to get the "correct final answer" to a problem. This
set the needed orientation towards successful completion of the task.
Due to the difficulties of such a situation, subjects were told they
would be working together and would be allowed to exchange information
with each other as to their initial decision on a two—step problem.
This set up the collective orientation of subjects to each other.

Subjects were told the "critical choice" situation would be
simulated by the "Contrast Sensitivity Task." This consisted of a
series of slides which contained two rectangles. Each of the rec-
tangles was composed of various arrangements of white and black areas.
Subjects were instructed to make a binary choice as to which of the
two rectangles had the greater degree of white area. In actuality,
the test was constructed to be ambiguous. Following the Berger and
Fisek procedure, the order of presentation of slides was randomized
to control for lack of homogeneity between slides. However, like
Berger and Fisek, the actual sequence of the slides was maintained.

P and 0 were told to communicate their choices to each other
only by means of the panel in front of them. Subjects were told
they would see a slide for five seconds and then be asked to make
an initial choice as to which of the two rectangles contained the
larger white area. Subjects were told they would then see their
partner's choice. Five seconds after this, they were told, they

would make their final choice. All communicated information was

23

done by use of the panels. To clarify this procedure, subjects were
given two demonstration slides.

In the actual experiment, the exchange of initial choice informa—
tion was controlled. That is, we were able to control the as of
Figure l, forcing agreement or disagreement between subjects, a
necessity of our experiment. Since we were interested in the relation
of influence patterns to the Cis, we needed to force a situation that
followed the path of a2 in Figure 1, that is, a situation that forced
subjects to use information about their and their partner's state of
the Cis relevant to the group task. Therefore, the majority of our
manipulated feedback took the form of controlled disagreement. These
were called critical trials. In all, twenty~five trials were run;
that is, the subjects saw twenty—five "Contrast Sensitivity" slides.
Twenty of these twenty—five trials were critical trials; the five non—
critical trials, called agreement or neutral trials, took place on the
second, seventh, thirteenth, nineteenth, and twenty-fifth trials.ll
It was necessary to introduce these controlled agreement trials to
allay the possible suspicion of subjects of continual disagreement,
given equal amounts of information about each other in the p:HL, o:LH
and szH, o:HL cases.

Before the actual presentation of the "Contrast Sensitivity"
task, the crucial manipulation of the experiment took place: subjects
were given information about their and their partner's levels of the

two abilities tested in Phase 1. Subjects were told they would be

 

11This is slightly different from the Berger and Fisek procedure

of randomly placing an agreement trial in each block of five trials.
Such a procedure may have as little as none or as many as eight
critical trials before an agreement trial.

24

given their and their partner's scores on the tests. The scores,
they were told, were standardized on a scale of zero to twenty such
that zero to ten correct is a low individual score and a low state
of the tested ability; eleven to fifteen is an average score, an
average state of the ability most subjects would score; and sixteen
to twenty a high individual score which ”clearly reflects" a high
degree of the ability.

When the scores were presented, subjects were told to pay atten—
tion to each other's score, since the scores would be the information
they had about each other and they should know about each other for
working together in the ”Contrast Sensitivity" task. In order to
establish the direct relevance of the "Meaning Insight Ability" and
"Relational Insight Ability" as specific characteristics to the
"Contrast Sensitivity" task, subjects were told that both of these
abilities somehow contribute to solving "Contrast Sensitivity" problems.

Our relevance is less explicit than Berger and Fisek's. In
their study, subjects were told high levels of "Meaning Insight
Ability" and "Relational Insight Ability" were highly correlated
with each other and with successful performance on the "Contrast

. . . . 12
Sens1t1v1ty" task.

 

l2We originally thought this would force the Berger and Fisek
results to indicate a combining mechanism, since the subjects are
told the abilities are associated with each other. This would tend
to make subjects view these abilities as part of a common set. However,
the processual section of this paper indicates the crucial question is
the nature of the univalent evaluation, not the difference between the
similar combining and balancing mechanisms. The Berger and Fisek
Phase II instructions can be found in Appendix D.

25

After the manipulation took place, the actual experiment was
conducted. After seeing the twenty-five slides, subjects were inter-
viewed and debriefed.13

With the exceptions noted above, the Berger and Fisek experiment
and our experiment were the same. However, the Berger and Fisek experi—
ment was concerned with three conditions: (a) p:HH, o:LL; (b) szL,
ozHH; and (c) p:HL, o:LH. The first and second conditions were simul-
taneously run, randomly assigning subjects to each of the conditions
on the basis of assignment of fictitious scores. High manipulations

were represented by a nineteen out of twenty on C "Meaning Insight,"

l,

and an eighteen out of twenty on C "Relational Insight." A nine on

29

C and an eight on C represented low manipulations. In the third con-

1 2

dition, scores were randomly assigned. Such scores were reported as
being unusual.

In our replication, we were interested only in the third condition,
the inconsistent symmetric condition. For practical purposes two sub-
jects were run at a time, so we actually have two sub-conditions:
p:HL, o:LH and szH, ozHL. In both our sub-conditions a score of
eighteen out of twenty is a high manipulation and nine a low manipula-
tion. Thus, unlike the Berger and Fisek study, our subjects received
symmetric scores on the zero to twenty scale. Subjects were told
that being high on one characteristic and low on another was an un-
usual condition.

Before we present our results, we should remember that our experi-
mental replication differs from that of Berger and Fisek in the follow-

ing manner:

 

3COpies of the interview form and debriefing appear in Appendix B.

26

(1) We removed what we thought would be a "bias" towards a
combining mechanism in the Phase II manipulation.

(2) Our agreement trials were not randomly interspersed with
our critical trials in Phase II.

(3) The Phase I scores used in our experimental manipulation
were mirror images of each other.

(4) Our "Relational Insight Ability" task was different.

(5) Our subjects were females at a midwest university, whereas
Berger and Fisek's subjects were males at junior colleges in the San

Francisco Bay area.

 

14It has recently come to our attention that our experiment
differs from Berger and Fisek's in one other major way. Berger and
Fisek administered the Phase I tests via paper and pencil technique,
where we used slides and subjects recorded their responses on panels.

RESULTS

Subjects

Subjects were all female undergraduates recruited from an intro- isti

ductory social problems course. We used only one sex to minimize the

I: ‘ "‘“"’_.'

basis of differentiation between subjects. Thirty-four subjects took

part in the study. One subject was eliminated due to failure to use

the information given about her partner's states of the C15. Three

pairs of subjects were eliminated since one member of each pair was

black; it was felt this would provide an extraneous basis of differ-

entiation for subject's formation of performance output expectations.15
In all, twenty-seven subjects remained. Fourteen were in the

p:HL, o:LH condition; and thirteen in the szH, ozHL condition.

Comparison of the Original Experiment and the Replication

 

Before we investigate the nature of the influence process in the
inconsistent symmetric case, we need to see if our procedure results
in findings different from Berger and Fisek's. If our findings are
similar, we can combine the sets of data.

Our chief concern is a measure of the influence process. This
can be measured as the probability of changing one's decision on a

binary task from the initial to the final part of a two-step decision

 

15A list of the criteria used for inclusion or exclusion from
the sample is found in Appendix C.
I

27

28

process. This is the P(S) we discussed in our section about processual
considerations. Table 1 presents our data in our two conditions, and
the Berger and Fisek data for the inconsistent symmetric condition.
More explicitly, the data show the mean and variance of the number

of self-responses over the twenty critical trials and the P(S)s.

TABLE 1

‘ P(S), MEAN, AND VARIANCE FOR INCONSISTENT SYMMETRIC STUDIES*

 

 

 

Q‘Trfi—m
Number of
Number of Critical
Study Subjects Trials Condition P(S) Mean Variance
erger 8 (u,u5)**
Fisek 26 20 p:HL, o:LH .661 13.23 5.62
Tress in 20 p:HL, o:LH .658 13.14 ”.51
Tress 13 20 szH, ozHL .650 13.00 n.55

 

 

 

 

 

 

 

 

 

*The mean is equal to np and the variance equal to npq, where n
is the number of critical trials, p is P(S), and q is l - P(S) or P(O).

**Use of npq gives a figure of u.u6.

Our two conditions should not be significantly different if each
Ci is equally relevant to the group's task. If our conditions are not
different, they can be combined and our combined result compared to
the Berger and Fisek finding for the p:HL, o:LH condition. If the
combined Tress data and the Berger and Fisek data are similar, both
sets of data can be pooled for further analysis.

The most direct way to see if data can be combined is to compute
a critical ratio, that is, a standard Z—score to compare two probabili-

ties. An alternative method is to compare the variances in the number

29

of self-responses, since these variances are direct transformations
of the P(S) parameters for our various conditions.

Table 2 shows that our two conditions can be combined and that
this combination can be fUrther pooled with the Berger and Fisek

condition.

TABLE 2

TESTS FOR COMBINING OF DATA

 

 

 

 

Condition Score/Ratio Significance Level*l
Z-Test
Tress condition .1759 n.s.**

Tress pool versus

 

 

Berger 8 Fisek .2u30 n.s.

F Ratio
Vress condition 1.01 n.s.
Tress pool versus 1.2” n.s.
Berger 8 Fisek l.01*** n.s.

 

 

 

 

 

*Two-tailed tests.
**Not Significant .

***Results if we use the Berger 8 Fisek variance as n.46.
F = 1.2% if their reported variance is used.

The Nature of Univalence

 

If we pool the Tress with the Berger and Fisek data we arrive at
a P(S) = .658. If we can somehow compare this probability with other
probabilities found in studies using specific characteristics, we can

arrive at some indication of the nature of the process of univalence:

30

is it a function of one or two P(S)s due to one or two CiS? Table 3

lists previous studies that have involved specific characteristics.

TABLE 3

PREVIOUS STUDIES OF SPECIFIC CHARACTERISTICS

 

 

 

 

 

Number of
Number of Critical Number of
Study Cis Trials Subjects Condition P(S) for p
Berger 8
Conner 1 22 28 p:H, o:L .79
32 p:H, o:H .66
31 p:L, o:L .67
28 p:L, o:H .us
Berger 8
Fisek 2 20 26 p:HH, o:LL .821
26 p:HL, o:LH .661
2” p:LL, o:HH .533
Simulation
4 Parameters
Berger,
Conner, 8
McKeown* 1 2O 5 p:H, o:L .8u6
p:L, o:H .320
u p:H, o:L .8u6
p:L, o:H .320
3 p:H, o:L .782

 

 

 

 

 

 

 

 

*The Berger, Conner, 8 McKeown study differs from the Berger 8
Conner and the Berger 8 Fisek studies in two respects:

(1) The parameter estimations in the simulation are not based
on a fixed perfbrmance expectation state. That is, the parameters
are based on a changing P(S).

(2) The study did not concern dyads. P and 0 form a three-
member group such that 0 consists of two people.

Note that our pooled P(S) of .658 is very near the results found
by Berger and Conner in the p:H, o:H and p:L, o:L conditions. This

would seem to imply that inconsistent symmetric characteristics act

31

like a single characteristic. However, as a consequence of the need
for a univalent evaluation of the Cis of either p' or o by p, the
influence process should be based on one or two Cis.

We can conduct such an analysis by use of a chi-square goodness
of fit. We need only compare our observed frequency of subjects making
zero to twenty stay responses with the expected number of responses
based on one or two P(S) values. Our observed frequency can be viewed
as an empirical probability distribution and our distribution based on
the P(S)s will be a Bernoulli distribution in the case of one P(S) and
a joint distribution in the case of two P(S)s. Comparison of expected
and Observed distributions will give us some insight into the nature
of the univalence process, the Bs in Figure 1. Figure 6 gives the

number of stay responses for the pooled data over twenty trials.

Number of Subjects

ll ' (Pooled P(S) of Berger 8
' Fisek and Tress of .658)
g .
7 .
t
5 r-
3 T

 

 

 

|,_a
o v
01%

7 9 ll) 13 15‘ l7 19 Number of Stay Responses

20

Fig. 6.—-Observed frequency of stay responses.

Since the process seems to be unimodal, we might expect the
theoretical process to be the result of a joint distribution of two

probability values which overlap or a simple Bernoulli distribution.

32

The former case could still result in the unimodality of Figure 6,
as the majority of the masses of the distribution functions are
located near each other.

Thus the argument that the unimodality of Figure 6 might be due
to the combined effects of two specific characteristics cannot be
denied. To complete such an analysis, a family of joint distribution
curves needs to be generated. Even if we computed such distributions
at increments of .10, 100 computations would be required. This is
beyond the scope of this thesis. Obviously, a joint distribution takes
on the properties of a simple binomial distribution in the trivial case
of the two parameters having near identical values.

Let's suppose the limits of the univalent values occur when p
and o are polarized on both of the Cis. For example, a survey of
Table 3 reveals extremes of P(S) occur in these cases. Using the
Berger and Fisek results, we can compare our distribution given in
Figure 6 with that expected given the joint operation of P(S) = .821

and P(S) = .533. Figure 7 compares these distributions.

Number of subjects

11 P
f observed

9 _ expected

 

    

0 5 7 9 ll 13 15 l7 19 Number of Stay Responses

20

Fig. 7.--Observed and expected frequency of stay responses,
given two parameters.

33

If we consider the joint case an expected distribution, we can
compare this to our observed data via a chi—square goodness of fit.
If the chi-square value is significant, we would assume our observa—
tions do not conform to the specific extreme limits given by Berger
and Fisek for polarized cases. After lumping categories to get a
frequency of at least five in each category, we are left with four
cells. Since our chi-square is based on two a priori parameters,
we do not subtract a degree of freedom for each parameter. Our com-
puted x2 with 3 degrees of freedom equals 17.019. This is highly
significant, implying that if the univalent process involves two P(S)s,
the P(S)s fall within the limits of .533 and .821. As shown in Figure
7, if the 2 parameters moved nearer to .66 our x2 value would be lower
and imply a better fit. The nearer the means of the postulated para-
meters are to the mean of our observation, the better the fit should
be. A trivial case occurs if the three means or parameter values over-
lap. This would involve a one-parameter test, the simple binomial
distribution. However, we have no way of knowing if our observations
are due to the effects of two parameters that are very close to each
other in value.

If we remember our pooled estimate of P(S) as .658, we can use
a similar reasoning to compare this distribution with other distribu—
tions due to l or 2 specific characteristics. Our value of .658 could
reflect a binomial with P(S) = .66 or the joint effects of the Berger
and Conner p:H, o:H and p:L, o:L distributions. Such analysis gives
a X2 value of 5.3uu.‘ In the test of one theoretical parameter, P(S)
estimated as .66, weitake away a degree of freedom for parameter

estimation. In the case of the two a priori P(S)s due to the Berger

31+

and Conner work, we do not take away any degrees of freedom. Com-
bining our categories to get five or more in each category, we have
five degrees of freedom in the first test and six degrees of freedom
in the second. Both these X28 are 22£_significant.

Unfortunately, our limited analysis does not allow us to specify
the number of Cis used by p in forming his evaluation of self and

other.

Processual Observations16

 

The above analysis does not investigate the nature of the influ—
ence process. Until now, such analysis has been used only in the
case of one specific characteristic. Our discussion of the univalent
process involving inconsistent and symmetric specific status character-
istics and influence processes interpreted the univalent process as
involving P(S)s. This implies two major questions:

(1) Is a response made on trial n+1 independent of a response
made on trial n?

(2) As the number of trials increases, does the pattern of
responses in the influence process reach a fixed value for P(S)?

In short, we are asking if the influence process in the case of two
Cis is an independent trials process.

The simplest way to test the independent trial question is to

make a contingency table analysis. Such a table would take the fol—

lowing form:

 

16Data analysis in this section involves knowledge of the actual

sequence of responses. Since we were unable to obtain this information
for the Berger and Fisek data, our analysis is restricted to the data
collected in the Tress experiment.

35

Number of Responses on Trial

n+1 of Type
Self Other
Number of
Responses Self A B
on Trial n
of Type Other C D

N

Notice A + B + C + D = 19N where N is the number of subjects, since
twenty critical trials implies nineteen possible transitions from one
trial to the next. By examining responses for subjects from one trial
to the next, we can fill in the cells of the table.

We can also conceive of the table as a probability matrix. Since
a self or other response must be followed by some kind of response,
the rows of such a matrix must sum to one. The matrix would take the
following form:

Type of Response on Trial n+1

Self Other
Type of Self 11 l—Al
Response
on Trial n Other 12 1-12

A significant X2 value on our contingency tables would imply that an
independent trial process does not take place. Figure 8 shows us the
contingency tables, probability transition matrices, and X2 values
when we subject our data to such analysis. Surprisingly, our p:HL,
o:LH condition is ngt_an independent trial process. We have no ex-
planation fOr this occurrence, since both conditions have similar
P(S)s.

By interpreting our pooled transition matrix as a simple two-

state Markov chain, the states being a self- or an other-response,

36

we can predict P(S) for any step of the influence process. Let “k
designate the probability distribution vector on the k+l step of the

process. More specifically, n is made up of two elements, the pro-

k

portion of subjects who make other responses on the k+l trial and

the proportion of subjects who make stay responses on the k+l trial.

 

 

 

 

 

 

flk is designated (P(S), 1-P(S)).
Condition ContingencygTable Transition Matrix
Trial n+1 Trial n Trial n+1
Self Other Self Other
p:HL, o:LH 105 70 175 .600 .400
65 26 91 .714 .286
266 x2 = 2.9127
N = 1n .10>p>.05
p:LH, o:HL 103 58 161 .640 .360
55 31 86 .640 .360
2u7 x2 = 0.0183
N = 13 n s.*
pooled 208 128 336 .619 .381
120 57 . 177 .678 .322
513 x2 = 1.u99
N = 27 n.s.

*Not Significant.

Fig. 8.--Processual matrices.

On the first trial, 92.6 per cent of the subjects made stay
responses. That is, “0 = (.926, .074). By post-multiplying powers
of our pooled transition matrix, the aggregate stochastic matrix, by
no, we can predict P(S) for any trial of the process. In general,

Wk gives the predicted distribution of P(S) and l-P(S) on the k+l

37

trial such that "k = Pkno, where P is the pooled transition matrix.

Table 4 presents our predicted and observed values fer P(S).

TABLE 4

DISTRIBUTION VECTOR AND OBSERVED VALUES FOR
CRITICAL TRIALS UNTIL THEORETICAL STABILITY

 

 

 

_k *1: ______j=
Distribution
Critical Trial Vector Observed P(S)
1 (.926, .074) .926
2 (.623, .377) .630
3 (.641, .359) .630
4 (.640, .360) .630
4+ (.640, .360) --

 

 

 

 

 

P, the pooled transition matrix, reaches a limit of (.640, .360).
That is, um, the equilibrium vector, is (.640, .360). However, Table
4 indicates that this value occurs on the fourth and subsequent trials.
Theoretically, this means the influence process stabilizes quite
rapidly, with a stable value of P(S) = .640.

In general, how does this projected stable value of P(S) = .640
compare to our observations? We can compare our sets of data by using
a chi-square goodness of fit test. 'After combining categories we end
up with a chi-square of 1 degree of freedom equal to 4.175. This is
significant at the .05 level. This implies our observed data do not
fit our expected distribution. This may indicate that the influence
process is ngt_a process with l P(S), more specifically a P(S) of
.640. However, this conclusion is not necessarily warranted, since

the sample size is small. The value of .640 is not significantly

38

different from the pooled Tress value of P(S) = .654. Comparison of
the two P(S)s via a critical ratio gives a Z-score of .4750.

If we plot the observed values of P(S) on each of the critical
trials, we can see if the process tends to stabilize as time goes
on. This curve is plotted in Figure 9. The dotted line in Figure 9

represents the values for the distribution vector presented in Table 4.

  
 

P(S)
1'0 P observed
' ____predicted
.8 r
6 L - P(S) = .640

 

A A )4 n j l a A e A A r A A A A l I A

0 1 3 5 7 9 11 13 15 17 19 Critical Trial
20

Fig. 9.—-Observed and predicted values of P(S) for
each critical trial.

Since our sample is small, Figure 9's plot of P(S) on a trial
by trial basis yields too great a variance. As Moore indicates, a
plot of a cumulative P(S) curve will obscure trends, especially if
they occur towards the end of the process (Moore, 1969, p. 149).
Following Moore, we reach the following compromise: P(S) is plotted
in blocks of seven trials. This is presented in Figures 10 and 11.

Specifically, each point in the curve represents the proportion of

39

stay responses made by subjects in a sequence of seven trials. For
example, the first point represents trials 1 to 7; the second, trials
2 to 8; the ith, trials i-6 to i; and the last point, trials 14 to

20.17

1.00<i;, observed

1.. ———- — —————— P@)=.am

JEDA A L 4 n j n A n L -_J L A

0 1- 3- 5- 7- 9- 11- 13-2-0
7 9 11 13 15 17 19

 

Critical Trial

Fig. lO.--Plot of trial blocks.

Figures 10 and 11 represent the reduced variance of the process.
This is seen if Figures 9 and 10 are compared. In addition, the
variance reduction implies a stabilization of P(S). However, the
asymptotic value reached by P(S) seems to be larger than our hypothe-

sized value of .640.

 

7Figure 10 is in the same scale as Figure 9. Figure 11 is a
more detailed plot of the same process, such that the X axis, the
trial blocks, is in the same scale as Figures 9 and 10, but the Y
axis, P(S), is five times larger in scale than the Y axis in Figures
9 and 10.

4O

P(S)

1.000 1 observed

.660

U

  
 

.640 4— ——————— P(S) = .640

.620 F

.600 t

?lln;jnlj_lllll

0 1- 3- 5- 7- 9- 11- 1320
7 9 11 13 15 17 19

 

Critical Trial

Fig. ll.--Detail of plot of trial blocks.

SUMMARY AND CONCLUSIONS

The general process that occurs in the event of two inconsistent
status characteristics seems to be universal. Such inconsistent evalu-
ations reflect themselves in the same type of influence process in
the group. Specifically, if conditions of less relevance of the char-
acteristics to the task of the group are made, the process will still
be of the same form.

This process is due to a member's evaluation of specific status
characteristics possessed by members of a group. At least one member
of the dyad evaluates himself or his partner in a consistent or uni-
valent manner. Thus we are concerned with the nature of a univalence
process. The nature of the process is a direct function of group
members' possession of states of status characteristics.

Unlike Berger and Fisek's initial investigation into this
process, we were not concerned with whether or not group members
combined or balanced their inconsistent characteristics. Balancing
results in univalence along the basis of one or two characteristics
in our example, and combining results in one univalent characteristic
or the creation of a new univalent level of the characteristics.

Such an interpretation ignores the idea that the combining mechanism
can be considered a special case of the balancing mechanism. Berger

and Fisek have ignored this and failed to investigate the nature of

41

42

the influence process in the case of inconsistent specific status
characteristics.

We conducted a partial replication of the Berger and Fisek
experiment. The general ferm of the experiment is to assign states
of fictitious abilities to subjects, the statuses, and then see the
influence pattern between subjects by controlling feedback between
subjects so that it appears they disagree. Resolution of the dis-
agreement in one direction or the other will measure the amount or
lack of influence of one subject towards another. By assigning fic-
titious scores on the abilities to subjects, we can set up cases of
inconsistent statuses.

Our data analysis showed our results to be strikingly similar
to those found by Berger and Fisek. Initially we thought the influ-
ence process would have the form of a simple Bernoulli process. That
is, the univalent process would consist of one probability parameter.
However, we argued that the joint effects of two parameters in a joint
distribution might confound the nature of the influence process. Our
limited data did not allow us to specify the number or actual values
of the parameters involved. In conclusion, we cannot state with
assurance the number of status characteristics involved in the uni-
valence process in the case of more than one specific status
characteristic.

The data in our replication showed the influence process is inde-
pendent from one trial to the next. In addition, the process theoreti-
cally becomes stable quite rapidly. However, the theoretical process
did not conform to our observed data. Analysis shows the asymptotic

value of the process is slightly greater than the theoretical value.

43

Our analysis is inconclusive about the exact nature of the
influence process. We can only say the process is an independent
trials process. However, this process is univalent and cannot be
conceived of as a balancing or combining of inconsistent states of
status characteristics.

Specification of the exact nature of the process is beyond the
scope of the thesis. Specification of a mathematical distribution
representing the process may allow us to extend our work to more than
two characteristics. At the same time, work should be done on multiple
diffuse characteristics. It is felt that work done along these lines
would be most fruitful, since such experimental work could be extended
to other areas of sociology, notably the issue of status inconsistency

in stratification.

BIBLIOGRAPHY

44

Berger,

Berger,

Berger,

Berger,

Berger,

Berger,

Conner,

BIBLIOGRAPHY

Joseph; Cohen, Bernard P.;Snell, J. Laurie; and Zelditch,
Morris. Types of Formalization. Boston: Houghton
Mifflin, 1962.

 

Joseph; Cohen, Bernard P.; and Zelditch, Morris. "Status
Characteristics and Expectation States." Sociological
Theories in Progress, Volume I. Edited by Joseph Berger,

 

 

Morris Zelditch, and Bo Anderson. Boston: Houghton
Mifflin, 1966.

Joseph; Cohen, Bernard P.; Conner, Thomas L.; and Zelditch,
Morris. "Status Characteristics and Expectation States:

A Process Model." SOCiological Theories in Progress,
Volume I. Edited by Joseph Berger, Morris Zelditch, and

 

Bo Anderson. Boston: Houghton Mifflin, 1966.

Joseph, and Conner, Thomas L. "Performance Expectations
and Behavior in Small Groups: A Revised Formulation."
Paper circulated to Department of Sociology, Michigan
State University, East Lansing, Michigan, 1970.

Joseph; Conner, Thomas L.; and McKeown, William L.
"Evaluations and the Formation and Maintenance of
Performance Expectations." Human Relations, 22 (1969),
481-502.

 

Joseph, and Fisek, M. Hamik. "Consistent and Inconsistent
Status Characteristics and the Determination of Power and
Prestige Orders." Sociometry, 33 (1970), 287-304.

 

Thomas L. "Continual Disagreement and the Assignment of
Self-Other Performance Expectations." Unpublished Ph.D.
dissertation, Stanford University, Stanford, California,
1965.

Moore, James C., Jr. "Status and Influence in Small Group

Interaction." Sociometry, 31 (1968), 47-63.

 

Moore, James C., Jr. "Social Status and Social Influence: Process

Considerations." Sociometry, 32 (1969), 145-168.

 

45

APPENDICES

46

APPENDIX A

INSTRUCTIONS GIVEN TO SUBJECTS
Phase I 5
Good morning. Thank you for joining us. I'm P

and this is . We are members of a team of

 

social scientists who are interested in studying the way in which
individuals and groups solve certain kinds of problems. We are
also interested in how these problems are solved in different

situations. So our work today will be divided into two phases

 

or parts, Phase I and Phase II. In each of these phases you will
be asked to solve problems, but under different conditions. I
will explain the nature of these problems as we go along.

Will you please sit here. You will be number one, and you
will be number two. Let's turn now to Phase I.

Within the last few years, social scientists have found that
individuals differ in their intuitive ability (pause) to perceive

and understand a set of objects or events. More simply, when some

 

people are presented with a group of unfamiliar events or objects

they are quickly able to intuitively understand or gpasp the overall

 

meaning or significance of these events or objects. Other people
do not seem to have this ability to the same extent. (from memory,

with emphasis) Social scientists call this ability to intuitively

47

 

li.ll.\l‘ .r ll" 1' la.

48

grasp or understand the meaning of gpparently unrelated things

 

"Meaning Insight Ability."

 

(begin from memory, switch gradually to reading) At this time,
we don't know all the answers as to why some people have this ability
more than others, although it may be related to background, training,
and possibly innate capacities. One thing we dp_know is that this
general ability is unrelated to most of the specialized skills a .fifiA
person may have. This means that a person with high mathematical A

skill, for example, does not necessarily possess high Meaning Insight

 

Ability. We have also found that whatever amount of Meaning Insight
Ability a person has, it is not affected by conscious effort. That
is, the amount of ability possessed by an individual improves ygpy_
little as a result of practice, and decreases hardly at all as a
result of fatigue. Consequently, a person's performance at Meaning
Insight Ability tasks remains very constant throughout his lifetime.
What we are going to do in this part of our work is to administer

a specially prepared test which is extremely accurate in measuring an

 

individual's Meaning Insight Ability. That is, this test distinguishes
those who have a great deal of this ability (pause) from those who
do not.

In order to solve the problems in the test, you must make use
of the Meaning Insight Ability that each of you possesses. The test
consists of a series of well-known English words, and a series of
words taken from a very primitive language studied by anthropologists.
(present slide, gesture with pointer) We will present on the screen
in front of you a slide containing one of the English words or phrases

and two non-English words marked A and B.

49

For the purposes of this test, the non-English words have been

written in simple phonetic form, using English script. One and only

 

gpg_of the two non-English words has the same meaning as the English
word presented.

Your task will be to determine which one of the non-English
words or phrases, A or B, ip_fgg£ has the same meaning as the
English word. After presenting a slide such as this, I will give
you five seconds to study it.

(slowly) We suggest that you first study the non-English words—-
A and B-- (pause) sound them to yourself, (pause) and try to
associate whatever meanings they call to your mind. Then, study the
English word or phrase (pause) and try to associate whatever meanings
it calls to your mind. Then decide which non-English word, A or
B, has the same meaning as the English word. (long pause)

In general, we have found that people with high_Meaning Insight

Ability consistently make correct decisions. Those with low Meaning

 

Insight Ability usually make incorrect decisions.

During the actual test, you will indicate which is the correct
answer by pressing the appropriate button on your panels. For this
phase you will push either the button labelled A or the button
labelled B on the bottom left side of the panels. (pause) (point)
Your decisions will be recorded by a member of our study staff in
the next room. At the end of this phase of our work, we will report
to you how well you have performed on the test; that is, on what your
level of Meaning Insight Ability is.

(start from memory) Before beginning the series of slides, let

me review the things you should keep in mind:

50

(1) You are about to take a specially designed test, which has
been shown to be an excellent indicator of Meaning Insight Ability.

(2) The test will consist of a series of twenty slides, each
slide containing one English word and two non-English words marked
A and B.

(3) Your task is to decide in each case which of the two non-
English words has the same meaning as the English word. You will

,1

indicate this choice by pressing the appropriate button on your panels.

..-‘;1 A _ 5.1
_l ,.
n

(4) Some of the slides may seem difficult to you, but there is
a correct answer to each and every slide. Persons with high Meaning
Insight Ability consistently make correct choices; persons with low
Meaning Insight Ability usually make incorrect choices. However, be
careful: guesses based on first impressions may often be incorrect.

(5) These choices which you make will enable us to measure the
level of your Meaning Insight Ability. At the end of this phase of
our work, we will report the correctness of your choices to you on
the scoreboard.

We are now ready to begin the series of slides. When the slide
appears on the screen, you will be given five seconds to study it, at
the end of which time I will call for you to make your decisions.

Please make your choice §§_soon §§_I call for them. Not before

 

(pause) and not long after. Is everything clear?
(show twenty slides)

This part of Phase I concerns your "Relational Insight Ability,"

 

This refers to the ability of an individual to understand the

relationship between unfamiliar objects or events. Social scientists

 

 

have recently found out some people have more of this ability than

51

others. Some people, when faced with objects or events that appear
on the surface to be unrelated, see a hidden relationship between
these objects more easily than others.

We don't know why this is so, but we do know that Relational

Insight Ability is unrelated to most specialized skills. This ability

 

is a unique attribute of individuals.

 

As you can see, there are many problems and questions posed by
this recently discovered property of individuals. What we are going

to do is to administer a test which is very accurate in indicating a

 

person's Relational Insight Ability. The results of this test show
if a person does or does not have a great deal of this ability.

This test also consists of a series of slides. However, in this
test some of the slides are repeated. Each slide will consist of a
geometric figure. In front of you, while you see the slide, will be
a standard. Your task will be to determine whether pp_pp£_the standard
is contained in the slide shown on the screen.

Please look at the standard we are now giving you. (Show sample
slide) As you can see, each slide is complex and contains numerous
patterns. After presenting a slide, you will have five seconds to
decide if the standard figure you are using is contained in the slide.
Is this clear?

We suggest you first try to move or rotate the standard around
in front of you or in your mind, and then do the same with the slide,
or imagine putting the standard on top of the slide and moving it
around to see if the slide and the standard overlap.

We have found that people with high_Relational Insight Ability

consistently make correct decisions about whether or not a slide

 

52

contains a standard and those with a 123 state of this ability
usually make incorrect decisions.

During the actual test we will give you a standard drawn on
a piece of paper at the beginning of each set of slides. There are
four sets of slides, with the same number of slides, five, and the
same types of slides in each set. The slides are repeated since each

slide is complex and may contain more than one standard pattern.

 

During the test you will push either the button labelled "yes"
if the standard is contained in the slide, or the button labelled "no"
if the standard isn't contained in the slide. The buttons are located
at the same place on your panel as the A or B buttons you used
before. Your choice will be recorded by a member of our study staff
in another room. After this test, we will tell you how you performed;
that is, what your level of Relational Insight Ability is. At the
same time you will receive your scores on the test you took before
so you will also know your level of Meaning Insight Ability.

Before we begin to test your Relational Insight Ability, let's
review some key things:

(1) This test is specially designed to measure a unique attribute
of individuals, Relational Insight Ability.

(2) The test consists of groups of five slides repeated four
times, a total of twenty slides. There is a different standard for
each of the repeated sets of slides.

(3) Your task is to determine if the standard is contained in
the slide. Some of the times it will be difficult to decide this,

but there is an answer in each case.

53

(4) Guesses based on first impressions may be false. Remember,
you have five seconds to study each slide, after which I will ask you
for your choice. Please make your choice as soon as I call for it,
not before and not long after.

(5) Your choices will be a measure of your level of Relational
Insight Ability. To repeat, persons with a high_state of this ability

consistently_are correct in their choices, and persons with a low state

 

are usually_wrong.

 

Are there any questions before we begin our test? If not, let's
start.
(show twenty slides)

We are now ready to begin Phase II of our work.

Phase II

This is a study of decision—making. More specifically, this
is a study in how individuals work together as a group to solve problems,
and how different types of organization affect the efficiency of
problem-solving. Today the two of you will be working together as
a team to make decisions about problems that we will give you. This
is a common yet important kind of situation, sometimes referred to as
a Critical Choice situation. The most important thing about a Critical

Choice situation is to make the right decision. For example, when a

 

doctor has to make a difficult diagnosis, the patient may die if he
does not make the right decision. But it is often difficult for the
doctor to recognize at first what is the correct decision. When con-
fronted with such a situation, the doctor will study the problem care-

fully, and, because the most important thing in a critical choice

 

54

situation is to get the right answer, he will usually seek advice

from others. Only then will he decide. Hg_will not care whether

 

h§_originally thought pf_the answer himself, pp_was assisted by_

 

 

 

someone else ip_recognizing the right answer. The important thing

 

 

is that he made the right decision.

Today we are interested in studying how people make decisions
in Critical Choice situations when they have advice from others.
Let me tell you something about the task you will be working on.

This perceptual task, which social scientists have labelled
Contrast Sensitivity, has been developed in recent years, and because
of its importance is currently being used in an extensive set of
studies in this part of the country and elsewhere. This Contrast
Sensitivity task involves judging color contrasts in figures or
objects. Today you will be judging a series of slides like the one
now on the screen.

(put demo slide on the screen) Each slide has two patterns,
which are made of smaller black and white rectangles. One of these
two patterns-~either the Epp_one or the bottom one--has more yhit§_
3232 than the other. Your task is to determine which of the two
patterns—-the top one or the bottom one--has the greater area of

(retract and advance slide) You may find that some of the
slides will seem difficult to judge as the difference in the area
covered by black and white rectangles is sometimes very small.
However, previous studies have shown that gpmg_individuals are

able to make correct judgments on the basis of very slight, almost

55

intuitive clues and feelings. That is, this is a fairly difficult

task on which individuals dp_differ markedly in their performances.

 

There are two things which make this task very important to
social scientists. First, we know from previous studies that the
capacity to make correct Contrast Sensitivity judgments is pp£_
necessarily related to specialized skills the individual might
possess, such as mathematical or artistic skills. That is, those
people with high mathematical or artistic skills are not necessarily
more accurate in their Contrast Sensitivity judgments. Those people
with low mathematical or artistic skills may in fact be quite accurate
at making Contrast Sensitivity judgments.

The second important finding is that the way individuals solve
Contrast Sensitivity problems strongly depends upon the kind of
situation they are in.

We know from previous studies that most individuals do con-
siderably better on these problems when they are working in a group
situation than they do when they are working alone. Since we are
interested in your making as many correct judgments as possible, you
will be working in a gpppp_situation today.

That is, because this is a study of how people make decisions
in a Critical Choice situation, the two of you will be allowed to
exchange information with each other as to what you think is the
correct answer befOre you make final decisions. Only your final
decisions will be recorded for your scores.

This is how it will work. First, I will present a slide on
the screen. After you have studied the slide for five seconds, I

will ask each of you to make an initial choice as to which pattern

56

contains the greater area of white, top or bottom. Five seconds
after you make your initial choices, I will ask you to make your
final choices.

will demonstrate how the equipment works.

 

(Boardman, equipment demonstration)

[Boardman: We'd like to go through one more practice trial
exactly the way it will be during the study.]

I will present a second demonstration slide so that you can
practice with this procedure. This will ppt_count on your score;
this is just for practice.

You will have five seconds to study the slide before I call for
your initial choice.

(show slide for five seconds)

Now make your initial choice.
(pause five seconds after both have chosen)
Now make your final choice.

(turn off projector)

(advance slide)

(press relay release)

(check off trial number on sheet)

In today's study there will be twenty trials exactly like this
one. For each of these trials we are interested solely in your making
the correct final decision. Therefore, you should not hesitate for
any reason to change your initial choice in order to make the correct
final decision.

During the slide series you may recognize certain similarities

between a rectangle that you are studying and one you have already

57

seen. This is due to the fact that some of the shapes which compose
the rectangles are repeated. It is the case, however, that within
each rectangle, these shapes are combined in a new fashion so that
no two rectangles are identical. Therefore, you should treat each

pair of rectangles as a new problem which requires its own study and

 

its own decision. Is everything clear?

 

Since you will be working together and helping each other, it is
important that you know about each other. This is the reason for the
two tests you took earlier. These tests measure two special distinct
abilities--Relational Insight Ability and Meaning Insight Ability.

As we mentioned, the Meaning Insight Ability is the ability to under-
stand the meaning of unfamiliar events or objects. The Relational

Insight Ability is the ability to understand the relationship between

 

unfamiliar objects or events. These abilities are quite different,
but both contribute to solving Contrast Sensitivity problems.

(pause) Mr. will now give you your scores

 

on the tests you took previously.

(Boardman Part II) (Boardman leaves)

Before we begin the series, I would like to summarize a few
points.

(1) You will be shown a series of twenty slides and be asked
to decide which pattern has the greater area of white.

(2) Only your final decision on each slide will count towards
your score.

(3) Each time you make the correct final decision, that adds
one point to your score; if you make the incorrect final decision,

that adds nothing to your score.

58

(4) We are interested in your making as many correct final
decisions as you can; therefore, you should not hesitate for any
reason to change your initial choice, if that helps you to make the
correct final decision.

(5) At the end of the study I will report to you how many correct
decisions you each made.

Finally, please hold your choices until I call for them, but
make them as soon as possible thereafter.

(show slide series)

This completes the series of slides.

We would now like to discuss your scores with you, and to talk
with each of you individually in order to get a further elaboration
of your feelings and opinions about the study. I will talk with you,

Number 1, and Mr. ‘ " will talk with you, Number 2.

 

Boardman: Part II

 

The Relational Insight Ability test and the Meaning Insight
Ability test have been given to many college students of your level
in this part of the country and elsewhere. The standardized scores
presented here are based on these previous studies. It was found that:

(Boardman points at scores)

11 through 15 correct of a possible 20 is the average individual
score. That is, most people will get 11 to 15 answers
correct.

16 through 20 correct is a high individual score and clearly
reflects a high degree of ability.

0 through 10 correct is a low individual score and reflects

a low degree of ability.

59

Previous studies have shown both abilities contribute to solving
Contrast Sensitivity problems.

(pause) I have scored your tests.

Number 1, you made nine correct answers on the Meaning Insight
Ability test. You made eighteen correct answers on the Relational
Insight Ability test.

Number 2, you made eighteen correct answers on the Meaning
Insight Ability test. You made nine correct answers on the Relational
Insight Ability test.

You can see how you've done by comparing your scores with the
national standards in the center column. (pause)

As you can see, we've had some unusual scores today.

Are the scores and standards clear?

APPENDIX B

INTERVIEW SCHEDULE
Before we discuss the results of the study, I'd like to get
your reactions to it. There are a number of things that affect the
results and I want to talk with you about some of them.
Your first name is? (pause) And what is your major field of

study, ? And your age? (general background,

 

probe in all cases)
(1) In general, what are your feelings about the study?
(2) (a) Have you ever participated in a study like this
one before?
(b) Have you ever read or heard about a study like
this one?

(if yes to 2a or 2b, probe for a description of the study)

Phase I

(1) When the Meaning Insight Ability test was first described
to you, how well did you expect to do on it? Why?

(2) At the same time, how well did you think your partner
would do? Why is that?

(3) How did you go about trying to get the correct answer
on the test?

(4) In general, how confident were you of your choices on

this test? Why (not)?

60

(5)

(continue

Now,

(1)

(2)

61

Did you think the first test was a good measure of
your Meaning Insight Ability? Why (not)?

Phase I for Relational Insight Ability, the second test)

Phase II
let's talk about the second part of the study, Phase II.
How well did you expect to do on the Contrast Sensitivity
task before you actually began taking it; that is, how
well did you expect to do after you received your scores
on the two tests we just talked about? Why?
At the same time, how well did you expect your partner

to do? Why?

(in l and 2 probe for combining versus balancing mechanism)

(3)

(4)

Let's look at your initial choices on the test.

(a) Do you happen to remember how many times you agreed
and disagreed with your partner on your initial choice?

(b) What did you think and feel when your partner disagreed
with you? Why?

(c) Do you have any ideas why you were disagreeing with
your partner?

Let's look at your fip§1_choices.

(a) Did you begin to feel that either you or your partner
was usually right or wrong? Who? Why?

(b) In those cases where you thought you were right on
your initial decision, what did you do? Did you
ever change your mind in these cases? Why?

(c) What about the times you thought you were wrong on

your initial choice?

62

(d) How confident were you of your final decisions

on the test? Why?

Debriefing

 

Now, , I would briefly like to explain in

 

a little more detail what we were trying to study in today's test.

We are studying the relationships between a person's abilities and

the way he behaves in a group, especially the decisions he makes.

That is, we are studying the effect upon a person's changing his
decision if another person with more, less, or equal abilities disagrees
with him on a decision. So as you can see, we were not testing Contrast
Sensitivity as such. Have I made sense so far? (give example of
people with different abilities interacting--e.g., a doctor and an
intern.)

To set up this type of situation there were two things we needed
to arrange: your abilities and your agreement with your partner.
Concerning your abilities, in Phase I of the study all of the slides
you saw were ambiguous and had no correct answer. For example, in
the test for Meaning Insight Ability each foreign word was a fiction.
The odds of picking each word was 50-50. By telling you your scores
on the tests, which were all fictions, we hoped you would naturally
assume that you had more or less of these abilities than your partner.
But really we know nothing about your abilities. I'm not even sure
there is such a thing as Relational Insight or Meaning Insight Ability.
Do you understand so far? (pause)

In Phase II, the Contrast Sensitivity test, we were interested

in a situation where you disagreed with your partner on your initial

63

decisions. The information you received about your partner's initial
decision was arranged so that you would disagree with him most of
the time. Each slide has equal amounts of black and white area in
the rectangles on the slide. Is this clear? For these reasons,
there are no scores on the tests. As in the first part of this
study, Contrast Sensitivity Ability is a fiction and doesn't exist.

This briefly is the reason we had to arrange the things we were
testing. Since the study involves some fictions, we would greatly h
appreciate it if you didn't tell anybody about it. They might be
tested later and such information might change the way they perform.
Can I have your word that you won't disclose this information?
(pause) If anyone asks you about the test, it's all right to tell
them it was a test concerning whether or not there were more black
or white squares on some pictures; but don't tell them about the
rest. Okay? Thank you very much.

We will pay you at the rate of two dollars an hour for your
participation today.

If you are interested in our results, please leave your name
and address on this sheet of paper so we can send you a brief report

of our findings. (give cash, cash receipt, and sheet)

APPENDIX C

REASONS FOR ELIMINATING SUBJECTS FROM SAMPLE
(l) Deliberately making wrong initial choices.
(2) Misunderstanding instructions.
(3) Prior acquaintance with other subject.
(4) Status differences based on physical characteristics,
e.g., sex, color.
(5) Suspicion (see below).
We used necessary and sufficient criteria in deciding the
suspicion of a subject in our post-experimental interview.
(1) Necessary criteria: The subject must have mentioned
one of the following:
(a) The scores on the abilities were false.
(b) There is no such thing as Meaning or Relational
Insight Ability.
(c) The stated purpose of the study was not its real
purpose.
Given one of these necessary criteria, a subject was classified as
suspicious if one of the following sufficient criteria is also true
(referring to a through a of the necessary criteria).
(2) Sufficient criteria:
(a) The belief is stated firmly and early in the interview.

(b) The belief is mentioned frequently in the interview.

64

65

(c) Such information is volunteered with little or no
probing.

(d) The belief is based on one of the following:
(1) prior experience in a deception study;
(2) prior knowledge about deception studies;
(3) having been told that this study involves

deceptions.

(e) The subject behaved in an unusual manner during
the interview.

Our necessary and sufficient criteria mean there must be positive
evidence of suspicion rather than positive evidence the subject was
not suspicious. It is not sufficient to say a subject is suspicious
if:

(1) the subject mentions suspicion late in the interview

or during the debriefing;

(2) the subject indicates suspicion after intensive probing.

APPENDIX D

BERGER-FISEK HOST PROCEDURE: PHASE II

Good . Let me introduce myself. I am 4

 

and this is . We would

 

 

like to thank you for joining us today.
This is a study of decision-making. More specifically, this is
a study of how individuals work together as a group to solve problems,
and how different types of organization affect the efficiency of
problem-solving. Today the two of you will be working together as
a team to make decisions about problems that we will give you. This
is a common yet important kind of situation, sometimes referred to as
a Critical Choice situation. The most important thing about a Critical

Choice situation is to make the right decision. For example, when a

 

doctor has to make a difficult diagnosis, the patient may die if he
does not make the right decision. But it is often difficult for the
doctor to recognize at first what is the correct decision. When con-
fronted with such a situation, the doctor will study the problem care-
fully, and, because the most important thing in a Critical Choice
situation is to get the pigh£_answer, he will usually seek advice

from others. Only then will he decide. Hg_will not care whether

 

 

 

p§_originally thought p£_the answer himself, pp_was assisted py_

someone else ip_recognizing the right answer. The important thing

 

 

is that he make the right decision.

66

67

Today we are interested in studying how people make decisions
in Critical Choice situations when they have advice from others.

Let me tell you something about the task you will be working on.
This perceptual task, which social scientists have labelled
Contrast Sensitivity, has been developed in recent years, and because

of its importance is currently being used in an extensive set of
studies in this part of the country and elsewhere. This Contrast
Sensitivity task involves judging color contrasts in figures or
objects. Today you will be judging a series of slides like the one
now on the screen.

(put demo slide on the screen)

Each slide has two patterns, which are made of smaller black
and white rectangles. One of these two patterns--either the tpp_one

or the bottom one-- has more white area than the other. Your task is

 

to determine which of the two patterns--the top one or the bottom one-—
has the greater area of 32232:

(retract and advance slide)

You may find that some of the slides will seem difficult to
judge as the difference in the area covered by black and white rec-
tangles is sometimes quite small. However, previous studies have
shown that ggpg_individuals are able to make correct judgments on
the basis of very slight, almost intuitive cues and feelings. That
is, this is a fairly difficult task on which individuals £2 differ
markedly in their performances.

There are two things which make this task very important to
social scientists. First, we know from previous studies that the

capacity to make correct Contrast Sensitivity judgments is not

68

necessarily related to specialized skills the individual might possess,
such as mathematical or artistic skills. That is, those people with
high mathematical or artistic skills are not necessarily more accurate
in their Contrast Sensitivity judgments. Those people with low mathe-
matical or artistic skills may in fact be quite accurate at making
Contrast Sensitivity judgments.

The second important finding is that the way individuals solve
Contrast Sensitivity problems strongly depends upon the kind of situation
they are in.

We know from previous studies that most individuals do considerably
better on these problems when they are working in a group situation
than they do when they are working alone. Since we are interested in
your making as many correct judgments as possible, you will be working
in a gpppp_situation today.

That is, because this is a study of how people make decisions
in a Critical Choice situation, the two of you will be allowed to
exchange information with each other as to what you think is the
correct answer before you make final decisions. Only your final
decisions will be recorded for your scores.

This is how it will work. First, I will present a slide on the
screen. After you have studied the slide for five seconds, I will
ask each of you to make an initial choice as to which pattern contains
the greater area of 223333 top or bottom. Five seconds after you make
your initial choices, I will ask you to make your final choices.

will demonstrate how the equipment works.

 

(Boardman, equipment demonstration)

69

[Boardman: We'd like to go through one more practice trial
exactly the way it will be during the study.]

I will present a second demonstration slide so that you can
practice with this procedure. This will ppt_count on your score;
this is just for practice.

You will have five seconds to study the slide before I call
for your initial choice.

(show slide for five seconds)

Now make your initial choice.
(pause five seconds after both have chosen)
Now make your final choice.

(turn off projector)

(advance slide)

(press relay release)

(check off trial number on sheet)

In today's study there will be twenty trials exactly like this
one. For each of these trials we are interested solely in your making
the correct final decision. Therefore, you should not hesitate for
any reason to change your initial choice in order to make the correct
final decision.

During the slide series you may recognize certain similarities
between a rectangle that you are studying and one you have already
seen. This is due to the fact that some of the shapes which compose
the rectangles are repeated. It is the case, however, that within
each rectangle, these shapes are combined in a new fashion so that

no two rectangles are identical. Therefore, you should treat each

70

pair of rectangles as a new problem which requires its own study

 

and its own decision.

 

Since you will be working together and helping each other, it
is important that you know as much as possible about each other.
This is the reason for the two tests you took earlier. These tests
measure two special distinct abilities--the Relational Insight Ability
and the Meaning Insight Ability. We have found that the results of
these tests are very accurate indicators of a person's perfbrmance on
the Contrast Sensitivity task. As we mentioned, the Meaning Insight
Ability is the ability to understand the meaning of unfamiliar events
or objects. The Relational Insight Ability is the ability to under—

stand the relationship between unfamiliar objects or events.

 

Meaning Insight Ability and Relational Insight Ability are two
quite distinct abilities since they involve different thought patterns.
Although these abilities are quite different, previous studies have
found that these two abilities are usually closely associated with
each other. That is, an individual possessing high Relational Insight
Ability will almost always have high Meaning Insight Ability. If a
person has a low degree of Meaning Insight Ability, he will usually
have low Relational Insight Ability as well.

As mentioned before, one of the particularly interesting things
we have found in our studies is that the capacity to make correct

Contrast Sensitivity judgments is very closely related to the possession

 

of these two abilities. It has been found that those individuals
with high Meaning Insight Ability and high Relational Insight Ability
usually make correct Contrast Sensitivity judgments; those individuals

with low Meaning Insight Ability and low Relational Insight Ability

 

JIIILHII I4

1.! Ill I.» ..

71

usually make incorrect judgments. That is, individuals who possess
these special and distinct abilities are also likely to be good at
the Contrast Sensitivity task and so usually make a high number of
correct judgments.

will explain in more detail what this means.

 

(Boardman, Part II) (Boardman leaves)
Before we begin the series, I would like to summarize a few points. 4
(1) You will be shown a series of twenty slides and be asked to }
decide which pattern has the greater area of white. 5-
(2) Only your final decision on each slide will count towards
your score.
(3) Each time you make the correct final decision, that adds
one point to your score; if you make the incorrect final decision,
that adds nothing to your score.
(4) We are interested in your making as many correct final
decisions as you can; therefore, you should not hesitate for any
reason to change your initial choice, if that helps you to make the
correct final decision.
(5) At the end of the study I will report to you how many correct
decisions you each made.
Finally, please hold your choices until I call for them, but make
them as soon as possible thereafter.
(show slide series)
This completes the series of slides. Now we would like you to
fill out a questionnaire. The questionnaire is an important part of

our work. Please read the questions carefully, and take your time in

 

II (III- l{ I AI. lll Al. Hull '11 |'|.l|

72

answering them. If there is something you do not understand, raise
your hand and I will come around and help you.

(hand out questionnaires) (erase the board)

(Boardman returns and collects them, checking for correct desk
number and completeness)

We would now like to discuss your scores with you, and to talk
with each of you individually in order to get a further elaboration

of your feelings and opinions about the study. Mr.

 

will talk with you, Number 1; and I will talk with you, Number 2.

 

II‘t‘Ifl‘n‘l

 

IGAN STATE UNIVERSITY LIBRARIES

mum lllllljlllllllllllll

1293 3 78 1028

”TIT