It.“ 'I.
. . ' "S‘.’
{I’MLJ‘M'
r‘! ' Ho"
1. “1' ‘|'<v‘.’ VI" '
. "Hg, 121313531" 1:. AI: a
l‘ I.“ 'u

I. ..
II n .
. .' I, I! ff?“
I’A. .'
{fir-H“ .. IE“. ‘5.‘.\
I..." "I”: (\E’t‘n ‘55:“...

‘6‘“.
'lisIfl': gait‘ffi: NI'JH ”It:

.' '.:‘Il' J¥
"I‘I-Ah.

'4...-

ti_o \ b

3 'IIVI 'f .
Er Ig'u'm.‘

'i‘l‘fi'

'..“ \I.I\{

. I a .. '-:~ MI
.I .m“ six ”Pl-N T-M.‘ N.‘ I.
52.... it!” I ." ' .. | t: '12:” to . 'I'.'.".':‘_: l.‘ '41. “5%" I
L' . ”Ir? ‘ ‘ . '5 fi'mz..t.rxhl;~ -"..‘f.sz 8’1“: *JJ‘IJJI IEi qm‘gl i \ll

. .LzzIszIwII; I

' é;1\\ "M“ ,.
‘ 1'": I \‘lofio' 'I::I"" :
In L.- ‘é~

kt”.

TN" 1335352 ,TI;:I.éo'-' ' > ""13.
' “II“? 33;. .II 5 " ""5“

I-”

‘r ' mum‘rau
. 4:2. .

2/-
m

I“!
. :~
”Li-I
. no"
:‘5 4-.
' L:.
I _'I'

 

II
1 {w ' '
X "I
1 V ‘- fl. ': "6b: Ill'f‘h
. . *1 If-Wx
J "51"
VI\.| . ~ 1% 'I. '3“ 5'! .
0&3.“ fit}. H‘: . ), r"i§'/\.tzi.:nu%'.|t
. fit I $9. 'II
Q‘I}.\¢;\\‘;‘¢. u:
IN“
.. .-
'I‘i‘ fit-\‘IIII' ‘ .‘nI ”6‘2“;
. "II :‘I N'I?‘ Is” “\l‘fisvvl‘f‘v‘q I ‘
5.3.. “3“ .-I.\ .f I. 4 H. .. | ,‘u \-"‘I.“ :‘
~.'. .91, ‘g, ‘J.’ , ’:I.;“t‘.:.;'- " ._":; 2.1, ,, ~. k.“ ‘ «EQ'M‘
""":I“".A,.','.N ' 'I:‘""":v'-':}e~' :‘V'n ‘VE'Y'I - “"1"."‘5;
\LIE" I." ' I!
I I
..'. 5‘ A :33‘ q. .4" l“'.l'o :fi:
‘5." I" "'. ._ ‘ ‘l {13555.5
I ’0 I ‘U '
: :r‘iz:}'n"-\IW:"W“IUI I 11:" ‘ R‘ W.
I . (.1. I""“‘ '. . '
I... 'I.' ”‘1‘...""' 5-". ’|I%.-O{IQ..!.A{':".. ' $3135)
I“ L' I.,,..1I'h "“" ""53 I“: "I': " 'I"-I'.“' ' All ' '1'” I -
.1” '. .' ‘ “My... $4...” Mil": I “In"; ‘ :5}; I"
I ' ‘. I. .'.‘ .,1. I l.
'5‘. LI "-1" m} pug-x _ . ‘.
EA I‘I """ - h. ‘ ' 'w‘ux' . ' 99'; ':'- I‘Qu‘uj
I. ‘ ““5 ‘H I '2} ‘.J % :ft‘z‘h

. _

‘I ”(7"

’ “M H- |

'I ‘.‘ I ‘v AI: I-"I
' M‘ q" ' " "3“"; 1“. ”14"“ " \l\.' H‘ AIM:

I l ) :‘4'1'QVM

' II'll “
I ('1': "‘. ."
"f"" "‘I . a ‘. r w ‘ ‘ |‘\ I") '3“.
l
I

I, n "'1 '.' c
'l..""c3€I-_,. " (‘3. ‘/'51.. 'ul'flk‘.‘ “I3“ 4:. fl 5"" kit:
I .‘ ".' q"ll '|L‘ "a y

3%
.x'.'

g

I .051]. I .4 I . \:‘3":'
I? n I .v ‘ “‘1‘ .v2'-'.‘v"f'/-a"‘ I ”'1‘“ H "I“ ‘. :“L'O Al: ' LI \'|"-'I £3"! :':‘\ ‘
.‘I‘, '9 ‘ ,‘ ‘5': 'N I 1;: Vi" 'I~"“I"‘“ 3" ‘ "'3' 'I,:' :3“... - W '1'") \c'“
_., .a I . " :foq'fi I‘IM‘ " I . .I I
k ( -.'4'I":'.")'1"~"I'LI ‘fl‘l‘o‘dl'vl‘ "' \: yfi' 'n‘l "1:"
n‘ . .‘I. I'-I.'I‘I' \' h‘ '- ‘:‘ "I" “5?." "'
‘ ' I.

"I
I <1..."I§ a If.-
II~|:..:.:1‘" . 't'} "| " k, "f. -' I' '
. ‘ ‘ I. I‘ . ' I. 9" r.
.I': {m 5|.” 1‘: . “43‘”? I ”'5“:
&‘ ~' ‘5 """ FIN":
‘33."; . . 4.8:]. 9|} ‘I‘ '):I I" . “I‘l'h :$"'.'-"
': I :‘-‘IJ.I1:.;E:‘“':ZI’:I‘IN‘:‘I ‘31:“; ._| ““ ,,,- NMIVM? .-
I; I I I- .I . - -

5- ; I - af-fiiqu‘Ij-z-‘z-a‘gg‘. :L'

‘5--

1'3
.; '

I I“
‘-‘ v4 .
,. I}. L. Ii} -.
W I}: :‘iu.!.' I" 'I,‘ ' '

In} I‘.

 

llllllmlilllllfllllll 1»

 

1293 01066 0953

 

This is to certify that the

thesis entitled

A PRELIMINARY EXAMINATION OF A MODEL
OF THE PROCESS OF SOCIAL INTERACTION
IN THREE SITUATIONS

presented by

Jerome David Johnson

has been accepted towards fulfillment
of the requirements for

Ph . D . degree inCommuni cation

A

//

(gm 7/ flew, [M2

 

Major professor ‘—//

Dam September 15. 1977

0-7 639

© 1978

JEROME DAVID JOHNSON

ALL RI GHTS RESERVED

A PRELIMINARY EXAMINATION OF A MODEL OF THE
PROCESS OF SOCIAL INTERACTION IN THREE SITUATIONS

BY

Jerome David Johnson

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Communication

1978

l/K' -..
<J<f
(.77 "l

.
(p

ABSTRACT

APWWIWOFAMEOF'IHE
PRCXZESS OF SCXIIAL INTERACI‘ICN IN THREE SITUATICNS

By

Jerare David Johnson

'Ihis dissertation proposes a nodel of the process of social inter—
action contaim'ng six categories: content, interpretation, errotion ,
communication, selection, and relationships.

These categories can be classified by their role in the process of
exchange and by their phenomenal level. IIhere are three phenomenal
levels: the surface level (content and communication) ; the mediating
level (interpretation and selection) ; and the underlying level (auction
and relationships). The categories can also be classified by their
roles in the process of exchange: content, interpretation , and emotion
can be viewed as the substance exchanged in the interaction, while com-
munication, selection, and relationships represent the form by which
this substance is expressed.

The relationships posited in the nodel are based on the classifica-
tion of categories. In general the substance exchanged in the inter-
action is viewed as determining its form of expression. In addition,
variables at a deeper phenonenal level are said to cause variables at a
more surface level.

(he hundred and twenty-four nail questionnaires obtained from a
random sanple of adults in Grand Rapids, Michigan were used to test the
nodel of social interaction in three situations—television, radio and

typical. While the characteristics of this sanple generally reflects

Jerome David Johnson
the nature of the Grand Rapids and the United States population and the
literature indicates that non-response typically has little affect on
relationships between variables, the low response rate, 22. 2% , suggests
that only limited generalizaticms can be made from this data.

Ordinary least squares multiple regression (018) was used to de-
termine the variance accounted for and to assess the significance level
of the paths in the model. When the altemative paths are included in
the model the individual multiple regressions account for at least 24%
of the variation in their dependent variables with p < .01.

LISREL, a computer program, was used to assess the overall goodness
of fit of the model to the data and to estimate individual parameters.
The radio situation was used to develop and to further refine the orig-
inal model presented in Chapter I. Tests in this situation indicated
that two additicmal paths—one between emotion and content and one be-
tween relationships and communication—should be added to the model .
'Ihis refined model was then tested in all situations. The x2 statistic
indicates that the model doesn't provide a better fit to the data than
would be expected by chance. waever, the ratio of degrees of freedom
to the x2 value would indicate that the model, with appropriate modifi-
cations, could eventually provide a good fit to the data.

In Chapter IV a modified model with two unobserved common variables
was tested in all three situations. The addition of these common vari-
ables was expected to ameliorate some of the problems with high zeta
variances, multicollinearity, and measurement errors fomd in the orig—
inal model. 'I’ney were partially successful in reducing these problems,

but their main effect was to reduce the residuals. The x2 values

‘ Jerome David Johnson
approach significance and the slight difference between them and a good
fit of the model to the data may be attributable to technical problems
with the data and to specification errors.

In sum, the results supported the addition of two paths to the
original model and suggests that all of the paths included in the orig—
inal model were meaningful. The results did not support the assurptions
that the values of parameters would remain invariant across different
situations and that interpretations and selection act as mediators.

In Chapter V a new model is proposed that incorporates the effects
of factors outside the process of social interaction, such as context,
and that collapses interpretation and content into one variable labeled
interpretation and reduces communication and selection to one variable
termed communication. There is reason to believe that a test of this

new model on a different data set would be successful.

ACKNOWLEDGMENTS

I would like to thank my committee members for their
contribution to this dissertation. Specifically I would
like to express my appreciation to Dr. Bettinghaus for his
continuing support, encouragement, and his substantial
editorial contribution to the various drafts of the disser-
tation. I would also like to thank Ed Fink for his substan-
tial contribution to my understanding of the methods used in
this dissertation and his suggestions concerning the presen-
tation of results. I would also like to thank Dr. Atkin and
Dr. Jacobson for their contribution to my substantive under-
standing of the areas in which I wrote my dissertation and
my preliminary papers.

I would also like to acknowledge the National
Association of Broadcasters for providing the funding to
conduct this research.

I also would like to thank Mrs. Ruth Langenbacher
for her assistance in preparing the copies of this disser-
tation and for her patience.

While my years in the Department of Communication at
Michigan State were sometimes frustrating, and I could find
things to quibble with in the emphasis given certain things
in the Department, I must also thank the Department for

ii

giving me a feel for how to do research-—something that in
its own way was more valuable than any other thing in con-
tributing toward my professional growth.

And finally I would like to express my appreciation
to the management of the Office of Research at the U.S.
Information Agency, and especially to Dr. Peter Janicki,
for providing the motivation I needed to complete this

dissertation.

iii

TABLE OF CONTENTS

Chapter Page
I INTRODUCTION . . . . . . . . . 1
Categories of Acts Within Social
Interaction . . . . . . . . 3
Content . . . . . . 3
Interpretation . . . . 4
Emotion . . . . . . . 6
Communication. . . . . . 6
Selection. . . . . . . 7
Relationships. . . . . . 8
Summary . . . . . . . 9
Overview of Previous Descriptions of the
Categories of Social Interaction. . . 9
Causal Relationships in the Model . . 12
Effects of Other Factors . . . . 17
Conclusion. . . . . . . . 22
II METHODS . . . . . . . . . . 23
Rationale for Selecting a Mail

Questionnaire . . . . . . 23
Response Rate . . . . . . 25

Effects of Nonresponse on the Rela-
tionships Between Variables . . . 28
Characteristics of the Sample . . . 29
The Mail Questionnaire. . . . . 32
Conclusion. . . . . . . . 37
III RESULTS OF TESTS OF THE MODEL . . . . . 38
Means and Standard Deviations . . . 38
Correlation Matrices . . . . 40
Multiple Regression Results . . . 43
Introduction to Model Testing . . 54

Advantages of LISREL over Multiple
Regression . . . . . . 57
Description of LISREL . . . . . 58

iv

Chapter III (cont'd.)

IV

V

Meaning of the X2 Text for Goodness

of Fit in LISREL . . . . .
‘-Operational Model Described in
LISREL Terminology. . . .
Identification of the Models. . .
Development of Final Radio Model. .
Results of Tests of Models in Radio
Situation . . . .
Comparison of the Radio Models . .
Results of Finally Chosen Radio
Model . . . . .
Results for the TV Model. . . .
Results for Model 11a in the Typical
Situation . . . .

Comparison of the Results of the
Tests of the Model. . . . .
Conclusion . . . . . . .

DISCUSSION OF THE RESULTS AND TESTS OF A

MODIFIED MODEL . . . . . . . .
Methodological Explanations of the
Results . . . . .
Modified LISREL Model . . .
Results of Model IV with the Common
Variable in the TV Situation . .

Results for Model IV with Common
Variable in the Typical Situation .
Results of Model IV with Common
Variable in Radio Situation . .
Comparison of the Tests of Model IV .
Comparison of Model 11a and Model
IV. . . . . .
Relationships Among True Variables .
Effect of Introduction of Common
Variable . . .
Methodological Explanations of the
Results . . .
The Role of Unspecified Factors in
the Results . . . . . .
Conclusion . . . . . . .

IMPLICATIONS OF THE RESULTS FOR FUTURE

RESEARCH

Methodological Implications . . .
v

Page

62
64
66
69

71
80

85
87

89

96
101

103
103
105
107
111

113
118

122
125

127
I30

132
135

136
136

Chapter V (cont'd.)

Page
Substantive Implications . . . 140
Suggested Model for the Process of
Social Interaction . . . . 142
Outside Factors in New Model . . . 145
Conclusion . . . . . . . . 149
APPENDICES . . . . . . . . . . . 150
Appendix A: Category Schemes for Social
Interaction . . . . . . . 150
Appendix B: Inclusion of Elements of Social
Interaction in Previous Category
Schemes . . . . . . . . 171
Appendix C: Instructions Given Telephone
Solicitors. . . . . . . . 173
Appendix D: Mailed Questionnaire . . . . . 177
Appendix E: Structural Equations for the Models
in Matrix Form. . . . . . . 189

Appendix F: Results of Tests of Model I in
Typical and TV Situation . . . . 193

Appendix G: Maximum Likelihood Solutions for
Models Tested in Dissertation . . . 195

REFERENCES . . . . . . . . . . . 220

vi

Table

10

ll

12

13

14

LIST OF TABLES

Response to Mail Questionnaire . . .

Characteristics of the Sample, of Grand
Rapids and of the United States . . .

Means and Standard Deviations for Variables
in Typical Situation. . . . . .

Means and Standard Deviations for Variables
in TV Situation . . . . . . .

Means and Standard Deviations for Variables
in Radio Situation . . . . . .

Pearson Correlations Among Social Inter-
action Variables in Typical Situation .

Pearson Correlations Among Social Inter-
action in Variables in Television Siutation

Pearson Correlations Among Social Inter—
action Variables in Radio Situation . .

Multiple Regressions for Paths in Model I
in TV Situation . . . . . . .

Multiple Regressions for Alternative Paths
in Model II in TV Situation . . . .

Multiple Regressions for Paths in Model I
in Typical Situation. . . . . .

Multiple Regression for Alternative Paths
in Model II in Typical Situation. . .

Multiple Regressions for Paths in Model I
in Radio Situation . . . . . .

Multiple Regressions for Alternative Paths
in Model II in Radio Situation . . .

vii

Page

27

30

39

41

42

44

45

46

48

49

51

52

53

55

Table

15

16

17

18

19

20

21

22
23
24

25

26

27

28

Page

Residual Matrix for Model Ia in
Radio Situation . . . . . . . . 73

Residual Matrix for Model 11a in Radio

Situation with Paths Between Relationship-
Communication and Emotion-Content with Psi

Matrix Elements Off-Diagonal Fixed at Zero . 75

Residual Matrix for Model IIIa in Radio

Situation with Paths Between Each of the
Underlying and Surface Variables and Off-

Diagonal Elements of Psi Fixed at Zero . . 77

Residual Matrix for Model 11b in Radio

Situation with Paths Between Relationship-
Communication and Emotion-Content with Psi 78
Matrix Estimated . . . . . . .

Psi Matrix for Model 11b in Radio Situation
with Paths Between Relationship-Communication
and Emotion-Content . . . . . . . 80

Residual Matrix for Model IIIb in Radio

Situation with Paths Between Each Underlying
Variable and Bach Surface Variable with Psi

Matrix Estimated . . . . . . . . 82
Psi Matrix for Model IIIb in Radio Situation

with Paths Between Each Underlying Variable

and Each Surface Variable. . . . . . 82
Comparison of x2 Values for Radio Models . . 84
Residual Matrix for Model IIa in TV Situation. 87

Determinants of Correlation Matrices . . . 91

Residual Matrix for Model IIIa in Typical
Situation. . . . . . . . . . 95

Goodness of Pit of Model IIa in the
Three Situations . . . . . . . . 97

Values of Paths Between Exogenous and Endo-
genous Variables in the Three Situations . . 98

Values of Paths Between Endogenous Variables
in the Three Situations . . . . . . 99

viii

Table Page

29 Zeta Variances for the True Variables

in the Three Situations . . . . . . 99
30 Scale Values for Ordinary Indicators

in the Three Situations . . . . . . 100
31 Measurement Error Variances for Multiple

Indicators in the Three Situations . . . 100
32 Scale Values for Common Variable in

Model IV in TV Situation . . . . . . 109
33 Residual Matrix for Model IV in TV

Situation with Common Variable . . . . 110
34 Scale Values for Common Variables in

Model IV in Typical Situation. . . . . 113
35 Residuals for Model IV with Common

Variable in the Typical Situation. . . . 114
36 Scale Values for Common Variables in

Model IV in Radio Situation . . . . . 115
37 Residuals for Model IV with Common

Variable in Radio Situation . . . . . 117
38 Goodness of Pit of Model IV with the

Common Variable in the Three Situations . . 118
39 Values of Paths Between Exogenous and

Endogenous Variables for Model IV with the

Common Variable in the Three Situations . . 119
40 Values of Paths Between Endogenous Variables

for Model IV with a Common Variable in the

Three Situations . . . . . . . . 120
41 Zeta Variances for Model IV with the

Common Variable in the Three Situations . . 120
42 Scale Values of Ordinary Indicators in

Model IV with a Common Variable in the

Three Situations . . . . . . . . 121
43 Scale Values for Common Variables in the

Three Situations . . . . . . . . 122
44 Measurement Error Variances for the Multiple

Indicators in Model IV with the Common

Variables in the Three Situations. . . . 123

ix

Table

45

1G

2G

3G

4G

5G

6G

7G

8G

9G

10G

Page

Differences Between Model 11a and Model
IVa in the Three Situations . . . . . 128

Maximum Likelihood Solution for Model Ia
in Radio Situation with Off-Diagonal
Elements of Psi Matrix Fixed at Zero . . . 195

Maximum Likelihood Solution for Model IIa

with Paths Between Emotion and Content

and Relationship and Communication with Off-
Diagonal Elements of Psi Matrix Fixed at

Zero in Radio Situation . . . . . . 197

Maximum Likelihood Solution for Model IIb

with Paths Between Emotion and Content and
Relationship and Communication with Off-

Diagonal Elements of Psi Matrix Free in the

Radio Situation . . . . . . . . 199

Maximum Likelihood Solution for Model IIIa

with Paths Between Underlying and Surface

Variables with Off-Diagonal Elements of

Psi Matrix Fixed at Zero in Radio Situation . 201

Maximum Likelihood Solution for Model IIIb

with Paths Between Underlying and Surface

Variables with Off-Diagonal Elements of Psi

Matrix Free in Radio Situation . . . . 203

Maximum Likelihood Solution for Model II
in Typical Situation . . . . . . . 205

Maximum Likelihood Solution for Model II
in TV Situation . . . . . . . . 207

Maximum Likelihood Solution for Model IV
with Common Variable in Radio Situation . . 209

Maximum Likelihood Solution for Model IV
with Common Variable in Typical Situation. . 211

Maximum Likelihood Solution for Model IV
with Common Variable in TV Situation . . . 213

Figure

10

11

12

LIST OF FIGURES

A Model of the Process of Social
Interaction . . . . . . . .

Classification of the Categories of
Social Interaction . . . . . . .

An Operational Model of the Process of
Social Interaction . . . . . . .

Results for Model Ia in Radio Situation . .

Results for Model IIa in Radio Situation with
Paths Between Emotion and Content and Rela—
tionship and Communication and the Off-
Diagonal Elements of Psi Matrix Fixed at Zero.

Results for Model IIIa in Radio Situation
with Paths Between Surface and Underlying
Variables and the Off-Diagonal Elements of
Psi Matrix Fixed . . . . . . . .

Results for Model IIb in Radio Situation with
Path Between Emotion and Content and Rela-
tionship and Communication and Off-Diagonal
Elements of Psi Matrix Estimated . . .

Results for Model IIIb in Radio Situation

with Paths Between the Underlying and Surface
Variables and the Psi Matrix Estimated . .
Results for Model IIa in TV Situation. . .
Results for Model IIa in TypiCal Situation .

Results for Model IV in TV Situation
Common Variable . . . . . . . .

Results for Model IV with the Common
Variable in the Typical Situation. . . .

xi

Page

13

67

72

74

76

79

81
88

94

108

112

Figure Page

13 Results for Model IV in Radio Situation

with Common Variable. . . . . . . 116
14 A Suggested Modifications in Model of the

Process of Social Interaction . . . . 145
15 A Suggested Expanded Model of the Process

of Social Interaction . . . . . . 148
1F Results for Model I in TV Situation . . . 193
2F Results for Model I in Typical Situation. . 194

xii

CHAPTER I

INTRODUCTION

Social interaction is the process by which interact-
ants express matters of substance to one another.* Many
category schemes of social interaction have been proposed in
the literature. However, no models of social interaction
have been derived from these schemes. This dissertation will
propose a model of the process of social interaction that
will be tested in three different situations.

This chapter will develop the model of social inter-
action. Chapter II will describe the means by which this
model will be tested. Chapter III will report the results
of the tests of the model in three different situations.
Chapter IV will discuss methodological and substantive ex-
planations of the results. Chapter V will conclude the dis-
sertation with a discussion of the implications of the
results and suggestions for future research.

Social interaction can be studied on at least four
levels: as a dependent variable, as an independent variable,

as an intervening variable, and in isolation. Examining the

 

*

This definition follows closely the definition pro-
posed by Ruesch and Prestwood (1949, p. 413). "Communication
(social interaction) is the process through which intention,
feelings and thoughts of one person are transmitted to
another."

1

2

effects of the situation, the characteristics of interact-
ants, the context of the interaction and the rules of pro-
cedure for social interaction involves analyzing the effects
of various independent variables on the dependent variable
of social interaction. Social interaction has also been
treated as an intervening variable through which various in-
dependent variables, such as group task or group structure,
act on dependent variables, such as group effectiveness and
group satisfaction. In this dissertation the process of
social interaction will be examined essentially in isolation,
although reference will be made to the possible effects of
various independent variables on the process. No attempt
will be made to examine social interaction as an intervening
or independent variable. The primary emphasis here is on
the development of a model detailing the relationships be-
tween the categories of acts contained in social interaction.

The model of social interaction that will be tested
is contained in Figure 1. This model contains six cate-
gories of the process of social interaction: content, com-
munication, interpretation, selection, emotion and relation-
ships. In the following section each of the categories of
social interaction will be defined, and its place in the
literature examined. Once the categories have been defined
the causal paths between them that constitute a model of
social interaction will be detailed. Chapter I will con-

clude with a discussion of the possible effects of the

 

CONTENT 4) COMMUNICATION

 

INTERPRETATION 3 SEL CTION

 

EMOTION RELATIONSHIP

V

Figure l. A Model of the Process of Social
Interaction

larger context within which social interaction is embedded

on relationships within the model.

Categories of Acts Within Social Interaction

Content: Content is the denotative meaning of sym-
bols expressed during an interaction. Appendix A contains
samples of category schemes that have been used in the lit-
erature to describe social interaction.* Most of the cate-
gory schemes reviewed do not contain all of the categories

of social interaction included in the model.** All six

 

*

Social interaction is used here as a generic term
that includes some terms, e.g. communication, other re-
searchers have used to describe what is meant here by
social interaction.

*
* See Appendix B. Thirty category schemes were re-
viewed. Category schemes are used here as generic term.
Some of the researchers intended merely to describe or de—
fine some of the elements of social interaction. They are
included here to demonstrate that other researchers have
frequently used these terms to describe aspects of social
interaction.

4
researchers who cited content as a category of social inter-
action also cited other categories contained in the model.
(Argyle, 1969; Bjerg, 1968*; Hare, 1958; Hawes, 1973;
Watson, 1958*; and Watzlawick, e£_al., 1967.)

In general, these researchers were interested in de-
signing category schemes for fairly abstract and widely
generalizable purposes, such as Hare's attempt to establish
a paradigm for the analysis of interaction or Hawes' attempt
to isolate the elements of communication processes. As a
result content was most frequently not defined nor elaborated
upon, but rather listed along with other categories of
social interaction.

Interpretation: Interpretation is the connotative

 

meaning associated with expressed symbols. In general, cate-
gory schemes that include interpretation have been designed
to describe specific situations and are more limited in

scope than those category schemes that contain content. Of
the category schemes that include a category similar to in-
terpretation (Auld and White, 1959; Bales, 1950; Borgatta,
1970; Crowell and Schiedell, 1961; Flanders, 1967; Gouran

and Baird, 1972; Lewis, gt_§l., 1961; Longabaugh, 1961;
McGuire and Lorch, 1968; Schiedell and Crowell, 1966;

Snyder, 1945; Steinzor, 1949; Strupp, 1960; and Weintraub

 

*
Bjerg (1968) and Watson (1958) called their cate-
gories tOpic agons and conversational resources respectively.

5
and Aronson, 1962) only five included other categories con-
tained in the model of social interaction.

The key difference between these category schemes
and the ones used for content is that they are generally
less abstract and often demand that an act be placed into a
more abstract category than the content that was contained
in the act. This placement often demands interpretation.1

The category schemes that include interpretation are
often intended for only a limited purpose (e.g. analysis of
group problem solving, Gouran, and Baird, 1972; analysis of
defense mechanisms, Weintraub and Aronson, 1962; analysis of

psychotherapy, Auld and White, 1959; and analysis of student

teacher interaction, Flanders, 1967). As a result they are

 

1For example, the content of the interaction might
be a group member saying "Joe, you really helped the group
stick together. If it wasn't for your leadership we would
have fallen apart." Now this manifest content has to be
interpreted to be placed in one of Borgatta's (1965) cate-
gories. If the speaker was the actual leader of the group
and there was a threat that the group would break apart if
Joe's ego wasn't raised, then this statement might fall
under category 2 (shows solidarity through raising the sta-
tus of others). If the speaker was justly praising Joe,
then it might be coded a 1 (common social acknowledgments)
given this was typical praise for a group member, or an 8
(gives Opinion, evaluation, analysis, expresses feeling or
wish). If Joe had fallen apart in this situation and the
speaker had spoken in a sarcastic tone, then this simple
utterance might be coded a 3 (shows tension release, laughs).
Given the outcome was favorable to the group, or a 17
(shows antagonism, hostility, is demanding) provided the
outcome was unfavorable. The point is that all of these
category schemes demand that a coder rely on other things
than the manifest content to place a symbol exchange in one
of these categories--the interaction must be interpreted.

6

often constructed in a manner that makes it unlikely that
they could be generalized to a wide range of social inter-
actions. Only Borgatta (1965), Longabaugh (1966), Steinzor
(1949), and Bales (1950) apparently intended that their
schemes be generalizable across a wide range of situations.

Emotion: Emotion reflects the affective states in
an interaction. Nine of the category schemes reviewed here
have included this category within their schemes (Argyle,
1969; Bjerg, 1968; Taylor, 1954; Reusch and Prestwood;
1949; Hare, 1958; Weintraub and Aronson, 1962; Carter gt_§l.,
1951; and Longabaugh, 1966). Only Taylor (1954), who was
attempting to describe the emotional dimensionality of
groups, and Reusch and Prestwood (1949), who were attempting
to define the structural components of interaction, fail to
include another category of social interaction contained in
the model in their category schemes. All save Weintraub
and Aronson (1962). Argyle (1969), and Carter gt_al., (1951)
intended their category schemes to be used for rather ab-
stract, general purposes. The greatest number of subcate-
gories within emotion identified by any of these category
schemes is three; the public dimension, dyadic dimension
and autistic dimension, by Taylor (1954).

Communication: Communication refers to acts of symbol

 

transfer. Twelve of the thirty category schemes reviewed
here contain categories that are similar to communication

(Watson, 1958; Pope and Siegman, 1972; Speier, 1973; Bjerg,

7
1968; Argyle, 1969; Bostrom, 1970; McGinnies and Altman,
1958; Weintraub and Aronson, 1962; Jaffe and Feldstein,
1970; Hare, 1958; Lewis et_al., 1961; and Amidon and Hunter,
1966). Only two of the ten categories, Jaffe and Feld-
stein's and McGinnies and Altman's, do not contain other
categories of social interaction used in the model.

The category schemes are either very specific, but
not exhaustive, in identifying subcategories, for example
Bostrom, or are very general in characterizing communica-
tion, for example, Watson's conversational style.

The category schemes represented here appear to be
equally split between those designed for specific purposes,
such as the verbal analysis of defense mechanisms, Weintraub
and Aronson (1962), and those designed for rather abstract
purposes, e.g. Hare's paradigm for the analysis of inter-
action. Most of these schemes include both verbal and non-
verbal acts of symbol transfer.

Selection: Selection reflects the interactant's levels

 

of attention to elements contained in the environment of the
interaction. Only four of the descriptions of social inter-
action reviewed here contains selection. Goffman's (1957)
scheme was designed exclusively to deal with attention.
Argyle (1969) included in his scheme a category for non-
verbal responsiveness, which reflect signals from one inter-
actant to another of attentiveness. Bostrom (1970) called

his category selectivity and described it in terms of relative

8
concentration, and Lewis et_al. (1961) included listening in
their category scheme. Goffman (1957) termed external pre-
occupation, self—consciousness, interaction consciousness,
and other consciousness as forms of alienation from inter-
action, or in other words forms of selection. All four of
these researchers use their concepts to indicate the inter-
actant's level of attention to a subset of the elements con-
tained within social interaction.

Relationships: Relationships represent the patterns

 

of response of the parties in the interaction to each other.
Seven of the schemes for describing social interaction that
have been reviewed here contain relationship categories
(Hawes, 1973; Hare, 1958; Bjerg, 1968; Pope and Siegman,
1972; Watzlawick, gt_al., 1967; Speier, 1973; and Argyle,
1969). All save Pope and Siegman (1972) and Argyle (1969)
intended their schemes to be applicable across a wide range
of social interaction situations, and conceived of their
categories at relatively abstract levels. All of these
schemes contain other categories of social interaction con-
tained in the models.

Some of the category schemes that contain relation-
ships deal with specific subcategories of relationships
(Bjerg, Hare, Pope and Siegman, Argyle, and Speier) and some
(Hawes and Watzlawick §E_§l.) just use the term relationships
to describe this category. None of the schemes that mention

specific subcategories would appear to have exhaustively

9
described the major types of relationships that could exist
in social interaction.

Summary: The model of social interaction posited in
this paper is composed of six categories: content, interpre-
tation, emotion, communication, selection, and relationships.

Content is the denotative meaning of symbols expressed
during an interaction.

Interpretation is the connotative meaning associated
with expressed symbols.

Emotion reflects the affectives states in an inter—
action.

Communication refers to acts of symbol transfer.

Selection reflects the interactant's level of atten-
tion to elements contained in the environment of the inter-
action.

Relationships represent the patterns of response of
the parties in the interaction to each other.

Overview Of Previous Descriptions Of The
Categories Of Social Interaction

Before the causal relationships between the categories
of social interaction are detailed, a brief overview of pre-
vious category schemes is in order. Appendix B contains a
listing of the category schemes indicating the categories of
social interaction contained in this model that they include.
Appendix A describes the schemes in more detail including

categories the author's described as being included in social

10
interaction that aren't contained in the model developed
here.*

When examined in their entirety, rather than category
by category, several criticisms can be leveled against the
category schemes contained in Appendix A.

First, they are often designed to serve only a limited
purpose, such as the analysis of teacher-student interaction,
psychotherapy sessions, or group decision making. As a re-
sult these schemes have only limited applicability to a wide
range of social interaction situations. For example, none
of the schemes reviewed here is suited for "aimless," non-
directed conversation.

Second, the schemes often fail to focus on the process
of social interaction. Traditionally studies of interaction
have followed a classic independent/dependent variable para-
digm where "inputs" of social interaction (e.g. setting,
context) are examined for their effects on "outputs" or the

effects of social interaction such as changes in self-concept,

 

*

Some of the schemes reviewed here include categories
not included in the model of social interaction presented
here. Two of these other categories are sequences (Argyle,
1969), an increasingly popular means of examining social
interaction (Stech, 1975), and rules (Bjerg, 1968). However,
the nature of the sequence is the important thing. Sequences
and rules should be viewed as means of analyzing various cate-
gories of social interaction, not as categories themselves.

The other categories used to describe social interaction
in these schemes are related to situational variables that
effect social interaction and which will be described later
in this chapter.

11
attitude change, information gain or need fulfillment.

Several recent criticisms of research on
face-to-face behavior have suggested that
designs are too often 'input-output' rather
than 'process' oriented. . . . That is
typically some independent 'input' variable
is manipulated (e.g., leadership style) in
order to see the effect on a cumulative
'output' variable (e.g. the quantity of work
accomplished), whereas the interaction it-
self is not examined (Roger and Jones, 1975,
p. 113).

Studies that focus on inputs and outputs neglect the process
of social interaction, they treat it as a black box.

Too often . . . small group research efforts
have purported to investigate 'process' by
subjecting groups to various 'input' vari-
ables of the HSM (Human System Model)

- e.g., members, group structure, task con-
ditions, and measuring 'output variables'
e.g. productivity, cohesiveness, efficiency.
The conclusion of the research infers that
the intervening interactive behavior account-
ed for the obtained relationship. But
measuring 'input' and 'output' does not

study process. Only direct investigations

of the intervening behavior--the group inter-
action--can claim to be a study of process
(Fisher and Hawes, 1971, p. 452).

This focus has resulted in categories that have only limited
usefulness in modeling the process of social interaction.
Three, quite frequently schemes have omitted elements
of social interaction mentioned by other researchers. Three
of the schemes, Argyle (1969), Bjerg (1968) and Hare (1958),
incorporate five of the categories. Of the thirty schemes
reviewed only these three identify a substantial number of
the categories of social interaction contained in the model.

In fact most (16) only identify one element of social

12
interaction. Of all of the elements of social interaction
interpretation is the most frequently represented (15),
followed by communication (12), emotion (8), relationships
(7), content (6) and selection (4). Lennarwi and Bernstein
(1969) sum up the current situation well.
A search through clinical and research
literature (on social interaction) in
the family and other social systems will
locate few comprehensive category schemes,
especially ones which are completely de-
scribed and complete with definitions and
indicators that make the system useful to
other students.

Four, category schemes, as a result of their limited
range, have often neglected the rich and simultaneous inter-
play of a number of elements of social interaction (Bjerg,
1968). One of the reasons for a dearth of models would

appear to be an incomplete description of the process,

which handicaps conceptual development.

Causal Relationships In The Model

The lines between the categories in Figure 1 detail
the proposed causal relationships in the model of social
interaction. A straight line indicates a causal path, with
the arrow indicating the direction of causality. For example,
the straight line between interpretation and content with
the arrow pointing to content indicates that interpretation
causes content. Only one relationship in this model is not
causal, that between emotion and relationships; this non-

causal relationship (correlation) is indicated by a curved

13
line with the arrows pointing to both variables.

Two means of classifying the categories of social
interaction will be used as the basis for determining the
causal relationships between variables in this model: the
role of the category in the process of exchange and the phe-
nomenal level of the category. Figure 2 presents a classi-
fication of the categories of social interaction by these

two means.

Role in the Process of Exchange

 

 

 

 

Phenomenal Level of »

Interactive Element Substance Form of Expression
Surface Content Communication
Mediating Interpretation Selection
Underlying Emotion Relationship

 

 

 

 

 

Figure 2. Classification of the Categories
of Social Interaction

The categories of social interaction can be classified
into three phenomenal levels. The first level is the surface
level which contains the most readily manifested and ob-
jective of the categories: communication and content. The
second level is the mediating level. The mediating cate-
gories, interpretation and selection, mediate the surface
and underlying categories. The underlying categories,

emotion and relationships, are the primary determinants of

14
the other elements.

The underlying categories cause the mediating cate-
gories which in turn cause the manifest variables that
serve similar roles in the process of exchange. Thus emo-
tion, an underlying category, is seen as causing both inter-
pretation and selection. It is a fairly commonplace notion
that the emotions that parties in an interaction feel toward
each other color their interpretations and they affect the
degree of attention interactants pay toward each other.

Relationship is an underlying category partly be-
cause "interaction grows out of the roles we play, the de-
fined relationships we have in various groups" (Ittelson,
§E_al., 1970, p. 127). Relationship is one of the determin-
ants of interpretations and selection. Naturally the inter-
pretations attached to the substance of the interaction are
dependent upon the relationships between interactants. In
addition, whom is attended to and what level of attention
is given is also heavily dependent upon the nature of the
relationship between interactants.

The mediating categories, in this framework, are the
mechanisms that translate the underlying variables into the
manifest acts in the interaction. Thus interpretation is
said to cause content and selection is said to cause commun-
ication. Selection in this model is one of the determinants
of whose messages are attended to and to whom messages are

given. Selection then is a necessary antecedent of any

15
communicative act.

The two underlying variables may be correlated in
this model. The nature of emotion and relationships and
their role as the exogenous variables in the process dic-
tate that there not be a causal path between these two
variables.

There are two primary roles a category can assume in
the process of exchange: substance and form of expression
(Hare, 1958). The content, interpretation and emotion
categories represent the substances that are exchanged dur-
ing an interaction. How the substance is exchanged is re-
flected in the form of expression elements: communication,
selection and relationships. Since it is expected that the
substance to be exchanged will determine the manner in which
the substance is transmitted, the model contains causal
paths from the substance to the form of expression elements
at the mediating and manifest levels. Thus it is expected
that the content that is to be transmitted will partially
determine the communication pattern in the interaction and
the interpretation of the symbols in the interaction will
partially determine the level of attention in the inter-
action.

The model presented in Figure 1 represents an elegant,
parsimonious representation of the relationships among the
categories. Some additional paths could reasonably be added

to the model. One set of relationships that would appear to

16
have some empirical support are causal paths between the
underlying and surface categories which share a common role
in the process of exchange. In the model it is suggested
that these relationships are mediated by interpretation and
selection, but a case can be made for a direct relationship
between these categories.

A causal path between relationships and communica-
tion has received some empirical support (Pope and Siegman,
1972). Roles have also been said to determine the commun-
ication systems of interactants (Kees and Reusch, 1970, and
Ervin-Tripp, 1964).

Another set of paths that could reasonably be added
to the model are crossed paths between relationship and con-
tent and between emotion and communication. The logic here
is that the underlying variables are truly determinants of
every variable within the system. Thus the relationships
between the interactants is seen as determining the content
of the interaction.

Similarly emotion and communication might be related.
Thus the level of emotion could determine communication
patterns within the interaction (Adams, 1967). For example
in a highly emotional interaction interactants may talk
more quickly, less fluently, and have a higher pitch in

their voice.

17
Effects Of Other Factors

Most studies of social interaction have been concern-
ed with examining the factors that affect social interaction
and their relationship to possible outputs of social inter-
action, such as attitude change, conflict resolution, and
decision making. Here the primary focus is on the process,
and outputs won't be discussed. However, inputs, those fac-
tors that affect the process of social interaction, may have
an effect on the tests of the model. HOpefully the basic
causal relationships in the model are invariant in direction
and magnitude across situations, but if they are not, the
source of the differences may be variables which, while not
directly involved in the process of social interaction, are
intimately related to it.

The literature has identified four major situational
factors that impinge upon social interaction: (1) cultural-
ly and/or biologically determined rules of procedure
(Speier, 1973; Malone, 1975; and Ittelson, gt_§l., 1970);

(2) psychological and physical characteristics of the inter-
actants (Hackman and Morris, 1975; Malone, 1975; Bales, 1950;
and Ittelson, gE_al., 1970); (3) the setting within which

the interaction occurs (Malone, 1975; Bales, 1950; Hackman

and Morris, 1975; Speier, 1973; and Ittelson, gt_§l., 1970);
(4) the context of the interaction, including its historical
antecedents, the purposes of the interactants, and the tasks

they are performing (Hackman and Morris, 1975; Malone, 1975;

l8
Lennard and Bernstein, 1969; and Speier, 1973).

In this section the possible effects of these fac-
tors on the model will be examined, with particular atten-
tion being given to their effects in the three situations
within which the model will be tested.

The three situations in which the model will be
tested-~radio, television, and typica1--primarily involve
"aimless" conversation in a familiar setting. Little is
known about the nature of interaction in the radio and typ-
ical situations, but a limited amount of research has been
conducted on interaction that is embedded in the television
situation. Forty percent of respondents in a national
survey (LoSciuto, 1972) reported interacting during partic-
ular television programs. In the same survey it was report-
ed that 32% of programs were watched alone; 43% in the com-
pany of one person; and 25% of the programs were watched
with more than one other person. This suggests that social
interaction in the presence of television occurs quite
frequently.

There is some suggestive evidence that rules deter-
mine the patterns of interaction in television situations.
For example, Maccoby (1951), in an early study, found that
58% of all respondents reported that they talked very little
or not at all when in the presence of television; 20% re-
ported that conversations occurred at certain specified

times (such as during commercials); 11% said that their

l9

conversations were limited to comments about the program-
ming; and only 11% reported that they talked quite a bit.
Social interaction in the presence of television tends to
be discontinuous, with more or less formal rules covering
such manifestations of interaction as content and communica-
tion (Johnson, 1976). The timing of communication is often
dependent on TV, conversation occurring mainly during com-
mercials or lulls in the show (Johnson, 1976). Social
interaction in the presence of radio tends to be very
similar to social interaction the typical situation. Dif-
ferences in rules would lead us to expect that the nature
of the relationships between communication and content and
the other categories would be somewhat different in the
television situation than it is in the other two situations.

Certain individual characteristics such as predis-
positions to attend to the media, tolerance to distractions
in the environment, and capacity to handle complex stimulus
inputs all would appear to have an affect on particular
social interactions in the presence of involving media.
The effect of television on these factors should be much
more pronounced, causing relationships between selection
and the other elements to be more unstable in the television
situation.

The context of the interaction is similar in all
three situations. The interactants are generally engaged

in sociable "aimless" conversation. In their national

20
surveys in 1960 and in 1970 Steiner (1963) and Bower (1973)
asked respondents how often they watched television to be
sociable with other people. In 1960, 17% reported that
they usually watched television to be sociable; 32% report-
ed that they occasionally watched to be sociable; 27%
reported they rarely watched to be sociable; and 24% report-
ed that they never watched television to be sociable. The
results for 1970 are very similar with 15% reporting usually,
29% reporting occasionally, 30% reporting rarely, and 26%
reporting never. For families the sociability motive would
appear to be especially salient. Television is often seen
by family members as an excuse to be with the family and
as a stimulus for interaction among family members (Lyle,
1972).

Television appears to frequently create a situation
of pseudo-interaction. While television increases the amount
of time that families spend together, "it appears that the
increased family contact brought about by television is not
social except in the most limited sense: that of being in
the same room with other people" (Maccoby, 1951, p. 424).
Other authors have also noted that while television appears
to bring the family together, it really doesn't enhance the
level or amount of their social interaction (Coffin, 1955;
Robinson, 1972; and Hamilton and Lawless, 1956).

This "aimless" interaction might act to reduce the

strength of the causal relationships between the elements of

21
the model. This is especially true in the case of relation-
ships. In contrast to a situation that demands a rigid role
structure such as the completion of a task, the roles in
these situations are rather diffuse. The situation them-
selves prescribe no clear, salient roles for the interact-
ants. The roles that they bring to the situation are carry-
overs from larger contexts, such as familial roles, and they
may not be salient in this situation.

Except for the presence of differing media or no
media the setting of these situations should be essentially
the same. Differences in settings, though slight, might
introduce more random error in tests of the model. Previous
research has indicated that there is a greater variety of
settings within which typical interaction can occur; radio
embedded interaction offers a slightly narrower range than
a no media setting, and television embedded interaction
appears to occur in only a limited range of settings (John-
son, 1976). However, these differences in variability are
inversely related to the potential effects of the media in
these situations on interaction. Thus it would be expected
that, even when differences in setting are taken into ac-
count, the television situation should be the most unstable,
followed by radio, and then by the typical situation.

Television appears to provide topics of conversations
or the content for much of the interaction that takes place

in its presence. Robinson (1972) reported that one half of

22
the conversations in the presence of television refer to the
content of the show.* Lyle (1972) reported that television
provides the interactants with topics of conversation.

In this section four major situational factors--ru1es,
context, setting, and individual differences--that might
affect the relationships in the model of social interaction
were identified and the nature of their effects were dis-
cussed for the situations in which the model will be tested.
It is assumed that any differences caused by these factors
are expected to be differences in degree and not in kind
for the relationships in the model across the three situa-

tions.

Conclusion

 

In this chapter a model of social interaction which
detailed the causal relationships among six categories of
acts composing social interaction was presented. All of the
categories--content, interpretation, emotion, communication,
selection and relationships--have been used in the literature
to describe social interaction. The chapter concluded with
a discussion of the effects of exogenous variables on the
process of social interaction in non-media and media

situations.

 

*
Results are from self-report data.

CHAPTER II

METHODS

This chapter describes the nature of the 124 mail
questionnaires returned from a simple random sample of adults
enumerated in the Grand Rapids, Michigan telephone book that
will be used to test the model prOposed in Chapter I. Spe-
cifically the following matters will be discussed in this
chapter: the rationale for selecting a mail questionnaire;
the response rate; effects of non-response on relationships
between variables; characteristics of the sample and the mail

questionnaire.

Rationale for Selecting a Mail Questionnaire

 

Several different techniques could be used to gather
data to test the model proposed in Chapter I. Each of these
techniques has its own peculiar advantages and disadvantages.
Self-administered questionnaires were used to gather the data
for this study. The advantages and disadvantages of self-
administered questionnaires vis a vis systematic observation,
the most acceptable alternative method, will be discussed in
this section.

Weick (1968) cites several disadvantages inherent in

the use of self-administered questionnaires in this context.

23

24
One, involvement in an activity may cause a respondent to be
unaware of crucial elements of the phenomenon. Two, when a
phenomenon constitutes a process between individuals the
relevance of individual reports is limited. Three, some phe-
nomenon may be too fleeting to be noticed by respondents. In
sum, the overall disadvantage of self-report techniques is
that they suffer from the limitations of respondents (Lennard
and Bernstein, 1969).

Self-report has several general advantages over system-
atic observation. One, it can provide information on a situa-
tion where it would be difficult for the researcher to obtain
access. Two, the participants in an event are more aware of
their aims and the aims of their fellows; thus they are better
able to interpret events in the interaction (Borgatta and
Crowther, 1965). Three, the very presence of an observer can
change the phenomenon that the observer is measuring (Crano
and Brewer, 1973). Four, self-report is generally less time

consuming for the researcher and the respondent.

Self-report was chosen as the data gathering technique
in this particular study because: (a) it allowed the research-
er to gather information from a wide range of respondents; (b)
social interaction in the particular contexts examined here
would probably be considerably changed by the presence of an
observer or videotape cameras; and (c) since the categories
are relatively global and abstract, respondent's reports
shouldn't be particulary affected by the problem of a fleeting

phenomenon.

25

Response Rate

 

One of the common problems associated with mail surveys
is a low response rate. This problem was exacerbated in this
survey because the task set for the respondents was a diffi-
cult one, because the questionnaire wouldn't apply to every
respondent in the survey, and because a substantial segment
of the U.S. population, 32%, don't watch TV in the presence
of another person (Lo Sciato, 1972) .

As a result of these factors every possible step was
taken to insure a high response rate. Regrettably the re-
searcher didn't have enough funds to provide each respondent
with some renumeration for completing the survey. However,
several other steps were taken to motivate respondents to
complete the questionnaire. The most important step was to
make a personal telephone call (the phone call transcript is
contained in Appendix C) to each respondent to get the re—
spondent to agree to complete the questionnaire before it was
mailed to them. A number of appeals were incorporated in the
telephone call to increase the respondents willingness to
participate in the survey (i.e., organizational affiliation,
advancement of science, the usefulness of the results, etc.).
An additional set of appeals were contained in the cover
letter attached to the mail questionnaire (i.e., organiza-
tional affiliation, advancement of science, utility of the
information, enhancement of respondent knowledge of self, and

importance of results for society). Finally, respondents

26

were called back at least once, both to encourage them to
return the questionnaire and to try to rectify any problems
respondents were having with the questionnaire.

Given the difficulty of the questionnaire, and the
probability that a substantial proportion of the population
couldn't respond to it, there was a reasonable response to
the mail questionnaire1 (see Table 1). The n of the survey
to start was 653. Of these, 108 (or 16.7%) were bad cases
(i.e., wrong numbers, dead, left town, etc.). This left an
n of 555. Of these, 481 were reached with the initial phone
call, 199 of which said they did not wish to complete the
questionnaire. Of the 282 respondents that agreed to receive
the questionnaire, 119, or 42.2%, eventually returned it.
Some respondents were sent the questionnaires without a
phone call. Only 5 of these 58 respondents returned the

questionnaire. The overall response rate was 22.3%.

 

1Some of the respondents, either on the phone or in
comments attached to uncompleted returned questionnaires,
indicated the reasons why they wouldn't complete the ques-
tionnaire. There were four frequently mentioned reasons for
failure to complete the questionnaire: The questionnaire
was too complicated for some respondents (7). Some of the
respondents didn't understand the questionnaire (4). Before
mailing out the questionnaire it was expected that some
respondents would fail to complete the questionnaire for
these reasons. These were not the most frequently cited
reasons for failure to complete the questionnaire, however.
To the researcher's surprise 11 respondents refused to re-
spond because they felt the questionnaire was too personal.
And, as was expected, some respondents (10) said that the
questionnaire didn't apply to them.

27

Table 1

Response to Mail Questionnaire

 

 

 

 

 

 

Returned Questionnaire

Condition Questionnaire NOt Returned = Row Total
Initial Call 55 199 254
Row % 21.6% 78.3%

Initial Call and

Callbacks 64 163 227
Row % 28.2% 71.8%
Ehmmemy'of Responses

with Initial Call 119 362 481
Row % 24.7% 75.2%
No Initial Call, No

Second Call, but Mailed 5 58 63
Row % 7.9% 92.1%
No Initial Call, but

Call After Mailed ll 11
Row % 100%

Summary Row for All

Responses 124 431 555
Row % 22.3% 77.7%

 

28
Effects of Nonresponse on the Relationships Between Variables

The rate of the return to this questionnaire raises
some question about the results reported here. The precise
estimate of population parameters from this data would be ex-
tremely difficult. The central concern of this study, how-
ever, is an exploratory examination of the relationships
among the variables constituting social interaction, not the
identification of the exact values of the parameters in a
given population.

What is the effect of nonresponse on the relationships
among variables? A number of studies have examined this
question for bivariate relationships in mailed self-report
questionnaires. Goudy (1956) found, in a relatively homo-
geneous population, that there were relatively slight differ-
ences in the strength of relationships between late and early
respondents to a mail questionnaire. Kivlin (1965), in a
study that compared respondents and nonrespondents by com-
pleting the total sample, found that there was no differences
in the relationships among variables between respondents and
nonrespondents. Lehman (1963), in a comparison of early and
late return, found that there was no difference in the rela-
tionship between variables. Schwirian and Blaine (1967) found
that while two of the questions they examined showed no dif-
ferences between early and late responses in the relationships
among certain variables, that for the third question they

examined there were significant differences, but this question

29

was tied to a characteristic of respondents that strongly
differentiated respondents from nonrespondents. Suchman
(1972), in an examination of numerous relationships, found
that these relationships remained unchanged between respon-
dents and nonrespondents. He went on to note that "these
data do appear strongly suggestive of a current overemphasis
upon 100 per cent response to a survey when one is concerned
with relationships rather than the description of frequency
distributions." These studies suggest that even when non-
response is a problem in a survey, the relationships among
the variables studied remain stable, and, at least for the
overwhelming preponderance of studies, nonresponse doesn't
significantly change the relationships among the variables
studied. However, the data reported here still should be
treated as only suggestive of the actual relationships among

the variables in the popualtion.

Characteristics of the Sample

 

Table 2 compares various demographic characteristics
of the sample with the distribution of these characteristics
within the Grand Rapids area and within the United States.
The questions used to gather this information are on page 8
of the mail questionnaire. The census figures for Grand
Rapids and the United States are for 1971 (Bureau of the
Census, 1973 and Bureau of Census, 1972) while the sample was
drawn in 1976, as a result slight discrepancies between the

sample and the 1971 figures can be expected. The sample had

30
Table 2

Characteristics of the Sample, of Grand Rapids
and of the united States

Characteristic Sample Grand Rapids" United States+
Sex
Male 42.5 47.8 48.4
Female 57.5 52.2 51.6
Race
White 97.5 92.9 86.7
Bladk 2.5 6.5 11.0
Other 0. 0 . 6 1. 4
Age
13+ 89.3 63.1 65.4
65+ 10.7 9.5 9.8
Average age 43.2 29.5 33.4
School (Years)
less than 5 . 8 2. 5 5.5
6-12 42.1 55.4 32.6
13-16 43.0 31.5 49.5
16+ 14.0 10.6 12.5
Income***
less than $3000 8.1 6.7 **
$3000-$4999 7.0 7.0
$5000-$6999 9.3 9.2
$7000-$9999 10.5 21.5
$10,000-$14,999 24.4 32.4
$15,000-$24,999 27.9 18.1
$25,000+ 12.8 5.1
Bkmtem'of Persons
in Hbusehold 2 8 3.2
Marital Status
Single 9.2 ** 16.2
.Married 81.7 ** 71.6
Divorced 2.5 ** 3.2

 

*Bureau.of the Census, 1973
*
Statistics Unavailable

**
Family
Bureau.of the census, 1972

*

*

31

a greater proportion of females and of whites, than the
United States and Grand Rapids populations. Since the ques-
tionnaire was not intended to be filled out by children it

is only natural that the average age of respondents is
greater than the U.S. or Grand Rapids average age, but the
percentage of respondents over 65 is quite close to the U.S.
and Grand Rapids figures. While there were differences in
the education of the sample compared to Grand Rapids, the
sample, in general, reflected the relative proportions of the
population contained in the various strata. While all

family income groups were represented in the sample and
groups with incomes under $7000 were essentially the same as
the Grand Rapids figures, the middle income groups from $7000
to $14,999 were underrepresented and the higher income
groups, $15,000 +, were overrepresented. There were fewer
single people and more married people contained in the sample.
However, married people might be more likely to return the
questionnaire, since they are probably more likely to watch
television with another person.

Several other demographic questions were asked in the
questionnaire. Respondent reported that they owned an
average of 3.7 radios and 1.9 televisions. Substantial
numbers, 73.3% of the respondents reported owning a color
television set. The respondents indicated that they felt
they belonged to the following social classes: upper class,

2.7%; middle class, 63.7%, working class, 35.1%; and no

32
respondents reported that they belonged to the lower class;
60.8% of the respondents reported that they were employed.
The respondents worked 49.6 hours a week on the average and

reported that they had 30.2 hours a week of free time.

The Mail Questionnaire

 

The mail questionnaire (see Appendix D) contained a
cover letter and 7 pages of questions. During a pretest it
took respondents 30 to 40 minutes to complete the question-
naire.

The respondents received the questionnaire in a hand
addressed envelope. A return, stamped, preaddressed envelope
was enclosed.

Before the questionnaire was sent to the respondents
it was pretested on several adult respondents. These respon-
dents were selected at random from the Lansing metropolitan
telephone book and the questionnaire was administered in
their homes. These sessions were taped recorded and respon-
dents were encouraged to make comments while filling out the
questionnaire. They were also asked several questions con-
cerning the questionnaire after they had completed it. As
a result of these pretests more detailed instructions were
included in the final questionnaire, the strength of the
appeals were increased, simpler wording was used, the order
of some of the questions was changed, and some synonyms were
substituted for certain difficult words. On the whole the

respondents in the pretest demonstrated that with the proper

33
effort they could answer the kinds of questions that were
used in the questionnaire.

In general, the questions were arranged in the ques-
tionnaire so that the easiest questions concerning the phenom-
enon of interest were at the beginning. This was done to get
the respondents involved in filling out the questionnaire be-
fore he came to the most difficult and potentially frustra-
ing questions. Personal questions came at the end of the
questionnaire. The most descriptive and concrete questions
also came at the beginning, with the questions getting pro-
gressively more theoretical and abstract as the questionnaire
continued.

Pages 2, 3, and 4 of the questionnaire contain ques—
tions relating to specific hypotheses, rates of behavior of
respondents, the respondents' evaluation of certain situa-
tions, the respondents' reactions to certain types of conver-
sations, the perceived effects of the media on the respondents'
conversations, respondents' degree of involvement in certain
situations, and the respondents' amount of attention to
certain features in the media situation.

Pages 5 and 6 contain questions 21 and 22* which test
the model of social interaction. Questions 21 and 22 are
designed to reveal the processual, variable nature of the

elements of the model of social interaction through the use

 

*
All respondents were instructed to complete both
questions 21 and 22.

34

of cross sectional questions. They do this in two ways.
First, they determine the degree to which particular elements
of social interaction change. Measuring elements in terms
of degree of change (or in terms of the degree to which they
reflect a particular property) permits the assessment of

the elements of social interaction as processual variables
in a way that the measurement of their presence or absence
does not. It allows the elements to be treated truly as
variables, instead of categories of acts, which is the tra-
ditional method of studying the elements within social
interaction.*

Questions 21 and 22 examine social interaction in three
situations--typical, radio, and television. These particular
situations were selected for the tests of the model for a
number of differing reasons. First, pragmatically, the
National Association of Broadcasters was willing to fund re-
search on social interaction in these settings. Second, re-
search has indicated that social interaction in the presence
of television may be so affected by the medium that it could
constitute a state of quasi-interaction (Maccoby, 1951) that
potentially could produce harmful effects. Thus comparing

the television situation with interaction in other situations

 

*

Once the extent to which a particular element changes
is determined, causality can be assessed by the multivariate
analysis of the resulting variables. This multivariate
analysis assesses the extent to which the change reported in
any variable is attributable to another variable when other
variables are controlled.

35

could serve as a preliminary indication of the nature of the
harmful effect, if any, of television on social interaction.
Three, these situations constitute a continuum of situational
impacts. Research has indicated that the effects of tele-
vision are pronounced, the effects of radio somewhat less so,
on social interaction (Johnson, 1976). Thus these three
situations permit an examination of the affect of various
levels of intensity of situational factors on social inter-
action. Fourth, these three situations are quite common and,
as a result, provide information on a substantial number of
interactions that occur society wide. Respondents were given
the following general instructions to be followed in answer-
ing each of the three sets of questions:

The following directions should be followed

when answering these questions. If zero (0)

is the complete absence of apparticular

property (e.g. importance, attention, change

over time, or effects) and 100 is the typical

or average amount of that property that

exists in a normal conversation, then rate

each part of the conversation for that prop-

erty. For example, if you feel a particular

part, say content, is nearly twice as important

as the typical part of a conversation, then put

down 195. If a part is nearly one half as

important, then put down 48. You can use any
number you wish.

 

 

 

 

Question 21 attempted to determine the relationships
between the elements of social interaction in a typical sit-
uation:

Parts of a conversation change over time.

How much do each of these parts change in
a typical conversation?

 

36

Question 22 attempted to determine nature of social
interaction in radio and television situations.

Parts of a conversation often change because
of the effects of things in the situation
upon them. How much are the parts of your
typical conversation affected by the presence
of: A television playing? A radio playing?

 

 

 

The following elements of social interaction and brief
descriptions of them followed Questions 21 and Question 22:
CONTENT: The things that you talk about.

ATTENTION:* How much you concentrate on any one thing.

EMOTION: Your level of emotional feelings, such as love,
fear, determination, etc.

RESPONSIVENESS:+ The degree to which you feel others
respond to what you say, are actually listening
to what you say.

CLOSENESS (INTIMACY):+ The degree to which you feel close
or distant to others in this situation.

PROCESS OF CONVERSING:#

Amount of conversation, pauses in
conversation, interruptions and ease of conversa-
tion.

MEANING:¢ Your understanding of the situation, your under-

standing of others, your understanding of what

 

*
Attention is an indicant of selection.

+ . . .
ResponSlveness, closeness, and roles are indicants
of relationships.

#

Process of conversing and conversation Options are
indicants of communication.

¢Meaning is an indicant of interpretation.

37
is said.
ROLES:+ Knowledge of who you are in terms of labels like
friend, mother, spouse, etc.

#

CONVERSATION OPTIONS: When to listen, when to speak, what

to listen to, and with whom to speak.

Conclusion

 

In this chapter the reasons why a mail questionnaire
was chosen over other data-gathering techniques were cited;
the response rate was reported; the effects of non-response
on relationships between variables was discussed; character-
istics of the sample were presented; and the nature of the
questionnaire was examined. The following chapter will use
the data elicited by these methods to test the model of social

interaction presented in Chapter I.

+ . . .
ResponSlveness, closeness, and roles are indicants of
relationships.

#Process of conversing and conversation options are

indicants of communication.

CHAPTER III

RESULTS OF TESTS OF THE MODEL

In this chapter the results of the tests of the
models of social interaction proposed in Chapter I will be
reported. First the descriptive statistics and correlations
for each of the models will be presented, then ordinary
least squares (OLS) multiple regression will be used to
assess the variance accounted for by the paths contained in
the models. Next the goodness of fit of the model to the
data will be determined by LISREL, a computer program that

estimates the parameters in a system of linear equations.

Means and Standard Deviations

 

The means and standard deviations for the variables
in the typical situation are contained in Table 3.* The
means for the variables range from a low of 83.23 for the
role indicant of relationship to a high of 116.23 for inter-
pretation. The standard deviations either center around 50,
as is the case for content, emotion, the process of convers-

ing, selection and responsiveness, or are greater than 90,

 

*
The skewness and kurtosis of the variables in all of
the models reveal that they do not depart markedly from a

normal distribution.
38

39

Table 3

Means and Standard Deviations for Variables

in Typical Situation

Variables

content

Interpretation

Emotion

communication:
Process of Oonversing

Conversation Options

Selection

Relationships:
Role
Closeness

Responsiveness

N=79

means

99.27

116.23

102.90

90.22

111.65

97.85

83.23
96.52

99.63

Standard Deviations

49.14

131.36

52.75

62.76

122.60

43.89

106.22
92.71

53.91

40
which is the case for interpretation, conversation options,
roles, and closeness.

The means and standard deviations for the variables in
the television situation are contained in Table 4. The
means here showed slightly greater dispersion than the means
for the typical situation ranging from 76.01 for roles to
162.00 for process of conversing. The standard deviations
are always greater than 60.87, with the standard deviation
for process of conversing being exceptionally high, 515.5.

The means and standard deviations for the variables in
the radio situation are contained in Table 5. The means, as
has been reported elsewhere (Johnson, 1976), for this situ-
ation indicate that radio doesn't have as great an effect
on the elements of social interaction as television does.
The values of the means range from 71.73 for role to 97.65
for conversation options. Standard deviations, save for the
standard deviation for conversation options (122.53) range

from 50.13 for responsiveness to 69.83 for interpretation.

Correlation Matrices

 

The correlation matrices reported here are used in three
ways. One, they assess the simple level of association be—
tween any two variables contained in model. Two, the corre-
lation matrices are the input into the LISREL program.

Three, the correlation matrices and calculations associated
with them are used as a diagnostic tool in determining short-

comings in a model tested by means of LISREL.

Means and Standard Deviations for
Variables in TV Situation

Variables

Content

Interpretation

Emotion

Communication :
Process of Cbnversing

Conversation Cptions

Selection

Relationships :
Ibles
Closeness

Responsiveness

41

Table 4

Means

107.20

103.87

89.75

162. 00
109. 92

119. 69

76.01
87.42

86.67

Standard Deviations

71.57

89.97

60. 87

515. 58

128. 54

95.73

70.14
71.30

71.19

Means and Standard Deviations for
Variables in Radio Situation

Variables
(bntent
Interpretation
Ehotion

Communication :
Process of Cbnversing

Conversation Options
Selection

Relationships :
Role
Closeness

Responsiveness

42

Table 5

Means

80.08

80.07

77.93

73.75

97.65

79. 72

71.73
76.68

75.38

Standard Deviations

50.75

69.63

55.24

53. 92
122 .53

54.99

58.13
57.93

50.13

43

The correlation matrix for the elements of social
interaction for a typical situation are contained in Table 6.*
The correlations reported in this matrix range from .0121
for content-role to .8958 for closeness-conversation options.
The correlations for meaning-conversation options, closeness-
meaning, conversation options-role, and closeness-role are
all greater than .8. Most of the correlations fall in a
range from .4 to .6.

The correlation matrix for the elements of social
interaction contained in the television situation are con-
tained in Table 7. The correlations in this matrix range
from -.0603 for role and process of conversation to .6868
for closeness and conversation options. Most of the correla-
tions fall in a range from .3 to .6.

The correlation matrix for the radio situation is con-
tained in Table 8. The correlations range from .2790 for
content-conversation options to .8237 for interpretation-
process of conversation. Most of the correlations fall in a

range from .45 to .7.

Multiple Regression Results

 

In this section OLS multiple regressions associated

with each dependent variable will be reported. These

 

*

The scattergrams for bivariate relationships among
all of the indicants in all of the situations indicate that
there are no substantial departures from linearity.

44

oooo.H

c0396

on n z

onmm. emmo. mmom. mmem. meme. mmmm. memm. some. compose
oooo.e oooo. mmmm. emmm. meom. some. ammo. mob». team
oooo.e mmmm. memo. mamm. omoo. mmme. ommm. mmoam>mmooomom
oooo.m mamo. mmmm. mosm. mmme. mmom. homeomomo
oooo.e mono. mamm. mmmm. memo. oaamhmoaoo
. Mo .3803
oooo.e mmmm. meeo. mmmm. whomuoo
Sflagca
oooo.m emee. meom. oomuamoae
oooo.m emom. oompaoo
884 ofiaomz

 

38 $9828 mmocomoau mfiwug 9638 combbmaom ucwucoo mama—mm:
Icommmm Isa mo coHuom
mmmoonm 1.85:8
southpaw Enema. 5 moanmmumkw comuomnmufi Hmm00m mg mcomumeHHoO cOmHmom

m manna

45

Ruz
oomo.e memo. memo. Homo. mmmo. meme. momm. mmmm. eomo. compose
oooo.e mmmm. mmee. mooo. omom. momm. Reem. mmem. omom
oooo.e emmo. omem. ooem. mmmm. mmme. omoo. mmoaoammaoommm
oooo.e mmmm. momo. eemo. mmoe. ommo. mmoawooeu
oooo.m moeo. emmm. mmmm. emom. oaammmoooo
mo "“80on
oooo.m mmoe. mmmm. mmom. maoeuoo
coflmmHQSO
oooo.m mmem. oeoo. aomoaouoe
oooo.e oeee. uaooooo
oooo.e ooaaomz

 

c0305 mag mmmcmSmm mmocwmofiu momma? mcoHu& cowucmuufi ucwucov mg
. Icommmm IGOU mo couflcm
mmoooum lumpcoo
cowumsumw acmmgwamn. 5 moanmmnmmw gggwucH Homoom mag wQOHuMHwHHOU comHmmm

N. Canon.

46

0000 .H

combed

mm n z
mome. oomm. emmo. mmoo. Homo. Smo. emeo. Zoe. 8395
884 mmee. oooo. emoo. mmmm. emmm. omme. Homo. 38
884 Boo. mooo. mmmo. ommo. momo. momo. moofiommooommm

884 oomo. mmmo. ammo. omme. eomm. 88808

884 mome. mmoo. memo. mmmm. mfimmmaaoo

mo mmmoonm

884 mmmo. mmmm. mmmo. m838

coflowumpcoo

884 emme. mmoe. 83884

884 momo. pooped

884 mamoomz

 

0H8 mmocgom mmdfimoao 93.985 9538 8303mm ucoucoo 92.3mm:
Icoouom :80 mo c033
980on L980
cooumouom 0.3mm cm mmanmoug coflbmmoufi Rocco 0:05 mcooumaomfioo comumom

m manna.

47
multiple regressions will be used as indicators of the sig-
nificance of individual paths in the models. They will also
provide information about how much of the variability in the
dependent variables in the model is explained by its assoc—
iated independent variables. The OLS multiple regressions
reported here will not be used to estimate values of para-
meters contained in the models. LISREL, by controlling for
many of the problems associated with estimation of parameters,
is a much more appropriate technique for this purpose.

The ordinary least square multiple regressions for
paths containedin.Model I, in the TV situation are reported in
Table 9. The multiple regressions for content, conversation
options, interpretation, and selection are all significant
at the .01 level. The process of conversing regression while
approaching significance (.06) is not significant at the .05
level. Save for process of conversing these multiple re-
gressions account for at least 20% of the variation in
their dependent variables. The multiple regressions for the
interpretation and selection dependent variables account for
more than 50% of the variation in these variables.

The alternative paths, discussed in Chapter I, in Model
II for relationships between the content and emotion, and
relationships and communication variables are reported in
Table 10. All of these multiple regressions are significant
at the .01 level and they account for at least 24% of the

variation in their dependent variables.

48

mm u o .9335?

 

 

 

 

 

 

mo. om.mm mm. mm. oe. compoumhououom
mo. mm.o em. em. mm. Ammoow>mmooommmc
om. mo.m mm. mm. em.u iomomv
em. oo.e em. em. mm. immocooomoc
oEmooflomom
mm. mm. om. om. mo.- compose
mo. em.mm mo. :oeoommom
mo. om.o mm. mm. mm. Ammocm>mmooooomc
mo. eo.o mm. om. om. Ammomc
mo. om.om mm. oe. om. Ammocwoomoc
monocomuoemm
em. om.m em. mm. em. compose
mm. em.om oo. compoumhohmoom
eo. mo.e om. mm. om. aoeuoomoo
mo. mm.e mm. oe. om. oomuaoo
mo. em.om om. locomooo
cooumammwcoov
c038;
mm. om.m mo. mm.m mm. aomuoomom
em. em. mm. mm. mm. oomoaoo
mo . mm . m mo . 353958
No mmooonmv
83805.80
mm. mm. oe.om mm. mm. me. oomuoumuohouom
8 . we . om acoucoo
883388 mmBES mmoBBEEEo mad m 38 28R§ EQRS
mmahooo . o ohmooooo oamooomooam autonomoo

coflgumm PH. c..m H ammonia 5 mficm How mcoommwuomm maniac:

m magma

49

mm u c .m>muomoaoo.

 

 

 

 

mo. em.om mm. mm. mm. momoooz
oo. mo.m mm. em. mm. compose
mm. oe.om om. oomucoo
oo. om.m mm. mo. em. Ammooomomoc
oo. mm. mm. mm. oo. phoneme
mo. om.mm mm. mm. oe. Ammoom>mmgoomomo
me. mo. om. om.a mo.u :omooomoo
mm. em. 2. em.. 8.. 83:
$838
Sou—mammfioov
S. omem oe. coflmoofiEho
mo. mo.o mm. mo.m om. Ammmowmomoc
mo. mm. mm. mm.: oo.: powoooo
mo. mm.o om. em.m mm. mmoomommcoommm
om. mm. mm. oo.- mo.u coohoomom
mo. mm.m om. mm.mu mm.u Ammomc
3:399:00
mo mmcoommmv
mo. mm.o em. 838558
methamomommm m metamoo mm moooommmommm momma m whom omnmmhm> omnmmum>
3890 a m 8885 €888€H ufiEHS

scour-3mm >9 cum HH .3on 5 93mm gougmuz How 903% camouflage
OH canoe

50

The OLS multiple regressions for the paths in Model I
in the typical situation are contained in Table 11. The
content regression is not significant and accounts for essen-
tially none of its variance. Conversation options, while
approaching significance (.07) is not significant at the .05
level. The regressions for the dependent variables of inter-
pretation, selection and process of conversing are each sig-
nificant at the .01 level. The variables in the equations
for process of conversing and for conversation Options ac—
count for small proportions of the variance in these variables.
Substantial proportions of the variance in selection (33%)
and interpretation (75%) are explained by their independent
variables.

The values of the alternative paths in Model II in
the typical situation (Table 12) are all significant at the
.01 level. The variables in the content equation account
for a moderate amount of the variance (26%) in this variable.
Substantial proportions of the variance in the process of
conversing (65%) and the conversation options (87%) are ex-
plained for by their independent variables.

The ordinary least square multiple regressions for
paths contained in Model II in the radio situation are pre-
sented in Table 13. All of the multiple regressions for
Model I are significant at the .01 level. All of these mul-
tiple regressions account for substantial percentages of the

variation in the dependent variables with a minimum of 30%

51

mm u a .33328.

 

 

 

 

 

 

mo. om. 3. oo... 3.- m38ooU
mm. om. oo. oo.- m3... mfi83
8 . mo . m S . mm . mm . "1883383
3. 83 mo. 3. om. 38
8. mm.o3 3. om. me. 8395
.8. om.m mm. genome?“
8. oo.mm S. 8.3 om. E38838
me. o. m3. m3. oo. 83838885
mo. om.m e3. em. mm. 39:
33.590333
om. o3. om. oo.. 8.. 839m
mo. mm.oo om. 8333333
mo . mm . o mm . om . mm . 83633
3. $3 mm. mm... o3- 383:8 $838
coflmagcoov
mm. oo.m 8. 8383888
8. m3.m 3. me. mm. 83633
mm. o3. o3. 3o. 8. 33:8 633380
mo mmooomnd
8. one 3. 83335580
mm . m3 . eo . o3 . eo . 8333333
mo. NH. oo. ucoucoo
3803383 m 3330 mm 883383 893 m 33 383.3, 383.3,
3330 a .3 33m 33335 38838

coflmsuom Havana. cum H Hmong 5” mfimm How 20333593 came»?
.3 dance

mm. H a .gflmdafiog

 

 

 

 

3. 33 3. 8... 3.- 8333335
. 8. 3.3 3. 3. mm. 8393
S 8.3 mm. 383:8
Hoe. hv.o¢ 33. on. mm. mmmcmefiu
8. mad 3. «a. mo. 83838
3. 33 3. 3. 8. 3393883
w mo. 36 3. mm... 3.- p.380
38. 3.3 S. 3. 3.. 30m Agog
COHHMWHQGBV
Ho . mm . mm 5 . 8330550
8. 8.8 8. 3. mm. 88888
S. 3. 3. mo. 8. 83838
8. 22mm 2. cm. 2. 333883
«m. 3. 3. mo... 8.- 8.83:8
8 . 8. mo. 8. 8. 38 333980
mo $8098
8. 3.3 mm. 83339528
85033.55 m H3390 mm mongflmgflm Mg m mumm $33.33, manmwbg
3396 h. .3 33m ”838883 38888

coflumsuflw Havana. 5 HH H89; 5 mfimm gaugwuz How scammmHmmm @3332

NH manna.

53

S u c .gflumgg

 

 

 

 

 

 

3. mm . 2. S. d . $89333“
mam. SA 2. S . I. mficmmz
mm . 8 . 2 . .2 . 2 . wwwcmmod
v3. 8. 2. mm. 8. .32
8o. 85 2. mm. mm. 8395
:5. Si 2‘. 838%.“
mm. cm. 2. 3. mo. Ammméflmcommmmv
8. mm.m~ S. S. E. Ammmawmodv
8 . fl . 3 2 . am . mm . ESE
mmfinmcoflmaa
mm. 8.. 2. 8.- 8... 839m
8. mm.om 8. 833855
8. £3 om. SA mp. coflfimflmm
mm. 34 mm. 3.. 3... E380 A2035
coflmmugcav
8. 22mm 3. 5385550
8. 85 3. mm. om. 8302mm
8. 3.3 S. 3. mm . 23:8 Bfiflggoo
mo mmmoonmv
8. Efi mm. 838%
8 . .2 . mm B . S. . «m . coflfimumumufi
8. 26m om. 23:8
mofioflflcmflm m 2896 m 8§oflflcmflm “Bum m 38 fining flange,
H390 A m 3ng ucmccwmmofi E8598

coflmsuflm 04.6mm 5 H H894 5 mfimm How mCmemmHmmm mamwpadz
m H manna

54
for content and a maximum of 60% for interpretation accounted
for by their independent variables.

The alternative paths for Model II in the radio situa-
tion are contained in Table 14. All of these multiple re-
gressions are significant at the .01 level. All of these
paths account for 50% or more of the variation in their

dependent variables.

Introduction to Model Testing

In the coming sections the models of social interaction
developed in Chapter I will be tested and refined by path
analysis (or more properly the estimation of the system of
linear equations which the models represent). An integral
part of path analysis is the post hoc refinement of a model
(Land, 1969).

"Path analysis . . . is a technique sometimes used to
assess the direct contribution of one variable to another in
a nonexperimental situation" (Joreskog, 1970, p. 248).

Path analysis attempts to estimate by means of a number of
possible statistical techniques, the parameters of a system
of linear structural equations which represent the model of

a process proposed by a researcher (Joreskog, 1970). In this
paper the parameters in the model will be estimated by means
of a maximum likelihood statistical technique contained in
the computer program, LISREL, developed by Joreskog and

Van Thillo (1972). One of the advantages of path analysis

gucfififiEEP

 

 

55

 

 

Ho. mm.oa am. we.: mm.u ucmugpu
Ho. om.m am. am. mm. mmmcmmoao
mo. m>.¢ mm. mm. mm. mmwcm>flmcommmm
Ho. mo.mm om. m~.H mm. :ofiuomawm
«m. ms.H ma. mm. HH. mama
Amccauao
:oflmagcav
8. 2.3 S. 838ng
mo. ~m.¢ Ha. «m. mm. ucwugou
Ho. qo.m~ mo. as. om. mmmcwmoao
Ho. mm. NH. oo.- mo.- mmmcm>flmcoammm
cm. mo.a mo. mo. mo. :oauomamm
no. mm.m mo. ma. NH. mama Amcnmum>coo
mo mmmoound
8. om.mm mm. 838%
Ho. mo.¢m mo. we. om. acacmmz
Ho. mq.ma mo. mm. mm. seduces
Ho. oo.ms om. ucwucoo
mogflm Hcmflm m HHMHgO NM GUfiMOHMHCDHm .HOHHM m 3mm wHQMHHMNV maflmflnflw
Hamumpo . m cumcamum unaccwamucH ucmncwama

coflumsuwm 0.6mm 5 HH H092 cw 93mm gflmfimug How mcoflmmmfimflm mama»?
3 manna

56
is that it compells the researcher to make his assumptions
concerning causal structure explicit (Costner, 1971, and
Kerlinger and Pedhazur, 1973). The use of path analysis to
test the models proposed here is particularly appropriate
since "path analysis is an important analytical tool for
theory testing. Through its application one can determine
whether or not a pattern of correlations for a set of obser-
vations is consistent with a specific theoretical formulation"
(Kerlinger and Pedhazur, 1973, p. 317). The values of the
paths coefficients are the basis for a causal inference in
path analysis. "Coefficients different from zero suggest
the presence of causality and the size of the coefficient
suggests the degree of causality" (Heise, 1970). Path an—
alysis has been used in genetics, biometrics, and economics,
and in the 1960's was introduced in a meaningful way to so-
ciology and psychology (Costner, 1971).

In this chapter path analysis will be used to develop
the most appropriate model in the radio situation. This
refined model will then be tested in the other situations.
The radio situation was chosen for the initial analysis pri-
marily because research has indicated that the effects of
radio on social interaction lie somewhere in between the
effects of television and a nonmedia situation (Johnson,

1976).

57

Advantages of LISREL over Multiple Regression

 

Earlier in this chapter results from multiple regres-
sions were used to assess the amount of variance accounted
for and the significance level of individual paths in the
models. It was said that LISREL would be used to test the
model as a whole and to estimate individual parameters.
LISREL is best suited for this purpose because it has the
following advantages over multiple regression. One, LISREL
estimates of parameters have minimum asymptotic sampling
variability (Hauser and Goldberger, 1971, and Werts,

Joreskog and Linn, 1973). Two, traditional multiple regres-
sion estimates don't provide an estimate of the goodness of
fit of the entire model to the data (Joreskog, 1973). Three,
the various parameters in the model aren't estimated simul-
taneously in traditional multiple regression estimates
(Joreskog, 1970). Four, multiple regression was not devised
especially for the analysis of causal relations (Wright,
1921, and Goldberger, 1973). Five, the use of multiple indi-
cators for latent variables is not handled well by tradition-
al multiple regression estimates (Werts and Linn, 1970).

Six, in traditional multiple regression, when multiple indi-
cators are present, no estimate can be obtained of the rela-
tionships between latent variables. Seven, LISREL permits
the simultaneous specification of theoretical and measurement
relations (Fink, 1978). In sum, LISREL allows for the

"parsimonious estimation and evaluation of complex theoretical

58

systems" (Fink, 1977, p. 13).

Description of LISREL

LISREL is a general computer program for estimating a
linear structural equation system (such as those found in
path analysis) involving multiple indicators of unmeasured
variables. In general, LISREL divides the model to be tested
into three parts: a simultaneous linear equation model re—
lating true exogenous and endogenous variables: a measure-
ment model relating observed exogenous indicators to true
exogenous variables, and a measurement model relating ob-
served endogenous indicators to true endogenous variables
(Wiley, 1973). Here only a brief description, enough to
acquaint the reader with the main logic of LISREL and its
associated terminology will be given, several other sources
(Joreskog and Van Thillo, 1972; Joreskog, 1973; and Stein,
1976) provide a more complete description of the mathematical
underpinnings of the program and its general operation.

The Operations of LISREL are based on several types
of parameters which are used to construct eight matrices.
Following Stein (1976) the various components of the matrices
are listed below:

True endogenous variables, eta (n), of which there

are m.
True exogenous variables, xi (c), of which there
are n.

Paths from xi to eta are gamma (y).

59
Paths from eta to eta are alpha (a),capital A.2*
Paths from measured variables to or from latent
(xi or eta) variables are lambda (A), capital
A, and are called "scale factors" (Stein, 1976).
Reflective indicators of eta are y.
Reflective indicators of xi are x.
Errors associated with measurement of y are epsilon
(8).
Errors associated with measurement of x are delta
(5).

Variances associated with eta, are zeta (C).

These components are used to construct eight differ-
ent matrices.

The first of these matrices are Ax (lambda x) and
Ay (lambda y). They are composed of the scale
factors of the observed variables.

8 (Beta) is the matrix that contains the values of
the paths (a) between endogenous true variables.

I (Gamma) is the matrix that contains the values of
the paths (Y) between the true endogenous (y)

exogenous (n) variables.

 

60
@(Phi) is the variance-covariance matrix of
the exogenous variables.
w(Psi) is the variance-covariance matrix of
the residuals of the true endogenous
variables.

6 (Theta Delta) and O8 (Theta epsilon) are the

6
diagonal error standard deviations* of the
x and y observed variables respectively.

It is useful to classify these matrices by whether
they are associated with exogenous or endogenous variables
and whether the variables are observed or true and what their
sources of errors are. Exogenous variables are those vari-
ables which are not dependent for their variation upon
another variable in the system. Endogenous variables are

dependent upon other variables in the system for their vari-

ation (Land, 1973, and Van de Geer, 1971).

 

 

Error Error
Observed True Observed True
Variable Variable (Measurement) (Residual)
Endogenous y n e ;
Exogenous x g 6 -

 

These matrices form the elements of the reconstructed

variance—covariance matrix 2 which is used to assess the

 

*Throughout the dissertation the standard deviations
reported by LISREL are converted to variances.

61

goodness of fit of the model to the data. The parameters of
these matrices can be "of three kinds (i) fixed parameters,
that have been given assigned values, (ii) constrained para—
meters that are unknown, but equal to one of the given para-
meters,* and (iii) free parameters that are unknown and not
constrained to be equal to any other parameter" (J6reskog
and Van Thillo, 1972, p. 2). These matrices also serve as
the elements of structural equations that compose the par-
ticular model to be tested. Thus the equation for the de-

pendent the variables in this system is:
Bn==F-+g-+;

The equations for y and x the observed variables are:

y=u+AYn+s
x==v + A g + o
x

The following assumptions are made by the LISREL
program:
(1) It is assumed that c is uncorrelated with g.
(2) B is nonsingular.
(3) The errors of measurement (2, 6) are uncorre-
lated with the true variates (n, c) and
with each other (Joreskog and Van Thillo,

1972, p. 2).

 

*
None of the parameters used here are of this type.

62
LISREL estimates the parameters in the matrices by
minimization of the derivatives associated with the compon-
ents of the model. This minimization is accomplished by
the Davidon-Fletcher-Powell method by successive iterations
of the relevant matrix (Joreskog and Van Thillo, 1972).
This method is applied in successive iterations until a

criterion is reached.*

Meaning of the X2 Text for Goodness of Fit in LISREL

 

One of the advantages of LISREL over traditional
means of estimating path models is that it provides a test
of the overall fit of the model to the data. This test in-
volves a X2 statistic; the degrees of freedom of this x2 is
equal to the degree of overidentification in the model (Werts,
J6reskog, and Linn, 1973). "The X2 test is a test of the
specified model against the most general alternative that z
is any positive definite matrix" (Joreskog, 1974, p. 4).

The probability level associated with the LISREL X2 test "is

 

*
There are two options in the program, accurate and

approximate solutions. An accurate solution means that the
program iterates until the magnitude of all the changes in
derivatives is less than .00005. This solution is usually
correct to three significant digits. For an approximate
solution the iterations terminate when the decrease in func-
tion values is less than 5%. The two different solutions can
produce substantially different results (Joreskog and Van
Thillo, 1972). All solutions reported here are accurate
solutions. In addition all solutions come from the 1972
version of LISREL which has been followed by other more ele-
gant versions that include, among other features, estimates
of the standard errors associated with particular parameters.

63

defined as the probability of getting a X2 value larger than
that actually obtained, given that the hypothesized model is
true" (J6reskog and Van Thillo, 1972, p. 32). One of the
limitations of this test is that it assesses the general
adequacy of the model, but that it doesn't test the model
against any specific alternative (Mayer and Younger, 1975).
Joreskog (1974) has cautioned that values of x2 should be
interpreted cautiously for, when sufficiently large samples
are obtained,* almost any hypothesized model is untenable.
He asserts that the real usefulness of the x2 test comes in
determining the number of parameters in a model that are
necessary for a good fit of the model to the data. The im-
portant thing, he asserts, is the differences in x2 values
for the same model under different assumptions, not neces-
sarily the absolute value of x2 itself. "In other words, the
problem is to extract as much information as possible out of
a sample of a given size without going so far that the re-
sult is affected to a large extent by 'noise'" (Joreskog,
1974, p. 4).

In testing a given model, especially in assessing
whether a particular set of parameters should be included in
the estimated model, the important consideration in adding

or subtracting parameters is that the reduction of x2 that

 

*

The X2 test statistic reported is only approximately
distributed x and approaches a true x distribution as n
increases.

64
is obtained by adding the parameters be large relative to the
degrees of freedom that are lost by estimating those para-
meters. Thus if a parameter when estimated results in a sub-
stantial drop in X2 relative to the degrees of freedom that
are lost, that parameter is adding substantially to the
amount of information needed to provide an adequate fit of
that model to the data. This feature of LISREL will be used
in the radio situation to compare several alternative versions

of the model of social interaction presented in Chapter I.

Operational Model Described in LISREL Terminology

Figure 3*contains the Operational version of Model II
developed in Chapter I with appropriate LISREL labels for the
parameters. This model and all models tested in this disser—
tation are recursive because "all the causal linkages run

'one way,‘ that is no two variables are reciprocally related

 

*

The following presentations of results relies mainly
on figures. The actual structural equations are contained
in Appendix E. The letters used to describe variables and
paths in the figures follow the nomenclature for LISREL that
was presented earlier in this chapter. The diagrams them—
selves use several other conventions: exogenous variables
are to the left and endogenous the right, straight lines be-
between variables indicate a causal relationship with the
arrow indicated the direction of causality. Curved lines
indicate that two variables are associated, but no causal
direction is specified between them. In addition to report-
ing the values of the paths each diagram will conta'n the
degrees of freedom associated with the model, its X value
and its probability level.

Appendix G presents the actual computer printouts
associated with all tests of the model presented in this dis-
sertation. These results are presented in matrix format
which the appendix explains in some detail.

65
in such a way that each affects and depends on the other,
and no variable 'feedback upon itself through any indirect
concantenation of linkages, however circuitous'" (Duncan,
1975, p. 251).

The two exogenous variables in the model are emotion
(£1) and relationships (52). Emotion only has a single ob-
served variable, x4. Relationship has three observed vari-
ables: closeness, xl; responsiveness, x2; and role, x3.
These ordinary indicators represent the scale factors of
these variables.

The true endogenous variables in the Figure 3 are
content (n1). communication (n2), interpretation (n3), and
selection (n4). There is only one observed indicator for
content (yz), interpretation (yl) and selection (y3). Com-
munication has two observed indicants; conversation options
(y4) and process of conversation (ys).

The basic model only estimates 4 of the 12 possible
paths between endogenous true variables in this model.
These paths are labeled a's in the model. The paths
between the true exogenous variables and the true endogen-
ous variables are labeled with y. 68 and 66’

represent the measurement error variance* of the

 

*
Elements of 6 reported in the body of the disserta-
tion reflect variances, not standard deviations.

66

observed indicants.*

Identification of the Models

 

Each parameter of a model must be identified for the
model to be identified. If a model is not identified the
unique estimation of one or more parameters is impossible.
"By identification we specifically mean that no two sets of
distinct parameters should be able to produce the same 2
matrix" (Wheaton, e£_31., 1977, p. 107).

The determination of identification for models that
contain multiple indicators of true variables is particular-
ly treacherous. Joreskog and others (e.g., Joreskog and Van
Thillo, 1972) that work with systems of linear structural
equations that involve multiple indicators of latent vari-
ables discuss the issue of identification in terms that are
much different from the traditional discussions of identi-
fiability in economics where there is usually no distinction
between observed and true variables.**

One of the traditional conditions set forth for iden-
tifiability is the rank condition (Duncan, 1975, and Koop-

mans, 1949). The rank condition specifies that for each

 

*The error variances for the single indicators of
variables will not be estimated. The errors associated with
these single indicators will be contained in the residuals
of the single indicators true variables in the Psi matrix
of the endogenous variables. If this procedure is not fol-
lowed there are problems with the identification since one
piece of information would be used to estimate two different
parameters.

**
See Koopmans, 1949, or Theil, 1971, for a more tra—
ditional discussion of this issue.

HH mexuzx.

 

836535 358 Mo $8on mfi no «882 afioflflmmo 2 .m 9:52

.1 .1 .1 .1 .1

J m

x58 mxmmmvEHmpommE memmfimod

 

 

 

 

 

 

68

equation in a model that the number of explanatory variables,
those variables on which the dependent variable directly de-
pends, must be less than or equal to the number of exogenous
variables and variables that are predetermined with respect
to that particular equation. This is a necessary condition
for identification (Duncan, 1975), and it is satisfied for
the relations between true variables in every model tested
here.

Another necessary but not sufficient condition sug-
gested for identifiability in the case of LISREL is the so
called "counting rule" (Stein, 1976). This rule simply
states that when the number of observed variables multiplied
by one more than that number and that product is divided by
two the result must exceed the number of parameters to be
estimated. This condition is met for every model tested
here.

Another definition of identifiability is suggested by
Joreskog for models tested using LISREL. He asserts that
"if a parameter has the same value in all equivalent struc-
tures,* the parameter is said to be identified. If all para-
meters of the model are identified, the whole model is said
to be identified" (Joreskog and Van Thillo, 1972, p. 4).

Models reported here have been subjected to this test. When

 

That is cases where different start values are set
for the various parameters in the matrices.

69
different structures are specified (in terms of substantially
different intitial values) and then tested the values of para-

meters of the models are identical.

Development of Final Radio Model

 

In this section the model proposed in Chapter I is
going to be tested by means of LISREL in the radio situation,
if the initial test results in a relatively high chi-square
value and low associated probability level, additional para-
meters will be added to the model to determine if they result
in significant drops in x2 relative to the degrees of freedom,
an indication that the alternative model provides a better
fit to the data (Joreskog, 1974).

Comparing alternative versions of the same model is
not always a clear out process. Several factors must be
taken into consideration in deciding which of several com-
peting models is superior. The criteria that will be used in
this section in selecting the model that will be tested in
the television and typical situations are:

1. Parsimony. Only the minimum number of parameters
should be estimated.

2. Consonance. The model should be as consonant as
possible with the models and framework proposed in Chapter I.

3. The model should minimize the chi-square value
relative to its degrees of freedom. A set of additional
parameters should be added to the basic model only if they

result in a considerable drop in the value of the chi-square

70
statistic. A small drop in the chi-square statistic when
additional parameters are added indicates that these addi-
tional parameters contribute very little new information to
the model (Joreskog, 1974 and Schoenberg, 1972).

4. The model should minimize the residuals* remain-
ing from subtracting the correlations generated by the model
from the observed correlation matrix. After LISREL generates
a solution it uses this solution to estimate the correlation
matrix of the original variables given the estimates of the
parameters and the structure of the tested model. The orig-
inal correlation matrix is then subtracted from this matrix.
The resulting residuals provide valuable indicators of weak-
nesses in a proposed model (Joreskog and Van Thillo, 1972
and Schoenberg, 1972). High residuals can be used to indi-
cate what paths should be added to a model and to indicate
the general adequacy of a given model. A relatively high
residual correlation would indicate that the model isn't
predicting or accounting for the relationship between two

observed variables (Costner and Schoenberg, 1973).

In the next section a final model of social inter-

action will be deve10ped through applying the preceding

*

The residuals are determined by subtracting the
correlation matrix, R, inputed into the program from
the reconstructed matrix. The resulting matrix will be
called the residual matrix here and its elements will be
called residuals.

To prevent confusion the residuals associated with
etas will be called zeta in the remainder of this disserta-
tion.

71

criterion to alternative models.

Results of Tests of Models in Radio Situation

 

The results for Model Ia (the basic model proposed in
Chapter I) with only the diagonal elements of the psi matrix
estimated* are contained in Figure 4* and in Table 15. The
ratio of the X2 value, 190.12, which reflects the overall
goodness of the fit of the model to the data, to the degrees
of freedom, 22, is 8.6 to l.** The residual matrix contained
in Table 15 indicates that high residuals are associated with
the content indicant, y2, and with the communication indi-
cants, y4 and y5. This suggests that the addition of causal
paths to these variables from the exogenous variables may
reduce the size of these residuals and result in a significant
drop in X2.

Figure 5 contains the results of Model IIa with paths
from emotion to content and from relationship to communica-
tion. The substantive reasoning behind these paths was dis-
cussed in Chapter I. Only the on diagonal elements of the
psi matrix are estimated in this model. With the loss of 2
2

degrees of freedom there was a substantial drop in the X

value, 79.11, relative to Model Ia. The ratio of degrees of

 

*
Models with only the on diagonal elements of the

matrix estimated will be identified by an a after the number.
**
A ratio of 5 to 1 usually indicates that a model

with appropriate modifications can provide a good fit to the
data (Wheaton, et al., 1977).

mm
Hooo. u Hm>mH auflaflnmnoua .mm n on .NH.oaH

II II
><

 

:oflumsfim 0.8mm 5 MH down: How magma ..v mun—mam

 

 

 

 

 

 

73
Table 15

Residual Matrix for Model Ia
in Radio Situation

Y1 y2 y3 y4 Y5 X1 x2 X3 x4
y1 -.000
y2 .000 .000
y3 -.000 -.249 -.000
y4 -.173 -.061 -.o22 -.018
y5 -.568 -.358 .009 -.o14 —.010
x1 -.046 -.057 .018 -.234 -.418 -.000
x2 .109 -.291 -.055 -.187 -.250 -.010 .000
x3 -.060 -.195 .040 -.094 -.337 -.023 .051 .000
x4 .000 -.406 -.000 -.121 -.202 .074 -.154 .067 -.000
n==88

freedom to the chi-squared value is 4 to 1.

pected with such a substantial drop in the X2

As would be ex-

value there is

a significant improvement in the residuals contained in Table

16.

However, again high residuals are associated with the

content indicators and the communication indicators.

Figure 6 contains the results of Model IIIa with paths

from emotion to communication and from relationships to

content.

of two degrees of freedom from the previous model.

the x2 value only drops by 4.3 to 74.82.

The ratio of the X

The addition of these two paths results in a loss

However,
2

74

wmuc

88. u Hm>mH 333308 .8. n 8 HHS u N 4

 

.83 0m. 1100me 58...: Ba Ho 3:288 858848 05
panasoflmoHfieEoO cam. QEmsoUmHmm can usmucoo paw soHuQeH
20038 mfimm fl? 83856 088m 1: mHH H032 new 8.882 .m 0.563
4 m m
w w

H
H. 3. mm. W. om. .@ 3.14

SH 962% m» ammoomm as mx .58 Nx mg; Hx mmmzmeU

84/ OH\ Hm. 4/8/31, A.“

 

 

 

mo V 65621128 A m m H m mHmmeoHefimm
m H w \H .
1 S
N: S .H
H . n
H:

 

 

 

 

75
Table 16
Residual Matrix for Model IIa in Radio Situation

with Paths Between Relationship-Communication and Emotion-
Content with Psi Matrix Elements Off-Diagonal Fixed at Zero

y1 y2 y3 y4 y5 x1 X2 X3 X4
y1 -.000
y2 -.000 -.000
y3 -.000 -.O60 -.000
y4 .129 .097 -.130 .006
y5 -.123 -.112 .116 .007 .008
xl -.008 .090 .024 .002 -.035 .000
x2 .106 —.l82 -.077 -.001 .067 —.035 -.000
x3 -.034 -.077 .041 .097 -.024 -.019 .027 -.000
x.4 .000 -.000 -.000 —.011 .052 .041 -.214 .036 -.000
n==88

value to degrees of freedom is 4.16 to 1. While some of the
residuals in the previous matrix decreased, some of the other
residuals increased.

Because of identification problems and theoretical con-
straints, this is the extent of the changes that can be made
in the paths between the true variables. However, there are
still substantial residuals contained in the model, especially

involving communication, content, and the exogenous variables.

mm n" c
88. u Hm>mH 333886 .1: n H16 .8.: u mx

k.

 

H685 x8282 H66 66
38.88 8:688:80 05 cs... 60H8H8> 653865 6:8 883m
1.6%me mfimm 5H3 soHumsuHm oHUmm CH mHHH H891 How mustwm .m 0553

v m

.1 1. :1 . 1

v» @4050 m» mmgm x 38 Nx meZMPHmHGmem Hx mwmzmg

.. /\ ./...1\.

N
Null 1652822218 :7) EmzoESMm
NH . I 1,

N:

 

 

5... Ho.

 

 

 

 

 

mv.H H

Table 17

77

Residual Matrix for Model IIIa in Radio Situation

with Paths Between Each of the Underlying and

Surface Variables and Off-Diagonal

Elements of Psi Fixed at Zero

y1 y2 y3 y4 y5 x1 x2 X3 X4
yl .000
y2 -.000 -.000
y3 .000 -.017 -.000
y4 .134 .112 -.152 .001
y5 -.095 -.082 .111 .001 .001
x1 -.045 .127 .021 -.030 -.051 .000
.111 -.120 -.051 .002 .088 -.030 -.000
x3 -.057 -.040 .044 .076 -.030 -.040 .037 .000
X4 -.000 -.000 -.000 -.045 .031 .074 -.153 .069 -.000
n==88

It is possible that if the covariances of the true endogenous
variables are allowed to vary in the psi matrix that these
residuals will be reduced enough to produce a substantially
better fit of the model to the data. Regrettably this step
involves the loss of 8 degrees of freedom. Each of the pre—
vious models will be reestimated with the only change being

that all of the elements in the psi matrix will be estimated.*

 

*
These models will be identified with a b after the
number.

free to vary produces a not-positive definite matrix.

78

The LISREL run for Model Ib with the psi matrix left

This

means that there was either an identification problem or that

the determinant of the sigma matrix was too low to allow the

program to calculate a solution for the model.

The results for Model IIb with the paths between emo-

tion and content and relationships and communication and psi

free are presented in Figure 7.

14 degrees of freedom.

is 5.3 to 1.

residuals,

see Table 18,

The x2 value is 74.07 with

The ratio of X2 to degrees of freedom

Again there is no clear improvement in the

compared to the other models.

Table 18

Residual Matrix for Model IIb in Radio Situation with
Paths Between Relationship-—Communication and
Emotion-Content with Psi Matrix Estimated

:3
ll

-.O4l

.106
-.061
-.000

88

-.000
-.000
.113

-.077

.118
-.l35
-.053
-.000

y3 y4 y5 x1 x2 X3 X4

-.000
-.157 .000

.107 .000 -.000

.020 -.028 -.046 .000
-.058 —.004 .085 -.009 .000

.038 .072 -.032 -.021 .048 -.000
-.000 -.o47 .032 .075 -.159 .064 .000

mmnc

H86. u H92 333886 .HH n H6 .8.: u mx.

 

.Eumfiumm 1355.1 H66 00 38.86 88665.80
68 88838.80 H68 62.68038 65... 28:8 65... 8308
5058 £86 .33 8886.8 9.68 5 an H08: How muHsmmm .5 056:

Vw mm
8.1. m1;
SH 962% m» mmmuomm

 

 

mm.1

 

 

 

NOIIIIVN>4||| E4 1

Hm. Hu

 

80
Table 19 contains the zeta covariances, only the one between

content and meaning, .21, is substantial.

Table 19

Psi Matrix for Model IIb in Radio Situation with
Paths Between Relationship-Communication
and Emotion-Content

C1 2:2 C3 C4
51 .51
52 .21 .03
£3 -.04 .07 .22
£4 .04 -.08 .01 .55

Figure 8 contains the last possible model (since there
are no theoretically or technically acceptable possibilities
remaining) with paths between all the exogenous variables
and all of the endogenous variables. The x2 value is 74.07
for Model IIIb with 12 degrees of freedom. The ratio of the
x2 value to the degrees of freedom is 6.17 to 1. Again the
zeta covariances, contained in Table 20, reveal no clear

improvement over the models with additional paths.

Comparison of the Radio Models

 

It should be clear that the "best" radio model is
Model IIa with paths between emotions and content and between
relationships and communication where only the on diagonal

elements of psi are estimated. Except for the Model Ia, this

81

mm " C
88. u Hm>0H 333880 .NH n 00 .8.: u x

 

 

 

 

 

N a
«888m 5.58: H8 05 08 8389.5 88.4.8.6. 05... 053805
05 806.8 658 58 88801.10 088 5 an H082 00.4. 8808 .m 8:06
50 m0 ms N0 H0
N81. 2; 8.1. N1; 2.1.
SH 908 mm 8808 5“. mx 88 Nx E888 Hx 082806
07 6H.H\ . mm. H64 8\\4
Nqulllv 856% A on . 82020348
0 I NC [IIIV M\A mm. \
”ml/l 41‘ vC
ON 7, [/8 m g 0H
2.- ,- . ./
I .m I. HNe
Ho. 8558 . .
\\ 17 /
N N H: \ «NV 8.-
ullv mall E 7? 3 N. 163.96
T
1. 1
Hm v 6
H0 x All 1.
NN.
Hw

 

82

Table 20

Residual Matrix for Model IIIb in Radio Situation with
Paths Between Each Underlying Variable and Bach
Surface Variable with Psi Matrix Estimated

y1 y2 y3 y4 y5 x1 X2 X3 X4
yl -.000
y2 -.000 .000
y3 -.000 .000 -.000
y4 .134 .113 -.157 .000
y5 -.091 -.077 .107 .000 -.000
x1 -.o41 .118 .020 -.028 -.046 .000
x2 .106 -.135 -.058 -.004 .085 -.009 .000
x3 -.061 -.053 .038 .072 -.032 -.021 .048 -.000
x4 .000 .000 .000 -.047 .032 .075 -.159 .064 .000
n==88

Table 21

Psi Matrix for Model IIIb in Radio Situation
with Paths Between Each Underlying
Variable and Bach Surface Variable

C1 C2 C3 C4
C1 .51
C2 -003 —001
:3 -.04 .06 .22
c .04 -.21 .01 .55

83
is the most parsimonious model. This model produces residu-
als that are much lower than those of the Model Ia, and that
are equivalent to the residual matrices of the other altern-
atives. Model IIa also has the best ratio of the chi-square
statistic to degrees of freedom.

The differences between the chi-square values of dif-
fering models can also be evaluated by a relatively simple
statistic that allows us to determine which of two competing
models is superior. The difference in the chi-square esti-
mates for two competing models is asymptotically a chi-
square whose degrees of freedom are equal to the corresponding
differences in degrees of freedom (J6reskog, 1977, and Wheaton,
et al., 1977). Table 22 presents the results of this test
for both the differences between the Model Ia and alternative
models, and Model IIa and alternative models. The differences
between Model Ia and the other models are significant at the
.01 level (xiO > 29.59 at .01 level). The differences between
the Model IIa and the other models, aside from the Model Ia,
are not significant at the .05 level x: .05 > 5.99, indicating
that the additional parameters estimated by the program in
these other models do not significantly improve the fit of
the model to the data.

Before Model IIa is tested in the other situations the
estimates of its parameters in the radio situation will be
discussed in more detail. For comparison the basic theoretical

model (Model I) proposed in Chapter I will also be tested in

8 4
Table 22

Carparison of X2 Values for Radio lVbdels

A)

FIXEI;

1. Nbdel IIa with Paths Between Emotion
and Content and Relationships and Carm-

unication with Off-Diagonal Elements of

Psi Fixed at Zero

2. lVbdel IIIa with Paths Between all of
the Underlying and Surface Variables and
Off-Diagonal Elements of Psi Matrix
Fixed at Zero

3. Nbdel IIb with Paths Between Emotion
and Content and Relationships and Com:-

unication with Off-Diagonal Elements of

Psi Matrix Estimated

4. Nbdel IIIb with Paths Between all of
the Surface and Underlying Variables and
Off-Diagonal Elenents of Psi Matrix
Estimated

B)

2
X22

190. 12

2
X22

190. 12
X3.

190. 12

2
x22

190. 12

Differences Between Nbdel IIa and Other Radio Nbdels

STATISTICS
2 _ 2
X20 ‘ X2
79.11 = 111.01
2 _ 2
X18 ‘ X4
74.82 = 115.30
2 _ 2
X14 ‘ X8
74.07 = 116.05
2 _ 2
X12 ‘ X10
74.07 = 116.05

Differences Between Model IIa with Paths Between motion and Content

and Relationships and Commmication with Off-Diagonal Elements of Psi

Fixed at Zero and the Other Radio Nbdels

REBEL

1. Nbdel IIIa with Paths Between all of
the Underlying and Surface Variables and
Off-Diagonal Elements of Psi Matrix
Fixed at Zero

2. Nbdel IIb with Paths Between Emotion
and Content and Relationships and Corm-
munication with Off-Diagonal Elements of
Psi Matrix Estimated

3. Nbdel IIIb with Paths Between all of
the Surface and Underlying Variables and
Off-Diagonal Elements of Psi Matrix
Estimated

Xi.

79.11

50

79. 11
5.

79.11

2
X18

74.82

x2

14

74.07

x2

12

74.07

N

4.29

5.04

5.04

85
the TV and typical situations. However, the results of these
tests won't be discussed; instead they are presented in

Appendix F.

Results of Finally Chosen Radio Model

 

Now that a final model has been selected, its results
will be discussed in more detail. The results are reported
in Figure 5 and in Table 16. All values of parameters re-
ported here are based on the maximum likelihood solution re-
ported by LISREL. The paths between true variables are all
substantial in the model. One of the applications of path
analysis is a process termed "theory trimming." In this pro-
cess paths that are considered not to be meaningful are
dropped from a model. Land (1969) recommends that paths less
than .05 be treated as not meaningful. All of the paths here
are greater than .1, so applying Land's criterion all of these
paths are meaningful. However, three of the paths fall in the
.10 to .15 range--emotion to selection, selection to commun-
ication, and content to communication--and could be seen as
only contributing marginally to the model. The remainder of
the true paths in the model are apparently major determinants
of their dependent variables.

In all cases the variances of the errors in measure-

ment are substantial with values ranging from .34 to .56.

86

The zeta variance for interpretation is .17; for con-
tent .51; and for selection .51.* The zeta variance for
communication is a -.12.

The scale factors for the ordinary indicators are all
greater than .82.

The residual matrix for the radio model is contained
in Table 16. As noted before substantial residuals are assoc-
iated with the content indicant and the two communication indi-
cants. One exception is the relatively high residual between
responsiveness and emotion -.21.

The probability level associated with this model is
less than .0001.** In this dissertation probability levels
less than .05 will be considered to indicate the model provides
a (Horse fit of the model to the data than would be expected by
chance. Thus the model does not provide a good fit to the
data. The chi-square value of this model was 79.11 and the

degrees of freedom were 20 for a ratio c>f about 4 to l.

 

*
These zeta variances include errors of measurement.

Given the substantial errors in measurement associated with
the multiple indicators it must be assumed that the zeta var-
iances for these variables are higher than they would be if
they had had multiple indicators.

*Remember the probability level associated with the
LISREL x value is the probability of getting a chi-square
larger than the one generated by the model, given the hypoth-
esis that the model is true. As a result probability levels
approaching 1.0 are indicants of better fits of the model to
the data for they indicate that the fit is better than chance
given n cases (J6reskog and Van Thillo, 1972).

87

Results for the TV Model

 

The results for Model IIa in the TV situation are con-
tained in Figure 9 and Table 23. None of the substantive
paths should be trimmed from this model, although the one
between emotion and interpretation approaches Land's criteri-
on. Four other paths appear to contribute only marginally in
determining their dependent variables: emotion and content,
.16; content and communication -.l6; selection and communica-
tion -.19; and emotion and selection -.12. The rest of the
paths between the true variables appear to contribute sub-

stantially to the variance in their dependent variables.

Table 23
Residual Matrix for Model IIa in TV Situation
y1 y2 y3 y4 Y5 X1 X2 X3 X4
yl -.000
y2 -.000 -.000
y3 .000 -.270 -.000
y4 .013 -.088 .042 .027
y5 .104 -.065 .049 .018 .011

x1 .019 -.154 .004 -.035 .129

x2 .000 -.167 -.067 .029 .021 .063 -.000
x3 -.105 -.140 .108 .061 .334 .022 .024 -.000
x.4 -.000 -.000 -.000 .083 .244 .007 -.066 -.174 -.000

n==93

66 m...
oo. 55
SH monEo m» wmmuomm
w .
oo.H v

Hooo.

u H93 338880 .8 n 8 .85 u x

 

88688 E. 5 6HH H86: .8... 8380 .m 05011.1

x 88 Nx mag/H9068 Hx

 

 

 

 

 

.1

1......

89

Except for conversation options,* all the variances
of the errors of measurement are substantial ranging from
.32 to .77.

The zeta variances of the true variables that have
only one indicator are all substantial ranging from .38 for
interpretation to .81 for content. The zeta variance for
communication is .36.

The scale factors of the ordinary indicators range
from .64 to .87.

The residual matrix of the TV model is contained in
Table 23. The high residuals are usually associated with
the content and communication indicants. However, there is
a high residual between role and interpretation (-.105) and
between role and emotion (-.l74).

The chi-square value for this model is 84.91, with 20
degrees of freedom. The ratio of chi-square to the degrees
of freedom is 4.25 to 1. The probability level is less than
.0001, indicating that Model IIa does not provide a good fit

to the data in the TV situation.

Results for Model IIa in the Typical Situation
The determinent of the correlation matrix for the

typical situation dictated some changes in Model IIa. The

 

*

Conversation option's measurement error variance
had a value of .00. This is sometimes indicative of an
estimation problem in the program. This model was run again
with this value fixed at .0. This procedure produced
identical estimates.

90
LISREL program will not function when a beta matrix or re-
constructed E matrix is singular. If R (the correlation
matrix) and the model reflect an approximately correct model
and are nonsingular, then 2 will also tend to be nonsingular,
thus one indicant of possible problems, when the model is
correctly specified, is a singular correlation matrix (one
that has a low determinant). A low determinant in a matrix
is an indication of a high degree of linear dependence be-
tween one or more variables. Another word for this linear
dependence is multicollinearity. Now a certain amount of
multicollinearity can be expected in a correlation matrix
composed of variables in the same causal model, some of
which may be indicants of the same true variable. However,
in this instance, in all of the correlation matrices, there
is almost perfect linear dependence among some subset of the
variables. Table 24 contains the determinants for the corre-
lation matrices, there is almost perfect linear dependence
among some subset of the variables. Table 24 contains the
determinants for the correlation matrices tested in this
dissertation. The television situation correlation matrix
has a determinant of .0080 and the radio situation has a de-
terminant of .0023. While these determinants are very low
they are sufficient for the LISREL program to function. The
determinant for the correlation matrix in the typical situ-

ation, however, is so low, .0004, that the program won't

 

91

Table 24

 

Determinants of Correlation Matrices
Correlation Matrix Determinants
Television Situation .0080
Radio Situation .0023
Typical Situation .0004

For Typical Situation with the Following

Variables Deleted:
Content .0006
Meaning .0022
Emotion .0010
Process of Conversation .0012
Conversation Options .0036
Selection .0008
Closeness .0028
Responsiveness .0010
Role .0019
Role, Responsiveness, and Closeness .0082

For the question with the Following Combin-

ations of Variables, treated as indices in

the correlation matrice:
Closeness, Responsiveness, and Role .0047
Closeness and Responsiveness .0011
Role and Closeness .0016
Role and Responsiveness .0009
Process of Conversation and Conversation

Options .0011

92
function* with it.**

Two strategies exist for correcting this problem: (1)
deletion of one or more true variables and (2) deletion of
multiple indicants of a true variable. These strategies
allow for the testing of the model, but they have the dis-
advantage of limiting the comparability of models. The re-
sults in Table 24 indicate the highest determinant resulted
from converting the individual indicants of relationship into
an index created by summing its three indicants. This corre-
lation matrix is the one that will be used to test the model
in the typical situation.

Table 24 also contains the determinants that result
from removing the single indicants of the other true variables.
While all of these determinants were greater than the determ-
inants for the complete correlation matrix, none was suffic-
iently high to justify excluding a true variable from the
model or was preferable to the relationship index.

The problems with the determinant which necessitated

the substitution of an index for the relationship indicants

 

*
One reason for the failure of the program in this

case is the higher correlations between variables in this
situation. A simple correlation greater than .80 has been
deemed sufficient to produce an unacceptable degree of multi-
collinearity (Rockwell, 1975). There are 5 simple correla-
tions in the typical model that exceed .80. No simple corre-
lation in the television matrix exceeds .80; and only one

correlation in the radio matrix exceeds .80.
*7:
Most multiple regression estimates require inversion

of a correlation or covariance matrix; a singular matrix (one
with a low determinant) cannot be inverted (Rockwell, 1975).

93
produced some slight modifications of the lambda x matrix
and of the theta delta matrix. These changes are reflected
in Figure 10 which reports the results of the model and in
Table 25 which contains the residual matrix.

Following Land's criterion the path between content
and communication should be dr0pped from this model. The
paths between emotion and interpretation, —.09; interpreta-
tion and content.-.l4; relationships and selection, -.14; and
selection and communication, -.l6, also appear to contribute
only marginally to this model. The rest of the paths all
appear to contribute substantially to the variation in their
dependent variables.

The errors of measurement variances were .41 for the
process of conversation and .14 for the conversation options
indicants of communication.

The zeta variances ranged from .13 for communication
to .69 for content.

The scale factor of process of conversation was .81.

The residuals are substantial for content and conver-
sation options, -.ll9; interpretation and selection -.ll9;
process of conversation and interpretation, -.087; and selec-
tion, -.101; and relationships and process of conversing,
-.l94.

The chi-square value for this model is 55.75 with 8
degrees of freedom. The ratio of chi-square to degrees of
freedom is 7 to l. The probability level does not provide a

good fit to the data.

an
88. ... H82 323880 .m n 00 62mm

II II
I:

 

203808 H881? :H 8 HH H86: 08 888.1 .2 856E

 

 

 

 

 

 

95

Table 25

Residual Matrix for Model IIIa in Typical Situation

.000

-.067

-.119

-.087

-.000

-.000

79

.000

-.119

.116

-.044

.036

.000

-.004

.033

-.101

.001

-.024

.019

.036

.012

.000

y5 x1 X2
.001
-.044 -.000
-.194 -.000 -.000

96

Comparison of the Results of the Tests
of the Model

 

The chi-square values (see Table 26) indicate that
all three of the models have a similar fit to the data. The
chi-square value for Model IIa in the typical situation is
approximately the same as that for radio and TV when the
loss of degrees of freedom for the relationships index is
taken into consideration. The degrees of freedom ratios
for the tests in all three situations indicate that the
models, while none of them are significant at the .05 level,
Model IIa could be at least approximately correct. While
these tests are disappointing, in that they provide no con-
clusive evidence that this model is a good fit to the data,
this disappointment is ameliorated somewhat by the realiza-
tion that J6reskog (1974) has indicated that the chi-square
test is often misleading in this regard and is really better

suited for comparing models.

97
Table 26

Goodness of Fit of Model IIa
in the Three Situations+

Degrees of

X Freedom Ratio
Radio Situation 79.11 20 4.0
TV Situation 84.91 20 4.2
Typical Situation* 55.75 8 7.0

 

*
Fewer degrees of freedom because of relationships index.

+All of the tests of the model had a probability level
less than .0000.

Table 27 compares the values of the true paths between
the exogenous true variables and the endogenous true vari-
ables. There is not a great deal of similarity in the value
of these paths from situation to situation. Only the y32,
Y22' and 731 paths exhibit much stability. The y32 path be—
tween relationships and interpretation is always greater

than .89,* indicating that relationShips have a substantial

effect on interpretations. The Y22 path between relationships

 

*

Following Fink and Mabee (1977) at least two inter-
pretations of coefficients absolutely greater than 1 are
possible. One, when sampling error is absent values greater
than 1 indicate that the rank of the correlation matrix
imposed by the model is too low for the empirical correla-
tions. Thus solutions absolutely greater than 1 may indi—
cate that there is specification error. Two, a state of
disequilibrium could exist among variables in some cross-
sectional units.

98

Table 27

Values of Paths Between Exogenous and Endogenous
Variables in the Three Situations

Radio TV Typical
y31 -.33 .05 -.09
y4l .10 -.12 f .51
Y32 1.37 .92 .89
y42 1.17 .68 -.14
Yll .51 .16 .60
y22 90 1.19 .78

and communication is always greater than .78 indicating that
relationships are powerful determinants of communication.
The y31 path varies from -.33 for radio to .05 for TV.

Table 28 compares the values of the paths between the

endogenous variables across all of the models. between

(1.21!
content and communication, and 024, between selection and
communication, have an absolute magnitude less than .2,
which indicates a relatively low causal relationship between

these variables. The 013 and 0 paths do not exhibit much

43
stability.

The zeta variances are contained in Table 29. The
greatest stability in zeta variances across all of the vari-

ables is exhibited by interpretation (.17 to .38) and

selection (.47 to .62).

99
Table 28

Values of Paths Between Endogenous Variables
in the Three Situations

Radio TV Typical
013 .31 .33 -.14
021 -.15 -.16 .02
024 .15 -.19 -.16
043 -.42 .28 -.72
Table 29
Zeta Variances for the True Variables
in the Three Situations
Radio TV Typical
Endogenous
Content .51 .81 .69
Communication -.12 .36 .13
Interpretation .17 .38 .28
Selection .51 .47 .62

The scale factors for the ordinary variables are con-
tained in Table 30. The responsiveness value is the same in
both the radio and TV situation, .87, the role scale factors
differ moderately and the scale factors for process of conver-

sation differ substantially.

100

Table 30

Scale Values for Ordinary Indicators in the
Three Situations

Radio TV Typical
Process 1.10 .64 .81
Responsiveness .87 .87 *
Role .82 .68 *

 

*Note: No values because of index used for relationship
variables.

The measurement error variances associated with the
multiple indicators are contained in Table 31. The measure-
ment error variance of the relationship indicators range from
.32 to .69. The measurement error variances of the commun-

ication variables are more unstable ranging from .00 to .77.

Table 31

Measurement Error Variances for Multiple Indicators
in the Three Situations

Radio TV Typical
Process .46 .77 .41
Options .56 .00 .14
Closeness .34 .32 *
Responsiveness .50 .47 *
Role .55 .69 *

 

*
Note: No values because of index used for relationship
variables.

101

Tables 16, 23, and 25 contain the residual matrices
for the radio, TV, and the typical situations tests respec—
tively. Arbitrarily it could be said that when the resid-
uals for the simple correlations between two observed
variables is greater than -.05 in all three situations the
relationship between those two variables is not satisfactor-
ily explained by the current model. This condition holds

true only for the residual between selection and content.*

Conclusion

 

In this section the results of the tests of the models
of the process of social interaction in three situations--
television, radio, and typical--were presented. First, the
means, the standard deviations, and the correlation matrices
were presented. Then, to assess the significance of and the
variance accounted for by the effects of the respective
independent variables on therespective dependent variables,
OLS multiple regressions were used. The tests of the overall
goodness of fit of the model was made by means of the LISREL

computer program, which has a number of advantages over

 

*
This path was added to the TV model and tested. The x2

value, with 19 degrees of freedom, was 68.16. This path re—
duces the residual to 0. The ratio of x to degrees of
freedom was 3.6 to 1; an improvement over the 4.25 to 1 of
the "best" model, but sti 1 far from the ratio that would
result in a significant x , and not enough to warrant dis-
turbing the theoretical symmetry of the model. The next
chapter will suggest alternative, and probably superior,
improvements in the model other than the addition of this
path.

102

multiple regression. The radio situation was used to refine
the model developed in Chapter I; this model was then tested
in the television and typical situations. This chapter con-
cluded with a comparison of the results of the models.

In the following chapter the results of this model will
be discussed on a substantive and methodological level.
First, methodological explanations of the results which sug-
get modifications in the LISREL model will be discussed.
This modified model will be tested and then compared to the
results presented here. The substantive explanations for
the results of the models will then be discussed. The dis-
sertation will conclude with Chapter V which will discuss
the implications of the tests of the model and suggestions

for future research.

CHAPTER IV

DISCUSSION OF THE RESULTS AND TESTS
OF A MODIFIED MODEL

There are two primary kinds of explanations for the
results presented in Chapter III: methodological explana-
tions and substantive explanations. This chapter will focus
primarily on the former, the latter will be discussed in
some detail in Chapter V. There are three primary method-
ological explanations of the results: the high zeta vari-
ances of the true variates; the high levels of measurement
error variances associated with the multiple indicators; and
the high levels of multicollinearity. These explanations
will be discussed in some detail initially in this chapter,
then a new LISREL model will be pr0posed that ameliorates
some of these problems. After the results of the tests of
this model are reported for each of the situations they will
be compared to each other and to the tests of Model IIa pre-
sented in Chapter III. The chapter will conclude with a

discussion of the role of unspecified factors in the results.

Methodological Explanations of the Results

 

The results reported in the previous chapter reveal
that there is a moderately high level of measurement error

variance associated with multiple indicators of true variables.

103

104

This indicates that there is an imperfect association between
the indicators chosen here and the true variables they repre-
sent. This imperfect association is at least partially re-
sponsible for the poor fit of the model to the data and
perhaps related to the instability of parameters across
situations.

The zeta variances, especially those found in the tele-
vision situation, and those for content and selection generally,
indicate a considerable proportion of the variation in the
true variables is caused by factors not included in Model IIa.
Again this is probably related to the poor fit of the model
to the data and the instability of parameters across situa-
tions.

The overall results of the models, which were quite con-
sistent, and the quite different estimates of the parameters
in the models, reveal a pattern similar to the recognized

effects of multicollinearity.* That is multicollinearity

 

*
In general, while multicollinearity affects the esti-

mates of particular variables, it doesn't have an effect on
the overall level of significance or variance accounted for
by all of the independent variables (Rockwell, 1975). When
multicollinearity is present there is a very high standard
error associated with the estimate of any one parameter in the
model (Rockwell, 1975, and Klein and Nakamura, 1962). In
effect when there is a high level of multicollinearity a
researcher is unable to distinguish the effects of any partic-
ular independent variable on a dependent variable (Theil, 1971,
and Althauser, 1971).

Wiley (1973) has indicated that a singular matrix in a
LISREL model has many analogues to the problems of multi-
collinearity in traditional multiple regression techniques.
When 2 is singular the derivatives used to calculate the

105

usually doesn't affect the overall test of a model, but it
does have a considerable effect on the precision with which
any one parameter can be estimated. This coupled with the
low determinants of the correlation matrices indicate that
there is a very real possibility that multicollinearity may

have affected the results.

Modified LISREL Model

 

In this section a modified LISREL model will be proposed
that overcomes some of the methodological problems found in
Model IIa tested in Chapter III. One of the unique advantages
of LISREL is that it allows the researcher to assess the im-
pact of an unobservable variable. This variable can be one
for which there are no direct or unique indicants. In this
case all of the observed variables will be made indicants of
common, unobserved variables. As a result this model (Model
IV) contains a new n and E that represent a common factor at
both the exogenous and the endogenous levels. In effect all
of the observed endogenous variables are, in addition to being
determined by their associated true variables, determined by
a common variable. Thus in lambdaX and lambday new para-

meters are estimated for each observed variable to determine

 

maximum likelihood function become suspect. The presence of
collinearity between the exogenous variables, which is evi-
denced by the high covariance between them, can cause special
problems in the identification of parameters within the model
(Wiley, 1973). As a result of possible problems with multi-
collinearity the estimates of the values of parameters in the
models tested in Chapter III must be viewed with extreme
caution.

106

the effect of the common factors upon it.*

There are no changes in the beta structure estimated
in this new model, although a new row and column representing
the common endogenous variable is added to the matrix.

The common exogenous variable is said to cause the
common endogenous variable, the path between them, y33, is
fixed at one.

The variance of the exogenous variable is fixed at one.

The residual of the common endogenous variable is

fixed at zero.

The thetas remain as before.

The effect of these changes in the model is to isolate
those common sources of variation in the estimations of the
parameters in the model and to allow for the true relation-
ships between the other variables in the model to be estimated
more accurately. This is done by allowing the observable
variables to be determined by both their unique true variables,
for which they are indicators, and to allow them to also be

determined by another true variable, which they all share in

 

. r v w 1.1

*
Because of the unique nature of the common variable,

causing as it does in this model all of the observed indicants;
because the other observed indicators are often reference
indicators of other true variables; because reference and
ordinary indicators are usually associated with a clearly de-
fined true variable with which they have a strong conceptual
tie; and because of the arbitrariness (and unknown implica-
tions) of choosing a reference indicator among the indicators
of the common variable it was decided to make all of its
indicators ordinary indicators.

107
common, which can represent the common sources of variation
in the model. These changes should allow a more accurate
assessment of the goodness of fit of the model by removing
the "noise" attributable to common factors.

Results of Model IV with the Common Variable in the
TV Situation

 

Since the only changes in the parameters to be estimated
in the Model IV is in the lambda matrices, the presentation
of results here will be in the same format as Chapter III.
The lambda factors for the common values will be presented in
column vectors in a separate table. Remember that there is
a path with a fixed value of 1 between the common variables,
and the phi and psi values of these two variables are fixed
at 1.0 and 0 respectively in all the models. For simplicity
these parameters and paths aren't included in the figures.

The parameters for Model IV in the television situation
are contained in Figure 11. The paths between the true vari—
ables in this model are substantial, ranging from .34 for
the path between interpretation and content to 2.79 for the
path between relationships and communication.

Except for conversation options (.00),* the variances
of the errors of measurement of the multiple indicators are
considerable with values ranging from .42 for closeness to

.55 for role.

 

*This value was fixed at .0 in another run which re-
sulted in identical values for the other parameters.

108

 

 

 

 

 

 

 

NN u :
NNoo. n Hm>0H 333880 .HH u H0 .NNNN u Nx.
mHanHm> .
«0 m0 CSEOU fiHz £030:un PH. 5 >H MHmfiQH How muHsmmm .HH wHDmHm H
N
v 8er mmm.1\ m mmoe mmmw/v H Nw.
1H 0.20 NH 8886 x 384/ Nx 81255688 mflvwamod
Nm \1 N N.N.NJ1 8. w...
8. H 10 \
NU NN.- . zoflguuéaaox NNN 8,1898
7 Nwlluvm 1H NH NON-
N: 1.
.l.
BEBE... N: N.H
SH
55.08885 N:
om... 1. 4. me.
NN. 3 H
8H-
N rmzx H: k .H.
N . . 4. mm. ”e. 70%
fl 1. No. 1.
mm. H0 H0 mu v c
X

109

The zeta variances range from .05 for interpretation
to .57 for content.

The scale factors of the ordinary indicators are all
greater than .59.

The scale factors for the common variables are pre-
sented in Table 32. The scale values for the lambda y's
range from -.35 for interpretation to .36 for process of
conversation. The scale values for lambda x are all negative

and range from —.07 for closeness to -.63 for emotion.

Table 32

Scale Values for Common Variable in
Model IV in TV Situation

5
1 -.35
2 .16
A = 3 .
y 18
4 .23
5 .36
3
1 -.07
A _ 2 —.13
X-
3 -.36

110

The residual matrix is contained in Table 33. The

greatest residual is a -.081 for content and role.

Table 33

Residual Matrix for Model IV in TV
Situation with Common Variable

y1 y2 y3 y4 Y5 X1 x2 X3 X4
Y1 .000
y2 .000 .000
y3 -.000 .004 -.000
y4 -.001 -.009 -.003 .011

y5 -.015 .010 .041 .007 .004

x1 -.002 .007 .017 -.007 .095 .000

x2 .006 -.035 -.066 .041 -.031 -.017 .000

x3 -.010 -.O8l .059 .010 .209 -.022 .011 .000

x4 .000 .000 -.000 .001 .027 .008 -.022 .023 .000
n==93

The x2 value for this model is 28.49 with 11 degrees
of freedom. The ratio of chi-square to degrees of freedom is
2.59 to l. The probability level for this model is .0027,
which, while approaching .05, is still indicative of a poor

fit of the model to the data.

111

Results for Model IV with Common Variable in the
Typical Situation

 

 

The results for the typical situation are contained
in Figure 12. None of the paths should be trimmed from the
model in this situation using Land's criterion. However,

4 paths--content and communication, .12; emotion and inter-
pretation and selection, .lO--have absolute values between
.09 and .14. The remainder of the paths are greater than
.26.

The measurement error variances for communication*
indicants are moderate: .42 for process of conversation
and .16 for conversation options.

The zeta variance for the sole multiple indicator
true variable, communication, is -.01. The other zeta var-
iances range from .12 for interpretation to .57 for selection.

The scale factor for the sole ordinary indicator,
process of conversation, is 1.25.

The scale factors for the common variable are pre-
sented in Table 34. The scale factors for the lambda y's
range from .90 for interpretation to -.27 for content. The
scale values for the lambda x's are .24 for emotion and .74
for relationships.

The x2 value for this model is 3.8798 with 1 degree

of freedom. The probability level is .0489. The model

 

*
As in the typical model in Chapter II the relation-
ship indicants were converted into an index.

mhflc

NNS. n H92 3338an .H u .6 .NNHNN n x

N...

 

«COHumHHuHm ASHES”.

9.3 :H 833.5, 5&8 2» fit. B H802 new 8.88m .NH «BEN

 

 

 

 

H
ww mw @U My
NH. W N. H
SH 99 o m» mmmuomm Hm Hx
731R . \
IHlonle 5562928 4 NN . . Emmzofl.
) 5mm
.lllllYm. .
Nc NH a/ < N
28.69% om.
NH. NH. NN.
onefimmmmEsz. NN.
mm. a /
H: 8
N N .2550 . 2059a

 

 

 

113

Table 34

Scale Values for Common Variables in Model IV
in Typical Situation

5
l .90
2 -.27
Ay = 3 .03
4 .85
5 .63
3
Ax = l .74
2 .24

approximates a good fit to the data.

No residual is greater than -.100, (see Table 35)

Results of Model IV with Common Variable in
Radio Situation

 

 

There is a slight change in the parameters to be
estimated in Model IV in the radio situation. The variances
of the true exogenous variables in the phi matrix were set at
a value of 1.00 instead of being estimated by the program.
This was necessitated by the failure of the LISREL program
to arrive at a solution for this model when these parameters

were estimated.*

 

*Three different sets of start values were used to
attempt a solution to this model; all of them reached an

114
Table 35

Residuals for Model IV with Common Variable
in the Typical Situation

y1 y2 y3 y4 y5 x1 x2
y1 .ooo
YZ .,000 -.000
y3 -.000 .000 .000

y4 -.002 -.000 .000 .000

y5 .007 -.000 -.000 .000 .000

x .000 .000 .000 .000 .000 .000

x .000 -.000 .000 .021 -.077 .000 .000

 

unsuccessful conclusion after all of the models had iterated
for more than 600 seconds of compter time. All of these
attempts, which were supplied start values from previous runs,
concluded there iterations with a message that IND = 2 fl IND
= x is used by LISREL to indicate the kind of conclusion that
a run has come to. A value of zero indicates a successful
conclusion. The conclusion that all other runs reported here
came to. A value of 4 indicates that the model hasn't arrived
at a conclusion; values from this run can be resubmitted for
continued iterations that may result in a value of O. A
value of l, 2, or 3 indicates that a "serious problem" (un-
specified as to type) has been encountered and the minimiza-
tion function cannot continue (Joreskog and Van Thillo, 1972).

In addition to merely feeding in the start values from
previous runs attempts were made to "guess" at start values
that would produce a conclusion. All such attempts ended in
failure.

Attempts were also made to estimate differing parameters
among the error terms and the residuals, one of these was
successful--when the whole psi matrix was estimated. However,
the gain in the chi-squre value was not sufficient to overcome
the loss in degrees of freedom when compared to the model
presented in this section.

115
Table 36

Scale Values for Common Variables in Model IV
in Radio Situation

5
1 .52
2 -.06
Ay 3 -.39
4 -.19
5 .30
3
1 .12
Ax 2 -.2o
3 .16
4 -.34

The results for the radio model with the common vari-
able with the variances in the phi matrix fixed at 1.0 are
presented in Figure 13. According to Land's criterion one
path, that between emotion and interpretation (.04), should
be trimmed from this model. Another path, that between con-
tent and communication (-.l4), would appear to have only a
moderate effect. The rest of the paths appear to have a
substantial affect on their dependent variables.

The variances of the errors of measurement for the mul-

tiple indicators range from .32 for closeness to .55 for role.

H08. u HN>NH 3233on .NH u H6 .83 u x

 

 

 

 

 

 

N a
8333, 5:86 £3 coHNNBHN 088m 5. B 88: 8N $3.8m .NH was:
N m
N N NN NN HN
NN. NN. «N NN.,” NN. NN. 3?
NH 2850 NH 880% x 58 Nx mgmimmmm x wmmémod
6 277 8. H\ NH. NH NN. 8H
1
Nuo Ilolllw 2055238 A NN. @39ng
N:
3.: N
H:
N N
N Illw .NTII 923.50
8% E
H v AIIIII v

 

 

117

The zeta variances for the single indicator variables
are .08 for interpretation, .50 for content, and .17 for
selection. For communication the zeta variance is -.06.

The scale factors for the ordinary indicators are .73
for role, .87 for responsiveness, and 1.03 for process of
conversing.

The scale factors for the common variables range from
-.06 for content to .52 for interpretation (absolute values)
among the lambda y variables. The lambda x's range from .12
for closeness to -.34 for emotion.

The residual matrix (Table 37) contains no consider-

able residuals.

Table 37

Residuals for Model IV with Common Variable
in Radio Situation

y1 y2 y3 Y4 y5 X1 x2 X3 X4
yl .157
y2 .122 .095
y3 .129 .095 .106
Y4 .168 .258 .104 .131

y5 .128 -.000 .133 .136 .140

X1 .238 .295 .174 .193 .182 .337

x2 .159 .014 .170 .220 .163 .230 .146

x3 .116 .020 .071 .166 .094 .184 .151 .103

X .136 .104 .112 .193 .066 .259 .078 .116 .113

118
The x2 value for this model is 40.32 with 13 degrees
of freedom. The ratio is 3.10 to l. The probability level
for this model is .0001, again indicating a poor fit of the

model to the data.

Comparison of the Tests of Model IV

 

In this section the results of the three tests of
Model IV will be compared. The following section will then
compare this model to the results for Model IIa presented in
Chapter III.

The chi-square values presented in summary form in
Table 38 indicate some slight differences, largely caused
by the differences in the degrees of freedom of the models,
in the goodness of fit of the model to the data. The fit in
the three situations, while not significant, approaches sig-
nificance. In general, when Joreskog's (1974) previously
mentioned caution concerning the chi—square value is taken
into consideration, and the demonstrable problems with meas-
urement error variances and multicollinearity are noted, it
would appear that Model IV provides a reasonable fit to the

data.

Table 38

Goodness of Fit of Model IV with the Common Variable
in the Three Situations

2 Degrees Probability
x of Freedom Ratio Level
Radio 40.33 13 3.10 .0001
TV 28.49 11 2.59 .0027

Typical 3.88 l 3.88 .0489

119

Table 39 compares the values of the paths between the
true exogenous variables and the true endogenous variables.
All the values of the paths for the TV model save for Yll
are much higher than the paths in the other two models. The
only path that appears to be somewhat stable across models
is Yll’ the path between emotion and content, which ranges
in value from .30 for the typical situation to .55 for the
TV situation.

Table 39
Values of Paths Between Exogenous and Endogenous

Variables for Model IV with the Common Variable
in the Three Situations

Radio TV Typical
Y31 .04 -1.00 .09
Y41 -.21 1.64 .39
Y32 .87 1.88 .30
Y42 -.73 -2.08 .47
Yll .40 .55 .30
YZZ .68 2.79 .26

Table 40 compares the values of the paths between the
true endogenous variables. Again there is little apparent
stability across situations. Only the relative magnitudes of
the -a13 paths for the radio and TV situation, the -a21 path
for the radio and typical situation, and the —a43 path for

the radio and the TV situation have any similarity.

120
Table 40
Values of Paths Between Endogenous Variables for

Model IV with a Common Variable
in the Three Situations

Radio TV Typical
-al3 .43 .34 1-55
-a21 —.14 -.60 .12
-a24 .32 -1.04 .14
—a43 1.82 1.67 .10

The zeta variances are presented in Table 41. The
zeta variances for interpretation are veyr similar across
all of the situations. The zeta variances for content and

selection in the radio and TV situations are somewhat similar.

Table 41

Zeta Variances for Model IV with the Common Variable
in the Three Situations

Radio TV Typical
True Endogenous
Variables
Content .50 .57 .20
Communication -.06 -.49 -.01
Interpretation .08 .05 .ll~
Selection .17 .19 .57

The scale factors for the ordinary indicators are
contained in Table 42. The responsiveness values for the

radio and TV situations, .87 and .88 respectively, are

121
Table 42

Scale Values of Ordinary Indicators in Model IV with
a Common Variable in the Three Situations

Radio TV Typical
Process of Conversing 1.03 .59 1.25
Responsiveness .87 .88 *
Role .73 .68 *

 

*
No values because index used for relationship indicators.

essentially the same. The role values for these two situa-
tions are also quite similar: .73 for radio and .68 for TV.
The process of conversation values reveal no clear similar-
ities.

Table 43 contains the scale values for the common var-
iables. None of these scale values, although some are quite
substantial, reveal a clear pattern.

Table 44 contains the measurement error variances.
Except for role, which has the same value in both the radio
and TV situations and process of conversation, none of these
values fit a discernable pattern.

Tables 27, 33, and 35 contain the residual matrices
for the radio, TV and typical situations respectively. The
only noteworthy finding here is that there is no residual

greater than -.100 in any of the matrices.

122
Table 43

Scale Values for Common Variables in the Three Situations

Radio TV Typical
yl .52 -.35 .90
y2 -.06 .16 -.27
Y3 -.39 .18 .03
Y4 -.19 .23 .85
y5 .30 .36 .63
x .12 -.07 .74+

1

x2 -.20 -.13 *
X3 .16 -.36 *
x4 -.34 -.63 .24

 

*
No values for these variables because relationship index
used for these variables.

+Scale value for index.

Comparison of Model IIa and Model IV

 

In this section Model IIa discussed in Chapter III
and Model IV presented in this chapter will be compared.
First, the tests for each of the situations will be dis-
cussed, then an overall comparison of the models will be

presented.

123
Table 44
Measurement Error Variances for the Multiple Indicators in

Model IV with the Common Variables
in the Three Situations

Radio TV Typical
Process of Conversing .36 .55 .42
Conversation Options .45 .00 .16
Closeness .32 .42 *
Responsiveness .36 .53 *
Role .55 .55 *

 

*
No values because index used for these variables in
typical model.

The paths between the true exogenous and the true
endogenous variables are different for the two radio models;
only the path Yll between emotion and content is similar:
.51 in Model IIa and .40 in Model IV. The -a13 path has
similar values, .31 and .43 respectively.* The paths be-
tween content-communication, -.l4 and -.15, and selection-
communication, .15 and .32, are also quite similar. The

zeta variance for content, .51 and .50; and communiction

 

*

In this section the first reported value will be
for Model IIa presented in Chapter III, the second value
will be for Model IV presented in this chapter.

 

124

-.12 and -.06 and interpretation, .08 and .17, are all quite
similar. The scale factors for ordinary indicators are also
quite similar: 1.10 and 1.03 for process of conversing; .87
for both models for responsiveness; and .82 and .73 for the
role variable. The measurement error variances are also
quite similar with those for Model IVa being uniformly lower
or equivalent to the values in Model IIa.

Of the paths between the true variables in the tele-

vision models only -cx has equivalent values, .33 and .34.

13
The zeta variances aren't similar. The scale factors of the
ordinary indicators are remarkably similar; process of con-
versation, .64 and .59; responsiveness, .87 and .88; and
role, .68 in both models. The measurement error variances
of the multiple indicators are similar, but they reveal no
uniform pattern.

The paths between exogenous and endogenous variables
in the typical model are somewhat similar with Y3l having
values of -.09 and .09 and Y41' 51 and .39. Only the -021
path, .02 and .12, and the -a24 path, -.16 and .14, have any
similarity among the true endogenous paths. The zeta var-
iances for communication and selection reveal some similari-
ties in the two models. The scale values of the two models
are dissimilar. The measurement error variances for process

of conversations, .41 and .42, and for conversation options,

.14 and .16, are very similar.

125

Across all the models, the y32 path, from relationship
to interpretation and Y22 path between relationship and com-
munication would appear to be particularly powerful determ-
inants, with values never lower than .30 and .26 respectively.

The -a21 path, between content and communication
appears, except in the case of Model IV in the TV situation,
(-.60), to be relatively weak with values ranging from -.16
to .15.

The scale factors for responsiveness and role appear
to be two of the most stable parameters across models.

The measurement error variances for the relationship
variables also appear to be relatively stable across models.

These are the only apparent similarities in the values

of parameters across the models.

Relationships Among True Variables

 

In this section the relationships between the true
variables in the model will be discussed. The emphasis here
will be on the identification of stable paths across the sit-
uations and on the strengths and weaknesses of the model
originally proposed in Chapter I.

Using Land's criterion none of the paths proposed in
the original model should be dropped. The results from both
the LISREL runs and OLS regressions presented earlier support
the addition of two paths to the model, paths between emotion-
content and relationship-communication. In fact, the most

stable path across all of the models is the one between

126
emotion and content. The paths between relationships and
interpretation and between relationships and communication
are also stable in the sense that their values are always
substantial.

The remainder of the paths between true variables in
the model are unstable either switching sign or producing
substantially different values across situations.

This instability points to the possible effects of
third factors, such as the situation, on the relationships
between variables across models. The possible sources of
the differing values of the paths between the true variables
will be discussed in more detail in the next chapter.

The addition of the path between the underlying and
surface variables and their relative magnitude and stability
call into question the assertion made in Chapter I that in-
terpretation and selection act as mediators between the
underlying and surface variables. The patterns and values
of the paths across situations would suggest that this is not
the case, although the model does suggest that these vari-
ables are powerfully affected by relationships and are for
the most part major determinants of the surface variables.
Still their effects would appear not to come from their role
as mediators.

The strength of the relationships between the exogen-
ous and endogenous variables would tend to support the argu-

ment that emotion and relationships are the key underlying

127
factors in social interaction.

In sum, there are relatively few stable paths between
the true variables indicating that there are powerful situ-
ational factors affecting these relationships. While the
model results suggest that no path in the original model
should be deleted, and that two should be added, it does not
support the notion that the mediating variables are indeed

mediators.

Effect of Introduction of Common Variable

 

What was the effect of the introduction of the common
variable into the model of social interaction? The most
noticable effect was a substantial reduction in level of x2
relative to the degrees of freedom (see Table 45). The
introduction of the common variable didn't, however, produce
common values for most of the paths between true variables.
While this wasn't necessarily anticipated, it was hoped that
a clarification of the measurement error variances and zeta
variances would produce more stable parameters across the
situations.

The effect of the common variable on the values of the
zeta variances was mixed. In the radio situation there was
a slight reduction in the zeta variance for content, commun-
ication, and interpretation; and a major reduction, from .51
to .17, for selection. In the TV situation there was a sub-

stantial drop in the zeta variances of selection and inter-

pretation, but there was a substantial increase for content,

128
Table 45

Differences Between Model IIa and Model IVa
in the Three Situations

 

 

 

Situation Model IIa Model IVa Differences
. 2 2 _ 2
Radio x20 x13 - x7
79.11 40.33 = 38.78
TV 2 2 = 2
X20 X11 X9
84.91 28.49 = 56.42
T ical 2
YP X8 - X1 = X7
55.75 - 3.88 = 51.87

 

*
x2 values greater than x2 = 24.32 and X3 = 27.87 respectively

are significant at the .05 level.

and a high negative variance for communication. In the typical
situation there was a reduction in the zeta variance for con-
tent, communication, selection and interpretation.

The scale factors of the ordinary indicators remained
stable, indicating that the scale factors of the common indi-
cators values were drawn from other parameters.

There was a drOp in the measurement error variance
associated with the multiple indicators in the radio situa-

tion. Two of the indicators in the TV situation dropped and

129

two of them increased, while the value of one remained the
same. In the typical situation the measurement error var-
iances are essentially the same for both models.

The scale values for the common variables themselves
reveal little in common, except the magnitude of the values
for content are relatively low and the values for interpre-
tation and process of conversation are relatively high
across situations.

Overall a comparison of the values of the other para-
meters with the scale values reveals that process of conver-
sations had :3 decrease in its level of measurement error
variance in the radio and TV situations and the highest posi-
tive scale values in the radio and TV situations. However,
no other clear, discernable relationship between the scale
values and the other parameters in the model is discernable.

In sum, the introduction of the scale values reduced
the level of the zeta variance; the scale values of the
ordinary indicators remained the same; the value of the paths
between true variables were still inconsistent across situ-
ations, and there was a slight overall drop in the level of
measurement error variance.

The major contributors to the scale factors for the
common variable, however, appears to be the residuals in the
correlation matrix that weren't accounted for in the para—
meters of Model IIa. While the common variable pooled some

of the measurement error variance and zeta variance, the

130
improvement in the x2 value of Model IV appears to result
from the reduction of the residuals in the correlation ma-
trix. Therefore, the major source of improvement in Model
IV is probably a result of the fact that the model is now
accounting for the variance of a factor (or factors) that
is determining to some extent all of the variables, but

which wasn't included in the model.

Methodological Explanations of the Results

 

The problem of high measurement errors variances,
high zeta variances and multicollinearity have all been dis-
cussed in some detail in this chapter. To a certain extent
the model with the common variable ameliorated these problems-
However, they are still present, although to a lesser extent,
in this model. It still must be noted that these flaws in
the model have substantially contributed to the difficulty
in finding stable parameters across situations and to the
failure of the model to provide a good fit to the data.

There is also a possibility that rather minor viola—
tions of the assumptions of linearity, additivity, etc. ac-
count for the rather small difference between the level of
probability of the tests of the common variable model and the
level of probability needed for a model that would provide
a good fit to the data.

To some extent the differences in parameters across
models can be attributed to the changes in some of the models

made necessary by technical problems in the LISREL program.

131
The conversion of the relationship indicants into an index
in the typical situation is likely to have contributed to
the differences in the parameters in this situation compared
to other situations. The problems with the estimation of
the variances of the exogenous variables in Model IV in the
radio situation also probably contributed to the instability
in the parameter estimates across situations. It has been
the author's experience (and to some extent this is born
out by the results contained in Appendix F and in the various
tests of the models in the radio situation) that even minor
differences in the parameters estimated in a model can re-
sult in substantial differences in the values of the para-
meters within even the same situation. Regrettably for Model
IV the tests in each one of the situations was slightly
different.

Other technical problems exist with the measuring
instrument itself. The same respondent answering contiguous
questions probably contributes substantially to the level of
multicollinearity in the data (Shoenberg, 1972). The wording
of the questions probably also made the measurement of the
variables more sensitive to situational effects than they
otherwise might have been. The lack of multiple indicators
for some of the variables probably contributed to their high
zeta variances, and may have distorted the nature of the re-
lationships among the true variables, because the indicators

failed to cover the entire range of the true variable. On

132
the positive side most of the variation in the indicants,
as reflected by the scale factors, appears to be attribut-
able to their true variables. Finally, the scaling procedure
probably contributed to the substantially high variances
around most of the means.

In sum, these various technical flaws by themselves
may account for the slight difference between the manifested
probability levels associated with Model IV and the level
necessary for a good fit of the model to the data. In addi—
tion they probably contributed substantially to the insta-

bility in many parameters across situations.

The Role of Unspecified Factors in the Results

 

The high zeta variances and the behavior of the
common variables point to the existence of factors that weren't
specifically included in the model that may have affected the
results. In this section the effects of these unspecified
factors will be examined. In Chapter V the possible nature
of the factors that produce these effects, and the possible
role of these factors in an enlarged model of social inter-
action, will be discussed. This section is primarily con-
cerned with two questions. What is the nature of the effects
of these factors? Are the effects uniform across variables
and relationships, and how do they act on the model as a whole?
The descriptive statistics reveal substantial differ-
ences in the variables across situations. The simple correla-

tions between the variables were extremely high in the

133

typical situation, somewhat lower in the radio situation and
much lower in the television situation. This indicates that
the relationships among the variables in the model of social
interaction are substantially affected by situational factors
that act to weaken the relationships in situations where
media is present, Apparently the more intrusive the media,
the more pronounced the effects.

There are substantial differences in the levels of the
zeta variances across situations.* The zeta variances for
communication have the lowest values across situations, in
most instances being zero or near zero in value, indicating
that communication is primarily determined by its associated
parameters in the model. Interpretation also has a low zeta
value, again indicating that its primary determinants are
specified in the model. On the other hand selection appears
to be moderately determined and content appears to be sub-
stantially determined by unspecified variables not included
in the model. This pattern suggests that the elements of
social interaction are differentially affected by unspeci-

fied factors.

 

*

Only two of the true variables, communication and
relationships, have multiple indicators. The variance of
their errors of measurement are incorporated in the variances
of their zetas. This should act to increase the level of the
zeta variance for those variables with only one indicator.

134

The results for the common factor indicate that al-
though the direction of the relationships is uncertain, inter-
pretation and process of conversing are substantially affected 3
by it,* and that, save for conversation options in the
typical situation, the other indicants of the true endogenous
variables are relatively unaffected. There is no clear rela-
tionship between the exogenous true variables and the common
factors across situations, and it would appear that although
the relationship and emotion indicants are somewhat affected
by common factors, they are also affected by other unspecif
fied factors separately.

In sum, the common factor doesn't appear to account
for all of the undetermined variation in the true variables
and it would appear that some of the variables are determined
by unique causes.** Further, the effects are not uniform

across situations.

 

*
Remember that the common factor should include those

factors, or that factor, which all of the endogenous indi-
cants have in common.

**The possibility remains that two of the endogenous
variables may share a common cause. The only evidence that
speaks of this, since the covariance of endogenous variables
was not estimated in Model IV, is the covariance in Models
IIb and IIIb in the radio situation. None of these covaria-
tions were substantial, with only the covariation between
content and communication (.21) in Model IIb and between
selection and content (-.21) in Model IIIb being even
moderately high.

135

The failure to specify these factors in the model
accounts in some measure for the failure of the model to
provide a good fit to the data and for the instability of
parameters across situations. These results don't support
the assertion made in Chapter I that situational differences
might cause some slight differences in the values of para-
meters, but that they wouldn't contribute to major differ-
ences in the values of parameters or to differences in the

direction of relationships across situations.

Conclusion

 

Model IIa tested in Chapter III suffered from serious
problems with multicollinearity exacerbated by high measure—
ment error variances and the high zeta variances. A modi-
fied LISREL model was proposed to ameliorate these problems.
Although this modified LISREL model was somewhat successful
in this regard most of the substantial improvement in the
fit of the model to the data appears to come from the reduc-
tion of the residuals in the correlation matrices. The
pattern of the results indicated that there are factors not:
included in the original model that impinge substantially
on the variables and the relationships in the process of social
interaction. Given this, and the various technical problems
associated with the tests and with the data, there would
appear to be reason to be Optimistic that, with certain modi—
fications in data collection procedures and with the model

itself, the current framework can eventually provide the basis

for an accurate model of the process of social interation.

CHAPTER V

IMPLICATIONS OF THE RESULTS FOR
FUTURE RESEARCH

In this concluding chapter the implications of the
tests of the model will be discussed. First the general
methodological implications will be discussed, then sug-
gestions will be offered on how the methodological problems
found here may be rectified in future research. Next the
substantive implications of the results will be detailed.
This chapter will conclude with preliminary suggestions on

the form of future models of social interaction.

Methodological Implications

 

It would appear that a substantial part of the re-
sults are attributable to methodological problems; e.g.
multicollinearity, slight differences in the models tested
in varying situations, measurement error, etc. These re-
sults have some larger implications for the use of OLS
multiple regression and LISREL in other settings.

First, if only OLS multiple regression had been used
the model would have been substantially supported, and the
problems with measurement error and multicollinearity may
have gone unnoticed. LISREL provided a means of detecting

and then of analyzing these problems so that new directions
136

137

for future research could be identified.

Second, there is a paradox in the use of LISREL with
multiple indicators, the purpose for which LISREL was de-
signed. Multiple indicators of the same process are going
to be correlated, and hopefully highly correlated, but if
they are correlated, then the determinant of the correlation
matrix is going to be low. If the determinant is sufficient-
ly low, and, if the model is correct, then the program won't
function; if the determinant is slightly higher, then it
will work but the estimates of parameters will be doubtful.
The more successful you are in getting powerful, highly cor-
related indicators, the more likely it will be that the re-
sults of the program will be of little use in estimating
particular parameters.*

This paradox leads to the conclusion that LISREL may
not be well suited for models of this type, where all of
the variables are elements of the same process and where
there is a high degree of correlation between indicators of
these variables. LISREL is probably most successfully used
when there is a low degree of correlation between the causes

of a variable, and between indicators of a single true

 

*
In a curious way the low determinants of the correla-

tion matrices, which have created so many problems in testing
the model are actually very supportive of it. The determin-
ants indicate, in essence, that there is an almost perfect
linear dependence among some subset of the variables. Thus,
the determinants indicate that these variables are closely
bound together, exactly what would be expected for the
elements of the same process.

138
variable. These conditions, however, would seem to be met
in only a limited range of social phenomena.

The results suggest several ways that future research
could be used to clarify the issues discussed here. Perhaps
the most important initially is to overcome methodological/
technical problems so that the substantive paths in the model
can be clearly tested.*

The first thing that could be done to provide a clear
test of the model, is to attempt to create indicators that
are as orthogonal as possible. This is a very difficult
task when you are dealing with a model that seeks to explore
the relationships between elements of the same process. But,
to the extent possible, work should be done on developing
indicators for the different true variables that are as dis-
tinct as possible and that don't have correlated errors
built in by the method of data collection.

One of the ways to reduce multicollinearity is to
insure that different methods are used to measure the var-
iables. One solution to this problem is to test the model
in a situation that allows for the systematic observation of

the process of social interaction by differing coders.

 

*

The ultimate irony in the results is that the esti-
mates of the error terms, zeta variances, and scale values
reveal more stability and invariance across situations than
the more substantively interesting values of the paths be-
tween true variables.

139

Observation of interactants could also serve to con-
trol for some of the extraneous sources of variation in
social interaction present in the data used here. That is
the people that answered questionnaires engaged in inter-
action with members of differing sexes, with differing
numbers of people, in different physical surroundings, with
differing preceding events, and watched or listened to dif-
ferent media presentations. All of these factors serve to
increase the random error present in the measurement of
these variables.

The model should also be tested in a situation that
allows for the observation of social interaction over time.
The model tested here was static since it was felt that it
would be inappropriate to test a dynamic model of social
interaction with cross sectional data. Further it is un-
likely that any variable, at the same point in time, in this
model feeds back on a variable that causes it. It is more
likely that one variable causes another at time 1, say emo-
tion causing interpretation, and interpretation at time 2
causes a certain emotional level at time 3. Development of
a truly dynamic model and the test of the model over time,
not with cross-sectional data, is needed. In many ways it
is surprising that the model provided as good a fit as it

did, since it wasn't dynamic.*

*
A test of the model over time would probably also be
more sensitive to the effects of the situation, since studies

140

Substantive Implications

 

In this section the general substantive significance
of the paths between true variables in the model will be ex-
plored. The reader is cautioned that the results of the
tests of the model will be treated here as if they weren't
subject to the methodological problems that probably con-
tributed strongly to to the instability of the various paths
across models.

The covariance between emotion and relationships is
strongest in the radio situation, somewhat weaker in the TV
situation, and weakest in typical situation. Research
(Johnson, 1976) has indicated that radio often acts to set
a mood between interactants. As a result it might be ex-
pected that radio could result in a greater consonance be-
tween emotion and relationships. On the other hand the
weaker relationship between these two variables in the typ-
ical situation suggests that in the absence of competing or
mood enhancing stimuli emotions and relationships are less
related to each other.

The results suggest that emotion is a substantial
direct cause of content in all situations. This suggests,
given the manner in which these variables were operational-
ized, that the higher a person's level of emotional arousal

the more the content of their interaction changes.

 

(Johnson, 1976) have indicated that a television situation
has a dramatic impact on the continuity of social interaction.

141

In the television situation there is a strong neg-
ative causal influence from emotion to interpretation,
while in the typical situation there is a minimal positive
influence, and in the radio situation there is essentially
no causal influence. This suggests that a person's level
of emotional arousal has little bearing on his interpreta-
tion of the interaction when there is no strong competing
stimulus in the environment. However, in the presence of
a strong competing stimulus, TV for instance, the more emo-
tional someone becomes the lower is his interpretation.

Thus the more emotional a person is the less likely he is to
possess the capacity to sort out all of the conflicting
stimuli in a complex situation.

In each of the situations there is at least a moder-
ately strong positive causal relationship between relation—
ships and interpretation. This suggests that the closer to
others, the more responsive to others, the greater the indi-
vidual's interactants are knowledgeable of whom they are.
the more able they are to interpret the interaction. This
causal path is stronger in the presence of competing stimuli
in the environment suggesting that they act to narrow the
number of factors an individual depends upon for his inter-
pretations. As this happens relationships apparently be-
come more important in an individual's interpretations.

The path between relationships and communication is

strongest in the TV situation and weakest in the typical

142

situation. This indicates that the more involving the situ-
ation, the greater is the influence of relationships on
communication patterns.

The relationships of the exogenous variables to selec-
tion are interesting. The patterns suggest that relation-
ships are strongly negatively associated with selection in
the media situations and moderately positively related in
the typical situation, while emotion is increasingly posi-
tively related to selection from the typical to the media
situations. Emotions apparently act, expecially in the pre-
sence of involving stimuli, to increase the level of an in-
dividual's concentration. Relationships on the other hand,
in the presence of strong stimuli, act to decrease the level
of attention individuals pay to any one thing. Thus a high
level of relationships compels individuals to split atten-
tion between other interactants and the media. On the other
hand a high level of emotion appears to cause an individual
to focus his attention on one element of the environment.

Contrary to expectations relationships are a very
powerful predictor of other variables in these situations.
Thus either they are more clear cut in media situations than
originally anticipated or relationships are even a'more

powerful predictor than first assumed.

Suggested Model for the Process of Social Interaction

 

The results reported here suggest several changes in

the model originally proposed in Chapter I. Naturally the

143
suggestions made here are tentative, and are at times
based on fragmentary evidence in the tests of the models of
social interaction presented in Chapters III and IV.

It appears that the current categories of social
interaction should be reduced. Emotion and relationship*
should be retained in any model of social interaction that
is developed, since both of these variables appear to be the
primary determinants of the other categories of social inter-
action.

The reduction of the number of true variables that
lies within the proceSs of social interaction will come at
the mediating and surface levels. The results clearly call
into question the suggested mediating role of the interpre-
tation and selection variables. It would appear that the
underlying variables act directly on the surface variables.

If these variables aren't mediating variables, then

what are they? An argument could be made that the

 

*

The incorporation of other "true" exogenous vari-
ables should also reduce some of the technical problems
caused by these variables now. While in the current model
these variables are exogenous in the sense they are not
caused by other variables in the system, they are not exo-
genous in the sense of lying outside the system described
in the model--rather they are intimately associated with
other elements of the process. In fact this intimate asso-
ciation, as a result of the high degree of covariation be-
tween emotion and relationship, contributed substantially
to the problems of multicollinearity present in the tests
of the model. The addition of other true exogenous vari-
ables to the model should remove the major flaw in these
two variables.

144

distinction between the surface and the mediating variables
is really artificial, that these two classes of variables
are really manifestations of the same underlying process.
This is especially true for interpretation and content,
which are intimately associated with each other. Conceptu-
ally in fact it could be suggested that content is the ob-
servable manifestation of the underlying true variable of
interpretation. Thus, these two variables should be reduced
into one true variable called interpretation.

Similar reasoning holds for communication and selec-
tion. Selection could be viewed as just another manifesta-
tion of the processes by which substance is transferred from
one interactant to another. That is who an interactant
selects to talk to and how much attention he/she pays are
really just another observable manifestation of communica-
tion processes. Thus these two variables should be merged
into a new variable labeled communication.

The reconstituted model of social interaction is pre-
sented in Figure 14. The relationships between variables in
this model are essentially the same as before. In the next
section this model will be expanded to include other true
exogenous variables that represent the effects of situation-

al factors on the model.

   

145

INTERPRETATION 4—3’COMMUNICATION

 

EMOTION RELATIONSHIPS

V

Figure 14. Suggested Modifications in Model
of the Process of Social Inter-
action.

Outside Factors in New Model

 

The results point to the existence of outside factors
that affect the variables in the model. The failure to
specify these outside factors in the original model probably
contributed to the poor overall fit of the model to the data
and the instability of parameters across models, for as
Watzlawick, et_al. (1967, p. 20) point out, "a phenomenon
remains unexplainable as long as the range of observation is
not wide enough to include the context in which the phenome-
non occurs." The results also suggest that the effects of.
outside influences are not uniform across variables and in
fact some isolated factors may be major determinants of
single variables within the model. "It is obvious, however,
that the environment is not an undifferentiated medium in
which people are immersed; it clearly involves a variety of
active processes which selectively spur, guide and restrain

behavior" (Barker, 1963, p. 42).

146

In this section a new model will be proposed that contains
some preliminary thoughts on the nature of environmental
influences identified in Chapter I: individual character-
istics, rules, context, and situational factors.

One element of rules that apparently impinges upon
interaction in these situations is the topics that can be
discussed. These rules govern the kinds of substance that
can be made manifest in an interaction. The literature de-
scribed in Chapter I suggests that in television situations
in particular there are rules that set the agenda of tOpics
that can be discussed.* These rules related to agenda set-
ting probably account for the substantial variances of the
residuals for content that weren't related to the common
factors or to the covariance of content with other variables,
especially in the radio and TV situations.

For the moment there are two element of context that
would appear to impinge upon the variables within the model
of social interaction. The first of these is the purpose
for which interactants come to an interaction. It was noted
earlier that one of the primary purposes interactants come

to the television situation for is sociability. This factor

 

*

Agenda setting and some of the other concepts dis-
cussed in this section are not traditionally considered to
be variables, but they can be operationalized in variable
terms. For example, the range or number of tOpics discussed
are two ways of quantifying agenda setting.

147
could be seen as primarily affecting the emotion and rela-
tionship variables within the interaction, and it might ac-
count for some of the manifested covariation between these
two variables, especially in relation to the responsiveness
and closeness indicants of relationships.

The second contextual element that could be seen to
have an effect on the variables within the model is the
short term historical pattern of the interaction. The his-
torical pattern of the interaction should primarily affect
he interpretation variable, and to a lesser extent affect
emotions and relationships.

Individual characteristics can also be seen as affect-
ing the variables in each of these models. However, the
nature of these characteristics and their relationship to
variables in the model is too complex an issue to deal with
at this preliminary stage of modeling. HOpefully the effects
of these factors can be treated as random error that cancels
out over a large n.

Two situational factors are probably important for
the process of social interaction. One is the extent to
which there are compelling outside stimuli present. If any-
thing represents the common factor in the models of social
interaction tested in Chapter IV it is probably the extent
of sensory involvement of the interactants with stimuli
other than the other interactants. This variable can be

viewed as having a determinant effect on the reconstituted

148
variable of communication which includes selection, and
somewhat less of an effect on all of the other elements of
the model.

The setting in which the interaction occurs, that is
the physical nature of the surroundings, probably also has
an effect on the relationships between variables in any one
interaction situation, especially upon emotions.

The reconstituted model of social interaction is con-
tained in Figure 15. This model is suggested by the results
and it represents the preliminary thinking on what a modi-

fied model, based on the current evidence, would look like.

 

 

 

 

 

 

RU ES
\
INTERPRETATION ; COMMUNICATION
EMOTION R LATIONSHIPS
CONTEXT SITUATION

Figure 15. A Suggested Expanded Model of the
Process of Social Interaction.* .

The variables within the box are the categories of the
process of social interaction. The variables outside
the box represent exogenous variables that effect the
process of social interaction. Arrows pointing to the
box, instead of to one of the variables, are meant to
indicate that an exogenous variable effects all of the
elements of the process social interaction.

149

Conclusion

 

In this dissertation several categories of the process
of social interaction were isolated; these categories were
then used to construct a model of the process of social
interaction. A means of testing the model was described in
Chapter II. The model was tested in Chapter III. OLS mul-
tiple regressions of the individual dependent variables in
the model revealed that their respective independent vari-
ables accounted for quite substantial proportions of their
variation. However, the overall tests of the model by means
of LISREL revealed that the overall model provided a dis-
appointing fit to the data. A revised model designed to
correct some of the methodological flaws in the original
model and to account for common sources of variation in the
true variables was proposed and tested in Chapter IV. This
model produced a substantially better fit, although still
not a good fit, to the data. Given the low ratio of degrees
of freedom to chi-square exhibited by this model there is
every reason to believe that a test of the reconstituted
model proposed in this chapter will produce a successful

fit of the model to new data.

APPENDIX A

Category Schemes for Social Interaction

150

APPENDIX A
CATEGORY SCHEMES FOR SOCIAL INTERACTION

This appendix reviews thirty category schemes for
social interaction. It is organized by the categories of
social interaction used here. The categories are identi-
fied along with their authors and the purpose that the
scheme was designed to serve. Regrettably none of the
category schemes reviewed here have been used to construct

or to test a model of social interaction.

151

INTERPRETATION

Author/Scheme/Purpose

Categories

 

Amidon and Hunter (1966)

The verbal Interaction Category
System (VICS)

System For Analyzing Interaction

In Classroom Settings

1. Gives Information Or
Opinion

2. Gives Direction

3. Asks Narrow Question
4. Asks Broad Question
5. Accepts

6. Rejects

7. Responds To Teacher

8. Responds To Another
Pupil

9. Confusion

 

Auld and White (1959)

Analysis Of Psychotherapy

Patient's Sentences

 

Anxiety
Dependence
Hostility

Love

Mild Agreement
Resistance

Sex

Social Mobility

Therapist's Utterances

Demand
Interpretation

Reward

 

152

Author/Scheme/Purpose

Categories

 

Bales (1950)
Interaction Process Analysis
Description Of Interaction

In Groups

\OCDQO‘U‘uwaH

H P6 H
N +4 o
O O 0

Shows Solidarity
Shows Tension Release
Agrees

Gives Suggestion
Gives Opinion

Gives Orientation
Asks For Orientation
Asks For Opinion
Asks For Suggestion
Disagrees

Shows Tension

Shows Antagonism

 

Borgatta, E. F. (1965)
Analysis of Patterns
Of Social Interaction

(Especially In Groups)

Common Social
Acknowledgment

Shows Solidarity
Through Raising The
Status Of Others

Shows Tension Release,
Laughs

Acknowledges, Under-
stands, Recognizes

Shows Agreement, Con-
currence, Compliance

Gives A Procedural
Suggestion

Suggest Solution

Gives Opinion, Evalua-
tion, Analysis: Ex-
presses Feeling 0r
Wish

Self-Analysis And Self-
Questioning Behavior

Author/Scheme/Purpose

153

Categories

 

10.

11.

12.

13.

14.

15.

16.

17.

18.

Reference To The Extern-
al Questioning Behavior

Gives Orientation, In-
formation, Passes Com-
munication

Draws Attention, Repeats,
Clarifies

Asks For Opinion, Eval-
uation, Analysis, Ex-
pression Of Feeling

Disagrees, Maintains A
Contrary Position

Shows Tension, Asks For
Help By Virtue Of
Personal Inadequacy

Shows Tension Increase

Shows Antagonism, Hos-
tility, Is Demanding

Ego Defensiveness

 

Carter et al., (1951)
Analysis Of Group Interaction
Particularly As It Pertains

To Leadership

Proposes And Initiates
Action

Disagrees And Argues
With A Somewhat Neg-
ative Connotation

Leader Roles In Carry-
ing Out Action

Follower And 'Worker'
Roles In Carrying Out
Action

Abortive Or Non pro-
ductive Behavior Or
Problem

Miscellaneous

 

154

Author/Scheme/Purpose

Categories

 

Crowell and Schiedell (1961)
Development Of Ideas In

Discussion Group

Assertion
Information
Inference
Substantive
Procedural
Volunteered
Requested
Initiation
Restatement
Clarification
Substantiation
Extension
Simple Response To Request

Pro Modification (Revision
Of Prior Idea)

Con Modification (Revision
Of Prior Idea)

Stated Acceptance
Summary
Imperative
Question
Judgment
Synthesis

Delayed Relationship (To
Idea)

Delayed Self Relationship
(To Speaker's Idea)

 

Flanders (1967)
System For Analyzing Interaction

In Classroom Setting

1. Accepts Feeling
2. Praises Or Encourages

3. Accepts Or Uses Ideas
Of Students

4. Asks Questions
5. Lecturing

Author/Scheme/Purpose

155

Categories

 

8.
9.
10.

Giving Directions

Criticizing Or Justi-
fying Authority

Student Talk-Response
Student Talk-Initiation

Silence Or Confusion

 

Gouran and Baird (1972)
Comparison Of Problem
Solving And Informal

Group Discussions

Initiates And Develops
Theme

Agrees With Expressed
Position

Disagrees With Express-
ed Position

Gives Information

Asks For Information

 

Lewis et al., (1961)

Analysis Of Student-Teacher

Pupil Interaction

Asks For Information

Seeks Or Accepts Direc-
tion

Asks For opinion Or
Analysis

Gives Information
Gives Suggestion

Gives Direction

Gives Opinion

Gives Analysis

Shows Positive Feeling
Shows Negative Feeling

Perfunctory Agreement
Or Disagreement

 

156

Author/Scheme/Purpose

Categories

 

Longabaugh (1966)

Uncovering Structure
Underlying Interpersonal
Behavior (Particularly
Observation Of Children's

Behavior)

1. Gives Help

2. Suggests Responsibility
3. Reprimands

4. Attempts To Dominate

5. Acts Sociable

6. Calls Attention To
One's Self

7. Gives Support

8. Physically Contacts
9. Is Succorant
10. Assaults Sociably
ll. Assaults

12. Symbolic Aggression

 

McGuire and Lorch (1968)
Rules Governing Natural

Language Conversational Modes

Associational Orientation
(Casual Conversation)

Problem Solving (Convey-
ance Of Factual Knowledge)

Interrogation

Clarification Of Misunder-
standing -

 

Schiedell and Crowell (1966)

Group Discussion Behavior

Substantive Themes
Procedural Themes

Irrelevant Themes

 

Snyder (1945)

Description Of Non-directive
Psychotherapy

Therapist's Responses

 

Restating Content
Clarifying Feeling

Author/Scheme/Purpose

157

Categories

 

Interpreting
Structuring
Leading
Suggesting
Questioning
Persuading
Accepting
Reassuring
Approving
Disapproving

Client's Responses

 

Problems
Simple Responses
Insight Planning

 

Steinzor (1949)

Activate And Originate
Structure And Delimit
Diagnose By Labeling
Evaluate

Analyze And Explore

Express And Give Informa-
tion

Seek Information To Learn
Defend

Offer Solution

Conciliate

Understand And Reflect
Give Support

Oppose And Attack

Show Deference
Seek Support

158

Author/Scheme/Purpose

Categories

 

Conform

Entertain

Miscellaneous

Clarify Confusion

 

Strupp (1960)

Analysis Of Psychotherapy

\lO‘U‘lbWN

Facilitating Communica-
tion

Exploratory Operations
Clarification
Interpretive Operations
Structuring

Direct Guidance

Activity Not Clearly
Relevant To The Task
Of Therapy

Unclassifiable

 

Weintraub and Aronson (1962)
Verbal Analysis Of Defense

Mechanisms

\lO‘LflbWN
0

Direct References
(Setting Of Experiment)

Evaluators

Non-personal References
Shift To Past Tense
Negators

Qualifiers

Retractors (Detracts
From Previous State-
ment)

 

159

Author/Scheme/Purpose Categories

 

Watzlawick et al., (1967) Content
Examination Interactional

Patterns

 

160

CONTENT

Author/Scheme/Purpose

Categories

 

Argyle (1969)
Description Of The Ways
Coordination Is Important

In Social Interaction

Content

 

Bjerg (1968)

Interplay Analysis

Topic Agons

(What Is Said Or Done)

 

Hare (1958)
Paradigm For The Analysis

Of Interaction

Content

 

Hawes (1973)
Elements Of A Model For

Communication Processes

Content

 

Watson (1958)
Description Of Formal
Characteristics Of Inter-

action In Three Situations

Conversational Resources

(Topics Of Conversation)

 

161

 

Author/Scheme/Purpose Categories
Hare (1958) Social Emotional Behavior
A Paradigm For The Analysis A. Control

B. Affection
Of Interaction

 

Longabaugh (1966) Socioemotional Dimension
Structure Underlying Interpersonal

Behavior

 

Reusch and Prestwood (1949) Emotional Reactions (in-
ternal facet)
Structural Components Of

Interaction

 

Taylor (1954) 1) Public Dimension
Emotional Dimensionality 2) Dyadic Dimens1on
3) Autistic Dimension

Of Groups

 

Weintraub and Aronson (1962) Expression Of Feeling
Verbal Analysis Of

Defense Mechanisms

 

162

 

 

EMOT I ON

Author/Scheme/Purpose Categories
Argyle (1969) Emotional Tone
Description Of The Ways In

Which Coordination Is

Necessary For Social

Interaction
Bjerg (1968) Instinctual Agons (Love,

Interplay Analysis esteem, etc.)

Sessional Agons (i.e.,
damage, agon, agon of

pleasing, etc.)

 

Carter et al., (1951) Shows A Personal Feeling
Analysis Of Group Interaction,
Particularly As It Pertains

To Leadership

 

163

 

COMMUNICATION
Author/Scheme/Purpose Categories
Amidon and Hunter (1966) Initiates Talk To Teacher
The verbal Interaction Initiates Talk To Another
Pupil
Category System (VICS) Silence

Analyzing Teacher-Student

Interaction

 

Argyle (1969) Timing Of Speech
Some Of The Ways Coordination
Appears To Be Necessary For

Interaction

 

Bjerg (1968) Conversational Agons

Interplay Analysis

 

Bostrom (1970) 1 to l Sends

Analysis Of Patterns Of Centrality
l to Group Sends

Communicative Interaction .
l to l Receives

In Small Groups Receive Sent Ratio

 

164

Author/Scheme/Purpose

Categories

 

Hare (1958)
A Paradigm For The Analysis

Of Interaction

Communication Network

Interaction Rate

 

Jaffe and Feldstein (1970)

Rhythms Of Dialogue

Vocalization

Pause

Switching Pause
Speaker Switch
Simultaneous Speech

 

Lewis et a1. (1961)
Analysis Of Student-Teacher

Interaction

Inhibits Communication

No Communication

 

McGinnies and Altman (1958)
Group Discussion Behavior

(Attitude Change)

Verbal Output

Participation

Rate Of Response

Recruitment (time
entered discussion)

Spontaneity

 

Pope and Siegman (1972)
Description Of Psychoanalytic

Interview

Informational Exchange

1) Hesitation
2) Fluency
3) Verbalization

 

165

Author/Scheme/Purpose

Categories

 

Watson (1958)
Description Of Formal Characteristics

Of Interaction In Three Situations

Conversational Style

 

Weintraub and Aronson (1962)
Verbal Analysis Of

Defense Mechanisms

Quantity Of Speech
Long Pauses
Rate Of Speech

 

166
SELECTION

Author/Scheme/Purpose

Categories

 

Argyle (1969)
Important Elements Of
Interaction Where Coordination

Is Necessary

Nonverbal Responsiveness
(Signals From One
Interactant To Another
Of Attentiveness)

 

Bostrom (1970)
Analysis Of Group Discussion

Statements

Selectivity (Relative

Concentration)

 

Goffman (1957)
Description Of Forms
Of Alienation From

Interaction

External Preoccupation

Self-Consciousness

Interaction-Conscious-
ness

Other Consciousness

 

Lewis eta1.(l96l)
Analysis Of Student-Teacher-

Pupil Interaction

Listens

 

167

RELATIONSHIPS

Author/Scheme/Purpose

Categories

 

Argyle (1969)
Important Elements Of Interaction

Where Coordination Is Necessary

Dimensions 0f Relation-
ships
I. Role Relations
II. Intimacy
III. Dominance

 

Bjerg (1968)

Interplay Analysis

Implicational Agons
(Why Things Said Or

Done)

1) Instinctional
Agons (Power)

2) Sessional Agons
(Superiority)

 

Hare (1958)

Socioemotional Behavior

 

A Paradigm For The Analysis A. Control
Of Interaction
Hawes (1973) Relationships

Elements Of A Model For

Communication Processes

 

168

Author/Scheme/Purpose

Categories

 

Pope and Siegman (1972)
Description Of Psycho-

analytic Interview

Attraction To Interviewer
Interviewer Warmth

Interviewer Status

 

Spier (1973)

Membership Category

 

Invariant Features Of Devices
Interactions
Watzlawick et a1. (1967) Relationship

Examination of Interactional

Patterns

 

169

EXOGENOUS VARIABLES

Author/Scheme/Purpose

Categories

 

Bjerg (1968)

Interplay Analysis

Problem Agons (Salient
Problem)

Service Agons (Exchange
Of Services)

Agons Of Material Goods
(Exchange Of Material
Goods)

 

Hare (1958
Paradigm For The Analysis

Of Interactions

Task Behavior

 

Longabaugh (1966)
Uncovering Structure Under-

lying Interpersonal Behavior

Task Dimension

Social Activity

 

Spier (1973)
Invariant Features Of

Interactional Elements

Main Activities
Local Setting
Temporal Orientations

Spatial Orientations

 

170

SEQUENCING, RULES

Author/Scheme/Purpose Categories

 

Argyle (1969) Sequences of Behavior
Important Elements Of Interaction

Where Coordination Is Necessary

 

Bjerg (1968) Meta Agons (Bargaining
About Which Agons To
Be Activated)

Interplay Analysis

 

APPENDIX B
Inclusion of Elements of Social Interaction in

Previous Category Schemes

171

.coHuumumucH HMHoom mo
mucmgomuoo mo mQOHDQHHommo .mumcunmmmmu m50H>muQ on Homo“ on mam: mHmmooH com: mH mewzum muomoumu
k.

 

x x x x 6mm: 98:

 

x Ammmd EH3 mam. GMHDOO

 

x Ahmad Ema

 

x Ahmad Wampum-Hm

 

x :63: HvamHfiom USN. HHOBDHU

 

 

x x AHmNHV .Hm um uwfimo

 

x x 8;: 858m

 

x 88: 3838

 

x x x x ANNNHV 9.8.8

 

x Hommd mmem

 

x Emmi mug 0cm 63¢

 

x x Good NONE—Hm 9.8 £8.33

 

x x x x x ANNNHV NH?

 

mmHmwonBE 205m 76Ha<UHzH§8 20395 763129; .5550 56m $89.6
«mmewnum muovoumu m50H>oum CH coHuomumucH HmHoom mo mucmeme mo cOHmsHocH

m XHDszmé

172

 

ANNNHV 8802 HEN 885563

 

 

2..va .Hm um gogHNumz

 

8mm: comumz

 

SNNHV NoHNmH.

 

88: 83%

 

Eva: HONckum

 

ANS: SEN

 

Gem: 83%

 

Goad HHDBOHU can HHmmvafiUm

 

ANHNHV 8988a as... 888

 

1N8: 583m 8cm 88

 

8mm: nuns mono wHHHGUZ

 

Ammmd 525.2 can meSHHUQH

 

AHNNHV 888:3

 

 

HHmmHV .Hm um mHBQH

 

8th Gkummvam HEM mmmmb

 

x x

350.3 gm

 

mmngfiag Egan—Mm 753.3qu SHE van—Hana; Hag

Esau. E836

APPENDIX C

Instructions Given Telephone Solicitors

173
APPENDIX C

Instructions Given Telephone Solicitors

TELEPHONE APPEAL

Hello, is this (Resgondent's Name) . I'm (Your Name)

from the Department of Communication at Michigan State Uni-
versity. We are looking into the ways in which TV and radio
affect how peOple communicate with each other. This research
is supported by the National Association of Broadcasters,
whose professional code is subscribed to by CBS, NBC and ABC
and most of the television and radio stations across the
country. We need your c00peration and the c00peration of
others like you in our efforts. Your participation may lead
to valuable information on the effects of television and
radio on American life. We would like your permission to
mail you a questionnaire concerning this topic that you can
complete at your leisure in your own home. Would you be
willing to answer the questions and help us out? ......

 

IF YES, THEN: Thank you very much. We appreciate your co-
operation. You can expect to receive a c0py of the questions
in a couple of days. Its been very nice talking to you

(Respondent's Name) - GOOd bye.

IF NO, THEN: Is there someone else in your household who
might be interested in answering the questions? (IF YES, THEN
ask if they are home and if they can come to the phone, then
repeat the above appeal to them. If the other person is not
at home, then mail out questionnaire addressed to the person
that the respondent says will be willing to fill out the
questionnaire, and tell subject to tell other person to

expect a letter in a couple of days.)

IF STILL NO THEN: Thank you for your time.

174

STEPS TO FOLLOW WHEN MAKING PHONE CALLS.

1. Get card. (See attached sheet for information on how
to read the card.)

2. To dial phone number first dial 174, listen for dial
tone, then dial phone number that is on the card.

3. Read the telephone appeal to respondent.

4. Follow the following procedures depending upon their
response.

A. IF RESPONDENT DOESN'T AGREE TO ACCEPT THE QUESTION-
NAIRE, THEN write refused, time, date, and initials
below the address on the front of the card.

B. IF THERE IS NO ANSWER, THEN write no answer, time,
date, and your initial below the address on the
front of the card.

C. IF THE NUMBER IS OUT OF SERVICE, THEN write out of
service, time, date, and your initials below the
address on the front of the card.

D. IF THE RESPONDENT AGREES TO RECEIVE THE QUESTIONNAIRE,
THEN FOLLOW THESE STEPS:

1. Write their subject number in the space on page 8
- the back of the questionnaire - that says
questionnaire number.

2. Write the respondent's name after Dear on the
front page of the questionnaire.

3. Address the envelOpe to the respondent.

4. Fold up the return envelope and place it in the
envelope.

5. Fold up the questionnaire and put it in the
envelope.

6. Write down mailed questionnaire, date, and your

initials below the address on the front of the
card.
5. Go to next card.

175

HOW TO READ THE CARD

NAME - Found on the left hand corner.

PHONE NUMBER - Found on right hand corner.

ADDRESS - Second line, middle. If just a street address mail
to Grand Rapids. If another city is named after the street
address, then address the envelope to that city.

SUBJECT NUMBER - Right hand lower corner in red.

HELPFUL HINTS

Remember be polite, be considerate, be helpful, but also be
firm in getting a commitment from the respondent to receive
and fill out the questionnaire.

Sell them on the idea of receiving and filling out the
questionnaire.

If the respondent is in a hurry make an appointment to call
back at a more convenient time.

ANSWERS TO SOME QUESTIONS ABOUT STUDY

TIME - It takes about 30 - 40 minutes for most people to fill
out the questionnaire.

WHY SHOULD I DO THIS?
Repeat appeals in the letter.
Say that I need this for my dissertation. For school.

176

TELEPHONE CALLBACKS

Hello, is this (Respondent's Name) . I'm (Your Name)
from the Department of Communication at Michigan State Uni-
versity. Remember we called you a couple of weeks ago and
asked your permission to send some questions to you about
how television effects how you talk to other people. We
haven't received this questionnaire from you yet and we were
wondering if you had any questions about how to fill it out?

 

 

POSSIBLE RESPONSES
If they have't returned the questionnaire because they don't
understand it explain to them how to fill it out.

If angry at us for any reason try to pacify them and say that
you regret any inconvenience that we may have caused them.

If the respondent just hasn't gotten around to filling it out
yet, try to get commitment from them to fill it out soon.

If they haven't received a questionnaire yet say we will mail
them another one. Be sure to get their correct address.

 

STEPS TO FOLLOW WHEN MAKING PHONE CALLS

1. Get card.

2. To dial phone number first dial 174 then listen for dial
tone, then dial phone number that is on the card.

3. Read the above statement to the respondent.

4. On the back of the card write the respondenthsreply, any
action you took, the date and your initials. If you
can't reach the respondent write down why (e.g., no
answer), the date and your initials.

5. Go to the next card.

APPENDIX D

Mailed Questionnaire

MICHIGAN STATE UNIVERSITY 177

 

College of Carmunication Arts EAST LANSING -. MICHIGAN - 48824
quutmanzof<1mnunflxnion

Dear

Recently you received a phone call from the Department of
Communication at Michigan State University. In this phone
call we asked your help in finding out the ways that tele—
vision and radio affect how people communicate with each
other. Our study is supported by the National Association of
Broadcasters whose professional code is subscribed to by CBS,
NBC, ABC, and the majority of the television and radio sta-
tions across the country. Your participation may lead to
valuable information on the impact of television and radio on
American life. In addition, the questionnaire may help you
learn something about yourself and stimulate you to have some
fresh insights into the situations that may be included. We
really appreciate your agreeing to take some of your valuable
time to help us.

Please read the following instructions carefully. Some of
the questions that follow are difficult. They are difficult
because they ask you to think about situations in novel ways.
We would appreciate it if you would make every effort to an-
swer the questions. If a question is confusing to you, then
save the questionnaire and we will call you soon and give you
assistance in answering the question. But please make every
effort to answer the questions. When you finish place the
questionnaire in the enclosed envelope and mail it to us at
your earliest convenience. If you would like a report of the
results of the study, write your name and address on the
upper left hand corner of the envelope.

 

 

Again we would like to thank you for your willingness to take
part in this effort. We believe your answers will be of great
help in the continuing efforts to understand and solve im-
portant issues relating to the effects of television and radio
on our society.

Sincerely,

J. David Johnson
Edward L. Fink
Sherrie L. Mazingo

178

In general do you like talking with others while you are
watching television?

 

 

Yes No

What do you like about talking with others when you are
watching television?

 

What do you dislike about talking with others when you
are watching television?

 

In general do you like talking with others while you
are listening to the radio?

 

Yes - No

 

In general, on a scale of l (hurts a lot) to 10 (helps
a lot), how would you rate the effect of television on
your conversations with other peOple?

 

 

 

 

 

In general, on a scale of l (hurts a lot) to 10 (helps
a lot), how would you rate the effect of radio on your
conversations with other people?

 

 

 

 

Can you recall any habits that you have developed when
you talk to others when viewing television?

Can you recall any habits that you have developed when
you talk to others when listening to the radio?

179

A nurber of the questions that follow use a method for rating things with
nunbers that you may not be fanu'liar with. This method may be difficult
at first, but nest people get the hang of it after answering a few ques-
tions. Here we will provide you with an exanple of how this method is
used. In this exanple we will rate how comfortable people feel in differ-
ent situations. If zero (0) is a couplete lack of canfort and _1_.__00 is a
typical artmmt of comfort , then how canfortable are you in the follmring
situations? For this exanple we will use the feelings of a hypothetical
Person X who will rate his degree of comfort in these situations.

SITUATIONS

ONTHEJOB GIVINGASPEEGI ATHOMEWITHFAMILY
ANDLNT Person x likes his job. Person Xmas never Person X gets along

OF He has been at it a long given a speech. He well with his fami-
CDMFORI‘ time. He feels a nearly doesn't like large 1y. He feels very
average anount of com- crowds. He is shy. happy and secure
fort in this situation. So he rates this with them. He rates)
So he rates it as a 1_12 situation as an 8 this situation as a
in atromt of comfort. for axrount of 342 for amount of
canfort. canfort.

 

 

 

 

 

 

 

 

In answering these questions we will be just interested in the nurber you
use to rate the situation. You don't have to provide verbal explanations.
You can use any numerjou wish. In surmary, this is the way you should
answer these questicns. First think of a typical anount (100) of and the
absence of (0) what you are rating. Then ccupare the situation you are
rating with the absence (0) and the typical anount (100). Then rate it
(give it a nmber) based on this comparison.

 

9. If zero (0) is a ccnplete absence of attention and 100 is how much a__t-
tenticn you pay to things on the averag, then what number would you
use to __describe your level of attention to each of the elements in the

situations listed from left to right on the next page. Renenber, 1’22
can use any nunber you wish.

 

180

 

aggg
BémMHBEQE

 

 

 

 

 

 

 

 

 

 

 

 

 

.x .3 .> .D ..H.
g
.m .m .0 .m .0
73¢ :58 on.» :H munch
[no .3988 no.5 2H mad
8% wZHEm gunman mum.
.z .2 ..H .M .b
_ .0 /// a 89% NE.
1.: ./ U. 0H om.m . .//. .../
‘ ., ./_/ . 4 28mg NE.
Smomm ~5th
.m .D .0 .m .HH
wzbwsm mH WEE 92m: PH. 20 meemz PH. 20 E mHGHBéHm m0 g
CHER mo OHan OH. IZHEHzm 23mg lag
>9. 9H zmmz 22ntqu OHQdm OF E
EHH
monBSFHm

ZOHBdDHHw .mO WE DH. EHEHAN mo 82%

181

 

mcoflmduflm

«5. an 33.5.

no: mums 90% :95

9533»? 8&5 mo

58 fi 85%: umfi

mafia... ”Bonn madam»

xwmz umma 98mm sow
U3 muse: wan: 30m .ma

 

 

Ammflufifiuom mo .3955

mmcoflmsuwm 8%

mo comm an" 5 vmmmm

new wad—83>» so» am.

703 .mcfiummu .mfiuswm

"mung Hows mafia
$30M Hgo ham: 30m .NH

 

 

mgflmfiflm

mfizodom m5 .6 8mm

5 wfimm 5% 3c 832
acme .6: x83 #3 .2

 

 

 

mmcoflumsuflm 3% 5
now 5.5 3.33%”. man.
a Hmfio >58 30: .3

 

75 mH CHER CHE mm? 75 PH. PH. 8 @650

 

 

 

 

 

 

OHQmmmO EZOMRHZ Egg; 75E 25H.
PHEZHZSJH OBUZHZMFWHH.Q DHGZHzflamHH.U UZHE.m IEQE£

 

.mfiflgfim 158:: fi 98%: “E? SE» “938 .23 2.. mo amp:

in: “.23 xfifi £803 .850 :H .8336 Bu 5 Ema ma ”ES 5? 85.398 €58 Ea .39
5958 qu= 55% mun H0393 3 2mm: ms “was .mcoflmsuflm 05.8%.? $0 manna—mama $9.65 swamo mcoflmmzv
mmmfi on mumzmam “may @5385 Q: .Goflmsuflm 3393 m ca .3me magma 053033 may mo “895a <
.cowumsuwm 50mm Hem cowummsv m Hmzmcm B Hung: 96 Human mm: .wmmm m5 m0 mounm can: “so? may :0 963
Imsfim m5 mo 68 How :85» H965 cm? can mama mg no mam USE puma m5 :0 meoUmmfiv «5 0mm“ mmmwam

 

 

meOflumsufim

wmmnu CH mgowummnmficoo

ca AOflcmn ccw_>so maume

ms» no.905 mum Hmow

luauu 3mm..AHMOHUdHo::

Nmm>v.bH on AHMOfluwHU
Nmmz H m0 0.30m m 5 .5

 

 

 

 

meOHumsuflm mmmnu mo

nvmm cw mGOfluwmumpcou

H50» mumu som_uasoz

3mmv.AmH >Oncmc5

a £8 33m .5
v a «0 mamom m :0 .ma

 

182

 

.smw3,soxfluwnaac mum
mm: cmo.so» mmgofiu
nmsuam mama» mo numw
cw mHm so>_om>ao>cw

 

 

so: mmfiuommc .882»
”.23 85 .839»?
Hmomgu m 5 $3 so»
unmaw>ao>cw mo unscem
mmmumpm may mw_mmm.
can. agape: mo x03
38. m 282 83 fl .3

 

 

 

mmumnuo ou mucmaaoo mxma

0» “canny new cadmw>

Imamuv MHGME may :O_Hmm£

Ho wow so» mmcflnu may

3 End—aim uo 8x9
.mmmqsoa mum H50: HMO
.35 m g 5&0 so: .3

 

 

 

ZO mH Oflomm OHnmm HEB ZO kPB >B_ZD mZOHBmMDO
Ofiomm.m0 mmfi ZO mzmz BZMZZHafimmBzm ZO mZBZ BZHZZHdHMMH
‘>B Dz 2m 3v.m 09.02PZMHEHAIHD 09.02H2EHmHA .U UZHZBH> .m 12m 02H3mH> 91

 

 

 

 

 

 

183

 

mmgHumnuHm

wwmfi CH «5% sum», mum.

033 SE .350 93

no so» mum H8330

..mm Smofldscs

Nmmfi bH B AHmUHUHHU
NW5 m3 Hm mo mHmom m 8 .mH

 

 

 

 

EmH

OHngo
PH 8 E .m

 

0H2

3520952
QUEEN—”Ad

OHQdm 275

 

 

 

8mg .0 ug .m

>975

gums
.zME .m

 

mag

 

I‘W

 

184

 

000000002000 00 0000 0:0
5:030300005 50300003805
000300 5000300308 00 £592

"020ng .8 E0 .0

 

000003000 03..» :0 00050
8 0:330 0o 820 080

30> 523 3 8.68 9:.
“$0600.20 000.0000 .0

 

>8 30> 023 B

90.30003 >HH00000 000 >00
30> 30:3 0.0 9.00000 3050
080 99> .02: 8 $080 9:.
“Eng .0

 

.000 .83002000

:3 .0000 .0>0H 00 £030 .005
$000 Hmcoflofi 00 7.53 03>
"SHE .U

 

.0005 0:0
>00 :0 00000080000 30> :02: 30:
"60006.5( .0

 

1.5000 50» 30> 005 005.5 05.

 

 

 

 

 

 

 

 

"020.38 .4

0.80.0: 0.005.020 080008068 200.040.0300 < .8 0020
0.600 0 500000300. < . «800000000000 00005 000005 0 :0 30> 00
0 :0 00000 0005 00 2000 00000 00. .00 £000 000
"00 00:0000m 05.: 00 >00 30> 00 5.040050 0:000 . 30: .8300
Nm 0008mm 830000350 0000003 m830000>80 £030: 30: 00050 00 00 n00>:00 05 3 00330008
0 030> 00 00000 05 000 :02: 30: 00003 0 5 000050 ”$000 30> :05 5300002000 040 >05 :02: 30: .00 00:0.
.205 :00: 000303000 05 :0 00:05 0005 .00 £000 on £02: 30: 00 00000 0600 00 0000:0000 .00:0000n€0 0005 0.0 0:0
“.0 0.000000 05 00 0030000 0mc050 .050 00>0 $0050 :03 000: >00 >05 30> 000000 000000 00000 00 :00 :03
:0000 00300000600 0 00 3000 .3 n0000>c00 0 00 00000 .3. 002000 H0203 0 :0 .2 100002000 0 00 00.000 .3

Gun—<0 Dz< >0. 00 m2; ”550 H0250 SHELZ 258$

Eganmog

 

 

 

 

 

 

 

022.550 .0003 8» «00,32 02¢ 08 26 300 .3

$.00 .030 :05 5.0000095. 00 30:05 >H000: 00 0000 0 00 .9: :300 .030 00:0 500000002000 0 00 .0000 H0000>u 05 m0 0:00.805. 00 0003.0 >100: 0H ~000.0000
>00 .0000 0003000000 0 H000 30> f. .3958 000 .>u000000 005 000 000000090000 05 00 0000 5000 0000 :05 00300000350 0:00: 0 :H Emma 005 Nwmmmmmum
“09m 00 003% 000090 00 ammw 05 00 a 0:0 $000000 00 .050 .850 00:50 5000:0000 00:00.59: 6.0. gm 0030020 00 00.00000 00.0.?

05 m0 2: 0000 00 600000030 0005 0:000:05 :05: 0030300 00 230:0 20300000 0:030:00 05. .0000 £000 000 530030 .0000 00305... 8003 .000—0 0:... 00
000 0:0 00 0030 000 00000 0005 05.500050 08.30030 .0000 05 no 0000 0:0: 003 0:... :0 0000: 000 00300003000 0 .00 0&00 0000000 3300 000.0 05 :H

 

 

 

 

£0000 00 .523 503 0:0
.00 :00»: 00 00:3 3.0000
00 :05 .500: 00 :05.
"mag 55% .x

 

5:30:05 .00 0:000 000

30> 00:3 03000 0:955 000
30 0:0 .30> 00 00000000 m0>0
:0 5000:0000 00 000.00

0:0 0.8 30> .023 B 0:000
"Egg .0

 

185

400000 00

3.0.5005 .5550 0:000
00 0000000 00. 0300000000 00
£0500 0050000 00505080
00 80003000 0.5 0:038 05.
“Egon .H

 

.000

003000 .0050: 0:050 00.:
.2022 00 053 5 ea 2080
050 0:0 30> 0:3 00 00000305.
“g .=

 

 

.0000 3 00:3

mo 050.000.0005 50> .0050
00 050500005 Hao> £05
03000 05 .00 05500005 “30>
6sz0$. .0

 

88328. <

 

"00 g0 00 Em

 

 

. . .ZH Sham—>020

70.0003 .HN

 

. . .BSHEER

70.05. .ON

 

. . .hogb—OQLH

70.050 .mH

 

 

”SHE

70.050 83% < 00 W920

186

'me following questicms ask for your ideas about the relationships
between parts of conversations . For a acre catplete description of these
parts refer to the preceding section. We are particularly interested in
whether a change in one part of a conversation causes a change in another
Ert of a conversation. For instance, does a change in enotion produce a
change in closeness? Now if zero (0) indicates that a change in one
thing doesn't produce a change in' another and fl is a moderate amount of
chan in one thing produced by another (for example, in an increase in

strength of the wind may came a noderate degree of change in tenper—
ature) , then indicate the amount of change that each of the following
parts of conversation cause. You can use any number you wish. Please
answer each blank in each coltmn.

 

 

 

 

 

 

 

23. A change in EMJI‘ION causes what degree of change in:

A. PKEESS OF CQWEIBING (anount, ease of conversation, etc.)
B. MEANING (understanding what is said, the situation, etc.)
C. CINVERSATIQV OPTIONS (when to listen, when to speak, etc.)
D. CINI'E‘N'I‘ (the things you talk about)

E. IDLES (knowledge of who people are in terms of labels)

F. CLOSENESS (How distant you feel from others)

G. FORMALITY (degree of constraint or predictabiltiy)

H. ATTENTION (how much you concentrate on any one thing)

24. A change in (INTENT causes what 25. A change in MEANING causes

 

degree of change in: what degree of change in:

A. MEANING A. (IDNVERSATIW

B. WICN OPTIONS OPTIONS

C. PHDCESS OF (DNVERSING B. PIOCESS OF cou- ’

D. IDLES VEISING

E. CLOSENESS C. EDI-ES

F. mm D. CILBENESS

G. ATI'ENI'ICN E. FURMALITY
F. A'I'IEN'I'ICN

26. A change in CINVERSATION OPTIQIS 27. A change in PROCESS OF C(11-

causes what degree of change in: VEFBING causes what degree
of change in:

A. PIOCESS OF CINVEIBING

B. EDIE; A. IDLES

C. CLOSENESS B. CLOSENESS

D. FOIMALITY C. WWITY

E. A'I'I'ENI‘IQI D. A'ITEN'I‘IW

28. AchangeinmIEScauseswhat 29. AchangeinCIDSENESScauses
degree of change in: what degree of change in:
A. CLOSENESS A. FORMALITY

B. EDWIN
C. A'ITENI'ICN

30.

187

Achangeinl‘OHflLITYcauseswhat
degreeofdnngein:

A. ATTENTION

QUESTICNNAIRE NUMBER

 

Now please answer the following questions about yourself. The answers
to these questions will be kept strictly confidential and will be used
only for statistical purposes.

31.
32.
33.
34.
35.

36.

37.
38.
39.

40.

41.

42.

43.

How old are you? _YEAIB

Are you: MAI..E___ FW___

Are you: AN ONLY CHILD__ THE FIRST BORN____ IATER Bom—
Are you : SINGIE____ MARRIED___ WIDOWED__ DIVORCED___
How many peeple are you living with? ___NU1VBER OF PEOPLE

Indicate if you are the head of the household, or how you are
related to the head of the household.

 

How many radios are there in your household?
How many televisions are there in your household?

Who do you usually watch television with? (For example, son,
father, etc.)

 

Do you normally view television on a color set? YES NO

Approximately how many hours during the last week did you spend
talking with other people informally when neither the radio nor
televisicn was on?

HOURS

Would you call yourself a nenber of the: UPPER CLASS
mRKING CLASS

 

 

 

MIDDLE CLASS
LINER CLASS
What is your race? WHITE BLACK (NEGRD)
ORIENTAL O'H-IER (PLEASE SPECIFY

 

CHICANO (NEXICRN AMERIOXN)
NATIVE AMERICAN (AMERIOW INDIAN)

 

44.
45.
46.
47.

48.

49.

188

What is your occupation?

 

What is your annual incare?

 

Are you currently exployed? YES NO

 

How many hours a week do you typically work? (including housework)
__HOURS

How many years of schooling have you completed? __YEARS

Howmanyhoursof freetimedoyouhave duringatypicalweek?

HCIIIS

THANK YOU

APPENDIX E

Structural Equations for the Models
in Matrix Form

189
APPENDIX E

Structural Equations for the Models in Matrix Form

I. Equations relating observed and true variables in the

simple model (Model I).

A. The observed endogenous equations for all models.

 

 

 

 

 

 

 

 

 

yl U 0 0 1 81 n 0
1 l . I
yz U2 1 0 O 0 n2 0
y3 = u3 + O 0 0 1 n3 + 0
\Y41 1141 'o .1 o 0 V4] \84
A
y5 5 _D 1 0 0-1 as}

 

 

B. The observed exogenous in the radio and TV situations.

 

 

 

rxl‘ ’\3\ {O l 51 61
X2 = V2 + 0 *2 52 + 52
X3 V3 0 A3 63
\X‘” \v41 \1 o 0 l

 

 

 

 

C. The observed exogenous in the typical situation.

1 1 E1
= + +
E2

190

II. Equations for true endogenous variables in simple model.

A. Basic theoretical model (Model I).

 

 

 

 

 

 

 

 

 

P1 0 -a 0781 % 0.11?1 1c
13 1 1 ,1
‘021 1 0 ‘“24 n2 = 52 + C2
0 0 1 0 n3 Y31 Y32 Z;3
L° ° 1 :41 142111 Km

 

B. Model with paths between emotion and content and

between relationships and communication (Model II).

 

 

 

 

h 0 ”“13 ° ”1\ 1Y11 0 51 11A
"“21 1 0 “24 n2 = 0 Y22 g2 + 12
V 0 1 0 n3 Y31 Y32 \53
0 0 ‘a43 1 ”41 L:41 Y4EJ C41

 

 

C. Model with paths between all of the underlying and

surface variables (Model III).

 

 

 

 

 

 

'1 0 ’“13 ° 1 P1) Y11 Y12 g1 19d

'“21 1 ° ”“24 n2 = Y21 Y22 g2 1+ C2

0 ° 1 0 n3 Y31 Y32 C3

1 0 0 ’“43 1 11n41 Y41 Y42 1C4!
— .4

 

 

191

III. Equations relating observed and true variables in
model with common variable.

A. Observed endogenous equations for all models.

 

 

y (“1 {o o 1 o 11\ (n1 ,0
yz “2 1 o o o 12 n2\ 0
Y3 = H3 + o o o 1 A3 n3 + o
y4 u4 o 1 o 0 A4 n4 e4

 

 

 

 

 

 

 

 

|~<
-L_
I:
\i
O
>J
U1
0
O
>3
m
\
U1
‘_
('7
_J9

B. The observed exogenous in the radio and TV

situations.

 

 

 

 

 

 

 

 

 

T I .

X11 V1 0 1 17 151 (51\
v 0 A A E 6

X2 = 2 + 8 9 2 + 2

X3 V3 0 A10 111 f3 63

41 V41 0 A12} 1’ I

 

 

C. The observed exogenous variables in the typical

situation.
x V 0 l A g \ O
l = l + 7 1 +
x2 V2 1 0 A8 52 0
31

 

192
IV. Equations for true endogenous variables for all models

with common variable.

P— —'
1 0 -a13 o 0 n1 ,yll o o (:1 {;l
'“21 1 0 ’“24 0 n2 0 Y22 0 E32 + C2
0 0 ”“43 1 0 n3 = Y31 Y32 0 E3 53

 

 

 

 

 

 

 

 

 

 

APPENDIX F

Results of Tests of Model I
in Typical and TV Situation

88. v m .mm u “6 .852 u New

acoflmnfiflm PH. CA. H Hmong How mugmmm

 

 

 

H
an m b: $ng m me e
8. mm. 8. me._ me.
e
U
a» 28.50 m» 8893 m m H
0% 0
v mm.
mm Illlv Mk
3 V
m m 8. 4 szfiEBeoo

1 a

 

Ho.

 

Cl

47

 

mm.

 

 

88. v d .3 n me $92: u x

 

 

 

 

N ..
..coflmfinm Hmong E H Eco: now 338m
f mm mm magma He
8 I? mm 9 —
v» @6550 mm mmmoomm sag
ow/ no . J fl
me am - 953%
4
9
l 0
mo
on.
\\ . . I
Nu llllvms 1528 28.5%
H
mm . m m .
H .HH Ml VX V0

 

APPENDIX G

Maximum Likelihood Solutions
for Models Tested in Dissertation

195

Appendix G

Maximum Likelihood Solutions
for Models Tested in Dissertation

Introduction

 

In this appendix the computer printouts of the var—
ious tests of the models will be presented. Before the
printouts are presented each of the matrices contained in
them will be briefly explained. This appendix presents the
results of the LISREL tests reported elsewhere in the dis-
sertation in a slightly different format.

Figure 3 (in text) contains the operational version
of the model developed in Chapter I with appropriate LISREL
labels for the parameters. The two exogenous variables in
the model are emotion (:1) and relationships (52). The
observed variables that are indicants of the true exogenous
variables were described in detail in Chapter II. Emotion

only has a single observed indicant, x Relationship has

4.
three observed indicants: closeness, x1; responsiveness, x2;

and role, x This information is used to construct the ma-

3.
trix lambda x of observed exogenous indicants which will be
the same in every television and radio model, I through III.

Lambda x is presented below:

 

 

196

The 0's in all of the matrices represent zero's or parameters
that aren't estimated by the program. A6 and A9 in this model
are reference indicators for relationships and for emotion
respectively. Their values are always fixed at 1. A7 and
AB are ordinary indicators that are free to vary and thus
will be estimated by the program (Schoenberg, 1972). These
ordinary indicators represent the scale factors of these
variables.

The true endogenous variables in Figure 3 are content
(n1), communication (n2), interpretation (n3),and selection
(n4). There is only one observed indicator for content (y2),
interpretation (yl) and selection (y3). Communication has
two observed indicants: conversation options (y4) and process

of conversation (ys). The matrix lambda y which contains the

indicators of the true variables follows:

 

 

r— __.
O 0 Al 0
12 0 0 0
Ay = 0 0 0 l3
0 A4 0 0
.2 AS 0 g—
In all the models Al to A4 have a fixed value of l. 15 is
the only parameter in A y that is estimated.

The A(Beta in printout) matrix contains the paths be-

tween the true endogenous variables. The diagonal element

197

of each row of this matrix has a value of l. The basic model
only estimates 4 of the 12 possible paths between endogenous
true variables in this model. These paths are labeled a's

in the matrix. The actual A matrix is:

1-
1 0 a13 O
a l 0 d

A = 21 24
0 0 l 0
.3 0 0‘43 1 ._

 

 

The gamma matrix contains the paths between the true
exogenous variables and the true endogenous variables.
These paths are labeled with y. In Model I there are only 4
paths between true exogenous and endogenous variables; al-
though in some models that will be tested additional paths
will be added to this matrix. All of the paths in this model

will be estimated by LISREL. The gamma matrix is:

 

 

o o

o o
I":

Y31 Y32

Y41 Y42

L. ..

Phi (o) is the variance-covariance matrix of the ex-
ogenous variables. The on-diagonal elements of the matrix

represent the variances and the off—diagonal elements

198
the covariances. All of these parameters will be estimated
by LISREL where possible.

Psi (w) is the variance-covariance matrix of the resid-
uals, zeta (C), of the endogenous variables. In one set of
models (labeled a) only the variances of the residuals will
be estimated by means of LISREL. In the other set of models

(labeled b) all of the parameters within this matrix will be

estimated.
_' '7
02
C1
0 02
12
W = 2
0 0 0C
3
o o o o:
4
L... ..

 

 

The last two matrices, 68 and 65, represent the di-
agonals of the measurement error standard deviations of the
observed indicators. This feature builds into LISREL the
assumption of uncorrelated errors of measurement (Joreskog
and Van Thillo, 1972). The measurement error standard devia—
tions for the single indicators of variables will not be
estimated in this matrix in any of the models tested. The
measurement errors associated with these single indicators
will be contained in the residuals of the single indicators
true variables in the psi matrix. If this procedure is not
followed there are problems with identification, since one

piece of information would be used to estimate two different

199
parameters. The error of measurements standard deviations
estimated by LISREL for the observed endogenous indicators

will be the same in every model. Thus 68 will assume the

following form:

The errors of measurement standard deviations associated
with the exogenous observed variables will be the same for

the television and radio models and 66 will assume the

following form:

66 = [06 06 06 0]

200
APPENDIX G

Table 1G

Maximum Likelihood Solution for Model Ia in Radio
Situation with @ff-Diagonal Elements
of PSI Matrix Fixed at Zero

book)!“ buNl—J :5me Ul-bWNH

NH

hWNH

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 .763 0.000 0.000
LAMBDA X
l 2
0.000 1.000
0.000 .919
0.000 .831
1.000 0.000
BETA
l 2 3 4
1.000 0.000 -.536 0.000
-.054 1.000 0.000 -.656
0.000 0.000 1.000 0.000
0.000 0.000 .201 1.000
GAMMA
1 2
0.000 0.000
0.000 0.000
-.408 1.420
.131 .895
PHI
1 2
1.000
.606 .654
PSI
1 2 3 4
.713
0.000 .167
0.000 0.000 .218
0.000 0.000 0.000 .546

201

Table 16 (cont'd.)

THETA EPS
1 2 3 4 5
0.000 0.000 0.000 .603 .794
THETA DELTA
1 2 3 4

.589 .669 .741 0.000

202
Table 2G

Maximum Likelihood Solution for Model IIa with Paths
Between Emotion and Content and Relationship
and Communication with Off-Diagonal Elements
of PSI Matrix Fixed at Zero in Radio Situation

ubUJNH ubWNH (neuter—I

ubUJNH

bWNH

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 1.099 0.000 0.000
LAMBDA X
1 2
0.000 1.000
0.000 .867
0.000 .824
1.000 0.000
BETA
l 2 3 4
1.000 0.000 -.305 0.000
.192 1.000 0.000 -.146
0.000 0.000 1.000 0.000
0.000 0.000 .418 1.000
GAMMA
1 2
.510 0.000
0.000 .906
-.331 1.368
.103 1.166
PHI
1 2
1.000
.572 .664
PSI
1 2 3 4
.505
0.000 -.122
0.000 0.000 .166
0.000 0.000 0.000 .513

203

Table 2G (cont'd.)

THETA EPS
1 2 3 4 5
0.000 0.000 0.000 .747 .683
THETA DELTA
1 2 3 4

.580 .708 .741 0.000

204
Table 3G

Maximum Likelihood Solution For Model IIb with Paths
Between Emotion and Content and Relationship
and Communication with Off-Diagonal Elements
of PSI Matrix Free in the Radio Situation

thNI" bWNH U'lubWNl-J

ubWNI-J

waH

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 1.140 0.000 0.000
LAMBDA X
l 2
0.000 1.000
0.000 .910
0.000 .825
1.000 0.000
BETA
1 2 3 4
1.000 0.000 -.357 0.000
.591 1.000 0.000 -.373
0.000 0.000 1.000 0.000
0.000 0.000 .246 1.000
GAMMA
1 2
.487 0.000
0.000 1.047
-.399 1.402
.119 .945
PHI
1 2
1.000
.687 .661
PSI
1 2 3 4
.507
.207 .033
-.041 .067 .220
.039 -.081 .012 .548

205

Table 3G (cont'd.)

THETA EPS
l 2 3 4 5
0.000 0.000 0.000 .758 .668
THETA DELTA
1 2 3 4

.582 .673 .742 0.000

206
Table 4G

Maximum Likelihood Solution for Model IIIa with Paths
Between Underlying and Surface Variables with
Off-Diagonal Elements of PSI Matrix Fixed at
Zero in Radio Situation

bWNl-J .5me UI-hUJND-J

ubbJNH

ubbJNH

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 1.135 0.000 0.000
LAMBDA X
1 2
0.000 1.000
0.000 .922
0.000 .834
1.000 0.000
BETA
1 2 3 4
1.000 0.000 -.007 0.000
.170 1.000 0.000 -.138
0.000 0.000 1.000 0.000
0.000 0.000 .432 1.000
GAMMA
1 2
.293 .582
-.156 1.085
-.486 1.549
-.037 1.344
PHI
1 2
1.000
.605 .630
PSI
l 2 3 4
.487
0.000 -.123
0.000 0.000 .164
0.000 0.000 0.000 .511

207

Table 4G (cont'd.)

THETA EPS
1 2 3 4 5
0.000 0.000 0.000 .757 .670
THETA DELTA
1 2 3 4

.608 .682 .749 0.000

208

Table 5G

Maximum Likelihood Solution for Model IIIb with Paths
Between Underlying and Surface Variables with
Off-Diagonal Elements of PSI Matrix Free
in Radio Situation

NH :33me thH waH UlobWNi-J

hWNH

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 1.140 0.000 0.000
LAMBDA X
1 2
0.000 1.000
0.000 .910
0.000 .825
1.000 0.000
BETA
1 2 3 4
1.000 0.000 -.353 0.000
.135 1.000 0.000 -.571
0.000 0.000 1.000 0.000
0.000 0.000 .242 1.000
GAMMA
1 2
.486 .005
-.206 .700
-.399 1.402
.121 .941
PH I
1 2
1.000
.607 .661
PSI
1 2 3 4
.507
-.027 -.005
-.040 .059 .220
.039 -.209 .011 .548

209

Table SG (cont'd.)

THETA EPS
l 2 3 4 5
0.000 0.000 0.000 .758 .668
THETA DELTA
1 2 3 4

.582 .673 .742 0.000

210
Table 66

Maximum Likelihood Solution for Model II
in Typical Situation

UluwaH

DWNH

nwaF-J

ubbJNl-J

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 1.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 .806 0.000 0.000
LAMBDA X
1 2
0.000 1.000
1.000 0.000
BETA
1 2 3 4
1.000 0.000 .144 0.000
-.016 1.000 0.000 .157
0.000 0.000 1.000 0.000
0.000 0.000 .716 1.000
GAMMA
1 2
.598 0.000
0.000 .781
-.087 .892
.506 -.138
PHI
1 2
1.000
.567 1.000
PSI
1 2 3 4
.694
0.000 .129
0.000 0.000 .284
0.000 0.000 0.000 .618

THETA EPS

1

O . 000
THETA DELTA

l

0 . 000

211

Table 66 (cont'd.)

2 3 4 5
0.000 0.000 .369 .654
2

0.000

2l2
Table 7G

Maximum Likelihood Solution for Model II
in TV Situation

waH .5me U'lubWNH

mel—J

wal-J

LAMBDA Y
1 2 3 4
0.000 0.000 1.000 0.000
1.000 0.000 0.000 0.000
0.000 0.000 0.000 1.000
0.000 1.000 0.000 0.000
0.000 .640 0.000 0.000
LAMBDA X
1 2
0.000 1.000
0.000 .874
0.000 .681
1.000 0.000
BETA
1 2 3 4
1.000 0.000 -.328 0.000
.157 1.000 0.000 .192
0.000 0.000 1.000 0.000
0.000 0.000 -.277 1.000
GAMMA
1 2
.162 0.000
0.000 1.187
.045 .921
-.118 {676
PHI
1 2
1.000
.533 .682
PSI
1 2 3 4
.809
0.000 .364
0.000 0.000 .376
0.000 0.000 0.000 .473

213

Table 7G (cont'd.)

THETA EPS
1 2 3 4 5
0.000 0.000 0.000 .000 .768
THETA DELTA
1 2 3 4

.564 .692 .827 0.000

214
Table 83

Maximum Likelihood Solution for Model IV with
Common Variable in Radio Situation

mwaH ubUJNH U'lnwaH

thbJNH

111.5me

LAMBDA Y
1 2 3 4 5
0.000 0.000 1.000 0.000 .518
1.000 0.000 0.000 0.000 -.062
0.000 0.000 0.000 1.000 -.385
0.000 1.000 0.000 0.000 -.185
0.000 1.034 0.000 0.000 .296
LAMBDA X
1 2 3
0.000 1.000 .124
0.000 .866 -.202
0.000 .729 .164
1.000 0.000 -.337
BETA
1 2 3 4 5
1.000 0.000 -.431 0.000 0.000
.138 1.000 0.000 -.320 0.000
0.000 0.000 1.000 0.000 0.000
0.000 0.000 -1.819 1.000 0.000
0.000 0.000 0.000 0.000 1.000
GAMMA
1 2 3
.403 0.000 0.000
0.000 .681 0.000
.039 .868 0.000
-.212 -.733 0.000
0.000 0.000 1.000
PHI
1 2 3
1.000
.832 1.000
0.000 0.000 1.000
PSI
l 2 3 4 5
.498
0.000 -.061
0.000 0.000 .077
0.000 0.000 0.000 .165
0.000 0.000 0.000 0.000 0.000

215

Table 8G (cont'd.)

THETA EPS
1 2 3 4 5
0.000 0.000 0.000 .666 .595
THETA DELTA
1 2 3 4

.568 .596 .738 0.000

216
Table 9G
Maximum Likelihood Solution for Model IV
with Common Variable in Typical Situation

LAMBDA Y

00000 0

500000 500001 5 0
_

0 7 O 4
0 3 0 7
00000 01000 50
O O O O O I I O 0
400100 400010 4 00
.
0 45 9 01 0 0 5
0 43 4 00 0 0 1
00000 72 50010 00000 0 100
O O O O O O O I I O O O O O O O O O I
310000 300 310100 300001 3 l 3 000
_ _
03 0 0 040 7 l
05 AU 0 607 4 1
00002 00 00000 02340 4.0 0000
O O O O O O O O I O O O O O 0 O O O O O O O 0
200011 210 201000 200000 2 00 2 0000
.
0 0 01 6 43 53 4
0 0 02 9 99 49 0
00000 00 01000 20030 930 20000
I O O I O I O I O O O O O O I O O O O I
101000 101 110000 100000 1000 100000
.
X
A
D
P A
m T I I
E H S
L B G P P

123.45 12 12345 12345 123 12345

217

Table 96 (cont'd.)

THETA EPS

l 2 3 4 5

0.0 0.0 0.0 0.410 0.652
THETA DELTA

l 2

WNH U'IthNI-J WQWNH waH UlanNH

woman—-

LAMBDA Y

LAMBDA X

BETA

GAMMA

PHI

PSI

1

0.000
1.000
0.000
0.000
0.000

0.000
0.000
0.000
1.000

1.000

.602
0.000
0.000
0.000

1
.546
0.000
-.995
1.640
0.000

.606
.487
0.000

.573
0.000
0.000
0.000
0.000

218

Table 106

2
0.000
0.000
0.000
1.000
.588

1.000
.877
.682

0.000

0.000
1.000
0.000
0.000
0.000

2

0.000
2.788
1.883

-2.077

0.000

.578
0.000

-.492
0.000
0.000
0.000

3

1.000
0.000
0.000
0.000
0.000

-.074
-.l30
-.363
-.627

-.343
0.000
1.000

-1.666

0.000

0.000
0.000
0.000
0.000
1.000

1.000

.051
0.000
0.000

4

0.000
0.000
1.000
0.000
0.000

0.000
1.043
0.000
1.000
0.000

.193
0.000

Maximum Likelihood Solution for Model IV
with Common Variable in TV Situation

5

-.353
.161
.184
.229
.362

0.000
0.000
0.000
0.000
1.000

0.000

219

Table 10G (cont'd.)

THETA BPS
1 2 3 4 5
0.000 0.000 0.000 .000 .736
THETA DELTA
l 2 3 4

.645 .734 .774 0.000

REFERENCES

220

REFERENCES

Adams, B.N. Interaction theory and the social network.
Sociometry, 1967, 29, 64-78.

 

Althauser, R.P. Multicollinearity and nonadditive regres-
sion models. In H.M. Blalock (Ed.), Causal models
in the social sciences. Chicago: Aldine-Atherton,
1971.

 

 

Amidon, E.J. & Hunter, E. Improving teaching: Analyzing
verbal interaction in the classroom. New York:
Holt, Rinehart & Winston, 1966.

 

 

Argyle, M. Social interaction. New York: Aldine-Atherton,
1969.

 

Auld, F. & White, A.M. Sequential dependencies in psycho-
therapy. Journal of Abnormal Social Psychology,

Bales, R.P. Interactiongprocess analysis. Reading, Mass.:
Addison-Wesley, 1950.

 

Barker, R.G. The stream of behavior: Explorations of its
structure and content. New York: Appleton-Century-
Crofts, 1963.

 

Bjerg, K. Interplay analysis: A preliminary report on an
approach to the problems of interpersonal understand-
ing. Acta Psychologica, 1968, 28, 201-245.

 

Blalock, H.M., Jr. Correlated independent variables: The
problem of multicollinearity. Social Forces, 1963,
43, 233-237.

 

Borgatta, E.F. The analysis of patterns of social inter-
action. Social Forces, 1965 , 44,27-34.

 

Borgatta, E.F. & Crowther, B. A workbook for the study of
social interaction processes. Chicago: Rand
McNally & Co., 1965.

 

Bostrom, R.N. Patterns of communicative interaction in
small groups. Speech Monographs, 1970, 31, 257-263.

 

Bower, R.T. Television and the public. New York: Holt,
Rinehart & Winston, 1973.

 

221

Brittan, A. Meaning and situations. Boston: Routledge &
Keganpaul, 1973.

Carter, L.F., Haythor, W., Meirowitz, B. & Lanzetta, J.A.
A note on a new technique of interaction recording.
Journal of Abnormal Social Psychology, 1951, 46,
258-260.

Cicourel, A.V. Basic and normative rules in the nego-
tiation of status and role. In D. Sudnow (Ed.),
Studies in social interaction. New York: The Free
Press, 1972.

Coffin, T.E. Television's impact on society. American
Psychologist, 1955, l9, 630-641.

Costner, H.L. Utilizing causal models to discover flaws in
experiments. Sociometry, 1971, 34, 398-410.

 

Costner, H.L. & Schoenberg, R. Diagnosing indicator ills in
multiple indicator models. In A.S. Goldberger & O.D.
Duncan (Eds.), Structural equation models in the
social sciences. New York: Seminar Press, 1973.

 

Crano, W.D. & Brewer, M.B. Principles of research in social
psychology. New York: McGraw-Hill Book Company,
1973.

Crowell, L. & Scheidell, T.M. Categories for analysis of
idea development in discussion groups. The Journal
of Social Psychology, 1961, 54, 155-168.

 

Duncan, O.D. Introduction to structural equation models.
New York: Academic Press, 1975.

Ervin-Tripp, S. An analysis of the interaction of language,
topic and listener. American Anthropologist, 1964,
66, 86-94.

Farrar, D.E. & Glauber, R.R. Multicollinearity in regres—
sion analysis: The problem revisited. Review of
Economics and Statistics, 1967, 42, 92-107.

Fink, E.L. Structural equation modeling: Unobserved
variables. In P.R. Monge & J.N. Cappela (Eds.),
Multivariate techniques in communication research.
New York: Academic Press (Forthcoming).

Fink, E.L. & Mabee, T.I. Linear equations and non-linear
estimation: A lesson from a non-recursive example.
Unpublished paper. Department of Communication,
Michigan State University, E. Lansing, Michigan.

222

Fisher, B.A. & Hawes, L.C. An interact system model:
Generating a grounded theory of small groups.
Quarterly Journal of Speech, 1971, 51, 444-453.

 

Flanders, N.A. Interaction models of critical teaching
behaviors. In E.J. Amidon & J.B. Hough (Eds.),
Interaction analysis: Theory research and applica-
tion. Reading, Mass.: Addison-Wesley, 1967.

Gouran, D.S. & Baird, J.B. An analysis of distributional
and sequential structure in problem solving and
informal group discussions. Speech Monographs,
1972, 39, 16-22.

 

Goudy, W.J. Non-response effects on relationships between
variables. Public Qpinion Quarterly, 1976, 49,
360-369.

 

Goffman, E. Alienation from interaction. Human Relations,
1957, l9, 47-60.

 

Goldberger, A.S. Structural equation models: An overview.
In A.S. Goldberger & O.D. Duncan (Eds.), Structural
equation models in the social sciences. New York:
Seminar Press, 1973.

 

 

Hackman, J.R. & Morris, C.G. Group tasks, group interaction
process, and group performance effectiveness: A
review and proposed integration. In L. Berkowitz (Ed.),
Advances in experimental social psychology. New York:
Academic Press, 1975.

 

Hamilton, R.V. & Lawless, R.H. Television within the social
matrix. Public Opinion Quarterly, 1956, 30, 393-403.

 

Hare, A.P. The dimensions of social interaction. Paper
presented at meetings of the American Sociological
Society, Seattle, Washington, 1958.

 

Hauser, R.N. & Goldberger, A.S. The treatment of unobserv-
able variables in path analysis. In H.L. Costner (Ed.),
Sociological Methodology. San Francisco: Jossey-Bass
Publishers, 1971.

 

Hawes, L.C. Elements of a model for communication processes.
Quarterlnyournal of Speech, 1973, 59, 11-21.

 

Heise, D.R. Causal inference from panel data. In E.F.
Borgatta (Ed.), Sociological Methodology. San
Francisco: Jossey-Bass Publishers, 1970.

 

223

Hempel, C.G. Fundamentals of concept formation in
empirical science. In Neurath,0. Uh1)International
Encyclopedia of Unified Science, (2, no. 7).
Chicago: University of Chicago Press, 1952.

 

Ittelson, W.H., Rivlin, L.G. & Proshansky, H.M. The use of
behavioral maps in environmental psychology. In
H.M. Proshansky, W.H. Ittelson, & L.G. Rivlin (Eds.),
Environmentalypsychology: Man and his physical
setting. New York: Holt, Rinehart & Winston, 1970.

Jaffe, J. & Feldstein, S. Rhythms of dialogue. New York:
Academic Press, 1970.

 

Johnson, J.D. The effects of media presence and media
content on social interaction. Unpublished grant
report. E. Lansing, Mich.: Department of Communica-
tion, Michigan State University, 1976.

 

Johnson, J.D. The effects of media presence and media
content on social interaction. Unpublished pre-
iiminary paper. E. Lansing, Mich.: Department of
Communication, Michigan State University, 1976.

 

Joreskog, K.G. Analyzing psychological data by structural
analysis of covariance matrices. In D.H. Krantz,
R.C. Atkinson, R.D. Luce & P. Suppes (Eds.), Contemp-
ory developments in mathematical psychology measure-
ment, psychoPhysics, and neural information process-
iflg (Vol. II). San Francisco: W.H. Freeman & Co,
1974.

Joreskog, K.G. A general method for estimating linear
structural equation system. In A.S. Goldberger &
O.D. Duncan (Eds.), Structural equation models in
the social sciences. New York: Seminar Press, 1973.

 

Jareskog, K.G. A general method for analysis of covariance
structures. Biometrika, 1970, 51, 239-251.

 

Joreskog, K.G. & Sorbom, D. Statistical models and methods
for analysis of longitudinal data. In D.J..Aigner
& A.S. Goldberger (Eds.), Latent variables in socio-
economic models. New York: North-Holland Publishing
Co., 1977.

 

Joreskog, K.G. & Van Thillo, M. LISREL: A general computer
program for estimating a linear structural equation
system involving multiple indicators of unmeasured
variables. Princeton: Educational Testing Service,
1972.

 

224

Kees, W. & Ruesch, J. Function and meaning in the phy-
sical environment. In H.M. Proshansky, W.H.
Ittelson, & L.G. Rivlin (Eds.), Environmental
psychology: Man and his physical setting.

New York: Holt, Rinehart & Winston, 1970.

 

 

Kerlinger, F.N. & Pedhazur, E.J. Multiple regression in
behavioral research. New York: Holt, Rinehart
& Winston, 1973.

 

Kim, J. & Kohout, F.J. Special tOpics in general linear
models. In N.M. Nie, C.N. Hull, J.G. Jenkins,
K. Steinbrenner & D.N. Bent (Eds.), SPSS: Statis-
tical package for the social sciences. New York:
McGraw-Hill Book Company, 1975.

 

 

Kivlin, J.B. Contributions to the study of mail back bias:
Introduction. Rural Sociology, 1965, 39, 322-326.

 

Klein, L.R. & Nakamura, M. Singularity in the equation
systems of econometrics: Some aspects of the
problem of multicollinearity. International Eco-
nomic Review, 1962, 3, 274-299.

 

 

Koopmans, T.C. Identification problems in economic models
construction. Econometrica, 1949, ll, 125-143.

 

Land, K.C. Principles of path analysis. In E. F. Borgatta
(Ed.), Sociological methodology: 1969. Jossey-
Bass, Publishers, 1969.

 

Lehman, E.C., Jr. Tests of significance and partial returns
to mail questionnaires: The problem of partial
returns. Rural Sociolggy, 1963, 28, 284-289.

 

Lennard, H.L. & Bernstein, A. Patterns in human interaction.
San Francisco: Jossey—Bass, Publishers, 1969.

Lewis, W.W., Newell, J.M. & Withall, J. An analysis of
classroom patterns of communication. Psychological
Reports, 1961, 9, 211-219.

Im>Sciuto,InAu A national inventory of television viewing
behavior. In B.A. Rubinstein, G.A. Comstock &
J.P. Murray (Eds.) Television and social behavior:
Reports andypapers, volume iv. Rockville, Maryland:
NIMH, 1972.

 

225

Lyle, J. Television in daily life: Patterns of use over-
view. In E.A. Rubinstein, G.A. Comstock, & J.P.
Murray (Eds.), Television and social behavior:
Reports andypapers, volume iv. Rockville, Maryland:
NIMH, 1972.

 

Longabaugh, R. The structure of interpersonal behavior.
Sociometry, 1966, 29, 441-460.

 

McGinnies, E. & Altman, I. Discussion as a function of
attitudes and content of a persuasive communication.
Journal of Applied Psychology, 1959, 43, 53-59.

 

Malone, T.W. System simulation: Computer simulation of
two-person interactions. Behavioral Science,
1975, 29, 260-267.

 

Mayer, L.S. & Younger, M.S. Multiple indicators and the
relationship between abstract variables. In D.R.
Heise (Ed.), Sociological methodology: 1975.
San Francisco: Jossey-Bass, Publishers, 1974.

 

MacCoby, E.E. Television: Its impact on school children.
Public Opinion Quarterly, 1951, l5, 421-444.

 

McGuire, M.T. & Lorch, S. Natural language conversation
modes. Journal of Nervous and Mental Disorders,
1968, 146, 221-229.

Pope, B. & Siegman, A.W. Relationship and verbal behavior
in the initial interview. In A.W. Siegman and
B. POpe (Eds.), Studies in dyadic communication.
New York: Pergamon Press, 1972.

Robinson, J.P. Toward defining the functions of television.
In B.A. Rubinstein, G.A. Comstock & J.P. Murray (Eds.),
Television and social behavior: Reports and papers,
volume iv. Rockville, Maryland: NIMH, 1972.

 

Robinson, J.P. Television's impact on everyday life: Some
cross national evidence. In E.A. Rubinstein, G.A.
Constock & J.P. Murray (Eds.), Television and social
behavior: Reports and papers, volume iv. Rockville,
Maryland: NIMH, 1972b.

Rogers, W.T. & Jones, S.E. Effects of dominance tendencies
on floor holding and interruption behavior in dyadic
interaction. Human Communication Research, 1975,

l, 113-122.

 

226

Ruesch, J. & Prestwood, A.R. Interaction processes and
personal codification. Journal of Personality,
1949, 33, 391-430.

 

Rockwell, R.C. Assessment of multicollinearity: The
Haitovsky test of the determinant. Sociological
Methods and Research, 1975, 3, 308-320.

 

 

Scheidel, T.M. & Crowell, L. Feedback in small group com-
munication. Quarterly Journal of Speech, 1966,
33, 273-278.

 

Schoenberg, R. Strategies for meaningful comparison. In
H.L. Costner (Ed.), Sociological methodology: 1972.
San Francisco: Jossey-Bass, Publishers, 1972.

 

Schwirian, K.P. & Blaine, H.R. Questionnaire return bias
in the study of blue collar workers. Public Opinion
Quarterly, 1967, 39, 656-663.

 

 

Seibold, D.R. A complex model of multidimensional attitude
and overt behavior relationships: The mediating
effects of certainty and locus of control. Unpub-
lished Ph.D. dissertation. E. Lansing: Department
of Communication, Michigan State University, 1976.

Snyder, W.U. An investigation of the nature of non-
directive psychotherapy. Journal of General Psy-
chology, 1945, 33, 193-223.

 

Speier, M. How to observe face-to-face communication: A
sociological introduction. Pacific Palisades,
California: Goodyear Publishing Company, 1973.

 

Stech, E.L. Sequential structure in human social commun-
ication. Human Communication Research, 1975, 3,
168-179.

 

Stein, C.A. A simple user's guide to LISREL. Unpublished
paper. Amherst: Department of Communication
Studies, University of Massachusetts, 1976.

 

Steiner, G.A. The people look at television. New York:
Alfred A. Knopf, 1963.

 

Steinzor, B. The development and evaluation of a measure
of social interaction: Part 1, the development and
evaluation of reliability. Human Relations, 1949,
3, 103-121.

 

Strupp, H.H. Psychotherapists in action. New York:
Grune & Stratton, 1960.

 

227

Suchman, B.A. An analysis of "bias" in survey research.
Public Opinion Quarterly, 1972, 33, 102-111.

 

Taylor, F.K. The three-dimensional basis of emotional
interactions in small groups. Human Relations,
1954, 1, 441-471.

 

Theil, H. Principles of econometrics. New York: John
Wiley & Sons, 1971.

 

U.S. Bureau of the Census. County and city data book:
1972. Washington, D.C.: U.S. Government Printing
Office, 1973.

 

U.S. Bureau of Census. Statistical abstract of the United
States: 1972. Washington, D.C.: U.S. Government
Printing Office, 1972.

 

 

Van De Geer, J.P. Introduction to multivariate analysis
for the social sciences. San Francisco: W.H.
Freeman & Company, 1971.

 

 

Watson, J. Formal analysis of sociable interaction.
Sociometgy, 1958, 33, 269-280.

 

Watzlawick, P., Beavin, J. & Jackson, D. Pragmatics of
human communication. New York: W.W. Norton &
Company, 1967.

 

 

Weick, K.E. Systematic observational methods. In G.
Lindzey & E. Aronson (Eds.), The handbook of social
psychology: volume two. Reading, Massachusetts:
Addison-Wesley Publishing Company, 1968.

 

 

Weintraub, W. & Aronson, H. The application of verbal be-
havior analysis to the study of psychological de-
fense mechanisms: Methodology and preliminary report.
Journal of Nervous and Mental Disease, 1962, 334,
169-181.

 

Werts, C.E., J6reskog, K.G. & Linn, R.L. Identification
and estimation in path analysis with unmeasured
variables. American Journal of Sociology, 1973,
13, 1469-1484.

Werts, C.E. & Linn, R.L. Path analysis: Psychological
examples. Psychological Bulletin, 1970, 14,
193-212.

 

228

Wheaton, B., Muthen, B., Alwin, D.F. & Summers, G.F.
Assessing reliability and stability in panel models.
In D.R. Heise (Ed.), Sociological methodology: 1977.
San Francisco: Jossey-Bass, Publishers, 1977.

 

Wiley, D.E. The identification problem for structural
equation models with unmeasured variables. In
A.S. Goldberger & O.D. Duncan (Eds.), Structural
equation models in the social sciences. New York:
Seminar Press, 1973.

 

 

Wright, 5. Correlation and causation. Journal of Agri-
cultural Research, 1921, 20, 557-585.

 

 

“7'11’71‘1111111311111115