------_-I

THE CGNCEP? 0F PROCESS IN
HUMAN COMMUNICATION RESEARCH

Thesis for the Degree of Ph, I).
MICHIGAN STATE UNIVERSIIY
ROBERT BURTON ARUNDALE

1971

IIIIIIIIIII 1

 

 

I

This is to certify that the

thesis entitled

The Concept of Process in
Human Communication Research

presented by

Robert Burton Arundale

has been accepted towards fulﬁllment
of the requirements for

Ph 0D 0 degree in ngmgni cation

 

ngﬁoéuQ269ﬁzhAahu/

Major professor

Date February 25 , 1 971

0-7639

  

LI BRAR . 1
Michigan State
University

ABSTRACT

THE CONCEPT OF PROCESS IN
HUMAN COMMUNICATION RESEARCH

By
Robert Burton Arundale

The concept of process is frequently applied in discussing
conmmnication, apparently because it identifies the important dynamic
aspect of this fcmnlof behavior. However, the concept is only occa-
sionally operationalized in research on human communication. To
remove the discrepancy between discussion and researCh, it appears
useful to examine the concept of process and its explication in detail.,

.At a basic level, the paper is concerned with time as a funda-
mental dimension of events, with the study of suCh events using the
scientific method, and with the class of events termed "communication.
These concerns are unified in the study of the concept of process as
it relates to the discipline of communication. [The major emphasis in
the paper is to seek, as a step in theory building in communication,
the means of explicating the concept of process.

A definition of communication as "a process of transmission of
structure among the parts of a system . . ." appears both general and
useful in this context. The concept of process is central in this
definition, and if distinguished from related concepts can be defined
as "all Change over time of matter energy or infornation.;//Ihe concept

is important in its relationShip to research designs and in its utility

‘with regard to certain problems and concerns in human communication

Robert Burton.Arundale
research. Nevertheless, an examination of researdh reveals that the
concept of process has not been widely operationalized. An explication
of the concept appears needed to correct this discrepancy and to make
the concept more available as a tool in researdh and theory building.

The first step in explicating the concept is to provide a
constitutive definition. The paper does not consider a specific
theory, so that it is necessary to provide not a unique definition but
rather a set of approaChes to constitutive definition. Sudh approadhes
.may involve either verbal terms or mathematical terns, and the necessary
and sufficient conditions for using the ternl"process" are stated in
both forms. The approadhes to constitutive definition also carry a
number of implications fOr the form of theory WhiCh deals with com-
‘munication events as processes.

The second step in explicating the concept of process is to
provide an operational definition. Because no specific research
problem is considered, it is necessary to develop a set of approadhes
to operational definition, rather than a single definition. Examina-
tion of these approadhes results in a set of necessary and sufficient
conditions for describing (or measuring) a process. These conditions
have several direct implications fbr researCh techniques and certain
indirect implications for analysis techniques.

The result of seeking the means of explicating the concept of
process, i.e., of seeking approadhes to constitutive and operational
definitions, is a set of tools for dealing with the concept in specific

theoretic and researCh frameworks. Such researCh is exemplified by

RObert Burton Arundale
‘work in language development, interaction analysis, and attitude change.
In general, though, the principles discussed appear to apply fairly
broadly in human conmunication research.

The general implication of the paper is that if theory building
and researdh on human communication are to deal with events as processes,
it will be necessary to use and to develop not only appropriate fOrms
lof’theory, but also researCh and analysis techniques capable of dealing
lNith the dimension of time. The use and development of suCh teChniques
will present many problems to the communication scholar, but will also
‘provide an important class of information on Change over time that is
:not now available. The concept of process is therefOre inportant in
ihuman communication researdh, even though it is but one of several

concepts whidh require careful study and explication.

THE CONCEPT OF PROCESS IN

HUMAN COMMUNICATION RESEARCH

By
Robert Burton Arundale

A THESIS
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Communication

1971

CC") Copyright by
ROBERTIRHUUNIAWﬂﬂMﬂE

1971

Accepted by the faculty of the Department of Communication ,
College of Communication Arts , Michigan State University , in partial
fulfillment of the requirements for the Doctor of Philosophy degree .

ﬂuéwe OT ﬂw

Director of Thesis

 

Guidance Committee: @WQ 5? “WW; , Chairman
44% 76:42.... 4

4144/! 11/

 

    

 

ACKNOWLEDGI‘IENTS

In Book II of The Prelude, William.WOrdsworth invokes the

 

image of the teaCher in considering the sources of What he had been

taught:
Who knows the individual hour in whidh
His habits were first sown, even as a seed?
Who that shall point as with a wand and say
'This portion of the river of my mind
Came from.yon fountain?‘ (ll. 20u—208)

In acknowledgment of their'help, in many different ways, in
making this dissertation possible, I must mention several individuals.
In particular, Jay weston.was a source of information as the seeds of
the ideas herein began to form. David Berlo added critiques and
encouragement as the ideas took initial shape. James Noonan's views
on language as behavior in time revealed some of the value of the
growing ideas. However, it remained for Clyde Henson to make me
seriously question the status quo of communication researCh. He did
so by showing me poetry, live and dynamic. His encouragements, to-
gether'with those of Miles Martin, were most helpfu1. James Campbell,
as well, provided encouragement and many important comments. The
encouragements and help of all of these individuals has been invaluable,
though in the final analysis, the encouragement and help of one stands
paramount-—thank you, Wendy, fOr making possible this small accompliSh-

ment.

ii

Beyond all these I mmet acknowledge the work of my committee:
Randall Harrison, Erwin Bettinghaus, and Hideya Kumata, as well as
of many other friends both known and unknown who have spurred.ideas,
‘mentioned sources, been patient, and helped in other-ways. Again, as
WOrdsworth.suggests, it is difficult to point to all the sources of
one's thoughts and work. Nevertheless, he points to Coleridge, as one
of his teadhers, and says of him what should be said of all scientists:
Science appears but what is truth she is,
Not as our glory and our absolute boast,
But as a succedaneumn and a prop
To our infirimity. No officious slave
Art thou of that false secondary power
By which we multiply distinctions, then
Deem that our puny boundaries are things

That we perceive, and not that we have made.

The Prelude, Book II, 11. 212—219

 

iii

TABLE OF CONTENTS

LISTOFTABLES . . . . . .
LISTOFFIGURES . . .

INI'RODUCI‘ION........

CHAPTER 1: THE UNIVERSE OF DISCOURSE

1.1 Some Basic Assumptions and Directions

The plan . . . . .

1.2 A.Definition of'Commmnication .

A process . . . . . .
Of transmission . . . .
Of structure . .

Among the parts of a system whiCh are
identifiable in time and space

1.3 Criteria fOr Choosing the Definition

Generality . . . . .
Utility . . .

CHAPTER 2: THE IMPORTANCE OF PROCESS IN STUDIES OF

HUMAN OOMMLNICATION . . .

2.1 .A Closer'Look at the Concept of Process
2.1.1 Process and Communication . . .
2.1.2 Process and other Important Concepts
2.1.3 Process and ResearCh Designs .

Point designs . . . .
Difference designs .
Process designs .

O

2.2 Criteria fOr Choosing the Concept of Process .
2.2.1 Utility in TWO General Problems .

Spurious categorization

Prediction and understanding
2.2.2 Utility in Two New Concerns

HUman information processing

iv

Page
viii

ix

\IU‘I

11
12
11+
15

16

18
19
23

26

26
26
28
32
35
36
38

I41
I43
143
1+6
50
50

CHAPTER 2: (contd)

System development . . . . . . . .

3 The Current Place of Process in Research . .
2.3.1 A Classification of Research Designs . .
2 3.2 Relative Use of Researdh Designs . .

Time series designs . .

Separate-sample with time extension design

Multipleawave panel design . . . . .
2.3.3 A Bigger Picture . . . . . . . . .

CHAPTER 3: APPROACHES TO A.CONSTITUTIVE DEFINITION

3.1 Nature of the Definition . . . . . . . .
3.2 Verbal.Aspects of Constitutive Definition . .
3.2.1 The Verbal Definition of Process . . . .
The individual terms . . . . . . .
Additional comments . . . . .
3. 2. 2 Process in Relation to other'Terms . .
3. 2. 3 Types of Processes . . . . .

3.

.3 Mathematical ASpects of Constitutive Definition

3.3.1 The Mathematical Definition of Process .
The mathematical terminology . .
Additional terms . . . .

3.3. 2 Process in Relation to other Terms . . .

3.3.3 Process and its Mathematical Expression .
Discrete and continuous expressions . .
Finding an expression . . . .

3. 3. l4 Necessary and Sufficient Conditions for
Use of the Term."Process" . . . . .

u Implications fOr the Form of Theory . . . .
General implications and comments . . .
Specific implications, comparisons,

and comments . . . . . . . . .

In summary . .

CHAPTER u: APPROACHES TO AN OPERATIONAL DEFINITION

u.

1+.

1 Nature of the Definition . . . . . . .

2 General Principles of Operational Definition
u.2.1 Discrete Measurement of Fbrms of Variation
Frequency of measurement

V

Pretest-posttest with time extension design .

Page

52

65
56
59
62
62
67
67
71

77

77

81
82
8Q
87
89
9M

98
101
103
107
111
115
115
120

12H

127
128

133
138
1H0
1H0
1H2

1u3
11m

CHAPTER H: (contd) Page

Minimum.number of measurements . . . . . . 1N8
Length of measurement period. . . . . . 150

u. 2. 2 Necessary and Sufficient Conditions fOr
Description of a Process . . . . . . . . 151

u.3 Implications for Research Tedhniques . . . . . . 15H
u.3.1 Direct Implications fOr ResearCh Design . . . 156
Variable value and time . . . . . . . . 157
Measurement interval and resolution . . . . 158
Length of measurement period . . . . . . . 161

3.2 Some Specific Process Designs . . . . . . . 16H
.3.3 Implications for Measurement . . . . . . . 170

In summary . . . . . . . . . . . . . 176

u.u Implications for Analysis Tedhniques . . . . . . 178
U.H.1 Basic Distinctions and Considerations . . . . 181
u.u.2 Reduction Techniques . . . . . . . . . . 18h
Comments .‘ . . . . . . . . . . 185
u. u. 3 Testing Tedhniques . . . . . . . . . . 188
Comments . . . . . . . . . . . . . 192
In summary . . . . . . . . . . . . . 19H

An overview . . . . . . . . . . . . 197

CHAPTER 5: EXAMPLES . . . . . . . . . . . . . . 200
5.1 Comments on the Examples . . . . . . . . . . 201

5.2 The Examples, Individually and In Sum, . . . . . 205
5.2.1 Individual Examples . . . . . . . . . . 205
Brown and Bellugi (196A) . . . . . . . . 205

Harrison (1969b) . . . . . . . . . . . 208

Insko (196u) . . . . . . . . . . . . 210
Other‘researCh . . . . . . . . . . . 213

5.2.2 The Examples In Sum . . . . . . . . . . 21”

CHAPTER 6: CONCUUSIONS AND IMPLICATIONS . . . . . . . . 217
6.1 A Review . . . . . . . . . . . . . . . 217

6.2 Implications and Place of "Process" in the
Study of Human Communication . . . . . . . . 220

vi

Page
LIST OF REFERENCES . . . . . . . . . . . . . . . 228

APPENDIX A: Sdhematic Presentation of Experimental
Designs . . . . . . . . . . . . . . 2H1

APPENDIX B: Derivation of MinimumnNumber~of'Measurements . . 2M7

vii

LIST OF TABLES

Table Page
1 Scientific Goals and the Foci of Research . . . . . 1+7
2 A Twoéway Classification of ResearCh Designs . . . . 57
3 Outline and Glossary of Verbal Aspects of

Constitutive Definition . . . . . . . . . . 83
ll Outline and Glossary of Mathematical Aspects of

Constitutive Definition . . . . . . . . . . 100
5 A.Comparison of Analogic and Symbolic Forms of

Theory................137
6 Reduction Tedhniques fOr Analysis of Data from

Process Designs . . . . . . . . . . . . . 186
7 Testing Techniques for Analysis of Data from

Process Designs . . . . . . . . . . . . . 191
8 Summary and Comparison of Research Examples . . . . 2014

viii

LIST OF FIGURES

Figme Page
1 A Complex Signal and Its Components . . . . . . 1146
8—1 General Sinusoidal Signal . . . . . . . . . 2“?

ix

INTRODUCTION

 

Most scholars engaged in the study of human communication will
agree with Berlo (1960) in choosing the phrase, "the process of 9_o_r_n_—

munication," to describe the focal point of their discipline. They

 

will, of course, explicate the concept of communication in many
different ways, devoting a great deal of time and attention to con-
structing viable constitutive and operational definitions . In this
effort, they will frequently make use of the associated concept of
process, but will, for the most part, avoid the task of constitutively
and operationally defining it. This paper will attempt to offset this
lack of attention, both in seeking the means of explicating the concept
of process, and in considering its implications for theory and research.
Why study "process"? My choose this concept for attention from
among others? In very broad outline, because it appears to be a basic
concept in the study of human commumication, but one which is seldom
examined either constitutively for use in theory or operationally for
use in research. In narrower outline, there exists a large body of re-
view and discussion (see Barnlund, 1968, pp. 3-29; Berlo, 1960, pp. 23-
28; Bettinghaus, 1960, p. 16-28; Harrison, 1969a; among others) which
appears to agree both that communication behavior takes place across
time, and that the temporal aspect of such behavior can be meaningfully
expressed by identifying "communication" as a form of "process."

Theoretical formulations and research techniques in communication often

1

2
fail , however , to incorporate the concept in any particularly useful
manner. The result is that a potentially large and important class of
information cannot become available--information on how commmication
behaviors develop , evolve , and interact during the course of an event ,

or in short, on how they change over time. Because an explication of

 

m appears to be a useful step in making this class of information
more available, this paper will attempt to study "process," in particu-
lar, and to suggest a set of approaches and tools for incorporating the
concept in theory building and research .

Several comments need to be made regarding this central concern
with the concept of process. One comment is that the term "process,"
as it is frequently applied, has a very broad and complex usage (e.g. ,
Berlo, 1960, pp. 23-28). This paper will narrow the broad usage of
"process" by introducing a set of closely related concepts that are
helpful in isolating the several different aspects of events often en-
compassed by the term. Such narrowing is not an "oversimplification"
in any pejorative sense. It is, rather, an essential and useful step
in examining the concept of process , both because it provides needed
tools , and because it reveals that the broader usage of the concept of
process may "hide" a number of complex and problematic aspects of events.

The narrating of "process , " and particularly the introduction of
related concepts , leads to another comment . Although this paper will
focus entirely on the concept of process , the concept is only one of
many which require attention because of their importance in the conduct
of theory building and research. It is the particular utility of the

concept of process which dictates its choice from among others , but only

3
with the recognition that it is not , in the broader view, the only con-
cept to be considered. Similarly, while this paper will focus almost
entirely on that form of theory and research which incorporates the
concept of process, there is no assumption, whatsoever, that such theory
and research is the only form which has value. Indeed, many studies of
human commmication do not require the consideration or use of the con-
cept , and clearly much valuable work has been done which has not incor-
porated it. The intent of the paper is not to disparage such studies,

but to suggest additional approaches and tools for those areas of theory

 

building and research on htman communication where the concept of process
must be considered and used.

Although the paper will be cemrally concerned with only a single
concept, a comment is still needed regarding the level of that concern.
Quite clearly, the task of seeking the means of explicating the concept
of process will involve consideration of a number of different approaches
and tools for constructing constitutive and operational definitions .

With the exception of one discussion of specific applications , however,
this paper will not apply such approaches and tools, but will only suggest
them and consider their implications for the discipline of commmication
as a whole. The actual details of applying the tools must be left to

the individual scholar working in a specific framework of theory and
research. If the distinction between theoretical and experimental
studies is to be made, then, this study must needs fall with the former,
though it will in no case lose sight of the latter.

One final comment, which also regards the level of concern with

the concept of process. In a very broad sense, this paper will suggest

u

the need for extending current views on the conduct of theory building
and research. The nature of the needed extension will be most evident
in the discussions of the implications of "process" for theory and re-
search, and will in many cases involve the refinement or development of
techniques. Accordingly, while this paper will consider a number of
problers and questions regarding the place of the concept of process in
the study of human communication, it can be expected to uncover a great
many more problems and questions at a lower level. There will, indeed,
be much to add after the last sentence.

We shall begin studying the concept of process, then, in Chapter 1,
by explicitly identifying several basic assumptions and directions , as
well as by defining the concept of communication. This particular defini—
tion, like the others in the paper, is presented only as a point of
reference, not as the "correct" or "true" definition. It is from the
definition, however , that the paper develops its focus on the concept of
process, and more specifically, on its importance, explication, implica-

tions , and use.

CHAPTER 1
THE UNIVERSE OF DISCOURSE

 

"Space and time are the framework in which all reality is
concerned . We cannot conceive any real thing except under the
conditions of Space and time." So wrote Ernst Cassirer in An Essay
m (19%, p. l$2) in introducing a discussion of man's perception
of space and time in the world about him. To be completely precise,
of course, he would have had to speak not of man's perception of "space
gig time," but of his perception of "space-time," as Alfred North
Wnitehead showed, following the lead of Albert Einstein, some #0 years
ago (1925, Ch. 7; 1929, Chs. 2, 10). The relationship of space to time
does not , however, make it any less meaningful to emphasize either
space or time, alone, in dealing with specific problem, whether
problem of physics or of behavior .

1.1 Some Basic Assumptions and Directions

 

Accordingly , we shall be concerned primarily with time , and we
shall take as given the statement that "all phenomena, as perceived by
human beings, both occur, and can be measured, along a dimension of
time."1 Once again, in Cassirer's words, "EX/en time [like space] is

first thought of . . . as a general condition of organic life.

 

1That is, the assumptions made here will not be discussed at
length. See, particularly, Gibson (1966), Kelly (1963, Ch. 1), and
Whitehead (1925, Ch. 7; 1929, Chs. 2, 10).

5

6
Organic life exists only so far as it evolves in time. It [life] is not
a thing but a process—a never-resting continuous sheen of events. In
this stream nothing ever recurs in the same identical shape (19%, pp.
Ll9-50)." Life, then, is a moving succession of events in a dimension
or "condition" of time—a "process" in Cassirer's terms.

If one accepted these four sentences from Cass irer without
qualification , one would view life as a succession of unique events ,2
and would require that each event have a unique explanation. Such a
person could not be a scientist, for his viewpoint would prevent him
from accepting a single explanation for even as few as two events . The
scientist would , of course , agree that individual events are unique ,
but he would add the crucial point that there are properties and patterns

in events which are independent of those individual events and are there-

 

fore observable in whole classes of events (Dubin, 1969, p. 3'4). These
independent properties and patterns are the core of the scientist's
concern, both in his study of unique events , and in his attempts at
explanation of events through theory .

We shall take the position of the scientist , here, with regard
to uniqueness and repetition in events , and we shall assume that the
scientist's primary goal is the building of theory to aid in the expla-
nation, prediction, and/ or control of events . As Ridner has phrased it,
". . . it is an ideal of science to give an organized account of the
1miveree--to connect, to fit together in relations of subsumption, the

statements embodying the knowledge that has been acquired (1966, p. 11)."

 

2"Event" is used here in the sense of a bounded interval of
time.

7

Quite obvionely, "the universe" for which the scientist gives an account
in his theory is in most cases a "universe of discourse"; that is, some
limited class of events which is identified either implicitly or ex-
plicitly by the scientist as the detain to which his theory applies .
This somewhat smaller "universe" is frequently identified by a single
concept3 whose definition plays an important role in two early steps of
ﬁneory building. In the f_°_1rit step a scientist takes , the definition of
ﬁne concept functions in specifying the particular properties and
patterns in events which are of primary interest, ﬁne class of which
events constitutes the scientist's universe of discourse (Dubin, 1969,
(In. 5; Rudner, 1966, p. 35). Also, in what is frequently a m step,
the definition acts in specifying a set of key concepts which may be
central in the theory building and for which the scientist must provide
some form of explication .

The plan. - In ﬁne broad sense of an overall plan for this paper,
we shall be concerned with the above two early steps in theory building:
defining ﬁne universe of discourse and providing explication for key
concepts. The universe of discourse here is that specified by ﬁne
concept "communication," and we shall begin by devoting ﬁne reminder of

Chapter 1 to the first step in theory building by defining and discussing

 

aspects of ﬁnis concept , and by considering ﬁne criteria for the choice

 

3We shall follow Rudner's usage of "concept," i.e. , "tenm
denoting certain characteristics, or term applicable to entities said
to have certain characteristics (1966, p. 18n; cf. Kaplan, 1961+, pp. 146--
52)." In particular, we shall not distinguish "construct," or other
typologies of terms , from "concept ," noting both that such distinctions
are utilitarian, not ontological (Kaplan, 1961+, pp. 59-60), and ﬁnat
some authors (Berlo, 1967; G. R. Miller, 1967) use "concept" and
"construct" as virtually interchangable.

8

of the definition. An explicit definition is given not because the
universe of discourse specified by "commnmication" is in great need of
redefining, but because the pleﬁnora of such definitions makes a point
of reference necessary. The definition formulated here is therefore
not a ﬁneory or a model” of communication, but only a first step in
that direction. Once the universe of discourse has been defined, the
scientist may then move on in ﬁne direction of eiﬁner (l) the construc-
tion of parts of a ﬁneory, or (2), as is more likely for ﬁne communica-
tion scholar-scientist, ﬁne construction of a typology of events which
may eventually be incorporated into a theory (Harrison, 1967b, pp. 2-3;
Rudner, 1966, pp. 28-|+0).

Neither of these two general directions in theory building can
proceed , however , without the important second step of providing expli-

 

cation for key concepts specified by ﬁne definition (Berlo, 1967, p. 2).
We shall , in fact , beginning in Chapter 2 , devote our attention primarily
to ways of aocarplishing this specific task in theory building. At that
point we shall select from the definition of communication a single
concept which unites the concerns expressed above in time as a dimension,
with those expressed in the scientific study of events , particularly
events in ﬁne universe of discourse of communication. That single
conceptuthe concept of process--is chosen partly for reasons of length ,
but primarily because it is seen as a highly useful concept in the

 

l"'l‘bdel" is used here in Rudner's sense of ". . . an alternative
interpretation of the same calculus of which the theory itself is an
interpretation (1966, p. 211)." The definition might be said to be a
"descriptive model," but ﬁnis usage has rot been adopted here (of.
Brodbeck, 1959, p. 379).

9
conduct of research and ﬁneory building related to hunan communication .
The concept itself, ﬁne criteria for its choice , and its current place
in research will be discussed further in Chapter 2. As a major
enphasis, then, we seek as a preliminary step in theory building in
communication , the means of explicating a concept which is seen as use-
ful in the further conduct of ﬁnat theory building.

Because ﬁne general concern is with explication , it is important
to note two steps or requirements in this task before returning to ﬁne
definition of comnmication. A concept is said to be explicated , and
hence admissible for use in observation , experiment, and ﬁneory, if and
only if it has been provided with both an operational definition and a
constitutive definition. The operational definition specifies a pro-
cedure , under specific conditions , whose performance by an investigator
will identify (i.e. , measure or generate) in nature a situation which is
the referent of ﬁne concept . The constitutive definition , on the other
hand, specifies a linkage to oﬁner terms in the code, language, or
especially ﬁneory , which the investigator manipulates to produce propo-
sitions about ﬁne events under study (Berlo, 1967, pp. 24!, 8, 11).

Operational and constitutive definitions of a concept are normally
constructed and applied only when the concept is used wiﬁnin a specific
observational, experimental, and theoretical framework (cf. Rudner, 1966 ,
p. 20). However, because this paper will not deal wiﬁn a single research
and theory building situation, such as would be found in a report of an
experiment , the specific conditions under which operational and consti-
tnrtive definitions are applied will be absent. As a result, reﬁner ﬁnan

discuss a single operationalization for the concept of process , we shall

10

consider both ﬁne general principles for operationalizing it , and ﬁne
implications of these principles for research (and analysis) techniques .
Similarly , instead of linking ﬁne concept to oﬁner terms of a specific ,
but prerature , ﬁneory of communication , we shall link the concept to
oﬁner terms of the language which might be used in a future ﬁneory of
communication, and shall consider the implications of the linkages for
ﬁne form of ﬁnat theory.S These approaches to an operational and a
constitutive definition , respectively , constitute what was referred to
earlier as a search for ﬁne m of explicating the concept under study.

In brief sumnary, ﬁnen, we shall begin by defining ﬁne concept
of communication, and by considering the choice of the definition , in
order to identify both ﬁne universe of discourse and a set of key con-
cepts requiring explication ( Chapter 1). We shall select one of these
concepts, ﬁnat of process, for special attention, and shall discuss the
concept itself, ﬁne criteria for its choice , and its current place in
research in communication (Chapter 2) . Since ﬁne "special attention"
involves seeking means of explicating the concept , we shall consider
approaches to boﬁn a constitutive definition (Chapter 3), and an opera-
tional definition (Chapter I1:) . Finally, we shall turn to examples of

ﬁne use of the concept in research (Chapter 5) , and to conclusions
(Chapter 6).

 

sMnether a given term (concept) is a primitive , is derivable from
other terms, or is unnecessary in a particular theory cannot be com-
pletely determined until that theory is fully formalized (Rudner, 1966 ,
pp. 18-19, #7, 51-52).

11

1.2 A Definition of Communication

 

We shall use ﬁne term "communication" to refer to "a mess

 

of_transmissicn of structure among ﬁn: parts _<3f a system which are

identifiable in time and space (Krippendorff, 1969a, p. 1021"5 Since

 

the properties and patterns in ﬁne events specified by this definition
are rot immediately obvious , it will be useful to examine each phrase
and its key concept in detail . Each of these concepts should , ideally ,
be explicated,7 but as indicated earlier , space dictates that only one
concept be considered at length . Consequently , the examination of each
concept will be limited here to furﬁner definition, and to discussion
and examples of its scope.

One additional point must be emphasized before examining the
definition of communication in detail. Each definition in this paper is
provided primarily to establish a point of reference and to identify the
writer's usage. In no case is a given definition considered to be the
"correct" or "true" definition, nor is the assurption made that ﬁne
reader will adopt the usage indicated as the "correct" usage. Rather,
each definition is considered to be an "equivalent verbalization , " used

only to reduce ambiguity; that is, it is not "Eh: meaning" of the term.

 

5Definitions of communication abound. There is little value in
reviewing these definitions here when reviews already exist, and probably
even less value in creating a new definition (see Barnlund, 1968, pp.
3-1u; Krippendorff, 1969a, pp. 106-109, and 1969b, p. 5-6; see also
Barnlund, 1968, pp. 17-29; Bettinghaus, 1960, p. 16-28; Harrison, 1969a;
Sarbaugn, 1968, pp. 1-2, on models of communication). The definition
given here will be related to others, below, in the discussion of the
criteria for its dnoice .

7No claim is made for ﬁne exhaustiveness of ﬁnis set of concepts .

12

A process. - Krippendorff, in commenting on his definition,
notes ﬁnat "The notion of 'process' subsumes that of behavior or changes
over time (1969b, p. 7)." More formally, "process" is used here, in
J. G. Miller's sense, to refer to "All change [including movement] over
time of matter—energy or information . . . (1965a, p. 209)." Hence,

a communication event is seen as occurring along a dimension of time ,
and as having ﬁne characteristic ﬁnat changes in the properties of ﬁne
event take place between the points in time which are ﬁne boundaries of
the event. There are, of course, a number of other possible uses of ﬁne
term "process," many of which have a wider scope than the definition to
be used here. Attention will be given to ﬁnese other uses, as well as
to ﬁne particular definition used here , early in Chapter 2 .

Notice ﬁnat ﬁne ‘term "event" is paired with the term "commmica-
tion" in ﬁne discussion above. Such usage is felt to be both useful,
and consistent wiﬁn Krippendorff's definition in two respects. First,
he uses the phrase "3 process," and hence identifies the properties of
a specific situation, or unique event, which may be termed a communica-
tion _e_y_e_n_t_. The properties identified, however, are also present in a

class of specific situations which may be termed communication events .

 

The addition of the term "event" or "events," then, serves to indicate
whether a specific situation or a class of situations , both labeled
"canmnmication," is being referred to (i.e. , a "communication experiment,"
involves an event, while "communication ﬁneory" and "discipline of com-
mnunication" involve classes of events).

Second, and perhaps more important, the use of ﬁne term "event"

recognizes an important aspect of the studyiof any process. As defined

13

above, a process has no bounds, but in fact may be (and often is) con-
sideed to have continuity over a_2_L_1_ time . Obviously , no scientist can
observe or study such a process, 1x932: as certain properties or
patterns in ﬁne process repeat themselves over intervals of time which
ﬁnat scientist can recognize and delimit. These bounded intervals are
events, as defined earlier, and it is ﬁnese events which ﬁne scientist
studies , wheﬁner or not he explicitly identifies the interval boundaries .
That the choice of interval boundaries is arbitrary, and is established
by ﬁne observer, not ﬁne event, we shall take as given. An event is
then a specific portion of a process which may have continuity over time ,
as in fact communication may have. Yet since events are ﬁne "obj ects" or
"situations" which a scientist studies, references to a "process" in this
paper will be to a process as it takes place wiﬁnin a bounded interval
(or event). 8 Again, these points and oﬁners concerning ﬁne concept of
process will be discussed further in Chapters 2, 3, and H.

Examples of processes range from trivial to highly sophisticated .
For example , the change in state from ice to water is a process (matter-
energy), as is the fading of a color photograph in the sun (matter-energy
3nd information) . Similarly, the release of energy in a controlled
nuclear reaction is a process (matter-energy) , as is the formation of a
single frame of a color television picture (matter-energy _a_r_nd information).
Note that in ﬁne second two cases, the process referred to is the actual

release or formation, not necessarily the complex nuclear power or

 

8Krippendorff's examples (1969a, pp. 108, 109; 1969b, pp. 5, 6)
make clear ﬁnat ﬁnis usage of "process" is consistent wiﬁn his, even
thougn he does not specify this use.

19
television operations necessary to bring ﬁnese processes about . But
this distinction raises an interesting point. Are these latter, complex
operations , incorporating m_a_ny_ changes , to be considered as many
individual processes (dnanges over time), or as one process? Usage here
will tend toward ﬁne latter as ﬁe more general case, since ﬁnis usage
also includes references to single processes . Note also ﬁnat in ﬁne
above examples only one of ﬁne four processes is associated wiﬁn com—
munication. The reason lies in the additional concepts which form the
definition of communication.

Of transmission. - The term "transmnission" is used here to refer

 

to ﬁne movemnt of matter-energy or information over space (of .

J. G. Miller, 1965a, pp. 1914-199). As such, transmission is a form of
process , but is more restricted in that mess includes 21th the change
of state and ﬁne movement of matter-energy (or information), whereas

transmission includes only ﬁne movement of matter-energy . Movenent over

 

space usually involves movemnt in time , as well (the space-time rela-
tionship), and in certain cases the time dimension of the transmission
is of primary importance.

Examples of transmissions are the ﬂow of heat in a home heating
system, ﬁne radiation of signals from a television transmnitter, the flow
of electrical energ in a power network , and the placement and discovery
of ﬁne Rosetta stone. In the latter example , time is certainly ﬁne
primary aspect of the transmission, althougln movement in space is involved,
as well. TWO of ﬁnese examples are transmissions associated with com-
munication, the reason being, in part, that ﬁney are transmissions of

"structure . "

15

Of structure . - Krippendorff notes in a highly technical , ﬁnough

 

very geeral definition ﬁnat ". . . structure should be understood in
the maﬁnematical sense: as a many valued relation, as a complex pattern,
as an above-chance distribution in a multi-dimensional space (1969a,

p. 107)." "Structure," then, is used here to refer to ﬁne pattern or
organization among ﬁne parts or units of some entity, as opposed to
chaos or carplete lack of organization. Krippendorff notes ﬁnat the
term "infornmtion," when used in sense of statistical information theory
(hence, ﬁne opposite of uncertainty), is a possible substitute for ﬁne
term "S‘tI‘UC‘hJI‘G. "

It is important to note , though, ﬁnat the transmission of infor-
mation (or structure , here), is always accompanied by the transmission
of matter-energy, since information is always borne on "markers" [e.g. ,
stones , ink and paper, chemical compounds , sound and electrical waves
(J. G. Miller, 1965a, pp. 199 and 1910], In the context of the defini-
tion, "structure" is used to refer to the informational or organiza-
tional aspects of a transmission, rather than to its matter-energy
aspects (the markers). Such a use of the term to refer to information
will henceforth be designated "structurei" to distinguish it from ﬁne
use of the term in Chapter 3 to refer solely to arrangements of matter
in space. This latter use will be designated "structurem. "

Ebemples of structurei are the arrangerent of signals or pulses
in a television transmission, and the linguistic patterns which appeared
an ﬁne Rosetta stone. Likewise, the "code" of a DNA molecule is an
example of structure: , as are the patterns formed in the wide range of

human bodily moveIents. Each of these four examples is one of

16
structurei associated with communication, but ﬁne reason is that each
situation also fulfills the final qualification in the definition.

Among the parts of a system which are identifiable in time and

 

_s_p§_c3. - The first part of this long phrase is the most important, and
revolves around ﬁne key concept of system. The term "system" is used
here in ﬁne same sense in which J. G. Miller applies the term in general
system behavior theory (1965a) . Since the concern in ﬁnis paper will
be wiﬁn "concrete," rather ﬁnan wiﬁn "concep " or "abstracted"
system, the term "system" is used to refer to . . . "a Ion-random
accumulation of matter-energy, in a region of physical space-time, which
is non-randomly organized into coacting , interrelated subsystem or com-
ponents (J. G. Miller, 1965a, p. 202)."

The enphasis on non-random organization in this definition of
"system" refers in part to ﬁne "common properties" among the subsystem
or components , some of which may be: proximity and similarity of parts ,
common information , fate , or duration , and maintenance of a single
boundary ". . . over which there is less transmission of matter-energy
or information than there is within the system . . . (J. G. Miller,
1965a, pp. 200, 2114)." The "non-random organization" of a system also
refer to ﬁne interrelationships between its sybsystens or components
(ﬁne units). In Miller's terms: "The state of each unit is constrained
by, conditioned by, or dependent on the state of other units. The units
are coupled. Moreover, there is at least one measure of the sum of its
units which is larger than the sum of ﬁnat measure of its units (1965a,
pp. 200-201)."

While this brief summary does some injustice to Miller's treatment

of the concept of system, it should suffice for the present purpose of

l7
examining ﬁne above phrase from ﬁne definition of communication. Given
Miller's me of ﬁne concept of system, it is evident ﬁnat the second
part of the above phrase ". . . which are identifiable in time and space,"
may be considered an addition to the definition wholly for the sake of

enrinasis, since in Miller's usage a concrete system must be located "in

 

a region of physical space-time." The first part of ﬁne above phrase,
"among the parts of a system . . . ," is less easily disposed of , however,
and raises an important point.

According to ﬁne first part of ﬁne phrase , communication takes
place E._th_1_n_ a particular system, a terminology which will be followed
here. This system may be, for example , ﬁne male and female of a species
and ﬁne embryo at the point of conception (DNA is the structurei trans-
mitted to ﬁne offspring), or a client and psychiatrist conducting an
interview (language, bodily movements, are ﬁne structuresi). Each of
these situations can be readily identified as a concrete system within
which there is transmnission among the parts or subsystem. In each case,
as well, the subsystems are also complex systems in themselves, at a ‘
lower level. On the other hand, it is somewhat less easy to identify
the concrete system within which ﬁne transmission of structurei takes
place in either the discovery of ﬁne Rosetta stone, or the formation of
a television picture. In these cases, the system which is most readily
identified is the archaeologist or ﬁne television set; however , these
system are only parts or subsystems (at a lower level) of ﬁne system
within which ﬁne transmission takes place . These latter systems are con-
pcsed of the maker and discoverer of ﬁne stone , and the television trans-

mitter and receiverb-system which , despite their temous appearence ,

18
are communication systems in ﬁne sense ﬁnat the concept of system is
used here.

These four very brief examples of communication system can also
be used in summarizing the key aspects of the definition of communica-
tion, before moving on from this consideration of specific phrases and
concepts to a consideration of the definition as a whole and ﬁne criteria
for its choice. Again, "commnication" is used here to refer to "a
process of transmnission of structure:-L among the parts of a system which
are identifiable in time and space." In particular, the transmission
of geretic material between generations of a species , ﬁne transmission
of linguistic and kinesic patterns between individuals , the transmniss ion
of a combination of linguistic codes between people over mnillenia, and
ﬁne transmission of a coded electro-magnetic signal from a transmitter
to a receiver, are all events whose properties and patterns qualify

under the definition to be termed "communication . "

l. 3 Criteria for Choosing the Definition

 

The choice of a particular definition or statement out of a set
of possible alternatives requires at least implicit consideration of two
factors. One factor is the range of alternatives to the definition or
ﬁne statement, and ﬁne second is the set of criteria on which ﬁne choice
is based. Because the choices made at certain key points in this paper
have an effect on its outcore , we shall explicitly identify at these
points boﬁn the range of possible alternatives , and the criteria for the
choice among ﬁnem. The first of these key points is ﬁne choice of the
definition of communication-a choice made on the basis of two primary

criteria, those of generality and utility.

19

Generality. - One aspect of the definition's generality can be

 

seen in the broad scope of phenonena which it includes . This aspect is
closely related to a second aspect of the definition ' s generality-~ﬁne
enconpassing of ﬁne universes of discourse of a range of alternative
definitions . The first aspect, ﬁne broad scope of phenomena covered by
the definition, can be seen in part in the examples given above, to
which might be added examples of the transmission of cultural patterns ,
of various persuasive messages , end of animal vocalizations , enong
others (Krippendorff, 1969b, p. 6). The reason for the definition's
broad scope is, of course, that it identifies a fairly small set of
concepts which specify only a limited number of properties end patterns
in events (Dubin, 1969, Ch. 5).

Certainly , many investigators would not wish to consider
such a broad scope of phenomna in their research, and would implicitly
or explicitly ne'row such a definition by restricting the scope of one
or more of its key concepts. Such restriction could take several form,
and would be entirely valid, as long as the universe of discourse of any
resultant theory were seen as similarly restricted. For example , an
investigator mnight restrict the scope of ﬁne transmission involved to
cover only "face-to-face" or "interposed" channels , or only decoding or
encoding. The structurei could be limited, say, to written, spoken, or
non-verbal, as well as to sub-categories wiﬁnin these (e.g., persuasive,
informative, consumatory) . Restrictions on the system are most frequent ,
however, and often begin with a restriction on type, as in human,
animal, or machine. System restrictions may also be in terms of number

of subsystem, as in "interpersonal? "mass," or "self"; in terms of

20

subsystem function, as in sender, receiver, or both; and in terms of
various system or subsystem "states" like meaning, intent , awareness ,
effect, and attitude. These examples are, of course, only a‘few of ﬁne
many possible ways of restricting the scope of the present definition of
communication. They serve to emphasize , however, not only the generality
of ﬁne definition itself, in terms of ﬁne phenomena which it includes ,
but also ﬁne generality of each of the concepts used in the definition.

The generality of the concepts used means , in particular , that
ﬁne universe of discourse specified by the definition is sufficiently
broad to encompass ﬁne domains of many alternative definitions of com-
munication. Krippendorff has reviewed a number of such definitions
(1969a, pp. 106-109; 1969b, pp. 5-6), end has found their domains to be
included in ﬁne universe of discourse of ﬁne definition which he
fornmnlates .9 It is worthwhile to extend ﬁnis review, as well as this
second aspect of the present definition ' s generality , by brieﬂy
exemining ﬁnree other definitions of communication.

The definition given by J. G. Miller in his discussion of general
systems behavior theory is simnilar in many ways to that used here:
". . . ﬁne change of information from one state to another or its
movement from one point to another over space (1965a, p. 198)."
"Information" is, of course, used here in the sense of structurei, and
since Miller does restrict "communication" to the movement of markers

in space (1965a, p. 191+), it is evident that his phrase "change of

 

gTlne definitions include ﬁnose of Ashby , Deutsch , Gerbner,
G. H. Mead, G. R. Miller, Wiener, L. White, and others.

21

information from one state to anoﬁner" must involve some form of trans-
mnission. The systems context of Miller's remarks makes clear, also,
ﬁnat ﬁne transmission he refers to is one occurring wiﬁnin or between
systems. He adds at a later point (1965b, p. 392-301)), however, that
en exchange of information between two systems is sufficient to create
an "information boundary in space-time," i.e. , a boundary which defines
the larger system wiﬁnin which the communication takes place (see the
discussion of ﬁne concept of system, page 16ff). When these factors
in ﬁne context of Miller's definition are taken into account , it be-
cones clear that the universe of discourse he specifies is essentially
identical with ﬁnat Specified by Krippendorff '8 definition.

A second definition of communication formulated recently by
Berlo10 specifies the conditions under which ﬁne term may be applied:

If: two X's (in which X is a system),

between which energy is transferred,
that energy being informational and

particularly a symbol,
Then: communication.
Provided that J. G. Miller's comments on an "information boundary" can
be applied here, this set of conditions resembles those given in
Krippendorff '8 definition, i.e. , "transmission of structurei among the
parts of a system." The one exception, of course , is that Berlo
specifies ﬁnat the structure;L have the characteristics of a symbol—-a
discriminable unit of matter-energl which bears information (hence , a

sign) and which is used in place of, or as a substitute for, some other

 

loPersonal commmnication, August 18,1969. See also D. K. Berlo,
_T_h__e Process o_f_’ Human Communication, forthcoming.

 

22

unit of matter-energy . In Berlo ' 3 treatment of the concept of symbol ,
however, it also appears ﬁnat the ability to symbolize is largely re-
stricted to human beings , so ﬁnat the effect is one of limiting the
type of system involved, as well. Given these restrictions on struc-
turei end on system, then, it is clear ﬁnat the universe of discourse
specified by Berlo ' 5 definition is encompassed by ﬁnat of Krippendorff '3
definition.

The third definition is one formulated by Ackoff (1958), and
shom a form of restriction beyond that of Berlo '5 definition:

If en individual Ib responds to a set of signs
selected by Ia in a purposeful state, ﬁnen Ia is the
sender end 113 is the receiver of ﬁne message.
Several aspects of ﬁnis definition should be noted.

First, Ia and lb may be the sere individual. That is,

a person may commicate to himself as in writing a

'reminder' to hinmelf. Secondly, ﬁne sender of the

message need rot intend or desire to communicate to ﬁne

receiver in order to do so. The interceptor of a message,

for example , is communicated to , although unintentionally .

Thirdly, ﬁne sender and receiver may be widely separated

in time and space (1958, pp. 229- 225).
Taking into account both these comments by Ackoff , and ﬁnose made above
concerning sign end symbol, it appears that Ackoff restricts structurei
end system in much ﬁne sene ways as Berlo does. Hmever, Ackoff chooses
to identify an even smaller domain by including in his definition a sub-
system attribute, i.e. , ﬁne requirement of a "purposeful state."11
The result is ﬁnat the domain which he specifies is a subset of ﬁne
universes of discourse identified by both Krippendorff '5 definition,

end Berlo '3 definition.

 

11"Response" is defined in terms of "purposeful state (1958 , p.
22%)," therefore making "purposeful state" the key restriction.

2 3
Were it not for limitations of space , it would be worﬁnwhile to

examine a number of additional definitions of communication (Barnlund ,
1962; Cherry, 1957; Fearing, 1953; etc.), both to expend the renge of
possible alternative definitions , and to establish more fully the
generality of the present definition. It would also be useful to
discuss one of a few definitions, such as that of Stevens ,12 whose
dorain is not encompassed by the present definition, but which instead
specifies enother set of phenomena. Space is limited , however, not only
because ﬁnis is not a review of definitions, but also because there is
a second criterion in the choice of ﬁne definition to be considered.
Utility. - The broad generality of Krippendorff '3 definition,
which is one factor favoring its selection, immediately raises the
question of utility-«a criterion which Berlo has termed ". . . the
primary criterion for the efficacy of eny scientific operation (1967 ,
p. 9)." Is ﬁne definition perhaps too general to be useful? As Kaplen
has pointed out, ﬁne making of a value judgnent of utility in the
"enalysis of meenings " requires attention boﬁn to the particular context
in which the "action" (ﬁne forming of a definition) is taken, as well as
to "ﬁne purposes which the action as a whole is meant [s_i_g] to achieve
(1961}, pp. 390—392, 1+2-|+6)." Consequently, if ﬁne present context were
that of a specific research program, the definition (as it stands) would
probably be too general to be of use, for it would fail in its purpose

 

12"Communication is the discriminatory response of an organism
to a stimulus . This definition says that communication occurs when
some environmental disturbance (ﬁne stimulus) impinges on an organism
end the organism does something about it (makes a discriminatory re-
sponse). If the stimulus is ignored, there has been no communication
(Stevens, 1950, p. 689)."

21+
of specifying a "workable" donain.

The context in which ﬁne definition is used in ﬁnis paper, how-
ever, is ﬁnat of ﬁne early steps in theory building , especially as that
ﬁneory building occurs in the discipline of commnication. The purpose
of ﬁne definition in ﬁnis context , ﬁnen , is ﬁne identification of a
nniverse of discourse, and a set of key concepts , which are broad
enough to encompass a large portion of the set of phenonena which have
in the past been labeled with ﬁne term "cormunication."l3 Phrased in
ﬁnis manner, ﬁne purpose both requires ﬁnat ﬁne definition integrate a
set of phenomena which might oﬁnerwise be seen as disparate , ﬁnereby
acting as a heuristic device, and requires ﬁnat the discussion apply
beyornd the limits of any one Specialized interest in communication .

The definition of communication chosen earlier is judged as highly use-
ful in fulfilling ﬁnis purpose.

Generality end utility, ﬁnen , are ﬁne two criteria in the choice
of ﬁne definition given above from the renge of possible alternative

11+

definitions . This choice of a general def inition , with its require—

ment of broadly applicable discussion, requires that one final comment

 

13It is exu'eIely important to note that the purpose of ﬁne
definition is not to identify a universe of discourse and a set of
concepts whichﬁcompass all phenomna labeled "communication . " The
definition chosen here , despite its generality , does in fact exclude
many phenonena so labeled, as Krippendorff has been careful to—po'lﬁt
out (1969a, pp. 105-109; 1969b, p. 6, and especially pp. 23-26).

1”"0bservability" of the concepts exployed, and hence of ﬁne
events specified, is anoﬁner possible criterion for consideration in
such a choice. Honever, while ﬁne concepts used, and ﬁne events speci-
fied, are considered observable, this does not distinguish ﬁnem from a
great many other definitions which also meet the criterion.

25

be made before turning in the next chapter to the concept of process
itself , the criteria for its choice , and its current place in research .

At points in ﬁnis paper references will be made , implicitly end
explicitly, to comunication events in which at least one subsystem is
a human being. Accordingly , references to "communication research" will
insonecasesbereferences tothemorerestrictedareaofhnmanconh'
munication research , even ﬁnough by definition, ﬁne observation of com-
munication events , end the building of theory to explain and predict
ﬁnose events, is a much broader endeavor (Krippendorff, 1969a). Such
restriction of reference is entirely a result of ﬁne writer's interest
in human communication research, an area of research for which the dis-
cussion of ﬁne concept of process is seen as especially relevant. These
references to hunan communication will , of course , be combined wiﬁn
references to oﬁner research areas such as telecommmication—-an area
for example , in which the concept of process is utilized almost
implicitly. Regardless of ﬁne area referred to, however, care will be
taken at all points, in keeping with the requirement above, to assure
ﬁnat ﬁne concepts and techniques discussed can be applied (in principle ,
at least, if not in utmost utility) to ﬁne range of phenomena encom-

passed by ﬁne present definition.

CHAPTER 2
'I'HE IMPORTANCE OF PROCESS 1N STUDIES OF HUMAN COPMJNICATION

 

In ﬁne previous chapter , the particular definition of conmunica-
tion to be used here was corpared to several other alternative defini-
tions . In each conparison , however , one key concept was inherent in
ﬁne definitions which did not receive any special recognition . The
concept was that of process, somtimes expressed as its subset, trans-
mission. With the definiticn of the universe of discourse complete,
it is possible in ﬁnis chapter to examine this key concept of process
in much greater detail. In so doing, it will be useful to examine not
only ﬁne concept itself more closely, but also ﬁne criteria for choosing
process from the alternative concepts , and ﬁne current place of ﬁne
concept in research on human communication. Such a discussion leads
directly to considering the means of explicating ﬁne concept in
Chapters 3 and in .

2.1 A Closer Look at the Concept of Process

 

Turning first to the concept itself, it appears worthwhile to
exemine its relationship to ﬁne concept of communication , to several

other important concepts, and to ﬁnree classes of research designs.

2.1.1 Process end Commmication
In the many definitions of communication extant today (Barnlund,

1968, pp. 3—l|+), it has become almost comonplace to at some point
26

27
acknowledge ﬁnat "communication is a process "--ﬁne definition chosen
here being no exception . The same aclcnowledgnent also appears frequently
in the discussions of the various general models of communication events
(Barnlund, 1968, pp. 17-29; Bettinglnaus, 1980, pp. 16-28; Harrison,
1969a). This continual re-emphasis, of course, lends a great deal of
the weight of experience to ﬁne identification of communication as a
form of process. A definition, however, is merely a statenent of
equivalence expressed in verbal form, and is not sufficient , alone , to
establish the relationship between the concept of commnication end the
concept of process .

Rather, ﬁne relationship between the two concepts will be estab-
lished by returning to ﬁne early, basic assurption ﬁnat "all phenomena,
as perceived by human beings, both occur, and can be measured, along a
dimension of time (page 5)." We shall assume ﬁnat communication is
included erong ﬁnese phenomena , or more precisely , following ﬁne con-
vention established earlier (pages 6n, 13) , shall assume ﬁnat communica-
tion events are part of the class of phenomenal occurring end measurable
over time. We shall also assume that all communication events involve
sore form of change, where ﬁne term "change" is used in its most general
sense to refer to both ﬁne presence and absence of variation in sore
relevant property or properties of an event (see discussion in section
3.2). On ﬁne basis of these two assumptions concerning time and change,
ﬁnen, it is evident that communication events are part of the class of

events termed processes, since the term "process" is used here to refer

 

1All perceived phenomena being some form of matter-energy or
information (see J. G. Miller, 1965a, pp. 193-190).

28

to "All change over time of matter-energy or information . . .

(J. G. Miller, 1965a, p. 209)." It is this means of relating the
concept of communication and ﬁne concept of process which forms the
basis for statements in this paper that "communication is a process , "
or, in the more technically correct form to be used here, "communica-

tion events are processes."

2.1.2 Process and oﬁner Important Concepts

Such a discussion of ﬁne relationship between ﬁne concept of
communication and the concept of process , however , may lead one to an
objection of the form: "Doesn't the consideration of 'process' in the
many definitions , models , and discussions of human communication involve
something more than just 'change over time of sone property or proper-
ties'?" The answer is "yes," for the term "process" is subject to a
broad range of uses , particularly in the discipline of communication.

It is worthwhile to review some of ﬁnese uses, brieﬂy, in order to
extend the examination of the concept itself by showing its relationship
to several other important concepts often "included" when the term
"process" is used in its broad sense. The review will also serve to
sharpen the definition used here. Once again, such a review is not
intended to isolate a "correct" or "true" definition, but instead to
provide a point of reference (see page 11).

It will be useful, first of all, to set apart from consideration
all uses of the term "process" in anatomy, law, and printing, as well as
uses as an adjective or as a transitive or intransitive verb. As a noun,
ﬁnen, "process" is often used in its broad sense to indicate that ﬁne

antecedents for an occurrence stretch back indefinitely in time, just as

29
its effects stretch forward indefinitely. In short, "process" may be
used to indicate that an occurrence "really" has no beginning and no
end (Berlo, 1960, p. 21+; Sarbaugh, 1968, pp. 2-3). This aspect of an
occurrence is frequently termed its "continuous" aspect, or, using the

term to be adopted here , its "continuity." Again , as noted earlier,

 

no scientist can observe or study an occurrence in its continuity,
except as he divides that continuity into discrete portions termed
"events" (page 13). "Process" will be used here, ﬁnen, to refer to
change over time wiﬁnin an event,2 while ﬁne "continuity" of an occur-
rence will be noted by ﬁne specific use of that term.

Closely related to the use of "process" in its broad sense to
indicate the continuity of an occurrence is its use to indicate two
other aspects. One of these aspects is that while ". . . certain
ﬁnings may precede others , . . . in many cases the order of precedence
will vary from situation to situation (Berlo, 1960, p. 25)." Rather
than use ﬁne term "process" to describe this aspect of occurrences,

the term "indeterminancy" will be applied, indicating that in most cases

 

no single causal sequence can be determined. The second aspect is
frequently expressed by Heraclitus' metaphor indicating that a man can
never step in the same river twice , since neither man nor river will be
the same again. Although decidedly unpoetic, this aspect of occurrences

or events will be termed their "irreversibility," thus further

 

 

2Note that this usage is not necessarily synonymous with the
use of "process" to refer to "a series of actions or operations
definitely conducing to an end," as in "process of tool or steel
making." Sequences of actions where the goal is of prime importance
will be termed "procedures . "

 

30
narrowing the range of usages for "process."

In its broad sense, "process" is also frequently used to indi-
cate that each elerent or property of an occurrence affects all the
otlners, i.e., they interact (Berlo, 1960, p. 21+, 26; Sarbaugh, 1968,
p. 3). This particular usage of "process" is especially evident in
discussions such as that by Dubin in speaking of "processes of inter-
action among variables": "My emphasis on interaction is identical with
that of G. Bergmann who Speaks of 'process knowledge' and its inter-
action feature, which is the complete knowledge of the interaction
among the variables of a system (1969, p. lOn)." Certainly inter-
action is an important aspect of an occurrence, and especially of the
system involved in that occurrence. Because of its importance, ﬁne
term "interaction," will be used as distinct from the term "process."3

Apart from "interaction" is another use of "process" in the
broad sense to indicate a second important aspect of both an occurrence
and the system involved in it. Berlo identifies this aspect in noting
that "As any good cook knows, it is the mixing process, the blending,
that makes a good cake; ingredients are necessary, but not sufficient
(1960, p. 26)," or in other words, a cake has properties which are not
evident from its ingredients. Clearly, such new properties may be
generated in an occurrence or a system by ﬁne "procedure" (see note 2

above) which produced it. Nevertheless, such properties are

 

3Closely related is the interaction between ﬁne properties of an
event and those of the observer. This particular interaction is usually
identified or considered under the heading "relativity." Such usage is
adopted here to avoid further confusion with either "process" or
"interaction . "

 

31
characteristics of the occurrence or the system itself, not of the
"process" of production. As Bertalanffy has expressed it:
The meaning of ﬁne somewhat mystical expression,

'ﬁne whole is more than the sum of parts' is simply

that constitutive characteristics are not explainable

from the characteristics of isolated parts . The char—

acteristics of ﬁne complex, therefore, compared to those

of the elements, appear as 'new' or 'emergent' (1968,

p. 55; see also Buckley, 1967, p. '42; J. G. Miller, 1965a,

pp. 201n, 217; Rappoport, 1968, p. xvii).

Accordingly, instead of ﬁne term "process," the term "emergence" will
be used to indicate that the whole of an occurrence or system exhibits
properties which are not evident from its parts .

Finally, the term "process" is also used in its broad sense to
indicate a third important aspect of both an occurrence and a system--
one which is closely associated with all the oﬁner aspects noted above.
'Ihat aspect is, of course, change over time in the properties of an
occurrence or a system. Certainly a great proportion of the uses of the
term "process" have this important aspect at their core, whether in

indicating change over time alone , or change over time associated with

continuity, or wiﬁn interaction, or with other aspects . It is this

 

single, key aspect of change over time which we shall adopt as the

 

rarrow sense or use of the term "process," a usage indicated, of course,

 

by ﬁne narrow definition given earlier. Henceforth, any use of the term
"process" in its broad sense will be avoided, unless indicated as such,
in favor of ﬁne use of "process" in its narrow sense, combined wiﬁn use
of the terms mentioned above to indicate oﬁner aspects of occurrences .
The attempt here , then, has been to systematically narrow ﬁne
sense or usage of the term "process" by introducing alternative concepts

which are closely related to change over time, but which are possibly

4.x

32
better adapted to isolating certain aspects of occurrences often indi—
cated by "process" in its broad sense. The narrowing has thus been done
with a knowledge of what has been omitted , ﬁnough wiﬁn the recognition ,
as well, that no such treatment can hope to examine the full range of
possible meanings for a term. The goal here is not exhaustiveness , but
establishment of a clearer reference point for later discussion.

'IWo additional points about the use of "process" need attention
before leaving ﬁnis review of the term's relationship to other important
concepts . One point is that the concepts noted above , e. g. , continuity ,
indeterminancy , irreversibility , interaction , emergence , and process in
its narrow sense, have all been mentioned at some time as characteristic
of art , or of its production or appreciation . One is led to speculate
whether in some cases "process" may have been applied in its broad sense
where a reference might more suitably have been made to ﬁne "'art' of
communication." The second point is that "change over time" can be dis-
cussed wiﬁnout using ﬁne term "process" at all. While this point is
recognized, it is felt that "process," in its narrow sense, is both a
highly useful organizing device in discussion, and fully in keeping with

an important aspect of the common use of the term.

2.1.3 Process and Research Designs

In addition to the above examinations of "process" as related
both to "communication" and to several other important concepts , it will
be helpful to examine one oﬁner aspect of the concept itself: its rela-
tionship to three general classes of research designs. This relationship
should indicate , in part, the nature of the concern with the concept of

process in this paper. A few technical points, first, as a preface to

33

discussing these general classes of research designs.

In order to record change over time, or process, in a research
situation, it is necessary (but not sufficient) to measure boﬁn ﬁne
value of ﬁne variable(s) of interest, and the point in time at which
that value is recorded." The time point is measured with respect to
some convenient reference point, such as ﬁne time point of ﬁne first
measmement taken. Time therefore becomes a variable which is measured
with respect to other measurenents . 5

Ideally, a scientist might wish to record the change over all
time during ﬁne interval of concern by measuring the value of ﬁne vari-
able(s) continuously in that interval. But continuous measnrrenent is
rot always necessary , or more importantly, not always possible for sore
Variables. In either case, the alternative is to measure at a number of
discrete points in time (usually regularly spaced) during the interval.
The resulting information (data on the variable (8) and time) will pro-
Vide eiﬁner an approximation of continuous measurerent , or , depending on
the form of variation of ﬁne variable(s) , ﬁne equivalent of continuous
me‘E‘S‘Jir‘ement. In ﬁne latter case, assuming that the criteria are met for

fretlllency of measurerent wiﬁn regard to ﬁne particular form of variation,

\—

and u"Researeh situation" encompasses both "observational" research,
the exPerinental research. The former case includes Situations 1n whidn
Scientist has re control or does not exert control over exposure to
Wm stimuli. The latter case covers all situations where sore
:nt‘rol over exposure (as well as some randomization) can be applied,
u e31 though a true experiment may not be achieved, as in the "pre-" and
quﬁSiJ experiments of Campbell and Stanley (1963).

u 5Time is measured as a linear flow, not as an "amount" or
quantity" (see Heirich, 198a, pp. 388-390).

31+
ﬁne data contain sufficient information to boﬁn fully characterize and
allow reconstruction of ﬁne continuous variation (Cherry, 1957 , Chs .
l}, 5; Pierce, 1961, Ch. H). The criteria, and the required lcncwledge
of ﬁne form of variation, will be discussed at length in Chapter 1+. The
key point here is that there is an intimate relationship between measure-
ment at discrete (and measured) points in time, and the form of variation
or change over time which can be characterized by ﬁnat measurement.

With these technical notes as background, it is possible to look
more closely at the concept of process as it relates to certain general
classes of research designs. The three classes of designs to be dis—
cussed here have been established on the basis of a single characteristic
of individual research designs-ﬁne number of "observation points" pro-
Vided in ﬁne design. An observation point is a point in time in a
research situation at which the relevant variables are measured, regard-
less of wheﬁner the measurerent involves one or more specific applica-
tions of ﬁne measuring instrnment(s). That is, if both a pretest and a
POS‘t'test are given to two groups in a particular research design, there
may be as many as four separate applications of the measuring instru-
ment(s). There are, however, only two observation points (ﬁne pre and
poet points). The same holds true for replication of the design at a
later time, i.e. , ﬁnere are only two observation points in the design,
even ﬁnough there may be many applications of the instrument(s) over an
extended period. The three general classes of designs to be described
below have been termed "point," "difference," and "process" designs, each
term referring to ﬁne type of data which may be obtained from designs

ha"ilng, respectively , one , two , or three or more observation points .

35
It should be noted ﬁnat no attenpt will be made , here , to review
individual research designs and to identify them with one of these
three general classes. The nature of such a categorization should be
evident from the discussion, and will be indicated more specifically
later in the chapter. Of particular interest in considering each
class will be the formalized observations or data gathered under those
designs , the nature of ﬁne evidence which can be derived from those
data by analysis , and the senantic interpretation of that evidence
(Krippendorff, 1969b, pp. 2-5; see also Cocmbs, 1961+, Ch. 1).

Point designs.-- The first general class of research designs

 

includes those which measure variables at only one point in time in the
research situation (it is assumed ﬁnat the time point is recorded, also).
The data gathered using designs with only one observation point may, of
course, be subjected to a number of relevant analyses. Despite the
form of analysis, however, such data (termed "paint" data) can provide
explicit evidence on only a single "state," or set of values of vari-
ables at 91E point in time.6 The scope of the serantic or theoretic
interpretation of ﬁnat evidence is ﬁnus formally limited to a concern

With a single point in time.

\—_

6Except when a variable is known a priori as constant, or as

haYing linear variation with a known slope , in some interval . Research
“fhleh gathers data at one point , and depends on population norms , etc . ,
bib a reference point, probably involves "difference" data, as discussed

low. Sore research, however, gathers data at one point in time, and
i130 asks, "What was the state at another point?" The question of
whe‘t:her such research produces point or difference data is a methodo-
1C"gical and philosophical question beyond the domain of this discussion.

 

1)

tie

L"

‘I

41"

{9

1".

36

Difference designs.-- The second general class of research
designs--ﬁnose wiﬁn two observation points (e.g. , pretest and posttest,
with ﬁe time points recorded)--will produce data which are subject to
a somewhat broader range of analyses than are data from point designs.
The two observation points, of course, define the interval under
concern, and the data gaﬁnered at these two points can provide explicit
evidence not only on the individual states , but also on ﬁne differences
between those states (such data are thus termed "difference" data).
The semantic interpretation of ﬁne evidence produced by analysis of the
data can likewise be extended to cover boﬁn ﬁne individual states and
the differences between them.

Both point and difference designs are, of course, frequently
used in testing predictions about the effects of independent variable(s)
(experimental stimuli) on dependent variable ( s) . That is , given adequate
e”(Petrimental control , a difference in the dependent variable(s) between
the experimental and control groups at ﬁne final observation point (the
Posttest) may be said to result from ﬁne manipulation of the independent
var"ZI.—able(s) . Conclusions of this form can be drawn boﬁn from adequate
Point designs and from adequate difference designs. The effect of the
pr‘e‘test in the difference design is, of course, to increase precision
in ﬁne analysis. Note carefully ﬁnat the term "difference," as used
above, does not refer to ﬁne difference between the experimental and

m‘bmn groups at a single point in time, but rather to ﬁne difference
7

bemeen the states of a group or groups at _tw_o_ points g time.
\-

7Unless otherwise noted, ﬁne term "difference" will be used here-
in to refer to a difference between states at two points in time.

37

A difference between states at two points in time is, of
course, one instance of change over time or process. That is, if
difference data are presented on a graph which has time as ﬁne
abscissa and variable value as the ordinate , ﬁne representation takes
ﬁne form of a straight line over the interval defined by the two
observation points . 8 Such a graph illustrates a specific paﬁn or
form of variation, i.e. , one instance of change over time. Note,
however, that in this paper we shall be concerned with change over
time, or process, without restriction on the shape of ﬁne path or form
of variation. Alternatively, in terms of a graph, the concern with
process, here, will encorpass all possible paﬁns or forms of variation
which may be drawn between ﬁne two values of a variable at the end
points of an interval. The straight line path representing difference
data is, quite clearly, only one special case in this more inclusive
consideration of change over time .

It is apparent , then, that a difference design produces data
which are capable of providing explicit evidence only on states at
two points in time; i.e. , no evidence is provided on states wiﬁnin the
interval. 9 The scope of the senantic or theoretic interpretation of

difference data is therefore formally limited to a consideration of the

 

8Obvious 1y point data cannot provide such a graph of change
over time, unless the variable is known to be constant, etc. , as
described in rote 6, above. Note that only a single variable is
referred to here, although the principle applies to one or more
variables (see discussion in Chapter 3).

9Except when there exists: (1) a priori lcnowledge ﬁnat the
change in a variable is linear in the interval, or ( 2) a iori
knowledgg about certain characteristics of the form of variation

 

38
two states and the differences between them--a consideration in terms
of change over time or process is ruled out. More specifically,
difference data do not provide evidence on the time relationship of
ﬁne values taken by a variable, or on the time sequence of states
within the interval . Hence , any attempt at a serantic or theoretic
interpretation involving such a time relationship or sequence moves
beyond ﬁne explicit evidence which can be obtained from the data.

Process designs . —- The evidence provided by difference data

 

may, of course, be interpreted in statenents noting ﬁnat change (or
difference) _d_12 occur, but it may not be interpreted in statements
dealing with h_gw_ that change occurred. The latter class of staterents
includes those dealing with the time sequence of states wiﬁnin an
interval, with ﬁne path or form of variation (the time relationship of
values), and more generally, with the change over time or ﬁne process
involved. 10 Interpretations of this nature may be made , however,

from ﬁe information provided by research designs from ﬁne third
general class. These designs measure variables not only at ﬁne two
points in time which define the interval of concern, but also at one

or more observation points (whose times are recorded) within ﬁnat

 

 

involved. These characteristics are discussed in Chapter 1+, and
derive from the close relationship between measurement at discrete
points in time and the form of variation which can be characterized
by that measurement. Even provided ﬁe special characteristics are
known, measurement at two points in time is an absolute minimum.

10No criticism of , or deficiency in, current research is implied
here. See the comments above on the usefulness of point and difference
data, and the comments on current research in the last section of this

chapter.

39
interval. The data gathered using such research designs are termed
"process" data and are subject to a rather broad range of analyses,
ﬁne exact form of analysis depending in part on how ﬁne measurement
at discrete points relates to the form of variation involved. That
is , if ﬁne measurement approximates continuous measurement of ﬁne
variation, the relevant analyses of the data will provide explicit
evidence both on the sequence of states within the interval, and on
the differences between them. Should the measurement completely
characterize the variation (or be a continuous measurement of it),
however, more powerful analyses will provide evidence on ﬁne full
paﬁn or form of ﬁne variation within the interval .

In eiﬁner case , the evidence produced by the analysis of
process data permits a much broader scope of semantic or theoretic
interpretation than does ﬁne evidence produced from difference data.
Such interpretation includes consideration in terms of change over
time or process , and might involve statements dealing with the
sequencing of particular states and configurations , or with the
appearance, operation, and decline of certain constraints and rela-
tionships . Such interpretations would apply boﬁn to observational
and to experimental research , the latter providing additional infome-
tion ﬁnrough experimental control and manipulation.

Process designs produce evidence which may be interpreted in
terms of change over time or process because such designs are capable
of operationalizing the concept of process . Specifically, the measure-
ment or description of any path or form of variation, i.e. , of any

process or change over time, requires data on the variation to be

no
gaﬁered at several points within the interval of concern . Since
process designs (as ﬁney will be discussed in Chapter u) are capable
of specifying procedures , under particular conditions , whose perform-
ance will identify (measure and describe) in nature any path or form
of variation, then by definition ﬁne designs are capable of opera-
tionalizing the concept of process (see page 9ff, and Berlo, 1967, pp.
24}, 11). It is ﬁnis third general class of research designs-~ﬁnose
which operationalize the concept of process--which will be of primary
concern, here, particularly as those designs apply to the study of
communication events as processes .

Two further additions are needed in closing this discussion
of the concept of process as it relates to these three; general classes
of research designs . One of the two is a note that while the interest
in ﬁne concept of process does indicate an interest in time sequences,
it does not imply an interest in making determinations of causality.
Second, and more important, is a recognition that a fuller under-
standing and explanation of human communication events is not to be
gained solely through attention to change over time or process . To
say so would be absurd, since communication events, and especially the
systems involved, are characterized by interaction and emergence , as
well as by process. A fuller understanding will come only as these
alternative concepts or aspects of events begin to be considered in
combination in research on communication events .

In ackrnowledging this important point , however, it must also
be admowledged ﬁnat each alternative concept or aspect generates

unique problems when it is operationalized in research--problems

91

possibly better studied in isolation, before in combination. This
factor, together with that of limited space , suggests that one concept
be chosen from among the alternatives . That choice, in ﬁnis case, is
ﬁe concept of process in its narrow sense of change over time--a
clnoice based primarily on the utility of the concept in the conduct of

research and theory building in communication.

2. 2. Criteria for Choo_s_ing the Concept of Process

Like ﬁne earlier choice of a definition (section 1.3), the
choice of the concept of process as ﬁne focus for attention here is a
sufficiently important choice to require identification both of the
range of possible alternative concepts , and of the criteria for the
choice among them. The choice of the concept of process from its
range of alternatives involves, in effect, two choices from two dif-
ferent subsets of alternatives , although the criteria for the two
choices are essentially identical. The choice from first subset has
already been mentioned, i.e. , the choice of ﬁne concept of process for
attention when it is acknowledged ﬁnat alternative aspects of events
and of the systems involved, namely, interaction and emergence , are
also important in research on communication. The choice from the
second subset has been less explicit, i.e. , ﬁne choice of the concept
of process from the alternative key concepts in the study of communica-
tion. In the context of ﬁnis paper, this second subset is comprised of
the key alternative concepts introduced in defining the universe of

discourse, namely, structurei and system.11

 

11Note that other concepts were introduced earlier in ﬁne
initial considerations of the concepts composing each subset of

1+2

One criterion which affected the choice from ﬁne second subset
of alternatives is worﬁny of brief mention. Process was chosen from
this subset partly because the other concepts have received special
attention in the past: structurei (C. Morris, 1968; Shannon, l9u8)
and system (Bertalanffy, 1968; J. G. Miller, 1965a), to mention only
a few of the scholars . Beyond ﬁnis criterion, though , there is only
a single, primary criterion for choosing the concept of process from
the two subsets of alternatives : ﬁne criterion of utility in
corparison to the utility of ﬁne alternative concepts. In Kaplan' 8
terms, again (see page 23) , "Whether a concept is useful depends on
ﬁne use we want to put it to (1961+, p. 51)," that use, in ﬁne present
context, being use as a tool in the conduct of research and theory
building related to human communication. The basis for choosing a
concept is therefore its possible utility as a tool in advancing
knowledge about communication . Because the criterion is a broad one ,
it will be most expedient to deal wiﬁn utility in two separate respects ,

first, in regard to general problems arising in building theory on the

 

basis of research, and second, in regard to theory building and

research in areas of new concern in the discipline of communication.

 

 

alternatives. Of ﬁne concepts mentioned earlier (page 28ff) , in
conjurnction with the first subset, "relativity" is grouped here with
"interaction. " The other concepts , "continuity, " "indeterminancy,"
and "irreversibility" are highly descriptive of an event or system,
but are not characteristics or aspects directly affecting the values
taken by variables, as are the concepts or aSpects included here as
the first subset of alternatives . The other concept mentioned earlier
(page llff) in conjunction with the second subset is ﬁnat of "trans-
mission." As noted, however, the concept of transmission is less
general than process and is encompassed by it (see page 11+), so that
its inclusion in the second subset of alternatives would be rednundant.

1+3
2.2.1 Utility in Two General Problems
In ﬁne first regard, a concept would be judged useful if its
application were important in avoiding certain general problems
inherent in the building of theory from research. Two problens seem
especially evident here , namely, ﬁne problems of spurious categoriza-
tion and of prediction and understanding.

Spurios categorization. -- As background for ﬁne first problem,

 

it is important to recognize that ﬁne scientist, in working toward his
goal of building ﬁneory to explain and predict events , is at all times
imposing his structurings on the natural world. He is not discovering
structures "inherent" in nature (Berlo, 1960, p. 25; 1967, p. 2). It
is important to recognize, too, that a fundamental part of any
structure imposed by the scientist is ﬁne categorization which he makes
of events in the natural world based on the properties he observes in
ﬁnose events (Berlo, 1967, p. 2). The scientist often begins with a
fairly "primitive" categorization, but seeks to refine it ﬁnrough
research on ﬁne properties of ﬁne events under study , whether ﬁnat
research be observational or experimental (see page 33n). His goals
in ﬁne task of continually refining his categorization are to make it
more useful, more consistent, and more predictive of events in nature
(Berlo, 1967, p. 2). As the scientist progresses and his categoriza-
tion moves from primitive to "sophisticated," he also advances in the
direction of more comprehensive theory, and may eventually reach ﬁne

point of fully formalized theory--the ultimate structure which he can

impose .

nu.

Clearly, the social and behavioral sciences are , for the most
part, still engaged in ﬁne early steps in formulating and refining
ﬁeir theories and categorizations of events (Rudner, 1966, pp. 28-140) .
If this statement can also be considered descriptive of the current
status of theory building from research in the discipline of communica-
tion, then it is evident from the earlier discussion of the three
classes of research designs ﬁnat a potential problem exists in the early
steps of categorizing communication events. Specificially, any cate-
gorization constructed on the basis of research on properties of com-
munication events may be a spurious categorization if ﬁnat research
utilized designs which gathered data at only two observation points.
Such a categorization might be spurious in the sense ﬁnat it classifies
events as similar only on ﬁne basis of similiarity in difference data,
even though it is clear that a set of events having similar overall
differences may still exhibit several distinct paths or forms of
variation. While spurious categorization may not necessarily be a
frequent result of research which produces difference (or point) data,
its obvious ability to frustrate the early steps of theory building
derands ﬁnat it be avoided.

Two counterargunents can be raised to the above, one of which
states that the social or behavioral scientist carefully avoids spurious
categorization by taking "other factors" into account when categorizing
on ﬁne basis of research which provides only point or difference data.
This argnment of extra care is certainly one to be considered, but it
is also subject to all of the criticisms that may be applied to inter-

pretations beyond the explicit evidence provided by data. The other

us

counterargument states that the form of ﬁne variation or change over
time within an interval or event "makes no significant difference" in
ﬁne resulting categorization of events. Such an argument may well be
valid in a particular situation, but it is crucial to note that it
can be adeqtately supported only by first formulating an hypothesis
to the effect that "variation within an interval makes no difference
in categorization," and then conducting the necessary research--
research whidn will require ﬁnat data be gaﬁnered within the interval
or event under consideration .

It is, of course, fairly obvious ﬁnat the type of research
needed to determine the presence or absence of spurious categorization
is also ﬁne means of avoiding such categorization and its potential
frustrations . T'tat is , research which gathers data at points in time
wiﬁnin an interval or event provides exactly the type of evidence on
change over time which is needed to avoid the particular aSpect of
spuriousness considered here . As noted before (page 39), the par-
ticular designs used in such research would operationalize the concept
of process, thus making that concept a hignly central and useful one
in avoiding the general problem of spuriousness in the categorization
of events .

It should be noted that a parallel case for usefulness in
avoiding spurious categorization could probab 1y be built , as well ,
around each of the other concepts in the range of alternatives:
structurei, system, interaction, and emergence. Examining each of these
concepts with regard to the prob len of spurious categorization, then,

it is apparent that a possible source of spuriousness might be the

us
failure to take into account various distinct ions among events both
in structurei and in interaction. While such Spuriousness is possible,
its probability is seen as lower than in the case of process, primarily
because both concepts are frequently considered in categorization and
theory building: for example, on structurei see Deutschmann (1957)
and C. Morris (1968), and on interaction see Dubin (1969), among others.
A failure to consider differences in systems could be a possible
source of spuriousness, too, except that J. G. Miller has noted (1965b,
p. 337) that differences in the processes of systems are a primary
basis for categorizing them-~a point which reduces "system" spurious—
ness to "process" spuriousness. Finally, there is the possibility of
spuriousness arising from the failure to consider emergence in events
and systems . The concept of emergence presents a problem here , how-
ever, in that it requires examination beyond the little it has received
(J. G. Miller, 1965a, p. 201m) in order that its implications with
respect to spuriousness be clear. In ﬁne absence of such information,
erergence will be considered, along with the concept of process , as
a highly useful concept in avoiding the general problem of spurious
categorization that is inherent in the early steps of building theory
from research.

Prediction and understanding." Along with the problem of

 

spuriousness is a second general problem inherent in theory building,
especially as that theory building is conducted utilizing present
research. This second problem has been recently discussed by Dubin

(1969), and his words serve to express it best:

H7

Theories of social and human behavior address
themselves to two distinct goals of science: (1)
prediction and (2) understanding. It will be argued
that these are separate goals and that the structure
of theories employed to adhieve eaCh is unique. I
will not, however, conclude that they are either in-
consistent or incompatible. In the usual case of
theory building in behavioral sciences, understanding
and prediction are not often adhieved together, and
it therefbre becomes important to ask why. It will
be concluded that eaCh goal may be attained without
reference to the other.

I mean one of two things by prediction [my
italics]: (1) that we can fOretell the value of one
or more units making up a system; or (2) that we can
anticipate the condition or state of a system as a
whole. In both instances the focus of attention is
upon an outcome.

As I employ the term understanding, it has the
fOIlowing essential meaning: it is knowledge about
the interaction of units in a system. Here attention
is f0cused on processes of interaction among variables
in a system.12

The relationships between the goals of science and
the analytical fOCi of attention in achieving these goals
can be shown in a fOurfold table like [Table l].

 

 

 

Table 1. Scientific Goals and the Foci of Research (Dubin,

 

 

1989, p. 10)
Goals
Understanding Prediction
Interaction X I
Analytical Foci 1
Outcomes x2 —_I

 

 

 

 

12A "unit" in this instance is a variable, as in the quotation

given earlier on page 30. It is well to note again, as in the earlier
quotation, that Dubin's use of the term1"process" is consistently
confined to what has been termed here as "interaction."

l+8

At first glance one would normally be constrained
to argue ﬁat the four boxes of the table are simultan-
eously populated. That is , to achieve understanding
[my italics] of a social system, we need to know the
interaction processes in it and ﬁne outcomes generated
by these processes. Similarly,— if we are to make accurate
predictions [my italics] about social phenomena, we have
to know the [interaction] processes built into ﬁnese
phenomena and the characteristics of all possible outcomes
toward which—the [interaction] processes move. This
initial reaction is simply the assertion of a pious value
position ﬁnat bears little relation to ﬁne practices of
social scientists . They actually operate in ﬁneory building
and in doing research by working primarily in two of the
four boxes, as indicated by the X entries in [Table 1].
What seems, in a logical sense, to represent the closure of
the theory building-research cycle turns out to be largely
ignored in the actual practices of theory building or
researching (1969, pp. 9—11).

 

Dubin goes on to illustrate his statement about the actual
practices of social and behavioral scientists in theory building, and
then focusses on the two subproblens or paradoxes he sees associated
with ﬁnese practices. The first of these may be termed the paradox
of understanding, which notes that it is possible to achieve under—
standing in science through knowledge of the interactions in a system,
wiﬁnout being able to predict with precision the outcomes of events
involving ﬁat system. The second paradox is termed the paradox of
prediction, which notes that it is possible to predict with precision
the outcore of an event, without an understanding of the interaction
in the system involved.

Dubin discusses the reasons for the existence of each paradox,
and in doing so establishes a firm basis for his claim that the two
goals of science, understanding and prediction, are not often achieved
together; that is, understanding focusses on interaction (Xlin Table

1) and prediction focusses on outcomes (X2 in Table 1), while the

1+9
other two cells remain vacant. Although Dubin notes that it would
seem logical that these other cells be filled, and indicates by his
presentations of the paradoxes that there may be some value in doing
so, he does not discuss the characteristics of research and of the
associated theory building Which would fill the cells.

Certainly a detailed consideration of the Characteristics of
such research is beyond the scope of this discussion, too. However,
it is possible to suggest at least one important step in the direction
of gaining both (1) an understanding of the means by Which the possible
outcomes in events are achieved, and (2) the ability to predict with
precision the nature of the interactions in the systems involved in
those events (the two yagagt cells of Table 1). In particular, note
that theory building in these two cells requires, to a.greaternextent
than the other cells, evidence on the path or fOrm.of variation, i.e.,
on the change over time, in an event. That is, theory building re-
quires evidence on the processes (in the narrow sense) in a set of
events which lead to particular outcomes and Which are associated with
particular types of interaction (of. Buckley, 1967, p. 136). Research
Which is done utilizing designs that operationalize the concept of
process provides just such evidence on processes and allows theory
building to move in the direction of "filling" the vacant cells, or
of closing the "theory building—research cycle." The concept of
process , then, together with the concept of interaction, is vital in
research if that research is to lead to more complete theory. Quite
clearly, too, the concepts of process and of interaction are the most

useful of those in the range of alternatives in avoiding the general

50
problem of restricted prediction and understanding that is inherent

in building theory from present research.

2.2.2 Utility in Two New Concerns

In addition to ﬁne utility of concepts in regard to general
problems arising in theory building, ﬁnere is ﬁne equally important
utility of concepts as tools in theory building and research in areas
of raw concern in the study of huran communication. In this second
regard, a concept would be judged useful if its application were
central in development of ﬁneory and research relevant to ﬁne new area.
It is, of course, impossible to review all of the new concerns in the
field of communication in this limited space; hence, a choice is
necessary. The two areas of huran information processing and of
system development appear to be particularly important , and have been
chosen both because they are highly general in scope and because they
have received relatively little study in the discipline, even though

attention has been called to ﬁnem.

Huran information processing.—- It is worﬁn noting, initially,

 

that some question exists as to whether the universe of discourse of
this new concern is greater than, equivalent to, or subsured by ﬁnat
of huran communication. The question will be answered here by noting
that G. R. Miller (1969) has pointed out certain general problems
common to both areas. Concerning one of these problems, in particular,
Harrison (1967a, pp. 1-10) has emphasized that communication scholars
need to redirect their attention to the important decoding aspects of
human information processing, instead of concentrating so exclusively

on encoding skills as ﬁney have done in the past. G. R. Miller has

51
been more specific: "Little is known about the complex process
through which people select, interpret, and respond to information
(1969, p. 61)." Miller expands his statement by adding a wide range
of questions to be answered about how information processing occurs
in huran events.

The questions posed cover events such as ﬁnose of seeking and
dissemirating information; of using information to influence, to
decide, and to resolve conflict; and of using information to facil—
itate growth and development, among others (G. R. Miller, 1969, p. 62).
Such events have in common not only the presence of some form of
information, but also one oﬁner very basic property—each involves
complex variation or change over time within the event. That is,
information is input, "processed," and output, often wiﬁn high

frequency, during the course 21: any of these events. The corplexity

 

of this information processing during an event becomes evident, though,
only when one considers that the human organism has distinct limits

on ﬁne amount of information it can handle in a given time. In order
to work within these limits, the organism may, for example, segment

a larger input of stimulation or information into "chunks," and then

handle the chunks as basic units--all of this while the information

 

input is continuing (G. A. Miller, 1956, pp. 90-95; J. G. Miller,

1965b, pp. 399-350).13

 

13Note that "chunking" is only one of several ways which
' ms have for coping with large inputs of information. Platt
(1969), for example, discusses a number of alterrnative means appro-
priate to different situations .

52

Quite obviously , ﬁnen , huran information processing involves
change over time or process as an integral aspect, even for a relatively
simple event. Accordingly, mnuch of the research which is done on
information processing will have to erploy designs which operation-
alize the concept of process, just as the associated theory will have
to include the concept among its basic terms . It is this centrality
of the concept of process which makes it highly useful in the develop-
ment of theory and research on human information processing. In
addition, of the other concepts in the range of alternatives, it is
clear ﬁnat structurei, or information, is a central concept in ﬁnis
area, as well; hence, the two concepts are grouped here as equally
useful tools in this new concern in the study of hunan communication.

System development . -— The second of the two new concerns is

 

closely related to an important endeavor or goal in ﬁne discipline

of communication: that of attending to "problematic situations"
(Krippendorff, 1969a, p. 112), or of relating "the results of research
to the operational problems of society . . . (Berlo, 1969, pp. 7, 16)."
There are, of course, many different ways to attain ﬁnis goal, among
which is the area of new concern with the design and development of
more effective communication systems, whether new systens or old ones
(of. Harrison, 1967a, p. 10; 1969b, pp. l4-5). Communication scholars
have, of course, been called on frequently to consult on the problems
found in conmunication systems, or to design conponents for existing
or projected systems. The problers of the design and development of
communication systems {a systems have only begun to receive attention

from such scholars, however, perhaps in part because they present a

53
number of new, difficult tasks and questions to the designer.

For example, in dealing with any system, the designer is
immediately confronted.with the often time—consuming task of speci—
fying the goals of the system in a form suitable for later "performance
testing." Once this task is complete, he is faced‘with several
general categories of questions related to system design: HOW'will the
system maintain (or stabilize) itself? How will it terminate and what
will be its life cycle? (Campbell, 1969; Harrison, 1967a, p. 9; 1969a,
p. 88; 1969b, pp. 1-2). Such questions have not often been asked in
past attempts at the design of, say, a teadhing or interview situation
whidh will also be an effective communication system, Yet suCh
questions must be asked, and answered, if something more than specu—
lation is to be given in response to questions like: Are the system
outputs being properly minimized, maintained, or maximized? Or, is
the systemnefficient with respect to the goals WhiCh were set, or to
criteria like cost and effort?

These latter questions, and the hundreds of more specific ones
whidh accompany them, are frequently difficult ones to answer, even
when the system.under design is fairly simple in comparison to a.human
communication system. In fact, the study of relatively straight-
fbrward systems reveals that the ability to find any answer at all to
such questions depends heavily on the ability to describe or analyze
two key aspects of the system involved: the relationShips among system
components, and the Change over'time or process WhiCh each component
exhibits (Koenig, 1967). The concepts of system and of process are

integral in these key aspects of systems and hence are central in the

su
theory, analysis, and research associated with system design and
development . Quite clearly, considering the range of alternative
concepts, these two are also the most useful for theory building and
research in this new concern in the disciplire of communication.

Because this review of new concerns could not be exhaustive,
it has been the generality of these two areas, as well as the relative
lack of study which they have received in the discipline, which has
prompted their selection from among other foci of concern. It is
important to note , however, that while the concept of process has
utility in theory building and research in these areas of new concern
in communication, it has utility only in combination with one of the
other concepts from the second subset of alternatives: structure;
and system. Likewise , in the earlier discussion of general problems
in research, the concept of process had utility in avoiding each type
of problem, but again, only in combination with a concept from the
first subset of alternatives: emergence and interaction.

The fact that process occurs as a useful concept together with
these other concepts only more clearly underscores the comment made at
the end of the previous section. To wit, a fuller understanding of
communication events will come in large measure only as concepts begin
to be considered in combination in research. On the other hand, given
that attention must be directed, for the present , to only one concept ,
it is the consistent utility of the concept of process that leads to
its choice from among the other concepts as the most useful tool in the

conduct of research and theory building in communication (Smith, 1967) .

55
Beyond this discussion of the criteria for choosing the concept
of process as the focus of attention, however, there remains an
important nunanswered question: What is the current place of the
concept of process in research related to human communication? The
answer to this qLestion should not only expand on the importance of
the concept of process already indicated in this chapter, but also

open the way to considering the means of explicating the concept.

2.3 The Current Plage of Process in Research

 

Given that the concept of process appears to be an important
and useful concept in research and theory building in communication,
it is worthwhile to examnine the extent to which the concept has been
incorporated in research. In view of the earlier discussion of the
concept of process as related to general classes of research designs,
this task can be seen as one of determining the degree to which studies
concerned with human communication have used designs which operation-
alize the concept (see page 39). In other words , to what extent have
process designs been used in ccmmmunication research, with respect to
point and difference designs?

It is of course clear that a precise evaluation of the extent
to which a particular class of designs has been used involves a review
of research far beyond the scope of this paper and of most books.
Accordingly, a somewhat imprecise evaluation of extent must suffice--
an evaluation based, in this case, on an estimate of the relative
usage in communication research of a number of common observational
and experimental research designs. As an aid in making this evalua—

tion, it will be helpful to set up a classification scheme for the

56
basic designs currently used in researCh in disciplines like education,
psydhology, and communication. Such a scheme, and the resulting
classification, is presented in Table 2, page 57. .A brief description
of the Table's two basic dimensions is in order before returning to

the estimate.

2.3.1 .A Classification of Researdh Designs

The horizontal dimension of Table 2, "NUmber of Observation
Points," is drawn from the earlier discussion of three general classes
of research designs (see page 3uff). The basis for classification on
this dimension is, as before, the number of observation points pro-
vided in a particular design. Again, an observation point is a single
point in time in.a.research situation at which variables are measured,
regardless of whether the measurement takes place for one or more
groups. Thus, "after-only" designs have one observation point, While
"before-after" designs have two. Replication of a design at a later

point in time does not increase the number of observation points, even

 

when the replicated data is included in the overall analysis. Like-
wise, when a particular basic design is repeatedly conducted.with a
single group (designs 8 and 9, Appendix A), the number of observation
points is not increased, but remains as given in the basic framework.
As noted earlier, a point design with only a single observation
point (first column of Table 2) will produce what were termed "point"
data. .A difference design with two observation points (second column)
*will produce "difference" data. Process designs with three or more
observation points (third column) are those whidh are capable of pro-

ducing data that characterize the fOrm.of variation within an interval,

57

.39. .A .33 $9800 003 mcwﬂmmc 90 8.58093 Swuuodoo 8.3 o 033 m 5 Bunch.»

#0: 09m ombudsman“? owning £05m Amanda do £me .HHDHHMS .mﬁmendncm angm we on $902508 3ng was Amﬂmzﬂmcw
novomd c5 .Cmegmp 3332: .8338 H35 wag .893 Ham. hgov Rowena—3m gwcawhoo owﬁ. 302 .8303
53980 5032 833.500 m wﬁnmﬁnowvg .«o moomﬁa maom on». you 08:00:00 am owns 523 5 0:0 we cwwuoc Eavﬂgoo <o

#8 can. 2; ﬂamenco mo 939868 on 0» 8,839.00 our mcwﬁmmc 3.38.08 .< xi 5 gapdpcog
ogﬁm 05 B. 33 com 6539 .35 can. 3398 9.3 cu human manoﬁcmnwo 5 Agtod gen-E m5. :.m:ww$c= Hmwcmu

 

a? .165. one .195: mo 8% 05 535 .333 >05 cam H3350 so 53m #99 we Sag H8392, 05....
H .1 q
as .81.... .8 .82 .
. . .ﬁmﬁﬁoo 88m .58 .
3:3 .83 .gomgv . Ammnnm do .33 3.39958 . ”a 582 gwpooam .
8V Hume ..m>w3.3aﬁd£. . 9: 35¢ 39.02. . nobmm 9333.958. Heggmué

H 33 bgﬁgmacoamgmm ”
SC gucwﬁumm .
3326 gingham .

SC 995
35:8 5w: $3.58 .
uuwwumﬁm wag—Emanumumomm .

A03

 

 

 

. .mSJd umoﬁmomuumﬁva ” Aﬁﬁoa ” 356 H
:3 838$an maﬁa: . 395-8% . Sﬁgmﬂo 96. 92 as . 8s: .
Bud .3285 go. ﬁe. . Sc . ”gem 55 8335 . 8 .
emoﬁwomumuuwa ﬁguﬁmkamm . 996 35:8 239%; . QC cooﬁamaupﬁoo. 0% .
E 85mins n as £383. 5338 H - - - n 995
3v moagwlg #5Hm>gcm 5 Hugguﬂmﬂé
. so 996159 5.38 . .
3585 .3983 95 Eu . c: 92w Honcho . a: 995 35.80 .
H9560 ﬁe. p8t8m$mog . ﬁe, wwwﬁmomﬁmmuwa . £3 sEonpuéﬁwom. Hughes 25.
. . ” A8 89.8980 9.80 033w”
. AS pwotmomnymog 995:20 . A5. 35.5 0mg “55:08. Eggnmﬂm
P [PL
«Swans? . mmumwruocwummwn . Emma ~59? 1|” maﬁa
A... We» m new HUS . any .Huv . AHS 658m m0 232
£8 woven. .i 9F .p 8P L

 

358 Savages no .8952

 

 

.9538 5803. so SﬁSmummmd swabs < .N 33.

58
i.e., the process. It is important to note, though, that three
observation points are not always sufficient to adequately Character-
ize a particular form.of variation. The minimum.number of points
needed is, instead, intimately related to the form of variation
involved, and may be any number up to infinity (w), or continuous
measurement.

The vertical dimension of Table 2, "Nature of the Research
Design," is taken from Campbell and Stanley's (1963) discussion and
classification of common, basic researCh designs. Although these
authors are primarily concerned with designs which might be applied in
educational researCh settings, their compilation draws on, and appears
applicable to, many diverse research interests. While no claim is
made, here, for the eXhaustiveness of their presentation, the authors
do attempt to be moderately comprehensive, and have included the basic
designs most frequent in communication research.

The basis for classification on this dimension is the degree to
which a particular design approadhes a true experiment. The latter is
taken by the authors to be one in which the researcher has both full
control over the exposure to experimental stimuli (i.e., the manipula—
tion of independent variables), and the ability to randomly select Who
will be exposed (i.e., to establish adequate experimental control)
(1963, p. 3h). The true experimental designs are described briefly in
the second row of Table 2. The "pre-experimental" designs in the first
row are those which have ". . . suCh a total absence of control as to
be of almost no scientific value (Campbell, 1963, p. 6)." The "quasi—

experimental" designs indicated in the third row, however, include

59
some degree of experimental control, but fall Short of the full
control of the true experiment. The designs included in the fourth
row are discussed by Campbell and Stanley, with one exception, but are
not given a generic name. They have been termed "non-experimental"
designs, here, following common practice.”

The brief descriptive titles which have been given to each of
the designs in the body of Table 2 are those assigned by Campbell and
Stanley. .A full description of eadh design is given in their volume,
along with a schematic showing eaCh design's key features. For ease
of reference, these schematics have been reproduced in Appendix A, and
are keyed to Table 2 by means of the numbers in parenthesis. It is
worthwhile noting that Campbell and Stanley consider factorial designs
to be extensions of designs (u) and (6) (1963, pp. 27-31). NOte, too,
that they do not consider research "designs" whidh involve replications
to be separate, basic forms of designs, even When the data from.the
replications are included in the overall statistical analysis. For
this reason, together'with that above regarding observation points,

replicated designs are not included in the table.

2.3.2 Relative Use of Research Designs
The result of combining the Campbell and Stanley dimension

describing the nature of a research.design, with the dimension

 

luClearly, other classifications of designs are possible, as
for example those by Cox (1958), Kirk (1968), Lindquist (1953), and
Troldahl (in press). Each such classification, however3 reflects a
particular fOCus of interest in its Choice of dimensions. For the
purposes of the analysis to be made here, the dimensions Which have
been selected appear to be the most highly relevant and useful.

60

categorizing the number of observation points, is the particular two—
way classification of common, basic research designs shown in Table 2.
Again, the classification has been set up as an aid in determining the
degree to which studies of human communication have used designs which
operationalize the concept of process. More specifically, the classifi-
cation is to be used in evaluating the extent to which research has
used process designs, with respect to point and difference designs.
Such an estimate of relative usage can be made most directly by
examining Table 2 and noting the relative prevalence in research of
certain types of designs. Such an examination reveals that the designs
most frequently used in human communication research are point and
difference designs, rather than process designs.15

While a number of exceptions to this general pattern of use do
exist (and will be discussed below), it is evident that both past and
present comment on the "state" of comnmnication research lends support
to this estimate. In his 1959 article, for example, Berelson notes
that in it's first twenty—five years human communication research was
characterized by techniques such as content analysis, sample survey, and
both "quasi—natural" and laboratory experiment. An examination of the
research which Berelson considers in his comments reveals, however, that
these different techniques, whether in observational or experimental
research, were seldom employed with process designs.

The relatively small use of process designs is also evident in

Krippendorff's (1969b) recent discussion of communication research.

 

15Among the designs in Table 2, numbers n and 6 are certainly
heavily used in experimental research, as are correlational "designs"
in observational research.

61

In his article, Krippendorff analyzes the nature of the data which
has been produced by past communication research , and finds that the
bulk of such data do not fulfill even the minimum requirements he
establishes for "communication data" (1969b, pp. ll}, 31+). While there
are several different requirements for acceptability as "minimum" com-
munication data (Krippendorff, 1969b, pp. 23-33), it appears that one
which is only occasionally satis fied is that requiring the data to
provide information on ". . . the sequence of states of the system . . .
(1969b, p. 28; see also pp. 21, 23, 26)." Since process designs are
the only ones capable of providing such information, it is apparent
that such designs have been little used in communication research.

While Berelson' s and Krippendorf f ' 5 comments on research do
support the estimate of relatively small use of process designs with
respect to point and difference designs, they provide no basis for
evaluating extent or magnitude of use. One means of gaining such
information is to consider briefly various past uses in human commun—
ication research of each of the process designs. In examining these
exceptions to the general pattern of use, it will be helpful for
purposes of later discussion to distinguish both between observational
research and experimental research (page 33n), and between the data
collection procedure or design, and the analysis of the data which is

produced . 15

 

16The second distinction has already been discussed (see page
3l+ff), and is consistent with the distinction made by Coombs: "The
method of collecting data determines what information they contain ,
but)the method of analysis defines this information . . . (1953, p.
l#95 . " _—

 

62
Pretest—posttest with time extension design.-- The first design
noted in the third (or process) column of Table 2 is one which Campbell
(1963, pp. 31-32) associates closely with Hovland, Etna—1°, a point
which makes clear that the research reviewed by Berelson in 1959 did
include at least some studies using process designs. This particular

design has been termed here the "pretest—posttest with control group

 

and time extension" design, and has been used by Hovland and others

 

(see review in Hovland (1953, Ch. 8)) in experimental research on
attitude effects over time. A similar posttest only design has been
used by Insko (1961+) in a study of primacy—recency effects. Typically,
the data collection using this design involves three observation points ,
although four or more might be used in a large study. The analysis of
the data generally considers the time dimension inherent in. the data,
since it is usually of primary importance in the study.

Time series designs.-- The "time series" design noted next in

 

 

the third column, and the "mnultiple time series" design noted below it,

 

are sufficiently similar to allow joint consideration. While these two
designs appear to be little used in communication research in their
"pure" forms (those presented by Campbell and Stanley, 1963, pp. 37—u3;
55-57), there are several areas of research relevant to communication
which make use of closely related designs. Such designs can be found
particularly in the areas of interaction analysis and language behavior,
as well as in research in "ecological psychology," and in arousal. A
brief look at each area is in order before returning to the consideration

of other process designs.

63

Within the broad scope of interaction analysis research, there
are a number of observation techniques whose sampling schemes tie them
closely to the time series designs. These techniques include those of
Harrison (Verbal-Non Verbal Interaction Analysis, 1969b), Lennard and
Bernstein (Clinical Sociology, 1969) , Lindsley (Conjugate Programming,
1969) , Scheflen (Context Analysis, 1966) , Bales (Interaction Process
Analysis, 1950), Chapple (Interaction Chronograph, 191+8), Ekman (1957),
and Flanders (Interaction Analysis, 1967) . For the most part, the
research which has been done using these techniques has been observa-
tional research, owing to the purposes for which they were developed
(Frahm, 1969). However, work using the latter four techniques (Bales,
Chapple, Ekman , and Flanders) has included experimental research, and
certainly all of the techniques are potentially useful in such research.

Although there is some danger of overgeneralization when con-
sidering these interaction analysis techniques as a group, they typi-
cally involve a data collection procedure which samples an on-going
behavior at a number of points in time. The distribution of these
points varies from essentially continuous sampling (Chapple, Linds ley,
Scheflen), to short, regular intervals (Flanders, Harrison, 3 sec.;
Ekman, 12 sec.), to intervals determined by the characteristics of

the behavior itself (Bales, Lennard and Bernstein).17 Such sampling

 

17the that the sampling, not the recording, of the behavior is
of prime concern here. That is , on-going behavior is often recorded
on film or tape in such tecluniques for use in later study. While such
recording involves "sampling" (angle of camera, 16 to 30 frames per
second), this is not the sampling which produces data for analysis
(unless each frame of a film is analyzed separately). See Ekman
(1969, p. 298) for further discussion.

Bu
procedures clearly link the "designs" used in these techniques to the
time series designs. Not infrequently, however, the analysis of data
produced by these techniques eiﬂner partially or completely removes the
time dimension inherent in the data. This is especially true of the
interaction matrices produced using the techniques of Harrison and
Flanders, and is often the case with the analyses of data produced by
the other techniques. Of particular interest in the context of this
paper, however, are occasional analyses such as those by Prahm (1970),
Lennard and Bernstein (1969, pp. 70-77), and Lindsley (1969), which
directly take the time dimension into account.

The broad range of research in language behavior also includes
a number of studies which make use of designs closely related to the
time series designs. Of the two research interests which are of special
concern here, the area of develOpmental linguistics is perhaps the one
most fundamentally concerned with the dimension of time (Smith and
Miller, 1966). In particular, much of the work reported by Brown and
Bellugi (196M), Erwin (196“), and Templin (1966) has been in the form
of longitudinal studies, with the primary intent of observing and
describing various aspects of language development in children. Such
studies normally involve data collection procedures which sample the
developing behavior at intervals ranging from two weeks (Brown and
Bellugi), to six months (Templin) . The analysis of the resulting data
does, of course, generally take the time dimension into account since
it is a primary concern .

In addition, some of the research which has been done in the

area of extralinguistic aspects of language behavior has also involved

65
a concern with the time dimension. In particular, Mahl and Schulze

(196l+) have noted a number of both observational and experimental
studies concerned with the expression of anxiety through changes over
time in such non—verbal, vocal features as voice quality (pitch), inter-
ruptions of continuity, silent pauses (non-fluencies) , and speech rate ,
as well as in certain temporal characteristics of interaction. Studies
of these particular extralinguistic features have frequently involved
data collection procedures which sample the langnage behavior either

at regular intervals , or at intervals determined by certain character-
istics of the behavior. Again, it is the sampling procedure which links
the designs used in these extralinguistic studies, as well as those
used in the language development studies, to the time series designs.
Unlike the case for language development, however, the analysis of data
from extralinguistic studies often removes the inherent time dimension.
Among those studies which do consider time are ones like that of Mahl
and Schulze (196”, p. 60) using a device such as the Interaction Chron-
ograph, and ones like those of Mahl (1963), observing emotional expres-
sion; of Starkweather, e: 3.1; (1969), studying non-verbal, vocal
expression of mood; and of Matarazo (1966, pp. 157—158), experimenting
with interruption behavior.

In addition to the two areas of interaction analysis and language
behavior, ttere are two other areas of researcln which demand brief
attention because they include designs related to the time series
designs. One such area is that termed "ecological psychology" by its
proponents, Barker (1968) and Wrignt (1967) . Research in ecological

psychology is entirely observational and employs a technique which

66
attempts to record as much as possible of an on—going behavior in its
actual context. The recording of the behavior is therefore essentially
continuous , although the analysis divides or samples the behavior in
terms of natural "units" (Barker, 1968, pp. ll-l7). The analysis of
the data produced does frequently consider the time dimension, but
often in a qualitative rather than a quantitative form, as in the small
group studies of Jordan, 3211: (1963).

Perhaps more important with regard to the study of communication,
however, is the research in the area of human arousal, particularly as
evidenced in the work of Berlyne (1960, 1965). In very general terms,
much of the experimental research done in this area of human information
processing has examined the relationships between arousal (in several
forms), and various situations in the environment such as conﬂict ,
surprise, etc. Other similar research such as that by Greenberg (1970)
has examined changes in EEG activity during the learning of speech and
non-speech stimuli. The data collection procedures used in such studies
generally involve either continuous or small-interval sampling of the
(In-going behavior, a feature which of course links them closely to the
time series designs. The data analysis, as well, frequently takes the
time dimension into account.

This brief digression into research in the areas of interaction
analysis, language behavior, ecological psychology, and arousal reveals,
then, that research in human communication has made some use of research
designs which are closely related, though not identical, to the time
series designs noted in the third column of Table 2. In addition to the

time series designs, however, there are two other process designs which

67

require dis cuss ion .

Wte-sample with time extension design.-- The design

 

described in the third column by the title, "separate-sample pretest-

 

gnstest with time extension," is one which is occasionally employed in

 

sample survey research. This particular design is a variation of a
more basic design noted in the second column, and has an advantage in
that it may be used to control for the effects of history, maturation,
and other temporal trends (Campbell, 1963, p. 53). However, while the
more basic "separate-sample pretest-posttest" design has been used in
research in human communication (Star, 1950), the time extension design

appears to have found little application.18

Multiple wave panel design.—- The final design noted in the

 

third column of Table 2 has been termed the "'multiple wave' panel"

 

design. This particular distinction among panel designs is not a
common one, although the nature of such designs (as opposed to "mo
wave" designs) is evident in Lazarsfeld's early discussions of the
panel teChnique (1938, l9u8; see also Zeisel, 1957, pp. 215—25”). The
multiple wave panel design has been used in communication research
(Lazarsfeld, 191414), and is one of the designs included in the research
to which Berelson refers in his review (1959). Despite this use, how-
ever, the two wave design has clearly remained the predominant design
wherever panel techniques have been employed (see Campbell, 1963, p.

67; Lazarsfeld, l9u8; Zeisel, 1957, pp. 215-25u).

 

18Campbell (1963, p. 5b.) notes that these designs, as a group,
are frequently used primarily to gain generalizability or accessibility
to respondents.

68

Perhaps conspicuous by their absence in this consideration of
past uses of process designs in communication research are the two
broad areas of diffusion and psycholinguistics. Clearly the time
dimension is of fundamental importance in theory dealing with the
diffusion of innovations (Rogers, 1966, p. 30; 1967, p. 8; fOrthcoming,
Ch. 2). It is evident, however, that much of the research in dif-
fusion depends on designs involving one or at most two observation
points (Rogers, 1966, p. 30), often supplemented by recall (see page
35m and Rogers, 1967, p. 8).19 The same is true for studies of the
diffusion of news events, often patterned on the studies by Deutsdhmann
and Danielson (1960) which utilized a single observation point. While
an exception to this general pattern.appears in a.report by Kivlin,
et_§l: (1968), analyzing data from the third phase or Observation
point of a major study, it is evident that research in diffusion.has
made little use of process designs.

{A similar~ccmment might be made about several different research
interests in the area of psycholinguistics (excluding language develop-
ment and extralinguistic phenomena), particularly as that field has
been reviewed by Osgood, Sebeok, and Diebold (1965). While much of
the theory and research in this area is not inherently concerned with
the dimension of time, certain interest areas such as sequential psy-

cholinguistics would, at first glance, appear to hold time as central.

 

19Systems analysis has been proposed as a means of freeing dif-
fusion studies from dependence on one or two observation points (Carroll,
1968, p. 13). Systems analysis, however, is an analytic technique, not
a data collection procedure or research design.

69
Closer examination (Osgood, 1965, pp. 93-125 and pp. 228-23M) reveals,
however, an overriding emphasis on the statistical structure of messages
and particularly on transitional phenomena, rather than on time rela-
tionships. .Accordingly, the research done in this interest area is
frequently observational and analyzes Spoken or written records to
establish probabilities of occurrence and of transition. Experimental
work, as for example that by Howes (195”), is likewise often concerned
with establishing probabilities .20 A similar lack of emphasis on the
time dimension can be found in the psycholinguistic research interests
which Osgood, Sebeok and Diebold (1965) termed "synchronic" and "dia-
chronic" psycholinguistics. Exceptions do occur, of course, in which
the time dimension is considered in such research, but in general the
area of psycholinguistics, like that of diffusion, has made little use
of process designs.

Clearly this relatively brief discussion of past uses (and non-
uses) of process designs in human communication research is neither
exhaustive norncomprehensive. It is not exhaustive both because of the
limits of space, Which force a focus on broad areas of research rather
than on individual studies, and because of the difficulty of setting
limits on what is or is not relevant to the study of human comnnnnica-

tion.21 In this latter regard, it is important to note that while a

 

20It is worthwhile noting that the Cloze procedure (Taylor, 1953,
1956), Which is a frequently used tool in sequential psycholinguistics,
does, in effect, collect data over a time dimension in a manner not un-
like that found in the time series designs. However, the analysis of
such data typically removes any time dimension present in the data.

21It is evident, however, that a great number of the studies
mentioned above do not meet the requirements fOr producing minimum

70
distinction has been made between observational and experimental
research, both are considered relevant in this context. This brief
discussion of past uses of process designs is not comprehensive in the
sense that much more could be added to the consideration of each area
of research, especially with regard to the purposes for which the
research was performed. It is because such purposes often do not
require a consideration of the time dimension that a distinction has
been made between the data collection procedure used in research and
the analysis of the data produced, the latter being keyed to purpose,
rather than to research design.

These two broad limnitations do not , hmever, prevent the above
discussion of research from serving as a basis for evaluating the
extent of use of research designs from the general class of process
designs. That is, while it is evident from the above discussion that
process designs have been used in certain areas of communication
research, it is clear that these particular areas by no means comprise
the mainstream of such research. Within the larger body of human com-
munication research, then, the use of process designs can be seen to
be relatively small in comparison to the use of point and difference
designs--an evaluation consistent with the information provided by both
Berelson's (1959) and Krippendorff's (1969b) comments on the state of
communication research.

This evaluation of the relative use of process designs clearly
supports the estimate made earlier by examination of Table 2 that

communication research has depended to a large extent on point and

 

communication data, as identified by Krippendorff (1969b, pp. 23-33).

71
difference designs. In sligntly different terms, it is evident that
studies of human communication have made only small use of designs
which operationalize the concept of process.22 Why the "state" of
communication research should be thus restricted is a matter of sore
interest , given both that the concept appears to be a useful one in
research and theory building, and that a number of scholars have made
calls for research and theory which deal with commmnication events as
processes--Barnlund (1968, p. 23), Chronkite (1969, pp. 128-133), Krip-
pendorff (1969b, pp. 34-35), Scheflen (1968, p. |+5), and the Speech
Association of America (Kibler, 1969, p. 35), to mention but five.23
A brief consideration of why the concept of process has not been more
evident in communication research is therefore appropriate , and will

serve as a preface to an overview of the central concerns of this paper.

2.3.3 A Bigger Picture

The question of why communication research has made only small
use of designs which operationalize the concept of process is one
which can prompt lengthy speculation. It is not the purpose here to
delve into "causes ," except to suggest three very general, yet relevant
factors. The first, of course, is that a concern with the concept of
process is not necessarily central in all human communication research,

just as it is not necessarily central in all research done in the other

 

2‘2In fact, the use of designs which operationalize process is
somewhat smaller than it appears from the preceding discussion. That
is, not all of the above uses of process designs meet the m1nimum
requirements for characterizing a process (see page 33 and Chapter u) ,
though they are capable of doing so under certain conditions .

23One is reminded of Mark TWain's proverbial comment concerning
discussion of the weather and failure to act.

72
areas of communication (e.g., telecommunication) which are enconpassed
by Krippendorff's definition (see page llff) . In some cases, then,
point and difference designs may be entirely sufficient for research
in human communication. For those cases in which process _13 central
in research, however, it is quite apparent that the concept has not
always been operationalized.

This situation may exist in part because of a second general
factor; namely, the extreme complexity of human communication events.
Communication scholars have often noted the difficulty or impossibility
of adequately modeling communication events, especially with respect
to the "process," in its broad sense, which is involved (Berlo, 1960,
p. 25). From this point of view, change over time or process, in its
narrow sense, is only one of many aspects to be considered in modeling
a communication event , a point which may account for some of the lack
of attention to the concept in the construction of communication models .
Failure to adequately deal with the concept in building models would
then certainly be reflected in a lack of concern with process in
research based on such models.

A third and perhaps most cogent factor in the failure to consider
the concept of process, hwever, revolves around what Kaplan has called
"the law of the instrnunent"--the idea that a research technique, once
made available, will occupy the center of attention. Kaplan expresses
it delightfully in the sentence, "Give a small boy a hammer, and he will
find that everything he encounters needs pounding (1961+, p. 28)." That
is, the presence of a set of research designs of considerable power and

advantage over other modes of inquiry into hturan communication may have

73
so occupied researchers that designs which would operationalize the
concept of process may not have been sought out , or even recognized
as lacking. Certainly if words or concepts can limit what their users
perceive in events, the heavy use of point and difference designs may
also have limited researchers both in viewing and in studying commun-
ication events .

Whatever more factors may be adduced, it is evident that except
for a relatively small body of work, communication events have not been
studied as processes . More succinctly, the current place of the
concept of process in hnuran communication research is a small one.

Such an evaluation, based as it is on a consideration of the relative
use of research designs, is not of itself a sufficient reason for
selecting the concept of process for special attention in this paper.
This particular reason is, however, one of the four which together
have determined the selection. The other three reasons deserve brief
mention, once again.

A second reason for the selection, mentioned several times in
this chapter, is the important place of the concept of process in
definitions and discussions of communication. Scholars have continually
pointed to the concept as a vital one both when defining their universes
of discourse, and when discussing basic concepts, as well as research,
theory, and models. A third reason for the selection is that opera-
tionalization of the concept in research can provide information which
is needed in theory building . That is , operationalizing the concept
as indicated above can provide information on the sequences of states

occurring in an event , or, more generally, on the form of variation

71+
or change over time of key variables in that event. That such infor-
mation should be provided by the research upon which communication
theory is built is perhaps one of Krippendorff '8 primary concerns in
his recent article on data in communication research (1969b). The
fourth arnd primary reason for selectirng the concept of process for
special attention, however, is its utility in the conduct of research
and theory building. As discussed earlier, the concept appears useful
not only in avoiding certain problems (like spurious categorization)
encountered in research, but also in dealing with new concerns (such
as human information processing) arising in the discipline of
communication.

It has been the burden of these first two chapters, then, to
develop and consider the reasons for selecting the concept of process
for special attention. Again, the concept is but one of several aspects
of events which must be considered in research (see page 140). Given
the limits of space, however, it is the broad utility of the concept
in the conduct of research and theory building in communication which
has lead to its selection. It will be the burden of the remaining
chapters to focus on the concept of process , alone, and to consider
the means of explicating it for further use in research and theory
building (see page 9ff).

Because an explication requires both a constitutive and an
operational definition, the first step in what follows (Chapter 3) will
be to discuss approaches to a constitutive definition. More specifically,
consideration will be given to a number of possible links between the

concept of process and other verbal and mathematical terms which might

75
apply in a fUture theory of human communication. Also, since the
form of a theory depends on the nature of the terms which corprise it,
it will be appropriate to consider the implications Which the approaches
to a constitutive defintition have for the form of theories dealing with
crmmunication events.

The second step (Chapter M) will be to discuss approaches to an
operational definition of the concept of process. That is, drawing on
the infermation provided in Chapter 3, consideration will be given to
the requirements WhiCh.must be met if a research design is to produce
data.Which adequately characterize a given form of variation or change
over time. Designs which meet these requirements are, of course,
operationalizations of the concept of process. In addition, given these
requirements, it will be appropriate to consider the implications they
carry both fortresearch techniques, and fer analytical techniques.

Such discussions of'approaches to both constitutive and opera-
tional definition of the concept of process will require the considera-
tion of terms and techniques drawn from.a number of sources outside the
discipline of communication. While each of these research tools may
find some fUture place in the study of human communication events,
they will not be presented here as suggested replacements for'present
research techniques. Rather, they will be presented as potential
additions to present techniques in order to expand the range of those
techniques to fill the "hole" or vacant spot in the classification of
research designs (column 3,Tab1e 2, page 57).

In addition to the consideration of approaches to constitutive

and operational definition, it will be helpful to consider~both

76
examples of research which takes the concept of process into account

(Chapter 5) , and certain broader implications of the concept's expli-
cation and place in research (Clnapter 6) . These two chapters, together
with the two devoted to explication, will, in a broad sense, serve to
develOp the earlier point that research which considers communication
events as processes will provide information on 133w those events occur,
not just on whether or not they did occur.

It must be emphasized again, though, with respect to the phrase
"how those events occur," that the interest in the concept of process
in this paper is not motivated by a desire to determine how events
"really" occur, i.e., to determnine causal bonds. Instead, the interest
stems from a desire to determine how, or in what manner, events take
place, i.e. , to determine certain characteristics of the sequences or
"paths" over which they develop. This latter form of information is
seen as lacking to a large extent in huran communication theory and
research to date, yet at the same time as vital and useful in that
theory and research. The concern in this paper, then, is to provide
a basis for the development of research tools to be used in providing
this missing information. Such development and eventual use does not
promise to be easy, both because of the effort involved, and because,
like any change, the overall effect may be to alter both the ways in
which events are studied, and the very content of those studies. In
this case, however, the potential change should prove to be highly

beneficial to the study of huran communication.

CHAPTERB

APPROACHES TO A CONSTITUTIVE DEFINITION

 

The previous chapter has indicated that the concept of process
or of change over time is an important and useful one in the conduct of
research and theory building related to huran communication. The choice
of tle concept for special attention on the grounds of its importance
and utility leaves the task of explicating it for use as a tool in
research and theory building. As indicated, the first major step in
this explication will be to provide approaches to a constitutive defini—
tion of the concept of process . Chapter 3 will be devoted to that step,
as well as to an examination of the implications which the approaches
have for theory. Chapter 3 will lead to a consideration of approaches

to an Operational definition in Chapter 1+ .

3. 1 Nature of the Definition

 

As mentioned earlier, in Chapter 1 , a constitutive definition of
a concept specifies linkages to other terms in the code, language, or
theory which the investigator manipulates to produce propositions about
the events under study (see page 9ff and Berlo, 1967, pp. 3, 8). It
was noted in the same discussion, however, that a formal constitutive
definition can be formed only in the context of a specific theoretical
framework, a situation which does not apply here because no attempt is
made herein to build a theory of communication. As a result, rather
than attempt to link the concept of process to the terms of some specific,

77

78
but premature, theory of'ccmmunication, the concept will be linked
instead to a set of terms Which might have a place in future theories.
In other words, the attempt will be to suggest various approaches to a
constitutive definition by providing a set of links to various verbal
and.mathematical terms which might be utilized in future theories of
communication. These suggested approaches not only provide the tools
for formning constitutive definitions in specific theoretical frameworks ,
but also hold a number’of important implications fer the formmof theory
dealing with communication events .

Because the concept of process has received only relatively small
attention in the study of hnunan communication, it will be necessary in
suggesting approaches to a constitutive definition to draw on information
from a number of disciplines outside that of communication. Specifically,
disciplines such as biology, mathematics, sociology, thermodynamics, and
systemntheory and analysis will serve both as sources of information, and
as sources of terms with which the concept of process is to be linked.
The choice of information and of terms from these various disciplines
will be eclectic, but will in all cases be a choice based on the criteria
of consistency, generality, and above all, utility. The judgments on
all of these criteria, and especially on utility, will be made in the
context of potential use of the information or of the term.as a tool in
theory building related to human communication.

Despite this necessary eclecticism in choosing infermation, it
will be fairly obvious in what follows that a relatively heavy depend—
ence has been placed on the discipline of system theory and analysis.

The dependence is not complete , however, in the sense that system theory

79

and analysis might be said to form a higher level theoretical frame-
work underlying the discussion (despite the absence of such a frame-
work avowed above). The disciplines of communication and of system
theory and analysis do overlap, as is evident in the definition of
communication considered earlier (page llff) . Such a statement, how-
ever, does not indicate isomorphism.of the disciplines, as should be
evident from.a study both of the criteria below for using systemntheory
and analysis information, and of the "non—systemW infermation and terms
introduced in sections 3.2 and 3.3. The discipline of system theory
and analysis is, then, a separate body of information as is mathematics
or thermodynamics, but is more heavily exploited below because of its
particular consistency, generality, and utility in the present concern
with the study of human communication.

System theory and analysis is a broad-ranging discipline which
concerns itself with phenomena characteristic of all forms of systems,
both living and non-living. The discipline has drawn on concepts from
many different fields, including those mentioned above, and has shaped
this information into a single approach.1 Among the concepts Which are
central in system theory and analysis are several whiCh have been intro-
duced earlier, in particular, concepts such as process, interaction, and

emergence (page 33ff). That these important concepts from the present

 

1It is perhaps misleading to consider "system theory and analysis"
as a single approach, when in fact it appears to have at least three
distinct "sub—disciplines" (Arundale, 1968). These sUb—disciplines
share a number of key concepts and approaches, however, and it is in
this sense, as well as fOr clarity of presentation, that the discipline
‘will be treated herein as single entity.

80
paper are also central in system theory and analysis makes this dis-

cipline particularly consistent with and appropriate to the concerns

 

of this paper. In addition, the ability of system theory to deal, in
an internally consistent manner, with the many types of systems in
which communication occurs (page l6ff), reveals that it has the
generaliﬂ and sc0pe necessary for use as a tool in building commun-
ication theory. Finally, because it is extremely fruitful in genera-
ting theory and research, and because it provides a wide range of tools
for such study, system theory and analysis is an especially _nis_e_f_n£ tool
in the study of human communication.2 It is the particular importance
of system theory on the basis of these criteria which leads to the
sorewhat heavier dependence upon it , as a source of information, than
upon other disciplines. Nevertheless, the attempt to suggest approaches
to a constitutive definition of the concept of process will remain
eclectic, drawing information and key terms from a number of disciplines
on the basis of the criteria mentioned.

Again, as implied in the above, there will be no attempt in this

chapter to build a single constitutive definition, or a nmnique set of

 

2There are, of course, a number of more extensive discussions of
the overall utility of system theory and analysis as a tool in theory
building and research (Bertalanffy, 1962, pp. 1-20 and 1968, Ehs. 8, 9;
Buckley, 1967, p. 39; Harrison, 1967a; and Rappoport, 1968, among others),
all of which are relevant here. Note particularly that there will be no
discussion in this paper of the basic concepts of system theory, in par-
ticular, the concepts of system, subsystem, suprasystem, component,
matter-energy, information, interaction, emergence , etc . , except as these
have already been discussed in sections 1.2 and 2.1. Such concepts are
nevertheless of some importance, herein, and a familiarity with one or
more of the basic works on system theory has been assured (Bertalanffy,
1968; Buckley, 1968; and especially J. G. Miller, 1965a, 1965b).

81

links to other terms. Rather, a number of potentially useful approches
to such a definition will be suggested and discussed. To do so, it

will be helpful to divide the remainder of this chapter into three

parts . That is, since verbal terms are more common than mathematical
terms in theory building in communication, attention will be focussed

in the second section on the verbal aspects of constitutive definition.
However, because mathematical terms have a potential importance in
dealing with the concept of process, and because they provide a different
perspective, the third section will focus on the mathematical aspects of
constitutive definition. These two discussions of approaches to a
constitutive definition in two different "languages" will not only
provide a mumber of tools for dealing with the concept of process in
theory building and research, but also lead to a consideration of the
necessary and sufficient conditions for the use of the term "process."
In addition, since the form of a theory depends on the nature of the
terms which comprise it, the fourth section will focus on the implica—
tions which these approaches to a constitutive definition have for the

form of theory which deals with communication events.

3.2 Verbal Aspects of Constitutive Definition

 

Because theory in the discipline of human communication is most
often expressed in verbal terms, the first major step in approaching a
constitutive definition of the concept of process will be to consider
links between this concept and a number of verbal terms which might
have a place in future theories of communication. Quite clearly, the

various linkages considered below are a subset of those which might be

82

discussed. Hence, theparticular subset which has been included
represents a choice of terms, in this case on the basis of their
potential utility as tools in theory building related to communication.
The discussion of these verbal aspects of constitutive definition falls
conveniently into three parts. First, an examnination in some detail of
the verbal definition of the concept of process. Second, a considera-
tion of the ties between this concept and several other closely re—
lated terms. And third, a brief discussion of several "kinds" or
"types" of processes in an attempt to provide additional useful terms
for dealing with different aSpects of change over time. Because of
the large number of terms and definitions introduced in this section,
Table 3, page 83, has been included both as an outline and as a glossary

useful for reference purposes.

3.2.1 The Verbal Definition of Process

As indicated from the outset by the Introduction, the central
concern of this paper is with the concept of process , a concept which
was defined verbally early in Chapter 1 as "All change over time of
matter-energy or information . . . (J. G. Miller, 1965a, p. 209)."
This definition or verbal equivalence statement has , of course, had
the effect in subsequent discussions of making the terms "process" and
"change over time" almost completely interchangeable. As was noted in
early Chapter 2, the concept of change over time, as the central concern,
could be discussed without recourse to the term "process," a possible
advantage since the latter term is used in a nLurber of different ways.
However, because the element of change over time is common to these

many alternative uses of "process," the term has been retained as a

E33

Table 3. Outline and Glossary of Verbal Aspects of Constitutive Definition‘
3.2.1 The Verbal Definition of Process

 

ProcessAll change over time of matter-energy or information
(within a bounded interval).
T1meA continuum of one dimension whidn is divisible into equal
units, those units being crchred unidirectionally under
the assumption that the entrqny of the universe is always

 

increasung.
Matter—energy or. .. ..Defined as in system theory and analysis (see J. G. Miller,
iﬁformaﬁon 1965a, pp. 193-199).

OnanggnHWA difference (including a difference of zero) in the value
of a variable over two or more points in time.

3.2.2 Process in Relation to other Terms

 

m...wmmﬂe static arrangemnt of a system' 8 parts at a moment in
three dimensional space. Structure," is therefore a

system's configuration at any one instant or point in a

process.
Function. . . . . . . . . . . . . . . . .The trarsient and versible changes, often repetitive,
that constitute $175an or functioning. Function is thus

a subset of process involving reversible change.

m. . . . . . . . . . . . . .Tne internally determined control process of a system which
maintains at least one of its variables at a given steady
state value.

Histgy. . . . . . . . . . . . . . . . . .T‘he enduring and irreversible changes, often progressive,
that constitute Ecoung or develcpnent. History is thus
a subset of process involving irreversible change.

Evolution. . . .. . .A subset of history in which the basic event is one of

mutation.

3.2.3 Inns of Process
General Descriptors:

Reversible. . . . . . . . . . .A process, which once having takentplace, may be retraced so
as to leave no change in either e system or its
surroundings.

Irreversible. . . .. . . . .A process which may not be retraced without changing the
system or its surroundings.

Sta__t__ic. . . . . . . . . . . . . . .Processes in which the value of the variable under con—
sideration does not change over time.

Mistatic. . . . . . . . . .Prooesses in which ﬁne deviation from the static situation
is infinitesimal, so that the process may be treated as
static with essentially no error.

Cyclic ...... .. . . ..Processes in which a variable changes from sore Specific
value or state, only to return eventually to that same
value or state.

Sgcific Descriptors:

_Mor_phostatic. . . . . . . . .Processes which tend to preserve or maintain a system's
given form, organization, or state. Sudn processes
maintain the values of variables within specific "ranges
of stability," and are often termed "dynamic equilibrium"
processes. These processes are often convergent, and may
be either reversible or irreversible.

Mo enetic.. . .. ...Processes whidn tend to elaborate or change a system's
given form, stnennen, or state. Such processes move
the values of variables outside of ranges of stability
and hence are "non-equilibrium" processes. They may be
divergent, and can be either reversible or irreversible.

 

aI‘he sources of ﬁnese definitions are given in the discussion.

8L}

useful organizing device for discussion, to be teed only in the
narrower sense of its definition as "change over time."

As a first step in considering the verbal aspects of con-
stitutive definition of the concept of process, it will be helpful to
examine this verbal definition of the central concept in greater
detail. The effect will be to link the concepts of "process" and of
"change over time" to a number of other important terms. Note again
that no attempt is made here to identify the definition given as the
"correct" or "true" definition (page 11). That is, J. G. Miller's
definition is but one of several equally serviceable definitions for
the term "process" in its narrow usage. Other definitions could be
used in its place with simnilar results, as exemplified below. Miller's
definition has been chosen here because of its generality with respect
to other definitions of "process."

The individual terms. - If one examines each of the terms in

 

Miller's definition, it becomes apparent that the key word in this,

and certainly in mnost definitions of "process," is the term "£1113."
Above all else, the concept of process serves to call attention to the
dimension of time, from among the many different dimensions of phemonena
which mnight be attended to. The concept of time is a basic concern in
this paper, as indicated in section 1.1, and has been the focus of
considerable scholarly discussion (Reichenbach, 1956, 1958; Whitehead,
1925, Ch. 7, and 1929, Chs. 2, 10; etc.) Space prevents a review of
this lengthy and involved discussion concerning "time," but does not
preclude a more precise definition of the term. '"I‘ime" will be used

here to refer to what is sometimes called "astronomical time": a

85

continuum of one dimension which is divisible into equal units , those
units being ordered unidirectionally under the assumption that the
entopy of the universe is always increasing . This latter qualifica-
tion is an aspect of "thermodynamic time," as discussed by Bertalanffy
(1968, p. 231), and reflects Eddington's statement that entropy is

"the arrow of time (Bertalanffy, 1968, p. 151)."

While "time" is an important term in the defintiion of "process,"

it is not synonymous with it, in that "process" also refers to "change

. . . of matter—energ; or information . " The concepts of matter-energy

 

 

and of information, as used here, encorpass all entities in the
universe, and have, like time, been the foci of extended discussion,
much too voluImnous to review here. The two terms have been and will

be used as they have been defined in the discipline of system theory
and analysis (J. G. Miller, 1965a, pp. 193-199; see also note 2, above).
Of much greater importance in examining the definition of the concept
of process is the term "change."

As it appears in Miller's definition, "91—32%." is used in its
broadest sense; that is, the term refers not only to the presence of
change or difference, but also to its absence. To clarify this seeming
contradiction, it is helpful to consider again the representation of a
process on a graph; i.e., if time is the abscissa and variable value
is the ordinate of a graph, then a line on this graph represents a
process and indicates the successive values of the variable at succes-
sive instants in time. On such a graph, there will generally be some
identifiable change or difference between the values of the variable

at any two given instants in time. The amount of change or difference

86
would be found, of course, by subtracting the two values of the
variable. Note however, that in the special case where the values
at the two instants are equal and the subtraction yields zero, it
is still meaningful to talk of a "change" or "difference" between
the values. In this case, the amount of Change or difference is
zero, but the two terms "change" and "difference" have remained the
same, having been used in this second instance in their'brcadest
senses.

Exactly the same situation obtains in Miller's use of the termn
"change" in the phrase "change over time." That is, the usage covers
'not only the case where a difference in variable value is present over
two points in time, but also the special case where the difference is
zero. The potential confusion of this broad usage can perhaps be
cleared by considering briefly another definition of "process," in its
narrow sense, which does not employ the termn"change." This par-
ticular definition is drawn from.the discipline of thermodynamics and
utilizes both the concept of a path or f0rm.of variation, and the con-
cept of a "state" or a set of values of variables at a single point in
time (see discussion on page 35ff). According to this definition,
"process" refers to "the path of the succession of states through
whiCh the system passes . . . (Van Wylen, 1959, p. 17)," the dimension
of time being implied, as is frequently the case in the physical
sciences due to its ubiquity. When defined in this manner, the concept
of process can be seen to encompass all paths or lines whidh might be
drawn on a graph SUCh as that above, including paths or lines parallel

to the time axis. Such a parallel line, translated into terms of

87
Miller's definition, would be considered to show zero change or
difference. Thus, the two definitions of "process" are equivalent
in encompassing both the presence and absence of change.

Additional comments . - In addition to this consideration of
each of the terms in Miller's definition, it will also be helpful to
examine three other important points regarding the verbal definition
of the concept of process. The first point or comment is that the
term "process" is a completely general term, as evidenced by the
presence of the word "all" in the phrase "all change over time . . . ."
That is, the variable which is plotted on the ordinate, in the graphic
representation, can be any variable. Hence, if the variable is one of
space, then the process which is represented is one of "movemnen ."
Similarly, if the variable represents the total amount of sore sub-
stance, then the process is one of "grovth," and so on. The concern

here, again, will be with variables related t_o communication and par-

 

 

ticularly to human commuunication-—a bias mentioned earlier (page 25),
and reflected throughout the discussion.

The second comment is closely related to the first, and reflects
a different aspect of generality in the concept of process. As
mentioned in the discussion of "process" in section 1.2, the term is
used in this paper to refer both to situations in which a single vari-
able changes over time, and to situations in which multiple variables
change over time. The latter case is the more general one, and would
be represented graphically by several separate paths or lines, all
extending through the same time span on the abscissa but each repre-

senting a different variable plotted on the ordinate. Because of its

88
simplicity, the single variable case is the most appropriate for ’
explanatory purposes, and has been used above and will be used below.
The simplicity gained in these uses does not bring a loss in general-
ity, since the principles discussed for the single variable case can
be directly extended to the multiple variable case, and since the
primary interest in this paper is in the relationship of variables to
t_ir_n_n_e_, not necessarily to each other.

The third, and perhaps most important comment regarding the
definition of "process" is closely tied to the discussion in section
2.1 of the "continuity" of an occurrence. As defined by J. G. Miller,
a process has no bounds and may be viewed as extending indefinitely
in time, i.e., as having continuity. It is clear, however, that no
human observer can study a process in its continuity. The observer is
forced, instead, to divide that continuity into segments which he is
capable of observing, the division being either implicit, through the
limits of perception, or explicit through the identification of specific
bounds in time. The need to segment the continuity of occurrences is
particularly pressing for the scientific observer, who can discriminate
properties and patterns in a process only to the extent that such
properties and patterns repeat themselves over intervals of time which
he can recognize and explicitly delimit.

These implicitly or explicitly segmented intervals of time were
earlier termed "events (page 13) , and because this segmentation or
bournding of occurrences is basic to science, all reference to a

process in this paper will be to a process as it takes place within

the bounds of an event. Such a restriction of the term "process" to

89

events is at variance with Miller's definition, but is seen as useful
in its consistency with the behavior'of human Observers in dealing
with the continuity of occurrences. Note also that references in this
paper’to a process or Change over time, in the interval between two
time boundaries, assumes that those boundaries are set by the observer,
not by the event.3

These three comments on various aspects of the verbal defini-
tion of "process," together with the consideration of the individual
terms in the definition, constitute a somewhat more detailed examina—
tion of the verbal expression of the concept of process. Sudh an
examination is only a first step in considering the verbal aspects of
constitutive definition. A second step of at least equal importance
is that of considering the ties or links between the concept of process
and several other closely related terms whiCh.have potential importance

in communication theory .

3.2.2 Process in Relation to other Terms

Within the discipline of system theory and analysis, "process"
is one of two key concepts used in the characterization of systems
(J. G. Miller, 1965b). The other concept, that of "structure," is
closely related and is used to refer to "the static arrangement of a

system's parts at a moment in three dimensional space (J. G. Miller,

 

3Note that this comment is considerably more general in scope
than the discussion of Watzlawick, et al. , in considering the concept
of punctuation of events (1967, p.-5'l+:59) . These authors are primarily
concerned with a participant ' s punctuation or bounding of communication
events, whereas in the present case, a "participan " is only one type
or class of observer.

90
1965a, p. 211)." In other words, because an instant (moment) has no
dimensions in time, the arrangement of a system's parts in Space at
a particular instant can be considered as fixed and unchanging. It
is this arrangement which is the structure of a system. Any change
in this structure would involve some span of time and would become
a consideration of process (this use of the term "structure" is the
same as that by Rappoport, 1968, p. xx; see also Gerard, 1968, pp.
52, 56) .

The use of the term "5th " presents a potential confusion,
however, in that it was introduced and used earlier (page 15) to
refer not to an arrangement of matter-energy, but to information.

The distinction between these two uses of the term will be made by
attaching the subscript "m" or "i," so that "structurem" will refer
to "the static arrangement of a system's parts at a moment in three
dimensional space," while "structurei" will refer to the pattern or
organization among the parts of an entity which is the informational
aspect of that entity when it is under study by some observer. Note
that the presence of structurei implies the presence of structurem,
since information is slvgys going gs markers; however, the reverse is
not necessarily true. That is, a completely random distribution of
matter—energy is one possible form of structnmem, but is one which may
well bear no structurei for an observer (depending on his particular
frame of reference).

Because the relationship between structurem and process is very
close, in the sense that structurem can be identified only at some

specific instant in a process, J. G. Miller makes several important

91
clarifying points in his discussion of the two concepts. In particu—
lar, regarding the analysis of a system;

Any subsystem.. . . is identified by the process it
carries out. It exists in one or more identifiable struc—
tural units of the systems The specific, local, distinguiSh-
able structural units are called components. . . . There is
no one-to-one relationship between process and structureEmJ.
One or more processes may be carried out by two or more
components. Every system is a component, but not necessarily
a subsystem, of its suprasystem.(J. G. Miller, 1965a, pp.
218—219).

Structurem and process in a given system are therefore not completely
independent, but neither are they isomorphic. As Miller suggests,
"It is notoriously hard.to deduce process from.structure[mJ, and the
reverse is by no means easy (1965a, p. 219)."

Note particularly that in the above quotation, Miller is using
"process" in its most general sense, a usage WhiCh in this case
includes at least two other concepts whiCh also demand attention,
namely function and history. The concept of fUnction is often treated
as synonymous with process, although in Miller's usage it is not,
being somewhat more specific. Function is definitely a form or

aSpect of process, but carries an additional reference to ". . . the

transient and reversible Changes [my italics], often repetitive, that

 

constitute 'behaving' or fUnctioning (Gerard, 1968, p. 52, see also
pp. 55-56)." .Alternatively, in Miller's terms, "Process includes [as
a subset] the on-going function of a system, reversible actions suc-
ceeding each other from moment to moment (1965a, p. 209; first italics
mﬁne)." Again, this definition of "fUnction" is similar or identical
to that employed by Rappoport in his use of the term "functioning"

(1968, p. xx). The concept of function thus carries the added

92
specificity of "behavior," or of the "regulation" of system.states

over time by means of reversible Changss in some or all of the system

 

variables. This added specificity in the concept of function ties
it to the concept of purpose , but it is especially important to note
that the two terms are distinct. Miller defines "purpose" as " . . .
the internally determined control process [my italics] of the system
which maintains [at least] one of its variables at a given steady
state value (1965a, p. 232, see also Gerard, 1958, p. 128)."

This link between the concepts of process and of function is
important not only in itself, but also as background for the relation—
ship between process and the concept of history. History is also a
form or aspect of process, but again is more specific in that it refers

to situations in which there are ". . . enduring and irreversible

 

Eggs [my italics], often progressive, that constitute 'becoming' or
development (Gerard, 1968, p. 52, see also pp. 5n-55)." Again in
Miller's alternative terms:

Process also includes [as a subset, the concept of]
histo , less read11y reversed changes like mutations,
birth, growth, development, aging, and death; changes
which commonly follow trauma or disease; and the changes
resulting from learning which is not later forgotten.
Historical processes alter both the structure and function
of the system. . . . History, then is more than the
passage of time. It involves also [the] accumulation in
the system of residues or effects of past events . .
(1965a, p. 209; first italics mine).

The initiation, growth or development, learning, pathology or disease,
decay, and termnination of a systemuin short, the phases of its life
cycle--are therefore all forms of process. Much more specifically,

however, to the extent that they involve irreversible changes they

 

93
are also aspects of the history of a system1(J. G. Miller, 1965b, pp.
372-378). This additional specificity in the concept of history
ties it to the concept of evolution. Miller considers evolution to
be an aspect of history (and hence of process), but is quite specific
in noting that the basic event in evolution is one of ". . . mutation,
a process Which is ordinarily irreversible . . . (1965b, p. 370; see
also pp. 369—372)." Note particularly that Rappoport (1968, pp. xx)
employs the term1"evolution" to encompass hgzh_"history" and "evolu-
tion," as those terms are used here. The distinction made here be—
tween the concepts of history and evolution is seen as valuable and
useful, however.

Briefly, again, "History, or becoming, . . . is a.regu1ar
change, normally progressive, in a system along the time axis; fUnction,
or behaving, is a repetitive perturbation along this secular trend;
and structureEmJ, or being, is the instantaneous status (Gerard, 1968,
p. 5M)." These three concepts, and those tied to them, are all linked
to the concept of process or of change over time. They are highly
important and useful concepts in the Characterization of any systemu
and given that communication occurs within systems (see page 16ff),
are concepts whiCh are potentially usefu1 tools in commmnication theory
building, as well. EaCh of these terms and their links to the term
"process" could be considered at greater length, as they are in the
discussions by Gerard (1958, 1968), Rappoport (1968), and J. G. Miller
(1965a, b) whiCh have been Cited (see also Bertalanffy, 1968).

Because such works are available, however, it appears more useful to
move on to the third step in considering the verbal aspects of consti-

tutive definition of "process."

gm
. 3 Types of Processes

It will be helpful to examine briefly several different terms
_ch describe various "kinds" or "types" of processes in a system—-
a purpose being to provide means for dealing with different aspects

"forms" of change over time. Each of the terms to be considered

low serves to modify the concept of process in some manner, so that
ken as a group, they illustrate another aspect of the generality of
Le concept. Note that the descriptive terms chosen here will £13 be
nose which identify processes according to specific system functions,
1ch as "input," "internal," or "output" processes. Irstead, it
ppears more useful, in view of the concerns of this paper, to discuss
erms which describe the kinds or types of variation which may take
lace in the values of a system's variables. It will be helpful to
group such terms into two categories: general descriptors and specific
lescriptors.

Two of the five highly general descriptive terms have already

oeen used above, namely, "reversible" and "irreversible." The term
"reversible" is used in the discipline of thermodynamics to refer to
". . . a process, which once having taken place, can be reversed and
leaves no change in either the system or surroundings (Van Wylen,

1959, p. 127)." J. G. Miller sharpens this definition by alluding to
the mathematical description of a process: "If the equation describing
a process is the same no matter whether the temporal variable is
positive or negative, it is a reversible process: otherwise it is

irreversible (1965a, p. 209; see also the discussion herein, section

3.3)." Clearly, then, an "irreversible" process would refer in

95
ermodyrnamic terms to a process which could not be reversed without
anging the system or the surroundings. As noted above, it is
:versible processes which are associated with the concept of function,
tile irreversible processes are associated with history.

The three other highly general descriptors which deserve
antion are "static," "quasistatic," and "cyclic." "Static" processes
we already been alluded to in the discussion of the term "change."
ney are processes in which the value of the variable under considera-
ion does not change over time; i.e., in which the line or path of
ariation in the graph of a process is parallel to the time axis. A
quasistatic" process is one in which the deviation from the static
ituation is infinitesimal, so that the process may be treated as
;tatic with essentially no error (Van Wylen, 1959, p. 18). Lastly,

1 "cyclic" process is one in which the change in the value of the
Iariable exhibits "cycles," a cycle being a situation in which a vari-
able changes from some specific value or state, only to return event-
ually to that same value or state (of. Van Wylen, 1959, p. 18).

The remaining descriptive terms to be considered are somewhat
more specific than those mentioned above, and can themselves be use-
fully divided into two groups. The first group is descriptive of what
can be termed "morphostatic" processes--those which ". . . tend to
preserve or maintain a system's given form, organization, or state '
(Buckley, 1967, p. 58)." The second group is descriptive of "morpho-
genetic" processes--those which ". . . tend to elaborate or change a

system's given form, structurefm], or state (Buckley, 1967, p. 58)."

96

Morphostatic processes are frequently described by the terms
nomeostatic" and "steady state." That is, they are processes in
nich the values of variables are constantly fluctuating or changing,
ut are maintained within limits or "ranges of stability" by means of
egative feedback (J. G. Miller, 1965a, pp. 22u—229).” Such fluctu-
ting or changing variables are often thought of as remaining in a
dynamnic" or "flux" equilibrium. One specific type of morphostatic
»rocess is that which is termed "convergent." In this type of process,
he value of the variable under consideration approaches some equi-
Librimn value, whether by means of a decreasing cyclic oscillation
about the equilibrium, or by means of a cyclic or non-cyclic trend
toward that value from a single direction (see Lennard and Bernstein,
1969, pp. 13-1u).

Morphogenetic processes do not have as well developed a term—
inology as do morphostatic processes. They mnight be termed "non-
horeostatic" or "non-steady state" processes, in that they include
situations in which the values of variables move away from equilibrium
values and outside ranges of stability, often as a result of positive
feedback. Morphogenetic processes are "non-equilibrium" processes,

and mignt be more specifically termed "divergent" processes. The

”Space forbids a more detailed discussion of the basic concepts
involved here (see note 2). Complete discussions may be found in
Miller, as well as in Bertalanffy (1968), and Buckley (1968). It is
important to note, with regard to Miller's discussion, that the part
of section 12.1 which is contained on p. 221+ makes several statements
which are inconsistent with the remainder of his discussion, especial-

ly pp. 225-226. Reference to the other works is particularly helpful
in Clearing this inconsistency.

97
mm "divergent" would include situations in which the value of a
triable moves away from an equilibrium, either through an increasing
rclic oscillation about the equilibrium, or through a cyclic or non-
zclic trend away from it.

It is important to note that both morphostatic and morpho—
enetic processes can be either reversible or irreversible. That is,
espite the usual association of morphogenetic processes with irre-
ersible change (e.g., history, development, evolution, etc.), there
5 no necessary connection between the two: a morphogenetic process
nay be reversible. Likewise, despite the normal association of morpho-
static processes with reversible change (e.g. , function, regulation,
etc.), a morphostatic process may in some cases be irreversible (see
I. G. Miller, 1965a, pp. 209, 22a).

These descriptors of certain specific kinds and types of
processes , together with the various more general descriptors , might
well be discussed at somewhat greater length and could certainly be
expanded in number (see Buckley, 1968, Pts. V and VI). Some additional
descriptive terms will be added in the next section, but as potential
tools in theory building in commmication, the terms considered above
appear to be the most useful of the verbal terms which might be chosen.
This criterion of potential utility has , again, been used throughout
this section in choosing the various terms which have been linked to
the concept of process .

In summary, the discussion of the linkages which have been
selected has encompassed not only an examination of the verbal defini-

tion of "process," but also a consideration both of the ties between

98

113 concept and three other key concepts , and of the various kinds

7 types of processes. In short, the discussion has covered a number
‘5 different verbal aspects of constitutive definition. The task of
1ggesting approaches to a constitutive definition of the concept of
rocess cannot be restricted, however, wholly to suggesting linkages

:> verbal termns. Linkages to mathematical terms need to be considered
s well, since they, too, have a potentially important part in future

heories of communication .

1. 3 Mathematical Aspects of Constitutive Definition

 

While theory in the discipline of human communication has not
>ften been expressed in mathematical terms, such terms are important
and useful, not only for themselves, but also for the perspective they
ire capable of providing if they can be employed in theoretical formu—
Lations. The second major step in approaching a constitutive defini-
:ion of the concept of process, then, will be to consider linkages
>etween this concept and a set of mathematical terms which may have a
place in future theories of communication. In keeping with the pre-
vious section, the attempt in considering such linkages will be to
describe how "process" and its associated terms are generally formu-
lated and treated in a mathematical terminology. This attempt to
establish links to mathematical terms should yield a number of useful
insights into the constitutive definition of the concept of process.

Again, as in the previous section, the linkages considered below
are a subset of those which might be considered—-in this case, a subset
which has two inputs. Clearly, one group of terms which needs to be

considered in mathematical form is the set chosen in the previous

99
section on the basis of the potential utility of the terms as tools

in communication theory building. These terms will be considered

here, together with a small number of additional terms chosen by the

same criterion. It is important to note that a comprehensive dis-

cussion of many of the terms chosen would require rather extensive

mathematical derivation. Because such derivation is not in keeping

with the basic purpose of this section, these particular terms will

be discussed briefly, but with special attention to additional sources

of information.
The discussion of the mathematical aspects of constitutive

definition can be divided into three parts. First, an examination of

the mathematical definition of the concept of process, as well as of
the mathematical form of a number of old and new descriptive terms.

Second, a brief discussion of the mathematical aspects of the ties

>etween "process" and its related concepts. And third, a consideration

>f certain forms for expressing a process and of the means for finding

functional representation. Becatse of the large number of terms and

efinitions introduced in this section, Table u, page 100, has been
ncluded both as an outline and as a glossary useful for reference
lI‘pOSBS. In addition to the above parts of the discussion, it will
appropriate to summarize briefly the discussion of the verbal and
thematical aspects of constitutive definition. Accordingly, a

rrrth part will be included in this section, examining in both verbal

i mathematical terms the necessary and sufficient conditions for the

2 of the term "process."

100

Table k. Outline and Glossary of Mathematical Asgcts of Constitutive Definiticna

 

3.3.1 The Mathematical Definition of Process

 

Process..................A function of the general form v = f(t). (A variable as a
single-valued function of time.)
tTme a continuum of one dimension which is divisible into
equal units, those units being ordered midirectionally
mder the assumption that the entropy of the universe is
always increasing.

Reversible.......A process whose equation is the same no matter whether the
temporal variable is positive or negative.

Irreversible. . . . .A process whose equation differs depending on whether the
temporal variable is positive or negative.

3A variable: any property of a unit or relationship within
a system which can be recognized by an observer, which can
potentially change over time, and whose change can poten—
tially be measured by specific operations. The variable
may be a single variable or a vector of several variables.

f........ .. .. ...... ..A function: a correspondence between two variables such that

' a value of one depends on a value of the other, as
determined by some rule of relation. A function encompas-
ses both the presence and absence of change, and in this
context is considered to apply only within a bounded
interval, tx ; t _<_ ty.

Static...........v = k, where k is a constant.

Mistatic......v = c, where 0 < I c-k I < e, and k is a constant, with e being
some small deviation around k within which v can be
considered essentially constant.

gzclic...........v = f(t) where v has the same value at two points in time,
say ta and tb, and a < b.

Additional Terms:

 

Discrete. . . . . . . . . . . . .A process in which the variable t appears as an integral
multiple of a least quantity.

Continuots... ..... ...A process in which the variable t may take on any value in
its range.

whostatic. .. . . . . . .Processes, often including feedback "mechanisms," in which
the value of a variable is maintained within a range of
stability, 0 < I v-m I < r, where m is an equilibrium value
and r is the limit of deviations around m within which the
system remains stable. Processes may be reversible or
irreversible, and can be convergent.

Pbgmogenetic. . . . .. . .Processes in which the equilibrium value of a variable, and
its range of stability, are altered from previous values.
Processes may be divergent, and can be either reversible
or irreversible.

Stochastic. . . . . . . . . . .A time dependent probability process. Actually a Specific
class of functions which makes certain assumptions about
the events described.

3.3.2 Process in Relation to other Terms

 

Structggm. ..........The static arrangement of a system's parts at a moment in
three dimensional space. Structurem is the evaluation,
at a particular point in time, of the function (x,v,z)
= f(t) where x,_v,z are positions on the axes of a three
dimensional physical space.

Structurej . . . . . . .A many-valued relation, a complex pattern, an above chance
distribution in a multi-dimensicnal space.

chticn.............A form or subset of process in which the changes are

reversible (as defined above). Function encompasses the
morphostatic or regulatory processes of a system, to the
extent that they involve reversible change. Note this
use of the term is distinct from that used in defining
the symbol "f" above.

Histog..............A form or subset of process in which the changes are
irreversible (as defined above). History encompasses the
morphogenetic processes of a system, to the extent that
they involve irreversible change.

 

'I'he sources of these definitions are given in the discussion.

 

P". "_ 7; L“ _..r

 

101
3.3.1 The Mathematical Definition of Process

Again, the concepts of'process and of Change over time have
been treated as almost completely interChangeable in this paper. As
a result, the examination of the mathematical definition of the term
"process" will involve a consideration of the terms "Change over time,"
as well. The discussion will link both of these concepts to a.wide
range of'mathematical terms, and will result in a "translation" of a
number'of'the terms introduced in the previous section as descriptors
of a.process.

In order to formulate the mathematical definition of the
concept of process, it is necessary to refer to the representation of
a process as a lin§_on a graph in whidh time has been plotted as the
abscissa and in which the value of a variable has been plotted as the
ordinate. The mathematical representation of a "process," quite
simply, is the equation describing this line. The general fbrm.of
suCh an equation, whiCh would correspond to using the phrase "change
over time" as a general definition of "process," is:

v = f(t) (l)
where V'iS the variable under consideration, expressed as some.fUnction
of t, or time.5 The fUnction may take any form consonant with the
notation of calculus and analytic geometry (Thomas, 1953, pp. 1-20).

As does the line in the graphic representation, the fUnction serves

5Quite clearly, "function" is not used here as it was in
section 3.2.2. See the discussion below of the mathematical
definition.

102
to indicate successive values of v fbr successive values of t.
The mathematical definition of'"process" or of "Change over time"
is therefbre quite brief in its general outline, and.requires expan-
sion to make the value of this particular’representation more clear.

One important comment needs to be made before examining the
individual terms in this mathematical definition of the concept of
process. Specifically, it is important to distinguish between the
adoption of a mathematical termﬁrtﬂtpyg as above, and the use of a
mathematical model of some event or events. That is, the present
adoption of a mathematical representation of a process is, fUnda—
mentally, a translation of the verbal terms used previously into
mathematical terms. Such a step is no different fromlthe trans-
lation of verbal terms into graphic "terms," as has been done
already, and in particular, makes ng_assumptions regarding the nature
of any events whiCh may come under study.

The use of a mathematical model, on the other'hand, gg§§_make
the assumption that the specific function WhiCh is employed'has the
same Characteristics as the event under study. In this sense,
mathematical modeling is no different fromlany other~modeling, i.e.,
the assumptions of the model must be clear and must conformlto the
events. The function noted in equation (1), in particular, is not a
mathematical model. It is, instead, a general function, or perhaps
more clearly, an indicator of a general form which functions may take
using this particular*mathematical notation.

The above choice of the notation of calculus and analytic

_ geometry is, therefore, a Choice of a particular mathematical

103
"language" or terminology, as well as of a set of conventions for
forming "sentences" or expressions . The content of those sentences
or expressions is pg: restricted by the presence of assumptions,
although as in any Lee of a language system, certain contents may be
more conveniently expressed in a different language. The particular
choice of notation is based on the broad generality and especially
the utility of calculus and analytic geometry for the present concerns
with human communication.

The mathematical terminology. - As was the case for the verbal

 

definition of "process," one of the key terms in the mathematical
definition is the variable t, or in this case, time. Ordinarily, of
course, the mathematical notation would not restrict the variable t
to representing time, much less a particular "kind" of time. However,
given the concern in this paper with the concept of process , particu—
larly as that concept applies to the concrete systems in which human
communication takes place, the variable "t" will be defined here as
"astronomical time," exactly as this concept was defined earlier
(page 8h).6 The important consequence of this definition of "t," or
time, is that equation (1) expresses v as a "single-valued" function
of t (Thomas, 1953, p. 11+). That is, for each t, there will be asso-
ciated only a single (though not necessarily unique) value of v.
Aside from the restriction to time, there are no other restric-

tions on the variable t, as, for example, on its sign. Specifically,

 

6See J. G. Miller (1965a, p. 203) for a discussion of the
concept of "variable" as it applies to a "concrete" system.

10%
the presence of a plus or a minus sign in association with "t" does
not carry the meaning of time running forward or backward. The sign
may be particularly important in another sense, however, for "If the
equation describing a process is the same no matter whether the
temporal variable is positive or negative, it is a reversible process;
otherwise it is irreversible (J. G. Miller, 1965a, p. 209)." Thus,
the equation v = t2 represents a reversible process, since v = (+t)2
= (-t)2, while the equation v = t3 is an irreversible process, since
v = (+t)3 x (-t)3.

Along with the variable t in equation (1) , the variable v is
also of importance. Unlike the case with t, however, no restrictions
will be placed on the nature (or name) of this variable. Defined in
a manner consistent with the definition for t, above, "v" refers to
"Any property of a unit or relationship within a system which can be
recognized by an observer . . . , which can potentially change over
time, and whose change can potentially be measured by specific opera-
tions . . . (J. G. Miller, 1965a, p. 203)." As a consequence of this
definition, the variable is completely general, and may refer, for
example , to some matter—energy or informational property of a system,
or to some relationship between such properties. As a result, the
concept of process in its mathematical form, as in the verbal form, is
completely general, even though the concern here will be primarily

with variables related t_o_ human commmication.

 

As it is expressed in equation (1), variable v represents only
a single variable as a function time, rather than the more general

multiple variable case. To extend the mathematical definition of

105
process to this more general case, the variable must be treated as a
"vector" of several variables, each identified by a numerical sub-

script. Bquaticn (1) would thus be rewritten as:

= F(t) (2)

 

 

where the capital F represents a function which is understood to
contain vector components . The vector [V1,V2 , . . . ,vn] is likewise
normally represented by a capital V, unless the particular mathema-
tical operatiors require the enumeration of the individual variables .
Equation (2) might therefore be rewritten as V = F(t) . Note that
each of the variables vl,v2,.. ., vn is, nevertheless, a function of
the same variable t--a situation which would be represented, on a
graph, by 1,2, or up to n lines, all extending through the same time
span on the abcissa, but each representing a different variable plotted
on the ordinate. Becatse of its simplicity, the single variable case
will generally be used for explanatory purposes, given that the prin-
ciples for the single variable situation are directly extendable to
the multiple variable situation .

In addition to the variables t and v, the function itself is
of some importance in the mathematical definition of the concept of
process, since it is the function which describes the nature of the
change in the variable v which occurs over time. The term "function"
is defined in its mathematical sense as "A correspondence between two

variables such that a value of one depends on a value of the other,

106
as determined.by some rule of relation . . . (J. G. Miller, 1965a,
p. 202)." The nature of the function or rule is of course not re—
stricted, except that the functions considered here will be those
which fellow the notation of calculus and analytic geometry (see
discussion of fUnctions in Bellman, 1965).

As was the case in the discussion of the verbal definition of
"process," a mathematical function also applies in the broadest
possible sense, encompassing both the presence and absence of change
or difference in the variable v. That is, given the single—valued
fUnction, v = f(t), there will generally be some change or difference
between the values of V'at two points in time. In the special case
where the two values of v are identical, however, the fUnction is still
considered single-valued, but the two values are termed "non-unique,"
and of course exhibit a "change" or difference" of zero.

The concept of a fUnction continues to apply even when the
variable t is "absent" fromlthe equation v = f(t). That is, in the
special case of a static process, the representation of a process on
a graph.reduces to a line parallel to the time axis, normally expressed
mathematically as v = k, where k is some constant. The equation v = k
is, nevertheless, still considered to be a valid form of the equation
v = f(t). .A quasistatic process would generally be represented by a
similar equation, v = c, where 0 < I c-k I < e, the e representing
some small deviation around k, within which v can be considered
essentially constant. A cyclic process would be one in which some

value of'v occurred at two points in time, say ta and tb, where a < b.

107

Again, as in the discussion of the verbal definition, a mathe—
matical fUnction in a general form such as v = f(t) is not considered
to be bounded in time, but to apply over'all time (-0° < t < +w). In
practice, however, many functions are bounded, even though the bounds
may not be explicit, and many mathematical operations require the con—
sideration of bounds. For these reasons, and because the segmentation
or bounding of occurrences in time was seen earlier as basic to science,
all mathematical expressions of a process herein will refer to a process
as it occurs within some bounded interval. Such a bounded interval or
event is indicated by identifying the limits on the range of values t
may hold, as for example in tx < t < ty, or in tX é=t é.ty° The first
case is normally termed an "open" interval, while the second is a
"closed" interval (Thomas, 1953, p. 1H); the length of the interval, T,
is clearly T = ty — tx. The restriction of the mathematical definition
of "process" to bounded intervals is perhaps at variance with the gen-
eral treatment of functions such as v = f(t), but is seen both as use-
ful, and as consistent with operations in science.

Additional terms. - In addition to this consideration of the

 

terms which comprise the mathematical definition of process and which
are closely tied to it, it will be helpful to consider the mathematical
femmlof several additional descriptive terms, both new and old. Two
very important new terms, "discrete" and "continuous," are descriptive
of two kinds or types of process, or more specifically, of.the kinds
or'types of values which the variable t may take on. A "discrete"
process can be defined as "A process in which the variables appear as
integral multiples of a least quantity," While a "continuous" process

is considered to be "A process in which the variables may take on any

108

values within their ranges (Bryne, 1967, p. 5)." When applied to the
variable t in the equation, v = f(t), these definitions indicate that
in a discrete process, t will take on one of a set of values, t1, t2,
..., tk, tk+1,..., tn, where tk+l = tk + At, and At is an arbitrary,
but constant interval of time . In a continuous process, on the other
hand, t may take on any value within its range, or in this case, with—
in the interval tl ; t :3 tn.

Whether the mathematical representation of a process is termed
"discrete" or "continuous" depends not on any inherent characteristic
of the occurrence under study, but on the manner in which the vari-
able v is measured. That is, measurement of v at a number of separate,
but regularly Spaced points in time requires representation in a dis—
crete form, whereas measurement at all points in time during the
interval of concern will permit representation in a continuous form
(as well as in a discrete form). Accordingly, as is often the case
in studies of human communication, the occurrence under study may be ,
in itself, a continuous process. However, because of the measurement
technique involved, the eventual mathematical representation of that
process may be in a discrete form. Functions of the general form,
v = f (t), can be treated as either discrete or continuous, although
the representations and techniques used in each case will vary some-
what (see discussion below in section 3.3.3).

Several other descriptive terms introduced in the discussion
of verbal aspects of constitutive definition also can be interpreted
in mathematical form. Morphostatic processes, for instance, were

described earlier as ones in which the value of a variable fluctuated

109
around some equilibrium value, but within a range of stability. Such
a range of stability might be indicated mathematically as, 0 < I v—m I
< r, where v is the variable value , m is the equilibrium value, and r
is the limit of deviations around m within which the system remains
stable. While the actual forms are too complex to treat here, a func-
tion representing a morphostatic process would generally incorporate
some type of feedback "mechanism" to maintain the stability of vari-
able v. This mechanism would "feed back" part of the system's output
to its input, and thereby act to counter any deviation of v from m,
as long as the deviation remained less than r. Beyond r, the feedback
mechanism might fail to operate or another mechanism might come into
play, depending on the particular function (see discussions of the
mathematical treatment of feedback in Koenig, 1967, pp. 366-“02, and
in Milsum, 1966, Ch. 2, passim, and Ch. 11). The more specific type
of morphostatic process which was termed earlier, "convergent," like-
wise has a mathematical interpretation, but one which is too lengthy
to develop here (see Thomas, 1953, Ch. 16; Spiegel, 1958, Ch. 5).

Like morphostatic processes, morphogenetic processes also can
be interpreted in mathematical form, although the interpretations are
possibly more complex and varied, and are in some respects less well
developed (Maruyama, 1963). More specifically, the concept of a range
of stability about an equilibrium value is not of particular importance
in the fonmlation of morphogenetic processes , except in the sense that
such processes may come into operation if a variable moves outside its
range. For example, a given function might incorporate both a feed-

back mechanism to maintain a given equilibrium, as well as an

110
additional mechanism to establiSh a new equilibrium, mhewv and a new
range, rhewv should the variable v ever move outside its original
range. The latter mechanism, if brought into use, would qualify as
a morphogenetic process.

Because of the complexity and variation in.representations of
morphogenetic processes, certain sub-classes of them have been treated
separately. Perhaps most notable in this respect is the work by
Thompson (19u2; see also Bertalanffy, 1968, pp. 60-63, 171—18u), con—
cerned in part with the mathematical treatment of growth or development.
Learning (WhiCh is not fbrgotten) is likewise a sUb—class of morpho-
genetic processes Which has been treated.mathematically (see Luce,
2:21;, 1963; v. 1, Pt. 2; v. 2, pp. 1—126, 206-265; v. 3, pp. 101-
203). Those morphogenetic processes which.ame specifically termed
"divergent" also have a definite mathematical interpretation, XiETéT
yi§_convergent processes, but again, development of the interpretation
is too lengthy to be usefully included here (see Thomas, 1953, Ch. 16;
Spiegel, 1958, Ch. 5).

This discussion of the mathematical form of these several
additional descriptive terms could be expanded to include a wider
range of terms whiCh describe processes by reference to their specific
mathematical characteristics. Among the most important of these terms
would be "stodhastic," a stodhastic process being one whiCh.may be
". . . intuitively regarded as a time-dependent probability process
(famaro, 1968, p. 2H6)." In this case and in others, however, an
extensive discussion.would be required in order to move mMCh beyond

an intuitive mderstanding (see the excellent discussions of stochastic

lll
processes by Fararo, 1968; Parzen, 1962; and Snell, 1965). Note, too,
that stochastic processes are a class of specific functions which d_o_
make assumptions about the characteristics of the events under study
(see page 102).

Again, it is not the purpose of this section to reproduce the
derivations of the various mathematical terms used to define or de-
scribe a process . Rather, the attempt is to indicate how the concept
of process and its associated terms are generally formulated and
treated in a mathematical terminology. Note , too, that the above
discussion of the mathematical definition of the concept of process
and of the mathematical form of sevéal different descriptive terms
is only a first step in considering the mathematical aspects of
constitutive definition. The second step of considering the mathema-
tical aspects of the ties between process and its related concepts is

also important .

3.3.2 Process in Relation to other Terms

In the earlier discussion of verbal aspects of constitutive
definition, the concept of process was linked not only to the descrip-
tive terms considered in their mathematical form above, but also to
three other closely related terms of importance in system theory and
analysis. These three terms, "structureln," "function," and "history"
need to be considered here, too, both for their own importance, and
because their ties to the concept of process reveal some interesting
points about the mathematical expression of processes in systems.

The first of the three terms, "structurern" was defined earlier

as "the static arrangement of a system's parts at a moment in three

112

dimensional space (J. G. Miller, 1965a, p. 211)." In a concrete
system, therefore, structure“ is the arrangement of matter—energy,
apart from any considerations of information or structurei. If one
wished to consider the structurem of a given system, he would be
forced to describe the system by a function of time such as:

(x,y,a) = f(t) (3)
where x, y, and a are positions on the axes of a three dimensional
physical space . Such a function represents the process normally
termed "movement," in this case, movement in all three dimensions.
Given such a function, the structurem of the system would be described
mathematically by the values of the variables at a specific point in
time, say ta. Such valLes would be fixed or constant at a particular
point in time, but not necessarily over an interval of time, hence
the confining of structurem to particular instants in time.7

In basic outline, then, the concept of structurem has a fairly
straightforward relationship to the concept of process , even though
the study of mechanics, which treats the "statics" and "dynamics" of
matter-energy arrangements , can be extremely complex. It is important
to note that the distinction between structurem and structurei is
often somewhat blurred. Structurei, as noted in section 3.2.2, is
always embodied in some form of matter-energy _ma_r_]_<_e_r_, but clearly for

many purposes the matter-energy aspects of the marker are of little

 

7See Koenig (1967, Ch. 14) for a different, though compatable
treatment of the "structure" of a system. Koenig's treatment focusses
primarily on structurei, even though it derives such information from
"physical systems."

113

or no concern. Indeed, many discussions of the various aspects of
system organization are primarily considerations of structure;L , almost
totally without the observer's consideration of any matter-energy
embodiment (Rappoport, 1966, pp. 8-9; Ashby, 1968). Again, structurei
was defined "mathematically" in section 1.2 "as a many valued relation,
as a complex pattern, as an above-chance distribution in a multi-
dimensional space (Krippendorff, 1969a, p. 107)," but a more detailed
mathematical development is clearly not within the scope or purposes
of this paper (see Shannon, l963, pp. 3—91).

"Function," the second of the three terms to be considered,
was discussed earlier as a form of process, but one which carried the
additional specificity of ". . . the transient and reversible changes,
often repetitive, that constitute 'behaving' . . . (Gerard, 1968, p.
52)." Again, the distinguishing aspect of the concept of function is

this association with reversible change (see page 101+). "Function"

 

is therefore a subset of the more general term "process." In addition,
to the extent that reversible change is involved "function" encompasses
the morphostatic processes which regulate and control system operations .
The importance in a system of these morphostatic aspects of function
has meant that their formulation in mathematical terms has received
considerable attention, not only from the specific point of view of
concepts such as feedback, homeostasis, purpose, and equifinality (see
Bertalanffy, 1968, pp. 75-80, 12n-12u), but also from the general
point of view of the concept of control (see Milsum, 1966).

The third of the three terms, "histogy," was also considered

earlier as one form of process. "History" is more specific than

11H
"process," and forms a subset of it, in that it refers to the ". . .
enduring and irreversible changes , often progressive, that constitute
'becoming' or development (Gerard, 1968, p. 52)." The key aspect of
the concept of history which distinguishes it from function is the

association with irreversible change (see page 10%). As a result, to

 

the extent that irreversible change is present, "history" encompasses
the morphogenetic processes of initiation, growth, learning, decay,
etc. The particular importance of growth and learning as aspects of
the history of a system has singled them out, in particular, for form-
ulation in mathematical terms (see Bertalanffy, 1968, p. 60-63, 171-
18u; Luce, it. al_., 1963, passim; and Thompson, 191+2). A detailed
discussion of the mathematical formulation of these two aspects of
the history of a system is , however, beyond the scope of this paper
--a comment which applies to the various aspects of function, as well.
Briefly, again, history, in its association with irreversible
change; function, in its association with reversible change; and struc-
turem, in its association with single instants in time, are all tied to
the concept of process. That is, each of the concepts, when considered
in its mathematical form, involves a function of the general form,
v = f(t). The concepts are distinguishable from one another by the
results of a particular operation on t; in particular, isolating a
single value of t to determine structure“, or substituting positive
and negative values of t to separate reversible from irreversible, and
hence function from history. The concepts of structure,“ function,
and history have other . important characteristics , in addition, but

operations such as these on the variable t reveal much about the

115
complexity of the mathematical expression of the concept of process,
despite its seemingly very simple form, v = f(t).

This discussion of the ties between the concepts of process
and the related concepts of structurem, function, and history, is one
part of the attempt to indicate how the concept of process and its
associated terms are treated mathematically. The earlier discussion
both of the mathematical definition of "process" and of the treatment
of the different descriptive terms is part of this overall attempt,
as well. There remains, in addition, a third part to the task of
considering the mathematical aspects of constitutive definition. In
particular, it will be helpful to consider two broad aspects of the

mathematical expression of the concept of process.

3.3.3 Process and its Mathematical Expression

While the verbal expression of the concept of process is rela—
tively straightforward, its mathematical expression involves certain
"difficulties" which should be discussed briefly. The two aspects of
expression to be considered below both concern the "forms" of expres-
sion of a process. The first aspect deals with the expression of time
as discrete or continuous , and with the relationship between the two,
while the second aspect concerns the general problem of finding a
mathematical equation or function which will express a given process.

Discrete and continuous expressions. - As noted in the discus-

 

sion of the mathematical definition of "process," the variable t in
the equation, v = f(t), may be expressed in either a discrete or a
continuous form. The form chosen has an effect both on the nature of

the functional representation, and on the techniques used to manipulate

116
the function. If time is treated as discrete, the variable t may take
on one of a finite set of values, any two of which are related by the
expression tk+l = tk + At, where At is the interval between successive
time points . If one were to evaluate the function to obtain values of
the variable v in this discrete situation, he would make an evaluation
only at each individual time point, a situation which might be, general-

ly expressed as follows:

v = f(t )

"t1 1

V132 = f(tz) OI” Vt2 = f(t]. + At)

vtk = f(tk)

vtk+l = f(tk+1) or vtk+l = f(t}< + At)

th = f(tn) (14’)

The conseqLence of treating time as discrete , therefore , is that the
variable v also takes on discrete values.

Very freqLently in dealing with a process, however, one is
faced not with the problem of evaluating a function at different
points in time, as above, but with the problem of predicting the value
or "state" of the variable v at the next point in time, given the
value or state at the current point (see Krippendorff, 1969b, pp. 23-
28, and Carroll, 1968, pp. 35-37). Such prediction involves finding
a new function which will express Vtk+l = f(tk+1) = f(tk + At) as a
function of vtk = f(tk), or, in a considerably simplified, but equi—
valent notation, v(tk+1) = v(tk+At) as a function of v(tk). Such a

new function may be represented in general form by the function g in

117

the equation:

v(tk + At) = g [mp] <5)
WhiCh, in verbal terms, would be a prediction of the state at the next
point in time, given the present state (and assuming suCh a.prediction
is viable).8 An equation such as (5) is termed a "difference" equa—
tion, and is treated.or'manipulated using a special class of'mathe-
matical techniques appropriate to sudh discrete time representations
("difference" equations have no necessary relationShip to "difference"
designs). The scope and application of these techniques cannot be
discussed here, but are treated in introductory form by Bellman (1965,
pp. u91—u96) and Goldberg (1958), and in more detailed fOrm, especially
as applied to systems, by Cuénod (1969) and Kbenig (1967, pp. 179-183).

On the other'hand, if time is treated as continuous in dealing

 

with a given function, the representation of that fUnction and the
techniques used to manipulate it will be somewhat different. In this
case, the variable t may take on any value in its range, t1 égt égtn,
with no restriction on the interval between values. If one were to
evaluate the function, v = f(t), for the value of v in this continuous
situation, he would face no restriction on the time point or points he
selected. The consequence of treating time as continuous, then, would

be that the variable v might well also exhibit continuous variation.

 

8Note the fOllowing very important point. Equation 5 is,
effectively, a model or s ecific class of functions. As sudh, it
makes assumptions about tEe event under study, as described. These
assumptions are not the only possible assumptions, and specifically
are ngt_inherent—iﬁ the notational systemn ﬂany_othertmodels can
be fOrmed having different assumptions.

118
The treatment of time as continuous also brings certain changes
in representation when one is concerned not with the problem of evalu—
ation, but with the problem of "prediction." In this case, the nature
of the prediction itself is altered slightly as a.resuﬂt of the change
in the treatment of time. That is, the situation of predicting the
value of a fUture state given the current state becomes, in the con-
tinuous time case, a situation of determining the rate of change in
the value of a state given the current state. The latterrdetermination
involves a function WhiCh will express the rate of change in v, or v'
= f'(t), as a fUnction of v = f(t), i.e., in the simplier notation,
v'(t) as a function of v(t). Such a function is represented in general
form by the function Dtin the equation:
v'(t) = Dt[v(t)]
or, with a.more common notation added:

v'(t) = dv = D [v(t)]. (6)
a? t

In verbal terms, the function Dt is the "derivative" of v with respect
to t, and expresses the rate of change in v at any given instant in
time.9 An eqLation such as (6) is termed a "differential" equation,
and is treated or manipulated using the techniques of differential and
integral calculus . The scope and application of such techniques is

much too broad to consider here, and will be left to introductory texts

 

gNote, again, that equation 6 is, in effect, a model or
Specific class of functions, and assumes that the event under study
has characteristics Which allow "rates" to be determined in this
manner. Other fUnctions could be fbrmed.to fit other characteristics
(see note 8, and page 102).

119
such as Thomas (1953). Additional discussions of interest can be
found in Bellman (1965, pp. 1496-502), and with particular application
to systems, in Koenig (1967), and Milsum (1966).

Tth, the variable t in functions of the general form, v = f(t),
can be expressed as either discrete or continuous. The two expres-
sions require separate techniques for handling the different represen-
tations, although in large part these techniques parallel one another.
Such a parallel is not particularly surprising when one considers

that the two expressions 9: time actually merge in the basic defini-

 

tion and operation of differential calculus. It will be helpful to
examine this relationship between the two expressions.

Very briefly, in the representation of a process on a graph,
where time is treated as discrete, the slope of a line connecting any
two adjacent points can be found by dividing the change in v between

the two points by the corresponding change in t, i.e:

 

tk+1 - tk " ‘5? (7)
However, Av can also be expressed in terms of the function

v = f(t), so that Av = f(tk+1) - f(tk), or, alternatively, Av = f(tk
+ At) - f(tk). Substituting this expression into equation (7), and
dropping the k at the same time, would result in an expression for the

slope entirely in terms of the function, t, and At:

slope : f(t + At) - f(t) . (8)
At

 

If the size of the At were very small, the discrete time repre-
sentation of a process would closely resemble the continuous time

representation, and the slope expressed by equation (8) would apply

120
over only a very small interval of time. In the limiting situation,
wTen the size of At approached zero (At + 0), the discrete representa-
tion would merge with the continuous representation , and the slope
would become the slope, or rate, at a single point in time. Such a
limiting process, in fact, defines the derivative and is usually

expressed as follows:

v'(t) = dv = 1m f(t + At) - f(t) .
at At + 0 At (9)

Equation (9) is the basic definition or operation in differential
calculus (Thomas, 1953, pp. 23-25), and is ﬁne tie between the ex-
pression of time in discrete and in continuous form.

The manner in which time is expressed is clearly an important
aspect of any mathematical expression of a process, but it is at
least equalled in importance by another aspect: the form of the
equational representation itself. The task of finding an appropriate

equation or function to express a given process is often a difficult

one, and requires comment, as well.

Finding an expression. - There is, first of all, no single,

 

well-defined procedure for generating an equation to describe a given
line in a graphic representation of a process. Accordingly, an
investigator must often depend primarily on his knowledge of the par-
ticular process as a guide. There are, however, certain general
approaches to generating equations or representations of processes ,

at least one of which has a high probability of success, though
occasionally at some expense of effort. In discussing these approaches,

it will be convenient to refer at times to the line in a graphic

121

representation of a process not as a "line," but as a "signal," a
term frequently used to refer to a variable which changes over time.

The first general approach perhaps depends most heavily on
the investigator's knowledge of the process he is studying. Here
the attempt is to find a representation for the given process in
the form of an alEbraic expression, a very simple example of which
is finding the equation of a straight line. More generally, it is
possible to find a polynomial of degree n which will pass through any
n points on a graph, and to fit this polynomial as closely as possible
to the line or signal by the method of least squares (see Gabcr, 19%,
p. H30).10 In many cases, however, a somewhat simplier function may
adequately represent a given signal, e.g. , a step function, exponential
function, logarithmic function, sianoidal function, etc. In addition,
certain signals such as rectangular and triangular waves have fairly
well known representations, and it is sometimes feasible to represent
these signals or others by an average or root-mean—square value.
Representations such as these are discussed in more detail in algebra,
analytic geometry, and calculus texts, although somewhat shorter
discussions of a number of these can also be found in Koenig (1967,
pp. 20—25, 30-33), and in Milsum.(l966, Ch. 3).

In numerous cases, however, it is not possible to find an
algebraic expression like those above which will adequately represent

the given line or signal. In these cases, a second general approach

 

10In their general forms, a straight line would be expressed
at + b, while a polynomial of degree 3 would be v = at3 + bt2

asv=
+ct+d.

122
will almost always succeed in producing a viable representation,
though sometimes with an expense of effort. This approach is to
find a "Fourier series" representation for the signal; more specif-
ically ". . . any function [or signal] defined over a finite interval
of time t1 < t < t2 can be represented by means of an infinite series

of sinusoidal functions of time known as the Fourier series (Koenig,

 

1967, p. 25)." The degree of accuracy in representing a particular
signal is determined by the number of terms in the series which are
considered. Obviously an infinite number of terms is not needed in
any practical application, and generally a fairly small number is
sufficient. The scope and application of Fourier series representa-
tions has been extensively studied, as in Bracewell (1965), and many
others. Of particular interest, here, are the briefer discussions
of this representation both in terms of effort reducing approaches
and the use of the discrete form (Bergland, 1969; Hamming, 1962, Ch.
6), and in the context of system analysis (Koenig, 1967, pp. 25-30)
and of telecommunication (Clnerry, 1957, pp. 128-1143).

The third general approach to finding a mathematical expression
for a given process is in many ways related to the use of Fourier '
series representations . This approach would have particular applica-

tion in cases where the signal to be represented had statistical or

 

random characteristics, a situation not uncommon in dealing with

living, and especially human systems (see Milsum, 1966, Ch. 1”). In
many cases, statistical signals could be treated using Fourier series
techniques, but frequently the Special techniques and functions which

have been devised for representing such signals may be more useful

123
(see Milsum, 1966, pp. 370-390).

These three general approaches to generating eqtations or
functions to represent given signals or lines constitute the primary
means by which an investigator might find a Specific equational form
in which to express a process he is studying. Again, the task is often
difficult, and lacks a unified, well-defined procedure. Note, too,
that while each of the three approaches may produce an expression for
a given process, the models which result will make certain assumptions
about the characteristics of the events under study . As always , the
scholar must identify ttese assumptions and be sure they are justified
in the particular situation (see page 102).

This discussion of the general problem finding a mathematical
expression for a process, together with the discussion of the expres-
sion of time as either discrete or continuous , constitute the examina-
tion of certain aspects of the mathematical expression of a process.
This examination , together with the consideration of the mathematical
treatment of the concepts of structurem, function, and history, as
well as the study of the mathematical definition of process, comprise
the discussion of the mathematical aspects of constitutive definition.
There are, of course, other aspects and terms which mignt have been
considered; however, in the context of use as potential tools in
theory building in commmication , the various aspects and terms which
have been chosen appear to be the most useful.

Note , too, that were this paper not concerned solely with the
concept of process, there would be a number of other terms which might

be considered. That is , if the paper were concerned with the concept

121+

of system, a great deal more might be added on the mathematical (and
verbal) formulation and treatment of systems . Such discussions may
be found in Bertalanffy (1965, Chs. 3, 5, 6), Hall (1968, pp. 81-92),
Koenig (1967, and more briefly, 1965, p. IJ.1-l+5), and Milsum (1966).
Despite the possible utility of such information for the study of
human communication, it cannot be considered here, since to do so
would be to change radically the level of discussion, and to obfiscate
the goal of seeking the mears of explicating the concept of process.

In review, this section has considered a number of linkages
between the concept of process and a set of mathematical terms which
may have a place in future theories of commmmication. The discussion
of linkages has encompassed not only an examination of the mathe-
matical definition of the concept of process and the mathematical
treatment of certain descriptive terms, but also a discussion of the
mathematical aspects of the ties between "process" and three related
concepts, as well as a consideration of the mathematical expression
of processes. In short, the discussion has covered a number of
mathematical aspects of constitutive definition. It is appropriate
therefore to summarize briefly the discussion of both the verbal and
mathematical aspects of constitutive definition, in that these aspects
lead to a statement of the necessary and sufficient conditions for

the use of the term "process."

3.3.14 Necessary and Sufficient Conditions for Use of the Term "Process"
It has been the burden of sections 3.2 and 3.3 to examine a

number of links between the concept of process and various verbal and

125
mathematical terms which might be utilized in future theories of
communication. Linkages to _ve;_rb_a_l_ terms have been considered because
of the central place of such terms in theory in the study of human

communication. Linkages to mathematical terms have been included

 

both because of the potential importance which these terms have, and
because of the perspective such terms can offer if they can be employed
in constructing theory . The mathematical formulation and treatment

of "process" provides, in addition, a number of useful insights into
the more Lsual verbal formulation and treatment of the concept .

The discussion of linkages between the concept and both verbal
and mathematical terms lnas been an attempt , again, to suggest a set
of potentially useful tools for constructing constitutive definitions
of the concept of process. There has been no attempt to build a
unique constitutive definition, though, because such a task can be
carried out only in the context of a specific theoretical framework.
Rather, the attempt here has been to suggest and discuss various
approaches to a constitutive definition, i.e. , to suggest linkages
between the concept of process and a set of verbal and mathematical
terms which might be useful in constructing theory. Clearly, other
approaches , or linkages to other terms , could have been included if
Space had not dictated a choice . The information and terms which have
been chosen are those which are particularly consistent, general and
above all useful, when judged as potential tools in theory building
related to human communication.

Because the attempt here has been one of suggesting potential

linkages or tools for use in forming constitutive definitions , rather

126
than one of constructing specific definitions, it is not appropriate
to judge these linkages by the normal criteria for a viable, fbrmal
constitutive definition. Such criteria are generally those of
clarity, Simplicity, consistency, fertility, and above all, utility
of linkages (Berlo, 1967, pp. 3—5). The linkages suggested here are
seen,though, as particularly capable of meeting such criteria if they
are used in a specific theoretical framework to constitutively define
the concept of process.

The approaches to a constitutive definition of the concept of
process which have been suggested in this Chapter have two broad con—
sequences. First of all, they lead to a statement of the necessary
and sufficient conditions for the use or application of the term
"process." Secondly, they carry a number of important implications
fOr the form of theory Which deals with communication events. This
second consequence will be the topic of the next section, and the
first‘will serve to summarize the consideration of the verbal and
mathematical aspects of constitutive definition.

Specifically, we shall adopt Kaplan's approach to the neces-
sary and sufficient conditions fOr the use of a term, as expressed in
his statement that "The definition formulates the conditions which
are both necessary and sufficient for the applicability of the term1
defined (196%, p. 72)." Hence, in verbal terms, the necessary and
sufficient conditions for the use of the term "process" are, that it
apply to "All change over time of'matter-energy or information, with—
in a bounded interval of time." Likewise, in mathematical terms,

the necessary and sufficient conditions fOr the use of the term

127
"process" are, that it apply to "A function of the, general form '

v = f(t), defined over an interval of time, t1 =<__ t ; tn."

3.1+ Implications for the Form of Theory

 

To this point, Chapter 3 has been concerned with a number of
suggested approaches to constitutive definition of the concept of
process. More specifically, it has considered both the verbal and
the mathematical aspects of such definition, and in so doing has
suggested a number of verbal and mathematical terms which have poten-
tial utility as tools in theory building in communication. As any
craftsman is aware , the nature of the tools one works with determines
in part what can be built. So too with theoretical tools: the nature
of the terms a scholar works with defines to some degree the form or
type of theory which he can construct. It is therefore appropriate
to examine the implications which the potentially useful terms noted
above have for the form of theory which deals with communication events .

It Should be noted that an examination of the forms which theory
may take is a formidable task in any discipline, especially in that of
human communication. Accordingly, we shall restrict the discussion by
identifying certain bounds——bounds which need to be made explicit at
the outset. In particular, the concern here will be entirely one of
outlining certain key implications for theory, particularly theory
dealing with human communication. That is, the discussion will focus
primarily on indicated modifications to theoretical forms , rather than
on a study of present forms, and will be more concerned with broad
classes of theories than with specific details or instances (except

as examples). The concern here will also be entirely with theory

128
itself—~the many and close relationships between theory and research
will be considered in Chapter u. In addition, in keeping with Chapter
1, the terms "theory" and "model" will be treated as virtually inter -
changeable, following Rudner's view of a model as ". . . an alter—
native interpretation of the same calculus of which the theory itself
is an interpretation (1966, p. 210." While some would reserve "model"
for more specific "interpretations," and "theory" for more general
ones (Lowry, 1968, pp. 55—56), such a distinction is not considered
useful in this discussion given the current level of advancement of
human communication theory.

In addition to these bounds on the discussion of implications
for theory, it should be noted again that the concern here is with
theoretical forms only as they incorporate the concept of process , as
distinct from other concepts such as interaction, emergence, or system
(see page L+0) . It should likewise be noted again that the considera—
tions below will not necessarily apply to all theory building within
the discipline of communication, since not all such theory need incor-
porate the concept of process (see page 71). For those bodies of
theory which do involve process, the implications noted here, as well
as the modifications to theoretical forms which they indicate, should
be of some importance. The various theoretical forms suggested below
are therefore presented not as replacements for present forms of
theory, but rather as highly useful additions to such theory (see page
75).

General implications and comments. —- Given these bounds for

 

the discussion, it is possible to ask what it is in the "nature" of

129
the many terms suggested earlier that bears implications fOr the form
or type of theory used in the study of human communication. .Again,
eaCh of the terms noted above is linked to the concept of process.
If those terms, and the concept of process itself, are to be applied
in theory dealing with communication events, then that theory must be
of a form or type capable of expressing the relationship between the
key properties of the events and the dimension of time. Mbre specifi-

cally, a theory (or model) must be capable of expressing the functional

 

relationship of properties or variables in events to the property or
variable of time. This primary implication brings with it a number
of more specific implications for the form of theory dealing with com-
munication events; however, several points about the primary implica-
tion need to be mentioned first.

The "functional relationship" between the properties of events
and time is of course a "function" in the sense of a mapping, or of a
rule of relationship or correspondence between values. The medium
through which this relationship is expressed is not a basic concern,
here, so that the function may be presented in verbal, logical, or
mathematical terms, by means of graphs or physical devices, etc.
Likewise, it is not a basic concern whether the properties or variables
(including time) are considered as discrete or continuous, or Whether
they are classed as independent, dependent, or intervening (although
time is in all cases an independent variable). Note, too, that it is
not necessary for all of the properties or variables of a class of
events to have a direct functional relationship to time in a theory

which deals with those events as processes. That is, it is entirely

130
possible that certain relevant properties or variables incorporated
in a theory will ( 1) have no inherent relationship to time, or (2)
be linked to time only indirectly, i.e. , by means of a functional
relationship to another property or variable which is itself linked
directly to time. Finally, it is important to note that the concern
here is with properties or variables of events only as they are related

to time, rather than to one another (except for (2) above). Clearly

 

the relationship of variables to one another is a key aspect of any
theoretical formulation . However, to consider such relationships here
would be to shift the focus of this paper from a concern with concept

pf process only, to a concern with concepts like interaction and

 

emergence, as well. The latter concepts do require study with regard
to the form of theory dealing with communication events , but obviously
cannot be examined in the space available here .

Once again, the primary implication which the suggested
approaches to a constitutive definition have for the form of theory
dealing with communication events is that such theory must express
functional relationships between the key properties or variables of
events and time. Taking into account the points jLst mentioned, it
is possible to trace a number of more specific implications for the
form of theory. As an aid in doing so, it will be helpful to distin-
quish among three general forms or types of theory: the iconic, the
analogic, and the symbolic (after Clarke, 1968, p. 33).

An 19215-9. theory or model is a ". . . coded isomorphic record
of observations (Clarke, 1968, p. um." Such theories or models are

primarily descriptions of events in the sense that the events are

131
documented in some medium (e.g., verbal, graphic, physical, etc.)11
Iconic theories or models for events may be formulated over'a wide

range of levels of abstraction, as for example the use of Who's

Afraid 2i: Virginia Woolf to describe or model interaction systems

 

 

(watzlawick, gt_al,, 1967, Ch. 5) versus Lasswell's "Whg_says what_in
Which channel to Whgm with what effégt} (l9u8, p. 37)." Because the
description of events is basic in any science, iconic theories or
models are widespread, especially in the discipline of communication.
Such description has limitations, however, one of the primary being
the difficulty or impossibility of making predictions from these
theories or models (Lowry, 1968, p. 5”). In addition, in the context
of the concern here, it is evident that there may be considerable
difficulty in expressing in an iconic theory or model a functional
relationShip between key properties or variables and time.

Sudh difficulties are often much less severe in the case of an
analogic theory or model, defined here as "a representation of events
employing a substitution of congruent structures."12 A "congruent"
structure is isomorphic to a structure in the events under study, so

that there is no basic distinction between iconic and analogic theories

 

llCertain authors dislike the terminology "descriptive theory"
or "descriptive model" (Brodbeck, 1959, p. 379). In general, the
same viewpoint is reflected in this paper, although an exception is
made at this point fOr the sake of continuity in the presentation.

12The term "analogic" and its definition are slight departures

from.C1arke (1968, p. 33) for the sake of clarity in this particular
presentation. The basic principle is nevertheless the same. A
similar comment applies to the term "symbolic" and its defintiion,
below. Note, too, that Clarke's distinction of iconic, analogic,
and symbolic theory is drawn from Ackoff, et_al, (1962).

132

regarding their degree of isomorphism. Rather, the distinction lies
in the use in analogic theories of verbal, graphic, physical, or
logical structures (and relationships) which are distinct entities
in themselves, but which have direct, well-defined correspondences
or cangruencies to the structures (and relationships) present in the
events. For the most part, analogic theories or models appear to be
formulated on higher levels of abstraction than are many iconic models.
They are also considerably fewer in number, particularly in the dis-
cipline of communication: the work of Katzer (19 70) and of Stanfield,
e}; 31. (1965, pp. 25—50) being examples of completed models, and that
of Borden (1970) of a model under develOpment. While not always easy,
it is nevertheless possible using analogic models such as these not
only to make predictions from the model, but also to express a func-
tional relationship between key variables and time .

In addition to iconic and analogic forms, it is possible to
distinguish a third class of symbolic theories or models, defined
here as "representations of events employing logical or mathematical
notations and systems." Like iconic and analogic theories, symbolic
theories must also maintain an isomorphism with the events represented.
However, the Specific logical or mathematical notations which form the
medium of expression need not be direct analogs of particular struc—
tures (and relationships) in the events under study. In part because
of the medium in which symbolic theories are expressed, they are usually
formulated on higher levels of abstraction than are either iconic or
analogic theories . Symbolic theories or models are also infrequent

in the discipline of communication, appearing normally only as lower

133
level theories or models within the framework of other theories , e . g. ,
the mathematical eqtations found in the work of Stanfield, 'e_1_:_ 31.
(1965, pp. 25-50), noted above. The work of Jaffe and Feldstein
(19 70) on dialogue is a notable exception, however. Despite their
relatively infrequent appearance to date , symbolic theories or models
are potentially the easiest of all from which to draw predictions and
in which to express a functional relationship between key variables
and time.

The distinctions which have been made above between iconic,
analogic, and symbolic theories or models quite obviously do not
define three mutually exclusive classes. Rather, the three general
forms or types of theory represent three points along what is essen-
tially a continuum. That is, there are some theories or models which
fall between the different classes, and perhaps a few which would be
difficult to classify at all. The distinctions have been made not for
the purpose of categorizing theories, but for the purpose of tracing
the more specific implications which the suggested approaches to con-
stitutive definition have for the form of theory which deals with
communication events .

Specific implications, comparisons, and comments. According to

 

the primary implication, again, a theory dealing with communication
events must be of a form which is capable of expressing a functional
relationship between key properties or variables of the events and
time. Of the three forms of theory considered above, iconic theory is
certainly less well suited to this task than are analogic and symbolic

theory, deSpite the fact that it encompasses much of the present theory

13u

in the discipline of communication. One implication, therefbre, is
that if communication scholars desire to build theories and models
which deal with communication events as processes, they will have to
devote someWhat greater attention to the analogic and symbolic fOrms
of theory. Note that this evaluation does not indicate any necessary
deficiency in current theory, but instead implies a need to expand
the scope of theoretical fbrmulations in the discipline to include
the other'two fOrms of theory. A.more detailed examination and com—
parison of analogic and symbolic theories or'models should indicate
more clearly what is implied by the need for added attention to the
two forms. In particular, it will be helpful to consider and compare
the two forms with respect to their inclusion of a time variable,
their relationship to computers, and their function and use as theories.

As they were defined above, analogic and symbolic theories or
models did not necessarily incorporate a functional relationship be—
tween properties or variables in events and time. In fact, no restric-
tion whatsoever was placed on the nature of the function. The primary
implication notes, however, that a fUnctional relationship to time
must be included in a theory or model which deals with communication
events. Accordingly, it will be useful in the present context to
redefine "analogic" and "symbolic" theories or models to take the
dimension of time into account.

An analogic theory or model WhiCh includes a time relationship
is one WhiCh specifies, fOr eaCh set of conditions in a given universe,
a single set of congruent structures Which operate on variables in a

specific time sequence. That is, for any one set of conditions,

135
there is only a single set of congruent structures operating in a
specific sequence. On the other hand, a symbolic theory or model
which includes a time relationship is one which, for all sets of con—
ditions in a given universe , expresses each key variable as a logical
or mathematical function of the variable time . Hence , for any one
set of conditions and any one point in time , there is only a single
set of values of the variables (see discussion of functions of time,
section 3.3.1).

The two definitions reveal that for one set of conditions and
one point in time in an event, the two forms of models would yield
the same results, though they would "produce" those results by dif-
ferent means. The definitions reveal, too, a difference in the nature
of the functional relationships to time . That is , analogic theories
express time by specifying sequences of operations in time, whereas
symbolic theories express time by specifying direct notational rela-
tionships to time. Nevertheless , both approaches qualify as func-
tional relationships to time in the sense of a mapping of properties
or variables onto time .

It is important to note that in the redefinitions above, ana-
logic and symbolic theories bear some interesting similarities to
computer programs (whether digital or analog), especially when the
latter are defined as "procedmres which specify, for each set of con-
ditions in a given universe, a single set of operations on variables."

Despite this similarity, it is crucial to note that analOgic and

 

symbolic theories are 2 ng Way restricted E implementation thrOugh

 

 

 

computer programs. To consider such a restriction as necessary or

 

136

implied is to neglect the broad range of highly useful flowcharts,
algorithms, games, manual and machine simulations, etc. , which meet
the definitions for analogic or symbolic theories, but which in no
way involve the use of a computer. In cases where a computer does
prove to be a useful tool, it should be noted that the program _i_t_s_e_l_._f;
is the theory or model under study, and that a similar situation pre—
vails if a flowchart or game is used as a theory. Scholars have not
been at all inclined to view programs, flowcharts, etc. , as theories
in the past, but it is becoming much clearer that they should perhaps
begin to do so (Gullahorn, 1965, pp. HHS—14%; Uhr, 1966, p. 367).

Irrespective of the particular manner of implementation, ana-
logic and symbolic theories do have considerable potential for use.
In particular, both forms will serve the four purposes which Deutsch
has noted for a theory or model: (1) providing a conceptual framework
for organizing data, (2) generating hypotheses, (3) providing actual
predictions , and (H) providing information relevant to measurement
techniques (1952, pp. 360-361). The means by which hypotheses are
generated and predictions are made from analogic theories clearly will
be distinct and perhaps less straightforward than the means employed
with symbolic theories. That is, both forms serve the same overall
purposes , but do so by somewhat different approaches , each approach
having inherent advantages and disadvantages . Because both analogic
and symbolic theories or models are potentially important in the dis-
cipline of communication, the advantages and disadvantages in using

the two forms have been compared briefly in Table 5, below.

137

Table 5. .A Comparison of Analogic and Symbolic Forms of Theory

 

Analogic Theory Symbolic Theory
Relatively complex-~many statements Relatively simple--few state-
often needed. ments often sufficient.
Requires low sophistication in Requires high sophistication
expression of relationShips and in expression of relationships
variables. and variables.

‘Minor changes generally easy Minor changes often difficult

to make. or impossible.

Derivation of hypotheses may be Derivation of hypotheses often

difficult . straightforward .

Non-analytic solutions generally Analytic solutions often

needed (see Carroll, 1968, pp. possible. Non—analytic

53-60; Lowry, 1968, pp. 58—59). solutions usually straight-
forward.

Manipuiative tools only Powerful manipulative tools

sketchily developed and often well developed and available.

unavailable.

Not often a relevant Certain tools restricted in

consideration application to classes of

expressions (e.g., linear) or
of behaviors (e.g., stOChastic
as Osgood, §t_al,, 1965, p.

230).
Investigator may become Investigator may become
"immersed" in minute details. "immersed" in mathematical
elegance.

 

A careful study of the advantages and disadvantages presented
in Table 5 reveals an important point regarding the use of analogic
versus symbolic theories or models in the discipline of communication.
That is, despite their desirability in terms of'ease in deriving hy-

potheses or availability of powerful tools, symbolic theories or

138
models for the most part require too high a level of abstraction and
sophistication in expression to find wide application at present.
Instead, it appears that the current state of theory relevant to
human communication is such that analogic theories or models are the
most useful and applicable form. The insights which may be gained
from symbolic theories related to communication are by no means to
be ignored, as Krippendorff has stressed (1969a, pp. 129-131) and as
will be evident in Chapter 14 . However, the largely untapped poten-
tial of analogic theory demands that it be more fully explored and
developed in building and advancing theory which deals with communica—
tion events as processes (see Katzer, 1970, and Scheflen, 1968).

In surmary.- As indicated in part by this examination of

 

certain aspects of analogic and symbolic theory, there are a number
of relatively specific implications regarding the form of theory used
in the study of human communication. These implications stem from
the more general implication that a theory or model which deals with
communication events as processes must include a functional relation-
ship of key properties or variables to time. This primary implication,
once again, results from the particular nature of the terms introduced
earlier in the chapter in suggesting approaches to the constitutive
definition of the concept of process. In short, those terms and
suggested approaches were ones appropriate for dealing with the dimen-
sion of time in events, and more specifically, with the change over
time or process which is seen here as an integral aspect of human

communication events .

139

Quite clearly, it has not been possible in a chapter of this
length to discuss all possible details either of the verbal and math-
ematical aspects of constitutive definition, or of the implications
which these approaches have for the form of theory. Recognizing this ,
the procedure in both cases has been to Specifically limit the range
of the discussion. That is, in considering the verbal and mathematical
aspects of constitutive definition, terms were chosen and discussed
only if they had potential utility as tools in theory building related
to communication. On the other hand, the approach to limiting the
discussion of implications for the form of theory was to identify in
advance the bownds within which the discussion would be placed. Both
of these techniques will prove useful in Chapter u, as well.

Clnapter 3 , then, has endeavored to provide approaches to a
constitutive definition of the concept of process, as well as to con-
sider some of the implications which these approaches have for theory.
But constitutive definition of a concept is but one step in the
explication. Operational definition is needed also, and approaches

to this step in explication will occupy the next chapter.

CHAPTER 4

APPROACHES TO AN OPERATIONAL DEFINITION

 

As in the previous chapter, the concern here is also with the
concept of process or of change over time—-a concept seen earlier to
be important and useful in the conduct of research and theory building
related to human communication. This importance and utility has led
to the task of explicating the concept for use as a tool in research
and theory. With the first major step in that explication complete
(Chapter 3), there remains the second step of providing approaches to
an operational definition of the concept of process . Chapter u will
be devoted to this step, as well as to an examination of the implica-
tions which these approaches have for research and analysis techniques.
Chapter u will lead to a consideration both of examples of research

in Chapter 5, and of broader implications in Chapter 6.

14.1 Nature of the Definition

 

As mentioned in Chapter 1, an Operational definition of a con-
cept specifies a procedure, urnder specific conditions, whose perfor—
mance by an investigator will identify (i.e. , measure or generate) in
nature a situation which is the referent of the concept (see page 9ff
and Berlo, 1967, pp. 2-3, 11). As noted in the same discussion, a
formal operational definition can be formed only within the context

of a specific observational or experimental framework. However,

1H0

1H1
because there is no attempt herein to design or carry out a specific
experiment or plan of observation, the present situation is not one
in Which a unique operationalization can be fbrmed. Accordingly,the
attempt in this paper'will be to identify the general principles fbr
operationalizing the concept of process. In other'words, the attempt
will be to suggest approaches to an operational definition.by pro—
viding a set of criteria.Which a particular operationalization must
meet if it is to produce data.WhiCh adequately Characterize a given
form of change over time or process. These suggested approaches not
only provide the tools for forming operational definitions in specific
observational and experimental situations, but also hold a number’of
important implications for'researCh and analysis techniques.

Because the concept of process has received relatively small
attention in the study of human communication, it will be again
necessary in suggesting approaches to an operational definition to
draw on information from a number of different sauces. Of particular
importance will be the information provided in Chapter 3, and especial—
1y in section 3.3. In addition, disciplines such as biology, cybernetics,
mathematics, systemltheory and analysis, and several others will serve
as important sources of information. The choice of information from
these various disciplines will be eclectic, but will be in all situa-
tions a choice based on the criteria of consistency, generality,
simplicity, and primarily, utility. The judgments on these different
criteria, and especially on utility, will be made in the context of
potential use of the information as a tool in researCh related to

hnman communication .

1142

Note again that there will be no attempt in this chapter~to
construct a.pgi92§_operational definition. Rather, a set of
potentially useful approaChes to such a definition will be suggested
and discussed. In doing so, it will be helpful to divide the
remainder of this chapter into three parts. In particular, Since the
general principles fdr operationalizing the concept of process are of
key importance, the second section will focus on them, as well as on
a consideration of the necessary and sufficient conditions for the
description of a process. These general principles carry a number of
important implications fOr research teChniques, especially fbr design
and measurement, so that the third section will focus on these partic—
ular implications. In addition, because research techniques are
closely associated with analysis techniques, it will be appropriate
to devote the fourth section to briefly examining implications for the
analysis of research information. In the normal context of observa-
tional or experimental research one would not normally consider
research technique apart from.analysis technique (of. Kirk, 1968, p.
11). HOwever, presentational economy makes such a separation useful
here, and perhaps justifiable given the absence of a single, specific

research framework.

H.2 General Principles of Operational Definition

 

Again, the first taSk is to suggest a set of approaches to an
operational definition of the concept of process. More specifically,
the task will be to develop a set of general principles fOr operation-
alizing the concept, i.e., a set of criteria or requirements Which a

research design must meet if it is to produce data Which adequately

1M3
characterize (or identify) a given form of change over time or process .
The development of such a set of criteria will involve the use of
information from a number of different sources , and in several cases a
detailed review of such information would require extensive discussion
or derivation. AS space is limited, such highly detailed discussions
or derivations will occasionally be bypassed, with special attention
given to appropriate references . In considering the general principles
of operational definition of the concept of process it will be useful
to divide the discussion into two parts. First, a consideration of the
relationship between measurement at discrete points in time and the
form of variation which can be characterized, and second , a statement
of the necessary and sufficient conditions for the description of a

process .

u. 2. 1 Discrete Measurement and Forms of Variation

In developing the criteria which a research design must meet to
produce data characterizing a process, it will be necessary to depend
heavily both on the concept of a path or form of variation , and on the
graphic approach to representing such a path.1 It will be necessary,
in addition, to restrict attention to the case of a single variable
changing over time within a bounded interval, recognizing that what
is said for the single variable case may be extended directly to the
multiple variable case (see pages 87ff and 101+ff). Finally, it is

important to note again that in order to record or characterize a given

 

lI'hese concepts were introduced in Chapter 2, pages 33 and 37,
respectively, and have been used several times Since, most notably in
section 3.3.

luu
form of variation, it is necessary to measure both the value of the
variable mnder study, and the point in time at which that value is
recorded (see page 33). In other words, the time of the observation
must be recorded along with the value of the variable.

In an ideal situation, the process or change over time being
studied would be indexed at all points in time by continuous measure-
ment of the variable. Such continuous measurement within a bounded
interval provides data which characterize the full form of the vari-
ation in that interval, i.e. , no consideration need be given to the
adequacy of the data in characterizing the process . Continuous
measurement is an ideal, however, and is not always necessary nor
always possible. In a more normal situation, the approach is to
measure or sample the variable at a nnurber of discrete points in time
(usually regularly spaced) during the interval. This latter form of
measurement raises the question to be considered here: Does the data
produced by the discrete measurement adequately characterize the form
of variation? The question may be broken down into three parts: How
frequently must measurements be made? What is the minimum number of
measurements? How long must measurement continue?

Frequency of measurement.-- That it is possible for a set of

 

measurements taken at discrete points in time to completely charac-
terize a continuous path or form of variation (a signal) in a bounded

interval has been lcnown for many years.2 The reason that the finite

 

2Complete characterization means that the data provide the same
information as continuous measurement . That is , even though the data
are finite, they allow reconstruction of the full form of the variation

 

1N5

number of data produced by discrete measurement can fully Characterize
the "infinite" number of data in a continuous signal is that only some
of the data in the continuous signal are independent of one another
(Cherry, 1957, p. lul). It is therefore necessary to establiSh What
data are independent, or in other~words, to determine how frequently
measurements must be made on a given signal or form.of variation. In
this particular context, such a determination can be made most readily
by considering both the technique of Fourier analysis and the Sampling
Theorem.

Fourier analysis has been introduced earlier (page 122), and is
a mathematically valid technique for "decomposing" agy_signal or foam
of variation within a bounded interval of time into a number of
component signals, each of Which is sinusoidal. In theory, an exact
decomposition might require an infinite number of such components, but
if one is willing to accept a small margin of error, the number be-
comes finite and often small. The advantage in decomposing a "complex"
signal into component signals is that each component, because it is
sinusoidal, can be characterized by only two data (G. A. Miller, 1951,
pp. 29—30). An example of a complex signal and its three component
signals, as might be determined by Fourier analysis, is presented in
Figure 1, page lu6. The sinusoidal components form the complex signal
when added algebraically, and a similar'relationship would hold between

any other form of variation and its components. For the purpose of

 

over the interval. Note that the term1"signal" will again be used here as
synonymous with "form.of variation," as in section 3.3.3.

1H6

Complex Signal

 

 

 

 

 

 

 

 

 

 

 

At . ‘t

Sinusoidal Component Signals

 

 

 

 

p ’ Highest
Frequency
Component
\/ y
w W Lowest
p Frequency
Intermediate Component
Frequen or
Lﬁanmﬁgy Fundamental
)—
t

Figure l. A Complex Signal and Its Components
(After Koenig, 1967, pp. 25-30)

111:7

determining how frequently measurement must be made on a complex
signal, it is the component with the highest frequency which is of
primary interest. 3

The high frequency component has a characteristic frequency,
f, in cycles per second, as well as a particular period, p, in
seconds, which is the time required for a single cycle (note p = 1/ f).
The Sampling Theorem reveals that discrete measurements spaced at
intervals less than or equal to one-half of the period p are suf-
ficient to conpletely characterize the complex signal. In other words ,
it is the period, p, of the high frequency component of a particular
signal or form of variation which determines how frequently the Signal
as a whole must be measured. Measurement which is spaced at intervals
(At) where At __<= p/ 2 produces data which characterize the signal over
the bounded interval and which allow its reconstruction through
Fourier techniques (see Cherry, 1957, pp. ltd—1143; Gabor, 19146, p.
430; G. A. Miller, 1951, p. 30; Pierce, 1961, pp. 272-273; and Shannon,
1963, pp. 53-5”). Measurement at intervals where At > p/2 can pro—
vide an approximation of the signal under study, but cannot completely
characterize it or allow its reconstruction in all detail.

This particular criterion regarding the frequency of measurement
is the primary criterion relating measurement at discrete points in

time and the form of variation which can be characterized by that

 

3Further information on Fourier analysis , together with
detailed derivations may be found in: Bergland (1969 , pp. |+l—52) ,
Bracewell (1965), Cherry (1957, pp. 128-1143), Gabor (191:6, p. l#30),
Hamming (1962, Ch. 6), Koenig (1967, pp. 25-30), and Pierce (1961,
pp. 30-3u).

1148
measurement (see page 34). It Should be noted though, that it is not
always necessary to identify the high frequency component and hence
the measurement interval , At , through Fourier analysis . That is ,
one can freqtently determine a maximum frequency or "bandwith" for a
given form of variation, or can identify a minimum response or reaction
time, etc. Both of these approaches would provide similar information,
and will be considered again, below. The Fourier analysis approach
has been used above because of its broad generality and power. Note
too, that in the absence of information on highest frequencies, it is
possible to Lee the same criterion to identify the forms of varia—
tion which _ca_rl be adequately characterized by a given measurement
interval, At. This point will also be considered, below, but it is
necessary first to complete the present task.

Minimum number of measurements . -- In addition to considering

 

how frequently measurements must be taken on a given Signal or form
of variation, it is useful to determine the minimum number of measure-
ments which might be required to characterize a form of variation.

To do so, it will be helpful to exclude from consideration the case
where the path or form of variation is known to be linear. In this
case, one measurement is sufficient if the slope is known (cf . page
35h) and two are sufficient if the slope is unknown (of. page 37n).U'
Given this exclusion, the simpliest possible measurement Situation is

that in which one attempts to characterize a sinusoidal signal.

 

‘4 . .

Note , however, that three measurements are not sufflClent to
determine whether the variation is linear. Linearity is , instead, an
independent determination .

1H9

That is, because any other form of variation can be analyzed into a
set of sinusoidal components, the sinusoidal signal is the basic or
simpliest ftmmn In.considering the number of measurements needed to
characterize suCh a signal, it will be helpfhl to focus again on the
high frequency component of a particular signal because of its cen-
trality in measurement considerations. More specifically, it will be
assumed below that the frequency, f, of this component is known.

Given this particular assumption, it can be shown (as in
Appendix B), that one measurement is insufficient to characterize a

sinusoidal signal in any case. Two measurements, however, may well

 

be sufficient to characterize such a signal (of. page 37n). However,
should the measured values be equal, two measurements are ggt_suf-
ficient, and it is necessary to obtain three measurements spaced at
intervals of At, Where At < p/2. Three measurements spaced at At< p/2
are sufficient, in all cases under the assumed conditions, to charac—
terize a sinusoidal signal (see Appendix B).

The minimumlnumbers of measurements mentioned above are derived
from an examination of the equation for a sinusoidal signal and
especially of the number of unknown variables in that equation for
whiCh infOrmation must be provided. The frequency of'a sinusoidal
signal is one of the variables in the general equation fOr suCh a
signal, but is assumed to be known in the derivation used here. Note
that if the frequency is ngt_known, a new unknown variable is intro-
duced-—one WhiCh makes three measurements an absolute minimum.number,
and WhiCh in all cases requires that the measured values be unequal

in order to Characterize a sinusoidal signal (see Appendix B).

150

Length of measurement period.-—- Again, the above analysis indi-

 

cates that three measurements spaced at At < p/ 2 are sufficient to

characterize a sinusoidal signal of known frequency. It is therefore

 

evident that if the signal under study is a cleex form of variation
having several sinusoidal components, it cannot be completely char—
acterized by three measurements. Other measurements are necessary to
provide information to characterize the other components . Since the
frequency of measurement is fixed at At ; p/ 2, it becomes useful to
determine how long measurement must continue to characterize a complex
form of variation in a bounded interval . Such a determination can be
made by considering, once again, both Fourier analysis and the
Sampling Theorem.

As noted in the discussion of frequency of measurement, dis-
crete measurements spaced at intervals, At, where At ; p/ 2, are suf-
ficient to completely characterize a complex signal in a bounded
interval. If the bounded interval has a total time span, T, then the
number of measurements needed to characterize it is T/ At = T/p/ 2 =
2T/p = 2fT, since f = l/p. Not surprizingly, the quantity 2fT is also
the number of independent data in any continuous signal of duration T
(see Cherry, 1957, pp. lHl—lMZ; Shannon, 1963, pp. ss-su).

Thus, in order to completely characterize a complex signal in
a bounded interval , measurement must continue until 2fI' data
collected (cf. Pierce, 1961, p. 272). In practice, with measurements
spaced at or just under the interval At = p/ 2, measurement would have
to continue for the entire interval, T. However, if the complex

signal repeats itself regularly in the bounded interval under study,

151
someWhat fewer measurements may suffice. In this case, measurement
whiCh continued over a sub-interval equal to the period, P, of the
lowest or fUndamental frequency component would be sufficient to
completely Characterize the complex signal (see Figure 1, page 1M6;
G. Am Miller, 1951, p. 30). While these latter two criteria are
perhaps slightly less general than that above, they will be used here
as a matter of convenience.

A number of other details could be added to the above dis-
cussions of the frequency of measurement, the minimum.number of
measurements, and the length of the measurement periodg'however, the
intent here is not to reproduce detailed information available from
other sources. Rather, the intent is to develop and explain a set of
criteria fOr determining Whether the data produced by the discrete
measurement of a signal adequately Characterize that signal. The
various criteria considered are useful in suCh determinations, but
have a somewhat broader use in the present context. That is, the
criteria suggested above are also the criteria whiCh a researCh de-
sign must meet if it is to produce data.WhiCh adequately Characterize
a given fOrm of Change over time or process. These criteria can be
most conveniently summarized by assembling them into a set of

conditions for the description of a process.

u.2.2 Necessary and Sufficient Conditions for Description of a Process

An examination of the criteria discussed above reveals that the
necessary and sufficient conditions for the description or measurement
of a signal, of a form of variation or Change Over time, or of a

process, are these:

152

(1) Measurement of the value of the variable(s) under

study, and of the points in time at WhiCh suCh
measurements are taken.

(2) Measurement at at least 3 points in time Where:

(a) the measurement interval, At, is At < p/2,
and p is the period of the highest frequency
sinusoidal component of the signal, or:

(b) the measured values, vh, are unequal, v1 #
v2 ¢ V3, and the frequency of the highest
frequency sinusoidal component is unknown.

(3) Measurement over the bounded interval, T, or over

a sub-interval, P, where P is the period of the
lowest frequency sinusoidal component of the signal.

Fulfillment of these necessary and sufficient conditions by an
observation procedure or researCh design assures that that procedure
or design will produce data WhiCh adequately describe, measure, or
Characterize a given form of variation or process in a bounded
interval. The conditions also constitute a set of general principles
for operationalizing the concept of process, in the sense that a
researCh design WhiCh meets these conditions would be an operation-
alization of the concept of process. That is, suCh a design would
specify a procedure, under specific conditions, Whose perfonmance
would identify (or measure) a situation WhiCh is the referent of the
concept (see page 9).

It has been the burden of section n.2, then, to develop this
set of general principles for operationalizing the concept of process.
Again, because the present situation is not one in whiCh a unique opera-
tional definition can be constructed, the attempt has been to suggest
a set of potentially useful approaChes to such a definition. These
suggested approaChes are, in effect, a set of tools WhiCh may be used

to construct specific operational definitions of the concept of

process. As noted earlier, the various approaches or principles have

153
been developed using information from a number of different sources--
infbrmation Chosen on the basis of its consistency, generality,
simplicity, and.particularly its utility When judged as a potential
tool in research related to human communication.

Because the attempt here has been one of suggesting approaChes
or principles for use in forming operational definitions , rather than
one of constructing Specific definitions, it is not appropriate to
judge these principles by the normal criteria fbr a viable, fOrmal
operational definition. Such criteria are generally those of formal
clarity, significance, correspondence to the concept, independence,
and above all, utility (G. R. Miller, 1967, pp. lH—ZH). HOwever, the
suggested approaChes or principles are seen as capable of producing
definitions whiCh meet these criteria, should they be applied in
specific research situations to operationalize the concept of process.

The approaChes to an operational definition considered above
clearly have a number of important implications fer researCh and
analysis techniques in the study of human communication. Specifically,
the principles for Operationalizing the concept of process have several
very gizggt_implications for researCh teChniques, i.e., for'researoh
design and for measurement. Section ”.3 will examine these direct
implications, and in so doing will attempt to tie the rather'abstract
necessary and sufficient conditions noted above to the more concrete
considerations of actual researCh.

Note, too, that the infbrmation produced in an actual researCh
situation usually undergoes some form of analysis, normally by means

of an analysis technique whiCh is closely associated with the

151+

particular research technique employed. As a consequence , the
principles for operationalizing the concept of process have a number
of indirect implications for analysis techniques. Section 14A will
examine these somewhat less direct implications, but in so doing will
create a distinction between research and analysis which is not
normally made (of. Kirk, 1968, p. 11). As noted earlier, it is the
absence in this paper of a single, specific research situation which

makes such a distinction necessary.

1+. 3 Implications for Research Techniques

 

To this point, Chapter 1+ has been concerned with a set of
suggested approaches to the operational definition of the concept of
process. In particular, it has considered certain general principles
for Operationalizing the concept and has provided a set of criteria,
in the form of necessary and sufficient conditions, which a partic-
ular Operationalization must meet. These approaches, and more
specifically the criteria or conditions , have a number of M
implications for research techniques, not only for research design,
but also for measurement. It is appropriate to examine these implica—
tions for research techniques, particularly as those techniques relate
to the study of human communication events .

An examination of the techniques of research design and of
measurement is clearly a lengthy and involved task in any discipline,
that of communication being no exception . The present limitations
of space dictate a somewhat restricted or bounded discussion, and it

will be helpful to identify those bounds in advance (see section 3.1+).

155
More specifically, the concern here will be entirely one of outlining
key implications for teChniques of design and measurement, especially
as these teChniques relate to researCh on human communication. In
other words , the discussion will focus primarily on indicated modifi-
cations in researCh teChniques, rather’than on a review of existing
teChniques. In addition, the concern will be more with general modi—
fications and broader classes of techniques than with specific Changes
or tools (except as examples). As a result, in some instances Where
detailed infermation might otherwise be useful, it will be necessary
to suggest outside references as more complete sources. .A similar
situation will hold in the separate discussion of analysis teChniques
in section u.u.

In keeping with Chapter 2, both observational and experimental
researCh will be considered relevant in the discussion below (see pages
33, 70; cf. Sidman, 1960, p. 237). That is, While certain researCh
teChniques lend themselves more readily to observational researCh than
to experimental researCh, or vice versa, suCh a distinction will not
be a primary concern in the discussion of those teChniques. In
addition to these bounds, it should be emphasized again that the con—
cern here is with research teChniques only as they relate to the
concept of process, not as they relate to other concepts suCh as inter-
action, emergence, or system1(see page HO). Likewise, it is important
to note again that the discussion below will not necessarily apply to
all researCh conducted in the discipline of communication, since not
all suCh researCh need incorporate the concept of process (see page 71).

However, for'researCh whiCh does involve process, the implications and

156
modifications suggested here should be of some importance. The research
techniques discussed below are presented therefore not as replacements
for present techniques, but as potentially useful additions to such
techniques (see page 75).

Acknowledging these bounds for the discussion, it appears appro—
priate to divide what follows into three parts. First, an examination
of the direct implications for research design of each of the necessary
and sufficient conditions. Second, a brief examination of several
potentially useful "process designs." And third, a consideration of

certain implications for the measurement of variables other than time.

L&.3.l Direct Implications for Research Design

As is clear from the discussion in section 14.2, a research
design which is to adequately measure a form of variation or process
must fulfill the necessary and sufficient conditions noted on page 152.
Designs which meet the conditions are said to be operationalizations
of the concept of process and will produce data which completely
characterize the form of variation under study. Because the necessary
and sufficient conditions are the criteria separating "adequate" from
"inadequate" designs, they quite clearly hold direct implications for
the development of research designs. Because each of the three con-
ditions holds distinct implications , it will be appropriate to examine
each of them in order. In doing so, it will also be possible to
express in more concrete form the various considerations to be made in
constructing process designs, i.e., designs capable of producing data
which characterize a form of variation or process (see pages 38-140,

56-58) .

157

Variable value and time. -- The first of the three conditions

 

for the description of a process notes that it is necessary to obtain
"Measurement of the value of the variables(s) under study, and of
the points in time at which such measurements are taken (page 152)."
While this first condition describes a fundamental operation in the
description of any process, it is perhaps the easiest of the three
conditions to fulfill. Its implications for research design are
therefore somewhat less far-reaching than those of the other conditions .

Very briefly, the first condition identifies an additional type
of measurement which must be made in conjunction with the measurements
taken on the variable(s) of interest. The additional measurement
indexes the times at which the other measurements are made, and re-
quires both the operational definition of an "observation point," and
the measurement of the time of this point from some reference point
(see pages 33-3u, 56). Time is capable of ratio measurement, and
the reference point may be either an arbitrary zero point, or the
point of the first measurement. These considerations apply both to
the measurement of variables at discrete points in time, and to
measurement over all points in time (i.e., continuous measurement),
although in the latter case the concept of a discrete observation
point loses its meaning. Note, again, that measurement at discrete
points in time approaches continuous measurement as the interval At
approaches zero (see section 3.3.3).

The actual implementation of the measurement of time with re~
spect to the measurement of other variables depends heavily on the

nature of the specific research sitLation. In general, though,

158
problems may arise in defining an "observation point" if the pro—
cedure for'measuring the variable(s) under study requires a large
amount of time in comparison to the interval At. The specific re-
searCh situation would indicate what type of operational definition
is most appropriate, as in the Choice, fer example, of defining the
Observation point as the beginning, middle, or end of the measurement
period. Other problems may arise, as well, if the measurement pro—
cedure interferes with the presentation of an important stimulus.
Consideration will be given to this problem in section u.3.2, below,
since the solution may require the use of a particular type of
researCh design.

Measurement interval and resolution.- The second of the three

 

conditions for the description of a process is somewhat more complex
than the first. It notes, in the first place (a), that measurements
must be made at at least three points in time Where "the measurement
interval, At, is At < p/2, and p is the period of the highest fre-
quency sinusoidal component of the signal (page 152)." This aspect
of the condition is by no means easy to fulfill, and carries a number
of important implications for human communication research. The
first aspect of the condition implies, in general, the need fOr
certain new considerations in developing adequate researCh designs.
That is, in constructing a design to measure a process, it is neces-
sary to consider'certain Specific Characteristics of the form of
variation in order to determine the appropriate span between Observa-
tion points. SuCh considerations are.not often made in human come

munication researCh, at present, and it will be helpful to examine

159
briefly What they involve.

As the first aspect (a) of the condition is stated, it requires
an §;p§ig§i_know1edge of the frequency or period of the highest fre-
quency sinusoidal component of the signal under study. SuCh informa-
tion can be obtained in advance only by depending on similar, earlier
studies whiCh indicate the expected form of the variation. If the
ftrmnis known in general (as in learning curves), or Specifically (as
in previous recordings), it may be possible by means of Fourier
analysis to determine directly the highest frequency component, its
period, and hence the appropriate measurement interval (see page 122).

However, it is not always necessary, or more importantly,
always possible, to obtain information on the highest frequency come
ponent as suggested above. In suCh cases it may be possible, as an
alternative , to determine a "minimum response time" for the particular
event(s) or individual(s) under study. If suCh a response time is the
minimum time required to initiate some Change in state, then that time
interval may well be a suitable measurement interval, At. More
generally, it may be possible to determine a.maximum frequency by
determining the "bandwidth" of the signal under study. As the term is
employed, here, the "bandwidth" of a complex Signal is the frequency
of the highest frequency sinusoidal component of the signal (Cherry,
1957, p. 303). In some cases, this maximum.frequency can be determined
‘without knowing (or anticipating) the general or specific form of the
variation, as above. That is, determinations of maximum "Channel" or
"input/outpu " rates or capacities, Whether for maChines or for indi-

viduals, may provide information on bandwidths (of. Shannon, 1963, pp.

160
66-68, 79—81; Pierce, 1961, pp. 36-38). If suCh a rate or capacity
can be found, it will provide the necessary information to determine
the appropriate measurement interval. In.more conCrete terms, if one
can answer the question "How fast can this state (e.g., an attitude)
possibly change?" he can determine how frequently to measure it in
order to record its variation.

Despite the existence of these alternatives for determining the
appropriate measurement interval, situations will still exist in.WhiCh
it is extremely difficult or even impossible to obtain the necessary
infOrmation on At. In suCh cases, the second condition remains
basically unfulfilled, as considered above, though it is by no means
irrelevant. Specifically, the condition notes, in the second place
(b), that measurements can be made at at least three points in time
where "the measured values, vh, are unequal, v1 i v2 f V3, and the
frequency of the highest frequency sinusoidal component is unknown
(page 152)." Note that this second aSpect of the condition is ngt_a
substitute for the first in providing the necessary information to
develOp adequate researCh designs. Nevertheless, this aspect is ime
portant and implies, in general, the need for additional new consider—
ations in developing research designs.

That is, in constructing a researCh design in the absence of
infOrmation on the appropriate measurement interval, it is neCessary
to consider What may be termed a design's "resolution capability" or
"resolving power" in measuring a process. In other words, the second
aSpect (b) of the condition indicates that if a design includes at

least three measurement points, it is possible to determine the fre-

161

quency of the highest frequency component which the design is capable
of measuring. Note that whether or not the design actually does
measure such a component depends on the presence of three unequal
values of Vn- However, a design would be capable of resolving or
measuring component frequencies up to a maximum of f = Up = l/(2°At),
where At is the smallest measurement interval in the design.

Considerations of the resolution capability of a particular
design do not often appear in research on human communication at
present, even though the resolving power of a design may place a con-
straint on the analysis of the data produced by the design, and on
the semantic interpretation of the resulting evidence (see page 3”) .
In addition, Sidman (1960, pp. 287-289) has suggested that in certain
research situations there may well be optimum resolutions which need
to be carefully examined and determined. Note that a design's
resolution capability need not be considered if continuous measurement
is involved, just as it is unnecessary to consider or determine the
appropriate measurement interval . The reason is that the second con-
dition is "automatically" fulfilled by continuous measurement.
Should the measurement be discrete, however, the first aspect (a)
focusses attention on the measurement interval, while the second
aspect (b) requires consideration of resolution capability, partic-
ularly if the appropriate measurement interval cannot be found .

Length of measurement period.-- The third of the three con—
ditions for the description of a process notes that the measurement
"TU-St continue ". . . over the bounded interval, T, or over a sub-

interval, P, where P is the period of the lowest frequency sinusoidal

163
component of the signal (page 152)." This third condition is fairly
straightforward, but is not always easy to fulfill. As such, it also
holds important implications for research design.

In brief, the third condition indicates that adequate measure—
ment of a form of variation or process within a bounded interval
requires that the measurement continue throughout that interval .
Exceptions to this requirement do exist, but only if the variation is
such that it repeats itself over the period of the fundamental and
within the bounded interval under study. AS in the second condition,
it is possible to take advantage of this exception only if there is a
m knowledge about the frequency or period of the fundamental.
Such information can be obtained through Fourier analysis of general
or specific forms of variation; however, in this case there are few,
if any, alternatives to the Fourier method.

The third condition implies the need for yet another consider-
ation in research design--one which is, in effect, the opposite of
resolving power and might be termed a design's "inclusiveness." That
is, whatever the length of the measurement period in a design
(whether T or P), the design is only capable of including or-measuring
component frequencies with periods shorter than or equal to T or P,
Whichever is shorter. Hence , the lowest frequency which can be
identified by a design is F = l/T or l/P, whichever is higher.

Considerations of a particular design's "inclusiveness" are

occasionally included in human communication research, at present,
Often in terms of a comment on a design's weakness in measuring a "long

term trend." Such considerations constitute an important constraint

161+
both on the analysis of research data and on the semantic interpreta—
tion of evidence , and should not be overlooked . Note that unlike the
case above, the third condition applies equally to discrete and to

continuous measurement of a process .

The third cond iti on, then, like the other two which comprise
the necessary and sufficient conditions for describing a process, holds
certain distinct and direct implications for developing research
designs. AS a group, these implications lead to a number of fairly
concrete considerations for constructing designs capable of producing
data which characterize a process . But the considerations above remain
at a fairly general level, and it will be helpful in studying the over-
all implications for research design to examine briefly several

Specific designs .

l4.3.2 Some Specific Process Designs

In particular, it will be helpful to review several designs
mentioned earlier, as well as to introduce one "new" design, all of
which are potentially useful as "process designs" in research on human
communication. Each of the designs below is capable of operation-
alizing the concept of process , though clearly whether it does so or
not in a given research situation depends entirely on whether or not
it fulfills the necessary and sufficient conditions. If the conditions
are fulfilled, then the data which the design produces will character-
ize the process under study.

Note that the discussion below is not intended to be exhaustive
in indicating the strengths and weaknesses of the various designs.

Many of the critical strong and weak points of each design have been

165

examined at length by Campbell and Stanley (1963), and it is clear
that many problems with specific designs arise from (or are solved by)
particular conditions of use . The discussion below is intended rather
to review a number of potentially useful designs in view of the con—
siderations above. Clearly, of all the research designs considered
by Campbell and Stanley (1963), and summarized earlier in Table 2,
page 57 (and in Appendix A), only those which include three or more
observation points are capable of meeting the necessary and sufficient
conditions. A brief review of each of the designs in the third
column of Table 2 is therefore in order, as only these can function
as "process designs."

Three of the process designs listed in column three of Table 2
(page 57) Share a common characteristic: they are direct extensions
of difference designs (column two) in attempts to index a time dimen-
sion. Specifically, the pretest-posttest with control and time
extension design (Hovland's design) is an extension of the basic "pre—
test-posttest with control group" design (H); the separate-sample pre-

test—posttest with time extension design (12b) expands upon a similarly

 

named difference design (12); and the 'multiple—wave' panel design (B)
is an extension of the "two-wave panel" design (A).5 The first two of
these three process designs, i.e., Hovland's design and the separate—
sample design, both have the somewhat unfortunate characteristic of
requiring one (or more). groups of subjects for each additional obser—
vation point desired. Such a characteristic multiplies the sample

Size, effort, and cost in using these designs for more than three

 

5A brief comparison of these designs in their schematic forms
(Appendix A) will make the nature of the extensions considerably more
eVident. The numbers in parentheses refer to Appendix A.

166

observations , although the high precision and control in the Hovland
design may be worth the expense. The multiple—wave panel design, on
the other hand, requires no additional groups (if used in its simple
form), but carries and compounds the many disadvantages of panel
designs noted by Campbell and Stanley (1963, pp. 67-68).

The other two process designs listed in column three of Table
2 (page 57) are distinct in that they are not extensions of simplier
designs, but instead are designs created especially to index a time

dimension. The two designs are the time series design (7) and the

 

multiple time series design (11+), and are sufficiently similar to be

 

treated together. These two time series designs require no additional
groups for additional observation points, and are especially flexible
with respect to the number and spacing of such points. In fact, as
the measurement interval in these designs becomes smaller, the measure-
ment involved approaches continuous measurement. These strengths , and
others, nevertheless may be offset by certain weaknesses, as Campbell
and Stanley (1963, pp. 37—u3, 55-57) have noted, among which is the
potential problem of repeated measures. This latter problem is not
often encountered in research in areas such as interaction analysis ,
however, a point which accounts in part for the many successful uses
of the time series designs in such research (see pages 62-614).

The five process designs reviewed above are ones which have
been mentioned earlier (section 2.3.2) as potentially useful designs
in research on human communication, but they do not exhaust the possi—
bilities for process designs. That is, the fact that three of the

five designs above are extensions of simplier designs suggests that

167
other process designs might be generated by using difference or even
point designs as bases. To exemplify such possibilities, it will be
helpful to present a single "new" process design, and to examine its
strengths and weaknesses, though again not in the same detail as the
Campbell and Stanley analyses (1963).

This new design draws in part on the pretest—posttest with con-

trol and time extension design (Hovland) , and on certain counter-
balanced designs (11). For lack of a more imaginative name it will

be termed the separated group delayed posttest design. Using the same

 

code employed by Campbell and Stanley (1963, p. 6; Appendix A) for

presenting designs, this new design may be presented schematically as

follows:
t1 t2 t3 tUr t5
Group A R Xa O Xb 0
Group B R Xa O Xb 0
Group C R Xa O Xb 0
Group D R X a O Xb 0

Very briefly, the R represents the randomization of subjects into
separate groups, the Xa and Xb represent the exposure of a group to
stimulus variable(s) or event(s), and the 0 represents some form of
observation or measurement. In this design, the posttest observation

is delayed for a different time span with each separate group, thus

giving rise to the name. Note that a pretest is possible in the
design if desired.
As presented above, the separated group delayed posttest design

would perhaps not be automatically classed with the other process

168
designs. However, if the four groups can be assured homogeneous due
to randomization, the design can be seen to produce data which are

equivalent to the following "design":

Group Ya xa o o 0 0
Group Yb xb o o o o
Quite clearly this latter equivalent design is a process design. Note
that the "Group Ya" observations are a composite of the observations
on Groups A, B, C, and D. Similarly, the "Group Yb" observations are
composed of the observations on the same four Groups , but following a
different stimulus. All of the observations on Group Yb actually
occur at time t5, but involve intervals which are the same as those
separating times t1 through tn.

Referring to this presentation, it is possible to examine some
of the strengths and weaknesses of this new design. It is apparent,
first of all, that the design requires as many separate groups as there
are delayed posttests, so that additional observation points do in-
crease the sample size, effort, and cost. However, this disadvantage
may well offset by the design's ability to accommodate situations in
which the measurement period is lengthy, or in which the measurement
procedure seriously interrupts or otherwise interferes with the pre-
sentation of the stimulus variable or event. For example, the design
would permit a complex measurement instrument to be applied to a con-

tinuous stimulus (e.g. , a speech), in effect without interrupting that

St‘imulus .

169

The separated group delayed posttest design may be subject, in
the second place, to weaknesses with respect to history, maturation,
and multiple treatment interference , as Campbell and Stanley have
defined these terms (1963, pp. 5-6). On the other hand, the design
has the advantage of not involving the problems of repeated measures ,
as do the time series designs, even though it produces similar data.
The design offers, as well, a considerable flexibility in the use of
its two stimulus exposures. That is, the design would allow Similar
testing on two different treatments , or would allow a wide range of
possibilities for incorporating controls in an experiment.

The separated group delayed posttest design clearly has certain
weaknesses and specific strengths, as do the other process designs.
It is presented here both as a potentially useful design in research
on human communication, and as an example of the possibilities in
developing new designs capable of operationalizing the concept of
process. Other designs could have been suggested, and more details
could have been added than have been included above , were the intent
here to present an exhaustive study of process designs and their use.
While designs above are capable of operationalizing the concept of
process, whether they in fact do so in a given research situation is
separate matter. That is, the designs must still meet the necessary
and sufficient conditions for describing a process (page 152) in order
to produce data which characterize the process under study. In general,
1Then, the necessary and sufficient conditions for describing a process
Carry the additional implication that if research on human communica-

tiOn is to adequately deal with the concept of process , it will be

170

necessary both to use existing process designs and to develop new ones.

As is perhaps evident in the above review of process designs,
the choice of any one research design depends, at least in part, on
the choice of the measurement technique to be employed. For example ,
choosing a time series design may bring the problems of repeated
measurement, while the nature or complexity of a measurement techrique
may dictate the choice of a design such as the separated groLp delayed
posttest design. The result of this interaction between research de—
sign and measurement technique is that the necessary and sufficient
conditions also have certain implications for measurement of variables
other than threw-implications which need to be examined as part of the

broader task of studying implications for research techniques.

u. 3. 3 Implications for Measurement

As the term "process" is used here, it refers to a "change over
time of matter—energy or information, within a bounded interval of
time." Measurement of the variable of time has been considered already
in section 14.3.1, so that the focus here is entirely on the measurement
of other types of variables, i.e. , those representing the states of
matter—energy or information under study. Again, the concern here is
with measurement in its interaction with research design, rather than
with measurement E3: s3, recognizing that changes in measurement may
well open new possibilities in the choice of designs. Note, too, that
because the details of a measurement technique are always closely tied
to a Specific research situation, it will be necessary in what follows

to consider only general aspects of measurement.

171

In other words , the intent here is not to review the many
different forms which measurement may take, ranging from psychophy-
sical to verbal, nor to consider the problems raised by levels of
measurement , whether nominal , ordinal, interval, or ratio. Rather,
the concern will be with those broader aspects of measurement which
require further development or which must be considered in choosing
research designs. In particular, it will be helpful to examine the
need both for development of unobtrusive measures and for considera—
. tion of individual or group measurement, as well as the need for
attention to two other, somewhat brie fer points .

One of the problems often encountered in communication research
is that of the effect which the measurement instrument or procedure
has on the research stimulus, on the subject(s), or on both. Such
effects may well be compounded in certain process designs by the
presence of repeated observation or measurement of some event, and
may well be disruptive enough to prevent the use of those designs.
While alternate designs may be applicable in some cases, an important

additional possibility is the use of some form of "unobtrusive" or

 

"nonreactive" measurement .

 

The different forms of and problems in using unobtrusive
measures have been discussed at length by Webb, g a}: (1966), and can-
not be detailed here . In general , such measures may involve the use
of "archival" records, or of simple or "contrived" observation, the
latter class including "hardware" such as the audio-video taping
equipment which has contributed to the recent advances in interaction

analysis (Ekman, _etgl.(1969); F‘rahm (1970); Harrison (1969b); cf.

172
Starkweather (1969)). Wlnile the measurement techniques discussed by
Webb, _el al. , are not always widely applicable, the importance in
finding or developing an unobtrusive measure for a given research
situation is that it may permit the use of a previously unsuitable
though desirable research design.

As a specific example, the availability of an unobtrusive
measure for a key variable might allow the use of a straightforward
time series design in a particular situation, in place of the more
complex and expensive separated group delayed posttest design. More
generally, development in the area of unobtrusive measures would
appear useful in broadening the application of existing process de-
signs, as well as in encouraging new designs. One must always be on
guard against the lure of "gadgets" and of "fads" in measurement, but
at the same time it is clear that developments in measurement tech—
niques (e.g. , the semantic differential) have occasionally brought
valuable changes in both the methods and the content of human com—
munication research .

In addition to the possible use or development of unobtrusive
measures as a factor affecting the choice of research designs, it may
be useful in certain communication research situations to consider

specifically the question of whether individual 2 group measurement

 

 

is to be employed. That is, is the measurement instrurent or tech-
nique to be one which is applied to subjects as isolated individuals

or to subjects assembled in some form of group?6

 

6Note that the concern here is with the "Situation" in which

173

At first glance, such a question may well seem trivial, Since
the answer would appear to be "determined" by the purposes underlying
the research, by the availability of subjects, by the nature of the
variable( S) under study, by cost, etc. Indeed, such factors do often
dictate the use of , say, group measurement over individual measurement,
but it is also the case that such a decision is sometimes made by
default, owing to tradition or to the availability of a particular
instrument or technique, for example. Particularly in the latter case,
it becomes important to consider whether individual or group measure—
ment is to be employed because of the strong tie between these forms
of measurement and research designs. That is, of the process designs
considered earlier in section l+.3.2, only the time series designs and
the multiple wave panel design appear readily adaptable to both forms
of measurement—~the other three designs are strongly associated with
group measurement.

Rather than implicitly accept such restrictions on the choice
of a research design as may be imposed by tradition or by an available
form of measurement , it appears important for the investigator
to consider of whether the "opposite" form of measurement might be used.
Specifically, if an investigator can recognize ﬁne need for and can
develop a group measurement technique where an individual technique had
previously been used, he may well find that he has a larger range of
process designs to draw on. On the other hand, development of an

‘

the instrument or technique is applied, not with the treatment of the
resulting data; i.e. , data gathered from isolated individuals might
Well be "grouped," while that gathered in a group might be analyzed
on an "individual" basis.

171+
individual measurement technique to replace a group technique might
allow, in a particular SitLation, more effective use of a time series
design. In general, it would appear useful to consider specifically
the question of whether individual or group measurement is to be
employed, since doing so may broaden considerably the range of appli-
cable process designs. .

Along with these considerations of the " orm" of measurement
and of unobtrusive measures, there are two other somewhat briefer
points which need attention and which should be mentioned in reviewing
implications for measurement. The first of these concerns both the

nature 2f variables other than time and their operational definition.

 

 

More specifically, ﬁne use of process designs which meet the necessary
and sufficient conditions discussed earlier will have certain effects
on the types of variables studied in research situations. Increased
frequency of measurement, for example, may make it difficult to employ
some variables which would be quite suitable in point or difference
designs (e.g. , Noble's m measure). New variables will have to be
developed, both to replace older ones and to utilize ﬁne added poten—
tials of increased measurement frequency. Such new variables will
require operational definitions, as do all variables (Berlo, 1967, p.
11), and in view of the possible complexity of some variables which
might be measured in a process design, it is important to note ﬁnat
there are other modes of operational definition besides the usual
verbal statement of procedures. In particular, analog or digital com-
puter programs are also statements of procedures ,- but wiﬁn the partic-

ular advantages of being able to define a wide range of variables, at

175
many levels of complexity, and with extreme uniformity.
The second of the two briefer points concerns the problems 31:

repeated measurement. In particular, ﬁne use of process designs in

 

research will result in sets of observations which will not always
satisfy the assumption of independence of observations required by
most statistical techniques . Note that while this point borders on
the concerns of section '4.” with analysis techniques, it deserves
mention here due to its intimate relationship to measurement.
Specifically, in the case where the measurement occurs over a total
interval, T, and where the highest frequency sinusoidal component has
a known frequency, f, it is evident that 2fT data ar_e independent (see
section l#.2.1). While such information will be of help in certain
situations, it is clear that many uses of process designs will require
a fuller consideration of the means for dealing with measurements or
data which are related in some degree. While it is quite clear that
the problems of repeated measurement have received a fair amount of
attention, as in Harris (1967) among oﬁners, it is also clear that
more work remains to be done (Campbell, 1963, pp. l+5-Ll6).

These two somewhat briefer points concerning variables and
repeated measurement could be considerably expanded, were the concern
here with ﬁne problems of measurement Er g2; The concern is , however,
entirely with measurement in its interaction with research design, and
ﬁnese two brief points are mentioned because they require at least some
consideration in using a process design. Of key importance in using,
and especially in choosing a process design are ﬁne two broader factors

of possible use of unobtrusive measures and of consideration of

176

individual or group measurement. Again, changes in eiﬁner of ﬁnese
two aspects of measurement may well open new possibilities in the
choice of designs. Because of this interaction between research
design and measurement technique, the necessary and sufficient con—
ditions for describing a process also carry implications for ﬁne
measurement of variables oﬁner than time. In general, ﬁnose impli-
cations are that if research on human communication is to utilize
research designs which meet the specified conditions, it will be
necessary to consider several broader aspects of measm'ement which
might normally be overlooked.

In summary.-- This examination of the implications for measure—

 

ment is only one portion of the study of implications attempted in
this section (14.3). That is, the review of implications for measure-
ment, in combination with the consideration of Specific process
designs and the examination of direct implications for research
design, as a group comprise the review of implications for research
techniques. Because many of these implications are relatively Specific,
it may be helpful to formulate a somewhat more general implication to
serve as a summary of the above. Namely, if research on human com-
munication is to deal adequately with ﬁne concept of process it must
use and develop process designs, considering carefully both the ways
in which the designs meet the necessary and sufficient conditions,
and ﬁne nature of the measurement which is employed.

If research can be conducted in ﬁne manner of this general
implication, it will provide data which adequately characterize a

process under study. Such data are clearly necessary if one is to

177
eventually make semantic interpretations of communication research in
terms of the concept of process (see page 35ff). However, while the
implications for research technniques discussed here do indicate what
is required of a research design which operationalizes the concept of
process, the implications also make clear that such research is not
necessarily easy. That is, in certain situations it may require far
too much effort to add even a third observation point to a difference
design,or to employ a process design in a manner which satisfies the
necessary and sufficient conditions. In such cases the conditions may
provide helpful information on the limits of ﬁne design in character-
izing a process, but they are also likely to indicate that despite its
desirability, a semantic interpretation in terms of "process" is
plainly not justified by the data.

It has not been possible in a section of this length to be
exhaustive in considering the implications for research techniques ,
even as those techniques may apply to research on human communication.
Recognizing this , the approach has been to examine the techniques of
research design and of measurement within certain bounds which were
identified in advance. A similar approach will be followed in section
Unit, as well. That is, while section 14.3 has considered the implica—
tions which the approaches to an operational definition have for
research techniques, it is evident that such techniques comprise only
one of the aspects of any research situation. Analysis techniques are
also a vital aspect of research, for as Coomnbs has noted: "The method

of collecting data determines what information they contain, but the

 

method of analysis defines this information . . . (1953, p. ”95)."

178

u.u Implication for Analysis TeChniques

 

To this point, Chapter u has been concerned both with a set of
suggested approaches to the operational definition of the concept of
process, and with the implications WhiCh these approaChes carry for
research teChniques. More specifically, the Chapternhas provided a
set of necessary and sufficient conditions fer describing a process,
and has considered the implications whiCh these conditions have for
researCh design and measurement. These approaChes, and more specific-
ally the criteria or conditions, have additional implications whiCh
extend beyond those for researCh teChniques. In particular, the
approaChes hold a number of indirect implications for analysis teCh-
niques—-imp1ications WhiCh it is appropriate to examine, especially
as the teChniques relate to the study of human communication events.

It is important to emphasize, at the outset, that the implica-
tions for analysis teChniques do not stem directly fromnthe necessary
and sufficient conditions, as in the case for researCh teChniques.
That is, a brief examination of the conditions presented earlier
(page 152) will reveal that they describe the "framework" within.whiCh
the measurement of a process is to take place. As a result, the
conditions provide information whiCh is directly relevant to researCh
design and to measurement. The resulting process designs, when
applied, produce data having a Special Characteristic WhiCh results
from the method of collection; namely, the data contain information
on the time dimension of an event. Although many different forms of
analysis may be perfOrmed on suCh data, we shall be concerned here

only with those forms whiCh take the time dimension into account.

179
Consequently, While the necessary and sufficient conditions have
direct implications for researCh teChniques, they have only indirect
implications fer analysis teChniques, since the latter in effect
derive from the former.

In addition, it is worthwhile noting again that the consider-
ation of analysis teChniques has been purposely separated from.the
discussion of researCh techniques. SuCh a distinction would not
normally be made, given that the method of analysis is often Chosen
and considered in conjunction with the researCh design (of. Kirk,
1968, p. 11). In the absence of a single, specific researCh Situation,
however, suCh a distinction appears necessary (see sections ”.1 and
u.2).

An examination of the analysis techniques applicable in any
discipline is a lengthy and complex task, especially in the discipline
of communication. The present limits of Space, in addition, dictate
a highly restricted or bounded discussion, and it is important to
identify those bounds in advance (see sections 3.” and H.3). In
general, the concern here will be to outline briefly the implications
for analysis techniques, especially as these teChniques relate to
researCh on human communication. In particular, because the implica—
tions are indirect, the discussion will fecus primarily on useful
additions to analysis techniques , rather than on a review of current
teChniques. Also, because the details of using anv particular
analytic teChnique are tied both to the purpose of the researCh and
to the specifics of the individual researCh situation, the concern

will be with general approaChes to analysis and with broad classes of

180

techniques. In cases where detailed information does appear useful,
special attention will be given to suggesting appropriate outside
references as more complete sources.

Again, in keeping with Chapter 2, both observational and
experimental researCh will be considered relevant in what follows
(see pages 33, 70; cf. Sidman, 1960, p. 237). That is, while certain
analysis techniques lend themselves primarily to one form of research
or the other, suCh a distinction will not be of concern below in dis—
cussing the techniques. In addition to these bounds, it must be
emphasized again that the concern here is with analysis teChniques
only as they relate to the concept of process, not as they may relate
to other concepts such as interaction, emergence, or system (see page
HO). Consequently, While many different analyses may be performed on
one set of data, and while many of the teChniques below'have broader
applications,the only analyses and applications considered here will
be ones relevant to the time dimension in the data, i.e., to the
process involved in the event(s) under study. It is important to note
again, too, that the discussion.here will not necessarily apply to all
researCh conducted in the discipline of communication, as not all suCh
researCh need incorporate the concept of process (see page 71). The
analysis techniques suggested below are accordingly not presented as
replacements for current techniques, but as additions with potential
utility in research which deals with human communication events as
processes (see page 75).

Taking into account these bounds on the discussion, it appears

useful to divide this section into three parts. First, an examination

181
of certain basic distinctions and considerations useful in discussing
analysis teChniques. Second, an outline of analysis teChniques WhiCh
serve primarily in "reducing" data. And third, an outline of teCh-

niques WhiCh function primarily in "testing" situations.

u.u.1 Basic Distinctions and Considerations

As noted above, the necessary and sufficient conditions for
describing a process hold both direct implications for researCh teCh-
niques and indirect implications for analysis teChniques. That is,
a researCh design WhiCh is to measure adequately a process or form of
variation must fulfill the necessary and sufficient conditions, so
that the conditions have a direct effect on the researCh design and
on certain related aspects of measurement. Designs WhiCh do fu1fill
the conditions are said to operationalize the concept of process, and
will produce data whiCh completely Characterize the variation under
study. The distinguishing feature of suCh data, for the present
discussion, is the association of a time measurement with eaCh obser-
vation or measurement on the variable under studv, and clearly any
analysis of this data with regard to the concept of process must take
this time dimension into consideration. The necessary and sufficient
conditions therefore have an effect on analysis techniques, albeit
indirect, resulting from.the nature of the data produced by the re—
searCh teChnique.

In discussing these effects on analysis teChniques, it is help-
ful, first of all, to distinguish two classes of teChniques WhiCh

serve different, though often closely related functions in the analysis

182
of research information. One class of teChniques serves the primary
function of "reducing" data to a more manageable form, in effect by
generating some smaller amount of information or data WhiCh symbolizes
or represents the larger body of original information. SuCh "re—
duced data" are frequently the input to another class of analytic
teChniques Which serves the primary function of "testing" data or
information, whether in testing predictions or in testing the per—
fOrmance of a model. More specifically, the teChniques fOr generating
various descriptive statistics are examples of analytic teChniques
WhiCh serve to reduce data, while the techniques for making signifi—
cance tests are key examples of teChniques for testing data. It is
evident that the functions of reducing and testing data or information
are not totally disjoint, since a descriptive statistic may be the
input to a significance test. Nevertheless, the distinction can be
useful, in this case as an aid in organizing the discussion of
analysis teChniques.

It will be helprl, secondly, to note in brief several other
considerations whiCh affect the discussion. In particular, it is
important to emphasize again that the concern here is primarily with
the case of a single variable changing over time within a bounded
interval (see section u.2.1), i.e., with a single path or form of
variation (see pages 33 and 37). That is, the analysis teChniques
considered below will be viewed primarily as they apply to the single
variable case, although in some cases they may be extended directly
to the multiple variable case (see pages 87ff and lOMff). There will

be no direct concern with the multiple variable situation, since to

183
do so would require an examination of individual analysis techniques
someWhat beyond the scope of the present discussion.

It is important to note, too, that the concern here is gply_
with variables as they are related to time, pgt_as they are related
to one another. Inter—variable relationships are clearly important
in much researCh, but to consider them here would be to Change the

focus of the paper from a concern with process only, to a concern

 

with interaction and emergence. It should be recognized, in addition,
that any one variable whiCh changes over time may be, in and of itself,
a measurement taken on a single individual or a composite of measure-
ments taken on a group (e.g., a mean value). While this distinction is
an important one in many researCh situations (Sidman, 1960, pp. us-su),
it will not enter into the present discussion of analysis teChniques.
Finally, it will be helpful to consider briefly the place in
the discussion of three different factors whiCh enter into the Choice
of a particular analysis teChnique. First, there will be no concern
here with.whether a variable is classed as independent, dependent, or
intervening, although it is clear that time is in all cases an inde-
pendent variable. Second, there will be no consideration of the dis—
tinction between discrete and continuous measurement of time, or of
other variables, as it may affect analysis teChniques. Discrete
measurement is the primary mode of measurement assumed in the teChniques
considered, however, and it should be noted that it is always possible
to adequately translate continuous measurement into discrete measurements
(section H.2.l) if a teChnique suitable for continuous measurement does

not exist. Third, there will be no concern below with the level of

H!

181+
measurement of time, or of other variables, as it affects the appli-
cability of various analysis teChniques. Clearly, the available level
of measurement is an important factor in the Choice of an analysis
teChnique, as are the other two factors; however, an examination of
individual teChniques taking level of'measurement into account would

move far beyond the sc0pe of this discussion.6

u.u.2 Reduction Techniques

In view of the considerations, above, it is possible to turn
to an outline of several analysis teChniques WhiCh serve the primary
function of reducing data to more manageable forms. Again, suCh re—
duction is accomplished by generating some smaller set or more con-
venient form.of data or information WhiCh represents or symbolizes the
larger body of original data. The reduction may be one of number, as
in generating the mean of fifty values, or possibly one of fermu as
in producing a graph of the same fifty values. In either case, the
reduced data may be an end in itself, or it may be the input to another
analysis technique, suCh as the testing techniques outlined in section
1+ . H. 3 .

As indicated earlier in the consideration of the bounds of the
present discussion, the attempt here is not to review current reduction
teChniques, but rather to outline useful additions to suCh techniques.
Accordingly, it is assured in what follows that the more common reduc-

tion techniques are known, as for example the techniques fOr

 

6Many of the techniques discussed do have several forms which
permit them.to be applied with different levels of measurement. Time,
of course, is capable of ratio measurement.

185

determining measures of central tendency and of dispersion, fOr con—
structing graphs, histograms, and frequency distributions, etc. SuCh
techniques are fUndamental ones, and are considered in most statis-
tical texts, as in MCNemar (1962, Chs. 2-H) and in Quenouille (1952,
Chs. l and 2), among others. Of Special interest here are certain
reduction teChniques whiCh may be generally less well known, and Which
are particularly useful in reducing data that contain a time dimension.
Five suCh "new" teChniques are presented in Table 6, page 186, namely:
pattern recognition, constraint analysis, signal analysis, stoChastic
processes, and trend surface analysis.

These five "teChniques" are not single, Specific teChniques,
but rather broad classes of teChniques, both as suggested in the earlier
consideration of the bounds of the discussion, and as evidenced by the
general descriptions included in Table 6. Clearly, this presentation
of techniques is not intended as a comprehensive or exhaustive summary,
but rather as a general outline, introduction, and guide. In partic-
ular, because Table 6 is to serve as a guide, special attention has
been given to references, both in including notes suggesting the use—
fulness of each reference, and in selecting sources whiCh provide basic
infOrmation or Which themselves fUnction as references. Again, it has
not been possible to be exhaustive.

Comments.-- EaCh of the classes of teChniques presented in
Table 6 has a distinct purpose or "output," even though all are poten-
tially useful approaChes to reducing data which contain a time dimension.
The attempt has been, by means of the general descriptions, to indicate

the nature of the various purposes or outputs so that the Table can

186

Table 6. Reduction Techniques for Analysis of Data from Process Designs
I

 

Technique and General Descriptiona

Pattern Recognition

Techniques for either identifying new
patterns or seeking specified patterns in
data. Patterns may exist in time or space,

 

with temporal patterns (and data) of special

interest here. Actually a broad area

containing many techniques, some Specifical-

ly adapted to certain types or classes of
patterns and data.

Constraint Analysis

Tachniques, often closely related to
pattern recognition techniques, for deter»
mining the bounds or constraints present in
a given set of data.
temporal, spatial, etc., althougn temporal
data, often termed "protocols" are of
special interest here. Individual techni-
ques vary, but are apt to be more generally
applicable than many pattern recognition
techniques.

Sigpé Analygis
e iques or deriving symbolic (mathe-

matical) representations of temporal data
or signals. Actually a broad area contain-
ing many different techniques.

 

Stochastic Processes

TEChrioues forHErivinp a specific class
of symbolic representations of data. Data
are often temporal, and the representation
differs from those above in that it descri-
bes the sequence or succession of individ-
ual data points.

 

Trend Surface Analysis

Techniques for deriving a symbolic
representation of a trend or movement in a
set of data or form of variation. Data are
often spatial, but may be temporal.
1y a broad collection of techniques and may
include regression analvsis.

 

Again, the data may be

Actual-

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

References and Notes

Uhr (1966, entire vol.)
Mason (1970)
Kanal (1968),

General information:

Comprehensive bibliographv:

Discussion of techniques:
Watanabe (1969)

Note that journals such as the IFTEE Transactions
on Systems Science and gbernetncs or on
Computers, as well as." e ourna of the
Pattern Recognition Society are alSB Valuable

sources a" information.

 

 

General information with Special reference

to human communication: Krippendorff
(1969b, pp. 2u—32)
General information: Ashby (1961;)

Potentially relevant sources on special aspects:
Ashby (1963), Klir (1967), Shannon and
McCarthy (1956)

General information: Gabor (191:6), Koenig
(1967, pp. 20-33), Milsum (1966, Ch. 3)

Specific information on Fourier analysis:
Bergland (1969), Bracewell (1965),
Hamming (1966, Ch. 6)

Basic information on statistical signals:
Milsun (1966, Ch. lu)

General information: Bartlett (1966),
Fararo (1969), Parzen (1962), Snell
(1965)

Detailed information, especiallv with
respect to spectral theory: Hannah
(1960, Chs. 1-3), cf. Milsun (1966,
Ch. 1”)

Detailed information with resoect to
multiple time series: Quenouille
(1957, Chs. 1-6)

General information: Chorlev and Haggett
(1968), Clarke (1968, pp. uﬁo-ugn),
Kim: (1969, pp. 152—15N)

 

W0 ranking or ordering of techniques is implied. The ordering is generallv alphabetical,
except for variations brought by relationships or similarities among ﬁne techniques.

187
serve as an initial guide to selecting an appropriate reduction tech—
nique. The question of "WhiCh specific teChnique do I use," cannot
be answered here except in general, both because it is clearly beyond
the present sc0pe and purpose to provide a sufficiently detailed out-
line of specific techniques, and because the answer to suCh a question
depends heavily on factors suCh as the available level of'measurement,
as well as on the purpose of the researCh.

The classes of teChniques presented in Table 6 are not to be
considered mutually exclusive. In many cases the classes border or
overlap one another in that they Share basic tools or methods, so that
they are perhaps best described as general emphases or directions in
reduction teChniques. In addition, all of these general teChniques
may involve the use of analog or digital computer programs. There is
of course no binding necessity that a specific teChnique be implemented
in a computer program, although complexity often makes it highly
expedient.

One fUrther’comment on Table 6 regards the absence of the class
of techniques generally termed "system analysis." These teChniques
have been omitted, here, in that they apply to situations where an
analysis (both reduction a2d_testing) of'system_data,is required.
System data are considerably more complex than the data considered
herein because they contain infermation not only on processes in events
but also on structurem_or structurei (cf. Keenig, 1967, Chs. l—N).

As such, system analysis teChniques encompass many of the teChniques
in Table 6, most notably signal analysis, and include other teChniques

as well. Because the concern here is entirely with analysis teChniques

188

applicable to the study of events as processes, rather than as systems,
the techniques of system analysis pe£:§e_have been omitted.

The general teChniques considered in Table 6, then, represent
a set of "new" analysis teChniques WhiCh are useful in reducing data
which contain a time dimension. The object of suCh data.reduction is
primarily to provide a more manageable fermnfor~the data, while taking
the time dimension into account, although the reduction may also pro-
vide an input to a testing teChnique, as considered below. Because
of their ability to deal with a time dimension in data, the reduction
techniques considered are all potentially useful techniques in the
analysis of data produced by process designs. As a.result, although
the implication of the necessary and sufficient conditions is someWhat
indirect, it can be seen that if research on human communication is to
deal adequately with the concept of process, it must use and develop

reduction teChniques such as those presented in Table 6.

u.u.3 Testing TeChniques

As suggested at several points above, reduction teChniques fu1-
fill only one of two functions of analysis techniques. It is important,
therefbre, to outline several teChniques whiCh serve the other primary
function of testing data or infOrmation. SuCh testing may, again, be

either a test of predictions or a test of the perfOrmance of a model.

 

 

Note that in both cases it may be difficult to separate the testing
teChnique from the reduction teChnique, since many statistical tests
in effect combine testing and reduction, and since tests of a model's

performance may require that large quantities of data be reduced. In

189
addition, it is clear that many testing techniques require some form
of reduced data as their "input." The distinction between testing
techniques and reduction techniques is a useful one, nevertheless, if
for no other reason than as an organizing device in discussion.
Within the general area of testing techniques, itself, it may
be difficult at times to separate a test of prediction from a test

of the performance of a model. Very briefly, prediction testing may

 

be thought of in the present context as a situation in which a set of
observed data containing a time dimension is to be compared to a set
of predicted data. The data appear in these situations as a sequence
of values in a time dimension, and it is specifically the variation
over time in the observed and in the predicted data which is to be
compared or tested. In a very general sense, then, prediction testing
may be viewed as a problem of "curve matching" in an attempt to answer
the question: "Is the predicted curve or data a 'reasonable' fit to
the observed curve or data?"

Performance testing, on the other hand, may be viewed as a

 

situation in which a model, in this case one incorporating the concept
of process, is manipulated to determine the characteristics of its
"operation." The manipulation may take the form of a mathematical
analysis or of an actual "run" or use of the model under specified
conditions . The characteristics sought may be various mathematical
relationships or the actual "outputs" or results of the model. How-
ever, it is at this point that the distinction between performance
testing and prediction testing breaks down, for obviously one extremely

valuable means of performance testing is to form a prediction from the

190
model, gather data, and test that prediction. Granting this exception,
the distinction between the two types of testing will still prove use-
ful as an aid in the discussion.

As indicated earlier, the intent here is not to review current
testing techniques, but to outline useful additions to those teChniques.
As a result, it is assumed in what follows that the more common testing
teChniques are known, as fer example the teChniques considered by
McNemar (1962, Chs. 5—16, 18), Ouenouille (1952, Chs. 3-10, 12), and
Siegel (1956). It will likewise be assumed that the problems of testing
predictions on change using difference designs are known, as considered
by Bohrnstedt (1968) and Lord (1967), fer example. Of particular
interest here are those reduction teChniques Which may be generally
less well known, and Which are particularly usefbl either in testing
predictions where process designs have been used or in testing models
in which the time dimension is central. Table 7, page 191, presents
five suCh techniques fer prediction testing and two teChniques for
performance testing.

These various "techniques" are not specific, individual teCh—
niques, but rather someWhat broader groupings of techniques, both as
suggested earlier and as indicated by the general descriptions in
Table 7. This particular presentation of sets of techniques is not to
be considered comprehensive or exhaustive, but is intended instead to
be an outline or guide. As in Table 6, particular attention has been
given here both to suggesting sources Which provide basic information
or Which serve as references themselves, and to including notes indi-

cating the usefulness of eaCh source. Nonetheless, it is impossible

191

Table 7. Testing Techniqnes for Analysis of Data from Process Designs

Technique and General Description
Prediction Testing!

 

Analysis of Variance:

Analyses and tests of "change" or "time-
curve" information by partitioning of
variance.

Trend Analysis:

Analyses and tests of the shape of the
relation between two variables. Closely
related to analysis of variance procedures.

Goodness of Fit:

Analyses and tests of the "fit" between
two series of data (often ﬁne predicted
and the observed).

Time Series Analysis:

(a) Analyses and tests of serial corre-
lation (autooor'relation), correlation,
partial correlation, and regression be-
tween time series.

(b) Analyses and tests with regard to
stochastic processes.

Sequential Analyses:

Analyses and tests where the number of
observations required by the statistical
procedure is not determined in advance.

Performance Tag!”

 

Sensitivity Testing:

A general approach to testing models in
which parameters or relationships are sys-
tematically varied over successive "runs"
of the model. The resulting differences in
outcones are compared to determine the
sensitivity or "strength" of the model.
Techniques are basically non-analytic.

Response Aralysis:

A collection of mathematical techniques
for manipulating mochls to examire charac-
teristics of their operation. Includes
analyses of stability or controllability
as well as examination of outputs given
specified inputs. Techniques are
basically analytic.

-Q¢--‘---Q-- C--‘¢¢--—-----O-¢ -¢----C---- C-CCQ----O----‘QCQQ-¢-

References and Notes

Univariate procedures:
(1967)
Multivariate procedures:

Gaito and Wiley

Bock (1967)

General information: McNemar (1962, On. 17)

Nonparametric trend analysis: Ferguson
(1965)

Related discnssion with special reference
to measurement: Sidman (1960, On. 10)

General information: Carroll (1968, pp. 71-
72), Jaffe (1970, pp. 83-8u)

Theoretical discussion: Hannah (1960, pp.
93-10%)

General information: Campbell (1967, pp.
22U-229), Holtzman (1967), Qnenouille
(1952, On. 11)

Detailed information: Bartlett (1966, On. 9),
Quenouille (1957, On. 7)

Detailed information: Bartlett (1966, Ch. 8),
Hannan (1960, Chs. '4-5)

General and detailed information: Wald (19147)

General information: Carroll (1968, pp. 73-
714), lowry (1968, p. 62)

Gereral information:
60)

Detailed information: Koenig (1967, Chs. 5-7),
Milsum (1966, Chs. 6-9, 11)

Carroll (1968, pp. 53-

 

Wo ranking or ordering of techniques is implied. The ordering is generally alphabetical,
except for variations brougnt by relationships or similarities among the techniques.

192
to be exhaustive.

Comments.—- Each of the groupings of prediction testing teCh-
niques presented in Table 7 has a slightly different purpose or'result,
as well as a someWhat distinct "input," even though all are potentially
useful in testing predictions where the time dimension must be con-
sidered. The two classes of perfOrmance testing teChniques likewise
have different purposes and results, and may also have someWhat dif-
ferent ferms of models as their "inputs." The general descriptions
provided fer both the prediction and the perfbrmance testing teChniques
attempt to outline these special purposes, results, and inputs, so that
Table 7 can function as a guide to testing teChniques. However, it
must remain a general guide to suCh teChniques, since the infermation
which would be necessary in a guide to specific techniques is beyond
the scope and purpose of the present paper.

One factor WhiCh does enter into the selection of specific pre—
diction testing teChniques should be mentioned briefly, though. That
is, each grouping of techniques has certain requirements and makes
certain assumptions regarding the data or information under study. A
specific technique may require, for example, a particular level of
measurement, or may assume that the data or Observations are indepen—
dent. Such requirements and assumptions must be checked before a
specific prediction testing teChnique is chosen and applied, and it
may be helpful to note one important point in this regard. That is,
the often troublesome assumption of independence may be Checked using
the teChnique of serial correlation or autocorrelation, as mentioned

in Table 7 in conjunction.with the time series analysis teChniques,

193
and as discussed by Quenouille (1952, pp. 165-167) and others.

As was the case for reduction techniques in Table 6, the
various groupings Of testing techniques presented in Table 7 are not
mutually exclusive classes. In many cases the prediction testing teCh—
niques share common methods and infOrmation, and much the same is true
fer the perfOrmance testing techniques. In addition, both prediction
and perfOrmance testing techniques may involve the use of analog or
digital computer programs. The high degree of complexity WhiCh may
be involved in some perfOrmance tests make suCh tools particularly
applicable, there, though they by no means lack importance in predic—
tion testing.

The groupings of techniques presented in Table 7, then, repre—
sent a set of analysis techniques WhiCh are useful in prediction testing
and in perfOrmance testing where the time dimension is a central concern.
The Object of prediction testing is, again, to compare an Observed event
with a predicted event, in this case taking the time dimension into
account. Because of their ability to deal with a time dimension, the
prediction testing techniques considered above are all potentially use-
ful in the analysis of data produced by process designs. The Object of
perfbrmance testing, on the other~hand, is to determine certain impor~
tant characteristics Of the use or Operation of a given model, especial—
ly a model in whiCh time is a key dimension. The performance testing
teChniques described in Table 7 are capable of dealing with suCh models,
and hence are potentially useful in analyzing models which consider
communication events as processes. This consideration of the two

classes of testing teChniques reveals an additional indirect implication

19”
of the necessary and sufficient conditions; namely, that if researCh
on human communication is to make proper'use of the concept of process,
it must use and develop testing teChniques suCh as those considered
in Table 7.

In summary.-- The above examination of testing teChniques is

 

only one aspect of the study of analysis techniques in this section
(4.9). The earlier discussion of reduction techniques is an equally
important aspect, and although these two classes of teChniques have
been distinguished here, as considered earlier (section N.H.l), it is
clear that the two functions of testing and of'reducing data or inforu
mation are not disjoint. Indeed, the distinction has been made here
entirely fOr ease Of presentation, fer the fUnctions of reduction and
testing normally are and Should be considered as one in any researCh
situation. Quite clearly, too, it has not been possible to be ex-
haustive in considering analysis techniques in a section of this
length. Recognizing this, the attempt has been to make clear the
limits of the discussion both by specific comments, and by placing the
discussion within a set of bounds identified at the outset.

The study of analysis techniques in this section has been
prompted.by the examination of researCh teChniques in the previous
section, both sections being considerations of certain implications
of the necessary and sufficient conditions. Again, while the implica-
tions of the conditions for researCh techniques are direct, those fOr
analysis techniques are relatively indirect, since analysis teChniques
must take into account the characteristics of the data produced by

researCh techniques. Because of this indirectness, it may prove

195
helpful to state a fairly general implication for analysis teChniques
as a Whole. That is, if human communciation researCh is to deal
adequately with the concept of process, it must use and develop tech-
niques capable of'reducing and testing the special data Which are
produced by process designs.

If the analysis of researCh information can be perfOrmed as
suggested by this general implication, it will provide evidence which
Characterizes the process under study. SuCh evidence is a necessity
if one is to move to the step of making semantic interpretations of
communication researCh in terms of the concept of process (see page
35ff). Note again, that While there are many different fOrms of
analysis WhiCh may be performed on a given set of data, the concern
here has been only with analyses which take the time dimension in a
given set of data into account. The implication above makes clear
that even if the data in a particular situation do adequately Char-
acterize the process under study, an analysis which does not consider
the time dimension is plainly not capable of semantic interpretation
in terms of "process."

Taking a someWhat larger perspective than this concern solely
with analysis teChniques, then, it should be evident that a comment
like that above on the distinction between reduction and testing tech-
niques may also be applied to the distinction between researCh teCh-
niques and analysis teChniques. That is, although researCh and analysis
have been discussed separately here, as mentioned earlier in the Chapter
(sections ”.1 and 9.2), it is clear that these two aspects of researCh

are not disjoint, either. The distinction between research and analysis

196
has been.made primarily as an aid in the discussion, because the
absence of a single, specific research situation makes it difficult
to consider research and analysis techniques in conjunction. In the
context of an actual research situation, these two important aspects
should be and normally are considered together because of their close
association.

This examination of research techniques and of analysis tech-
niques has had the purpose Of indicating the implications for each
which result from the necessary and sufficient conditions fOr describ-
ing a process. In a considErably more general fOrm, the implication
of those conditions is that research which deals with communication
events as processes must use researCh techniques and analysis tech—
niques whiCh take the dimension of time into account. This general
implication results both from.the necessary and sufficient conditions,
and, in a somewhat broader sense, from the principles and criteria
introduced in suggesting approaChes to the operational definition of
the concept of process. Once again, those criteria and suggested
approaches were ones appropriate for dealing with the dimension of time
in events, and more specifically, with the change over time or process
which is seen here as an integral part of human communication events.

It has not been possible in a chapter of this length to consider
all of the details either of the principles and criteria fOr Opera-
tional definition, or of the implications WhiCh they hold fOr research
and analysis techniques. Consequently, the procedure in both cases
has been to specifically limit the range of the discussion. In par—

ticular, in considering the principles of operational definition,

197
information and derivations were included only if they had potential
utility as tools in research related to human communication. The
approach to limiting the discussions of implications, on the other
hand, was to identify in advance the bounds within Which the discus-
sion would take place.

An overview.-- Chapter u has attempted to provide approaches

 

to the Operational definition of the concept of process, and to con—
sider some of the implications Which these approaches have for research
techniques and analysis techniques. Again, While research and analysis
have been considered separately here, they are not distinct in that
research is seen as leading to analysis. Similarly, research and
analysis are not distinct from theory, either, for there clearly exists
a "cycle" from theory to research to analysis and back to theory. Con-
sequently, the implications fOr research and analysis considered in
Chapter u may be grouped with the implications for theory examined in
Chapter 3 to fOrm a someWhat larger set, i.e., the implications Which
explicating the concept of process hold for theory and research on
human communication.

This essential unity of the implications fOr research, analysis,
and theory suggests that comments similar to those above on the dis-
tinctions made among various techniques may also be applied to the
distinction between constitutive definition and operational definition.
That is, While constitutive definition has been considered apart from
operational definition in this paper, it is clear that these two
aspects of the explication of a concept are not totally separate.

Here, too, a distinction has been made as an aid in organizing the

198

overall discussion, even though it is evident that in the larger
context of theory building and researCh, these two important aspects
must be considered together.

In fact, a concept can be said to have been explicated if and
only if it has been provided with both a constitutive definition and
an operational definition (see page 9; cf. Berlo, 1967, p. 2-u).
Again, such definitions are normally constructed and applied only
within the context of a Specific theoretical or research framework.
Because there is no attempt in this paper to build a specific theory
or to carry out specific research, the procedure has been to suggest
a set of potentially useful approaches to both constitutive and opera-
tional definition of the concept of process. These suggested approaches
not only provide a set of tools fOr explicating the concept of process
in specific theoretical and research frameworks, but also carry the
implications fOr theory, research, and analysis Which have been
mentioned above.

Chapters 3 and M, then, have attempted to provide approaches
to the explication of the concept of process or of change over time, a
concept Which was indicated in Chapter 2 as an important and useful
one in the conduct of research and theory building related to human
communication (of. pages 73 and 7H). This attempt to provide approaches
constitutes the major emphasis of this paper——an emphasis described
earlier as a.search fer the means Of explicating the concept of process,
a concept which is seen as a useful tool in the furthernconduct of

theory building in the discipline of commmmication (see pages 8 and 9).

199

The attempt of these second two chapters to develop and
consider approaches to explication can be materially aided by con-
sidering examples of research which takes the concept of process into
account. Accordingly, Chapter 5 will be devoted to a brief review Of
several specific bodies of research relevant to the study of human
communication. The attempt will be to provide a set of concrete
referents fer the use of many of the approaChes discussed in Chapters
3 and u. Following these examples, it will be apprOpriate in Chapter
6 to examine certain broader implciations of the explication of the
concept of process, as well as several aspects of its place in research

and theory related to human communication.

CHAPTER 5

EXAMPLES

Having seen in Chapters 1 and 2 that the concept of process is
an important and useful one in the conduct of theory building and
research on communication, and having considered approaches to the
concept's explication (both constitutively and operationally) in
Chapters 3 and H, it will be helpful in Chapter 5 to consider several
examples of research relevant to human communication Which take the
concept of process into account. Such an examination of specific
examples of current research will be useful both in providing a some—
what less abstract consideration Of many of the approaches than has
been possible to now, and in answering several general questions re-
garding the nature of research whiCh deals with communication events
as processes. In particular, questions like: How widely applicable
are the approaches considered in Chapters 3 and H? What is the fOrm
of researCh which makes use of these approaches, i.e., what does it
"look" like? What kind Of results will such research produce, and
how will that information differ from the results of othernresearCh?
Chapter 5 will attempt both specific and general answers to suCh
questions, and will Open the way to a consideration of conclusions

and implications in Chapter 6.

200

201

5.1 Comments on the Examples

 

It should be noted again that the purpose of this paper is not
to consider or develop a specific theoretic or research framework
(i.e., to build theory or to test hypotheses), but rather to consider
the means Of explicating the concept of process (see pages 9 and 10).
Given this important point, together'with the fact that examples of
researCh utilizing the suggested approaches have been presented in
the communication literature, it appears more appropriate to consider
several such examples with respect to the above questions than it does
to design a single, specific experiment.

Because there are a number of examples of researCh Which might
be considered in this context, it is necessary to Choose from among
them. The limits of space dictate a consideration of no more than
three examples in any detail, although a number of other possible
examples will be mentioned briefly. The three examples which are
presented have been chosen on the basis Of two criteria. First, the
examples must be relevant to the study Of human communication and
should be representative of larger bodies of research. Second, the
examples as a group should show variation over the fOllOWing six di-
mensions to assure a breadth of coverage:

(1) Content area of the research, within the scope

of human communication research.

(2) Type of research design, though it is clear that
the research must employ a process design (see
sections, 2.1.3, 2.3.1, 2.3.2, and u.3.2).

(3) Focus on observational or experimental research.
.Again, both fOrms of research are considered
relevant in this paper (see pages 33m, 69-70,
and sections u.3 and u.u).

(u) Type of measurement technique, given that
research design and measurement technique are

202
linked to one another (see section 9.3.3).
(5) Type of analysis technique. The techniques
mentioned fOr each example will be those which
deal with the time dimension, although others
may have been employed (see section 9.9).
(6) Focus on individual or group behavior. A
potentially relevant consideration in the use
of process designs (see section 6.2).
No particular ranking is implied in the ordering of these various
dimensions. Again, variation on these six dimensions and relevance
to the study of human communication are the criteria on which the
examples of research below have been chosen.
Two furtherncomments should be made before mentioning the
three examples, specifically, and considering their variation on the
above dimensions. It will be evident in the discussions in section
5.2.1 that the examples chosen have both strengths and weaknesses with
respect to the approaches considered in Chapters 3 and 9. In partic-
ular, it is clearnthat fer the most part, the research considered does
not derive fromntheory which deals with communication events as
processes, as such theory has been described in section 3.9. That is,
it is evident in some cases that the concept of process per:§e_was not
a primary concern of the researchers in either their theory or their
researCh and analysis techniques, even though the research does opera-
tionalize the concept. In addition, it is also clear that among the
examples chosen (and especially among those not reviewed), certain
research designs may fall short of meeting all of the criteria suggested
on page 152 for the adequate description of a process. Such research

may therefOre only approximate full Operationalization, again usually

because the concept of process was not a.primary concern of the

203
investigator. These various problems will be considered below, but
specifically dp_pgt_make a particular example inappropriate fOr con-
sideration here, in that the example still may reveal important charb
acteristics of research on events as processes and can produce infer-
mation WhiCh is obtainable by no other means.

It should be noted once again that the concern here is solely
with the concept of process. Other concepts such as interaction,
emergence, and system eventually must be considered and operationalized
in research, as well, in order to reach a fuller understanding of
human communication (see page 90). However, space dictates a focus
below entirely on those aspects of researCh which relate to the concept
of process.

Given these comments, it is possible to turn to a consideration
of the three examples. Very briefly, those chosen fOr consideration
are: (1) research by Brown and Bellugi (1969) on language development
in young children, (2) studies by Harrison (1969b) using his Verbal-
Nonverbal Interaction Analysis technique, and (3) a study by Insko
(1969) on primacy—recency effects as a function of timing of arguments
and measures. (Note that there is no ranking implied here--the order
is alphabetical). BefOre considering eaCh of these examples of re-
search in detail in the next section, it will be useful to summarize
their'aims and to compare their characteristics. These two fUnctions
are fUlfilled by Table 8, page 209, Which presents both a brief
summary of each research example and compares them on the basis of the
six dimensions noted above. Table 8 will be an important reference

in the discussion of eaCh example.

209

 

 

 

Table 8. Summary and Comparison of Research Examples
Brown and Harrison (1969b) ' Insko (1969)
Investigators, BEIlugi (1969) Studies Of’interu' A.study of pri-

 

Content Areaa,
and Summary of

A study in child personal intern

action , esp . the

macy vs. recen-
' cy effects as a

I
I
I
I
Research ment, esp. acqui—' use of verbal and' fUnction of the
sition of syntax ' nonverbal Chan- ' timing of mea-
frcmnlB to 36 ' nels of commun— ' sures and argu—
months. I ication. : ments.
Type of ' Time Series (7) ' Time Series (7) "MOdified fbrm
Research : (Longitudinal) ' I of separated
Designa ' I ' group delayed
posttest e31gn
: : 1 (section 9.3.2).
Observational ' ' '
or Experimental : Observational I Observational : Experimental
Researcha
I I I
Type of : Individual: : Group: ' Group: Verbal
Measurement ' "Unobtrus ive" ' Unobtrusive ' recall by
Techniquea ' recording and ' recording with I subject.
' transcription. ' sampling. '
Type of I Reduction to 1 Reduction to I Descriptive
AnalySis ' categories and ' categories and ' statistics; t—
TeChniquea ' graphs; Des— ' graphs; ' tests, trend
' criptive sta— ' Descriptive ' analysis.
' tistics. ' statistics. '
Focus on I Individual I Dyadic : Individual
Individual or ' behavior'with ' behavior. ' behaviors
Group ' comparisons . ' ' measured; Group
Behaviora ' ' ' responses used
in analysis.
I I I
Notes Analysis techs. 1 Analysis techs. Analysis techs.
I
I
I
I
I
I
I
I

“‘-.“‘C-.

language develop-

included are
only those for

do not include
those dealing

included are
only those used

I I

I I

I I

I this report. with inter- : fOr analysis of
' Other'work action mar. ' temporal

' uses addition— trices (Which ' effects.

' a1 techs. remove the '

' time dimen- '

' Sion). '

 

a C
These dimen81ons

are described more

fully in section 5.1.

205

5.2 The Examples, Individually and In Sum

 

The procedure fer presenting the examples in this section will
be first to present the three, as individual entities, along with
certain other potentially relevant examples, and second to consider
the examples as e.group with respect to the general questions mentioned

at the outset.

5.2.1 Individual Examples

In each case below, the plan in presenting an example will be
to begin with a surmary of the aims, methods, and results of the re—
search. Only a brief sketch will be provided, as in each case a more
detailed report is available and because the primary interest is not
to reproduce an existing report but to relate the research to the
present concerns. Next, the example will be considered as it incor-
porates the concept of process. In particular, it will be examined
with respect to the principles and criteria discussed in Chapter 9.
This step will suggest both strong and weak points in the research,
as well as certain advantages and problems.

Brown and Bellugi (1969).—- In their article entitled "Three

 

Processes in the Child's Acquisition of Syntax" (1969), Brown and
Bellugi report on one aspect of an extended researCh project on lan-
guage development in children. In this particular article, the authors
investigate the acquisition of syntax in children between the ages of
18 and 36 months, or approximately the time span during Which an
average or normal child moves from the point of basic, two-word utter—
ances to the construction of complete, simple sentences. This partic-

ular study is one out of a numbernwhich are based on Observational

206
research conducted by Brown over a several year period.
In particular, the data for'these studies were gathered long-

itudinally fromntwo children.(called Adam.and Eve). Brown describes
the data gathering and analysis as fellows:

Every second week we visited eaCh child for at least
2 hours and made a tape recording of everything said by
the child as well as of everything said to the child. The
mothernwas always present and most of the speech to the
child is hers. Both mother and child became very much
accustomed to our presence and learned to continue their
usual routine with us as the observers.

One of us always made a written transcription, on the
scene, of the speeCh of mother and child with notes about
important actions and objects of attention. From this
transcription and the tape a final transcription was made,
and these transcriptions constitute the primary data of
the study. For many purposes we require a "distributional
analysis" of the speeCh of the child. To this end the
child's utterances in a given transcription.were cross-
classified and relisted under suCh headings as: "A +
noun"; "Noun + verb"; "Verbs in the past"; "Utterahces
containing the pronoun it," and so fOrth. The categorized
utterances expose the syhtactic regularities of the
child's speech. (1969, p. 133)

Brown's description, together with the characteristics noted in Table
8, page 209, reveal the basic outline of the research.

The analysis of any such complex, developmental process re-
quires that particular features be Chosen for attention. In the study
cited here, the authors fOcus on three aspects which they term
"Imitation and Reduction" on the part of the child, "Imitation with
Expansion" on the part of a parent, and "Induction of the Latent
Structure" by the child. These three aspects of the child's syntax
acquisition cannot be considered at length, but Brown's concluding
remarks with regard to them are rather instructive:

we have described three processes involved in the
child's acquisition of syntax. It is clear that the

207

last of these, the induction of latent structure, is by

farnthe most complex. It looks as if this last process

will put a serious strain on any learning theory thus

far conceived by psychology. The very intricate simul—

taneous differentiation and integration that constitutes

the evolution of the noun phrase is more reminiscent of

the biological development of an embryo than it is of

the acquisition of a conditioned reflex. (1969, pp.

160-161)

The research reported by Brown and Bellugi incorporates the
concept of process in its measurement of the developing behavior at
frequent, equal intervals over an extended period of concern (basi-
cally a time series design). Whether the two week measurement interval
is the appropriate interval for the form.of variation undernstudy
depends on the particular aspect of language development being con—
sidered. Two weeks does appear to be suitable in this case, and it
would be easy to check such an assumption, as suggested earlier,
although this does not appear to have been done in this case. Note
especially that the method of collecting and.recording data in this
research is essentially one of "capturing" the event under study so
that it may be subjected later to a wide range of analyses at many
levels Of complexity.

Those analyses which take into account the dimension of time
may yield especially useful evidence on the development of language,
as Brown's semantic interpretation above suggests. Such evidence not
only raises questions about the nature of language learning, but also
is needed to distinguish among those theories of development and of
structure Which currently exist. Note too, that While Brown and

Bellugi are concerned primarily with how the child acquires syntax,

their research indicates how heavily that acquisition depends on the

208

communicative interaction between parent and child.1

Harrison (1969b).-- In his paper entitled "Verbal—Nonverbal

 

Interaction Analysis: The substructure of an Interview" (1969b),
Harrison reports on the development of a technique fOr measuring and
describing certain aspects of the communicative interaction of two
individuals. In particular, the technique indexes the presence or
absence of communication in both the verbal and nonverbal bands, and
has been used to examine, fOr example, the interaction between doctors
and patients.

Data collection using the Verbal—Nonverbal Interaction Analysis
technique (VNVIA) is basically a two-step process in which an on—going
interaction of some fOrm is first recorded (by audio—video taping),
and later sampled at three second intervals. The sampling is perfOrmed
by coders whose output is an indication of whether or not the verbal
and nonverbal bands are being used at eaCh sampling point by each
participant. The analysis of the interaction is performed by entering
the data into a matrierhich is so constructed that the relative pre-
sence or absence of entries in its various sectors indicates character-
istics of the interaction such as monologue, changes between communica—
tors, etc. Entering the data 2uTuO the analysis matrix thus provides
a characterization of the interaction period as a Whole, although at

the same step it removes the time dimension from:the data.

 

1Brown and Bellugi's work is representative of other longitudinal
studies of language development by such investigators as: Bellugi and
Brown (1969), Erwin (1969), Erwin and Miller (1963), Lenneberg (1966),
and Templin (1966).

209

The research Which has been conducted to the present has been
observational in nature, as indicated in the summary in Table 8, but
there is nothing to prevent the technique ficmubeing used in experi—
mental contexts, as well. In a related study by Frahm and Harrison
(1969), the authors have used the interaction matrices from successful
and unsuccessful doctorspatient interviews in a test of several hy—
potheses on the differences between such interviews. The authors
conclude their study with these comments:

While it is tempting to generalize from this data,

it must be remembered that the amount of data available

to us at this time is very limited. Essentially what

is now needed most is an exhaustive test of the analysis.

What this paper suggests is that VNVIA can be a viable

tool fer quantifying human interaction. It provides an

objective look at the structure of an interaction and can

help to isolate interaction typologies which will lead to
greater insights into the problems of human dyadic

interaction. (1969, pp. 11-12)

In a someWhat more general vein, Harrison concludes his report with
several comments about the value of the VNVIA technique:
Verbal—Nonverbal Interaction Analysis gives us a

new picture of the interview, or of dyadic interaction

generally. It provides a content-free 'x-ray' of the

substructure of the interaction, highlighting the

skeleton of verbal—nonverbal interplay. Like an x—ray,

it doesn't show us many things that we might also find

interesting; but it does provide a new research tool-—

and possibly some theoretic insights. (1969b, p. 18)

The research reported by Harrison incorporates the concept of
process in its measurement of the on-going interaction at frequent
intervals throughout the period under study (basically a time series
design). The adequacy of the three second measurement interval does

not appear to have been examined in the manner considered earlier

(section 9.3.1), although a recent study by Frahm (1970) has shown

210
the sampling scheme to be reliable. Also, as in the case of the Brown
and Bellugi study, the VNVIA procedure of recording data for later
examination is one whiCh permits a type of in-depth sampling and an—
alysis which would not be possible otherwise.

To the present, the analyses WhiCh have been employed on the
sampled data have been, for the most part, analyses whiCh have not
considered the time dimension and.which have therefOre not incorporated
the concept of process. Here again, however, Frahm.(1970) has made
preliminary analyses Which do consider time, and it appears likely that
further analyses of this sort could produce quite valuable information
on the development of an interaction. Questions about the various
ways or patterns by which a given interaction matrix might be produced,
for example, are ones Which can only be answered by considering an
2

interaction event as a process.

Insko (196”).-- In his article "Primacy versus Recency in

 

Persuasion as a Function of the Timing of Arguments and Measures"
(196%), Insko reports the results of experimental research WhiCh
examined certain predictions made by Miller and Campbell (1959) re—
garding the persuasive impact of a communication as related to its
retention. Insko's work differs from.the work previously done by
Miller and Campbell both in his choice of a measure of retention and

in his use of a wider range of intervals between communications and

 

2The VNVIA technique is representative of a number~of other
interaction analysis techniques, particularly Flanders (1967), but
also Chapple (19u8), Ekman (1957), and Scheflen (1966). One
important set of studies very similar to those using VNVIA, but Which
do consider the time dimension in analysis, are those reported in a

211
measures.
It is this latter innovation concerning the timing of measures
which is of interest here. Insko describes his work in the following
way:

In an after-only design u independent variables were
manipulated to test Miller and Campbell's theory of'pri-
macy versus recency in persuasion: time between communica-
tions (none, 2 days, 1 week, or 2 weeks), time between the
second communication and the measures of Opinion and re—
tention (none, 2 days, or 1 week), order of communications
(pro-con or con-pro), and order of measures (Opinion-recall
or recall—opinion). There were 2 dependent variables:
Opinion (measured on a rating scale) and retention (measured
through recall). Confirming Miller and Campbell, the longer
the time interval between 2 communications the greater the
recency effect in both Opinion and recall immediately after
the second communication; and the longer the time elapsed
from the second communication until measurement the less the
recency effect. Contrary to Miller and Campbell's pre—
diction, delayed measurement did not tend to produce primacy
in the case of the groups in which the second communication
followed immediately upon the first. The theoretically
predicted shape of the recency function over time was only
roughly supported. A correlational analysis of the relation
between opinion and retention called into question the
assumption that opinion is a direct function of retention
of message content. (l96u, p. 381)

This abstract, together with the characteristics noted in Table 8,
provide a brief outline of Insko's research.

Because of the complexity of the overall research design,
Insko's analysis of the results is also rather complex and cannot be
extensively considered here. The analysis does pay particular
attention to comparisons of the predicted and observed shapes of the

"retention curves," in this case through the use of trend analvsis

 

recent book by Jaffe and Feldstein (1970). A review of interaction
analysis techniques was presented earlier in section 2.3.2.

212
(see Table 7, page 191). Insko summarizes this particular aspect of
the research as follows:
The predictions concerning the shapes of the opinion

and recall functions over time are only roughly supported.

Many of the curves either are linear or have significant

linear components when they should be quadratic. It

should be noted, however, that the number of cases at anv

one data point is fairly small (each point plots the dif-

ference between two groups of 1M subjects eaCh), and that

the time intervals may not have covered a sufficient range

to show the true shape of the curves. (l96u, p. 390)

This particular experiment incorporates the concept of process
in its use of a process design. As noted in Table 8, the design was
a modified form of the separated group delayed posttest design (see
section M.3.2). Although too complicated to discuss at length, the
design involved the same principle of employing differing time Spans
between stimulus and measurement for different groups, then combining
those results to obtain, in effect, a sequence of measurements over
time. In this particular experiment, the measurement intervals are
not all equal, since measurements effectivelv occurred with no delay,
and 2, 7, lu, and 21 days delay. The adequacy of these intervals was
not examined in the manner considered here (section u.3.1), and Insko
indicates a possible problem in this regard in the second quote.

Although Insko does not say so in his report, it appears that
his choice of design was dictated at least in part by the problems
Which repeated measurement would have produced in the researCh--one of
the main reasons for favoring the separated group delayed posttest
design. Such a design, and the corresponding analysis, both of Which

take the time dimension into account, are by no means easy to employ.

On the other hand, given a prediction involving changes in effects of

213
communication over time, it is clear that no other fOrm of research
would be appropriate. Such research seems especially valuable when
one considers a recent remark by Insko (1970) on subsequent research
which has shown that, contrary to Miller and Campbell's assumptions,
subjects are not "passive" with respect to the stimulus materials be-
tween presentation and measurement—-instead, they actively consider

it during this time span (of. Chronkite, 1969, p. 128).3

Other researCh.-— In addition to this more detailed discussion

 

of three examples of research which incorporate the concept of pro—
cess, it will be helpful to mention very briefly certain other
examples of researCh which also incorporate the concept. Several of
these examples might have been used in the discussion above, but were
bypassed for one reason or another, often because the direct relevance
1x>crmmmmication (as defined here) was someWhat unclear.

One area of research which incorporates the concept of process
is that done on the "extralinguistic" aspects of language behavior,
as fOr example, on the relationships between changes in pitch, in
various non-fluencies, in speech rate, etc., and the expression of
emotional states. Such work generally examines verbal output over
time as an indicator of other behaviors, and may be observational or
experimental. The designs involved and the relationships studied in
such research have been discussed at greater length earlier in this

paper (see page 65).

 

3Insko's work is representative of other work Which has examined
attitude or opinion effects over time, as for example Hovland, Janis,
and Kelley (1953, Ch. 8).

21H

A second research area of some importance is that concerned
with human arousal, particularly as arousal relates to changes in an
individual's environment such as conflict, surprise, etc. The work
of Berlyne (1960, 1965) is especially important in this respect, and
is a primary example. Such research generally uses some measure of
arousal, such as EEG or GSR activity, WhiCh is obtained continuously
or by small—interval sampling during the on—going behavior. Other
recent, and not unrelated work by Greenberg and Graham (1970) has
examined changes in EEG activity during the learning of Speech and
non-speech stimuli. Greenberg's work is of special interest, here,
because of its use of Fourier analysis teChniques in analysing the
EEG record (see page 122; of. Table 7,page 191; and see related work
by Oppenheim, 1970, and Smith, 1967).

Finally, other work WhiCh might have been considered in this
context is that by Lazarsfeld (19mm) using a multipleawave panel
design, that by psychologists such as Barker (1968) and Wright (1967)
working in "ecological psychology," and that by Kivlin, et_al, (1968),
in the area of diffusion. The latter study is of interest in that it
represents a somewhat different content area in communication; how—
ever, it has been bypassed here both because it is an isolated case,
not representative of the larger class of diffusion studies, and be—
cause its approximation of the criteria noted on page 152 appears at

best minimal.

5.2.2 The Examples In Sum
The aim of this Chapter, then, has been to examine briefly

several examples of researCh relevant to human communication Which

215
takes the concept of process into account in ways similar to those
suggested in this paper. The attempt has been to provide both a some-
what less abstract discussion than has previously been possible, and
an answer to several general questions on the nature of research whiCh
incorporates the concept of process. The examples above have provided
fairly specific answers to those questions, and it will be helprl to
examine them once again from a more general point of view.

First of all, it is apparent from the examples that the ap-
proaches considered in Chapters 3 and u are applicable over a fairly
wide range. The processes which may undergo study may be slowly
changing language development, attitude, or diffusion processes,
rapidly changing human interactions, or extremely fast-moving EEG
activity. The events involved may be verbal and/or nonverbal. The
measurement techniques may range from psychophysical to "paper and
pencil." In short, the suggested approaches appear to apply in a
number of content areas of research, and with a number of different
kinds of activity and forms of measurement.

Secondly, it is evident that the form of any research which in—
corporates the concept of process is determined in large measure by
the necessary and sufficient conditions for describing a process (see
page 152, and section u.3). Research Which studies events as processes
must utilize some form of process design, those which.have mentioned
(section H.3.2) being only a few of the possibilities. If the results
of the research are eventually to undergo semantic interpretation in
terms of the concept of process, it is apparent, too, that the analysis

techniques employed must be capable of dealing with the dimension of

216
time in events (section u.u).

Third, and finally, the results provided by researCh Which has
at all steps taken the concept of process into account are evidence
and semantic interpretation on how events change over time. Other
evidence and interpretation may result from a given set of data, but
those relevant to the concept of process or of change over time can
come only from."process researchf' Interpretations in terms of
"process" may include, for example, those on the growth or history,
on the behavior or function, and on the statics or structurem of a
system or event. Information on structurem can be obtained, in many
cases, from other forms of research, but information on function and
history (as these terms are defined here) can come only from research
like that described in this paper. That such information is of value
irltumemlcrmmmmication research is, hopefully, an understatement (cf.
Krippendorff, 1969b). Again, information on how communication events
change over time is not the only type of information which is of
value, but in cases Where it is useful, the approaches suggested in
this paper for the explication of the concept of process should be of
some help.

Having examined the three general questions, again, in light
of the examples of research presented in this Chapter, it is appro-
priate to draw together the various aspects of the paper and to seek
an overview. To this end, Chapter 6 will briefly review what has been
accomplished, and will consider the broader implications and overall

place of the concept of process in the study of human communication.

CHAPTER 6

CONCLUSIONS AND IMPLICATIONS

 

In a general sense, the Introduction and Chapters 1 and 2
indicated that the concept of process was an important and usefu1 one
in the conduct of theory building and research in the discipline of
communication. Chapters 3 and H fulfilled the major goal of the paper
in presenting approaches to the explication of the concept, thereby
making it more available as a tool in theory building. Both chapters
examined certain implications of the approaches to explication, as
well. Chapter 5 presented examples of the use of the approaches, so
that it is apprOpriate in Chapter 6 to conclude with an overview and
a consideration of broader implications. We shall do so by attempting
to answer, in effect, the general questions: Where did we go? and

Where does the paper lead us?

6.1 .A Review

One means of considering the general question "Where did we go?"
is to reconstruct the basic argument of the paper and to trace its
development. At the outset, the paper identifed three main concerns,
in particular, with time as a key dimension in events, with the
scientist's approaOh to such events through theory building, and with
the universe of discourse identified by the concept "communication."

All three of these concerns are unified in the study of the concept of

217

218

process as it relates to the discipline of communication, thereby
making the major emphasis of the paper one of seeking, as a step in
theory building in communication, the means of explicating the concept
of process (section 1.1).

The definition of communication as "a process of transmission
of structure among the parts of a systemlwhich are identifiable in
time and space" (section 1.2) appears both general and usefu1 in the

context of this paper, even though the paper emphasizes human commun-

 

ication (section 1.3). The concept of process is a significant aspect
of this definition, and if distinguished from.a number of related con-
cepts, may be defined as "All change over time of'matteruenergy or
information" (section 2.1). The concept is an important one not only
because of its relationship to research designs (point, difference,
and process designs; section 2.1.3), but also because of its particular
utility both in solving general problems in research and as a key con-
cept in new concerns in human communication researCh (section 2.2).
However, despite this importance and utility, an examination of
the current place of the concept of process in researCh reveals that
its Operationalization and use have been small. In other‘words,a1-
though the concept is regarded as important and appears particularly
useful in human communication research, it has not been widely applied
in such research. An explication of the concept appears needed to
correct this situation and to make the concept more available as a
tool in the conduct of researCh and theory building in the discipline

(section 2.3).

219

The first step in explicating the concept of process is to
provide a constitutive definition. However, because no specific theory
is involved, it is necessary to provide not a specific definition but
a set of approaches to constitutive definition (section 3.1). SuCh
approaches may involve either verbal terms (section 3.2) or mathe—
matical terms (section 3.3), so that the necessary and sufficient con—
ditions fOr using the term ”process" must be formulated with both as-
pects in view (section 3.3.H). In addition, because theory is fOrmed
using verbal and mathematical tools, the approaches to constitutive
definition have a number of implications fOr the form of theory WhiCh
deals with communication events as processes (section 3.”).1

The explication of a concept is not complete without the second
step of providing an operational definition. Here too, the absence of
a specific research situation requires the development of a set of
approaChes to operational definition (section H.l). Examination of
these approaches or general principles fOr operationally defining the
concept of process results directly in a set of necessary and sufficient
conditions fOr describing a process (section u.2). These conditions
have several rather direct implications for research teChniques (i.e.,
design and measurement; section H.3), as well certain indirect impli—
cations for analysis techniques (i.e., reduction and testing; section

u.u).

 

1It is impossible to be more specific about these implications
in the space of this chapter; see the section indicated for more
infOrmation. The same is true for the implications noted in the next
two paragraphs.

220

In a general sense, then, Chapters 3 and H perfOrmlthe task set
out by Chapter 1 and 2, namely, that of seeking, as a step in theory
building in communication, the means of explicating the concept of
process. One of the results of seeking approaches to the constitutive
and the operational definition of "process" is a set of potentially
useful tools fOr dealing with the concept in specific theoretical and
research frameworks (as exemplified in Chapter 5). This latter result
stems directly from the result already mentioned, i.e. the set of
implications fOr theory building and.researCh related to human commun-
ication. Rather than reiterate here the many specific implications
which have been noted previously, it will be helpful in review to
fOrmulate a single, mmch.broader implication. That is, if theory
building and research on.human communication are to deal with events
as processes, it will be necessary to use and to develop not only ap-
propriate fOrms of theory, but also research and analysis teChniques
capable of dealing with the dimension of time. This broader implica—
tion, together with certain others Which stem from it, comprise the
concern of this chapter.

6.2 Implications and Place of "Process" in the Study of Human
CommunicatiOn

 

 

The implication considered above is one very general answer to
the question: "Where does the paper lead us?" A number of othervmore
specific answers are of considerable importance, as well, and an exam—
ination of several of themlshould not only indicate certain additional
implications of the concept of process, but also make clearer‘its place

in the conduct of theory building and research related to human

221
communication. The discussion will provide an opportunity, as well,
to consider once meme the various limits on the scope of this paper.

In particular, it has been noted at several points in the paper
that What is said.here may not apply to all theory and research related
to human communication, as it is evident that many studies do not re-
quire the consideration or use of the concept of process. Indeed, it
is quite clear that much valuable work has been done WhiCh has not
incorporated the concept. On the other~hand, it should be equally
clear that theory and research which does consider and utilize the
concept of process is valuable and vital, even though it is only occa—
sionally fOrmulated or perfOrmed. That is, although studies whiCh
incorporate the concept have been recognized as necessary in the dis—
cipline of human communication (see sections 2.1.1 and 2.3.3), neither
theory nor researCh in the ftmms considered here has been especially
evident (see sections 3.u, n.3, u.u). It is only as such theory and
research begin to appear that a potentially large and important class
of infOrmation will become available, i.e., information on how commun-
ication behaviors develop, evolve, and interact during an event, or in
short, on how they change over time.

Many of the tools needed to perfOrmlstudies such as those
suggested above are currently available, as noted in Chapters 3 and u.
Such tools have not been actually applied here (except fOr the examples
in Chapter 5), because there is no attempt in this paper to build
theory or to conduct researCh. As mentioned earlier, these tools are
not presented as replacements fOr existing theoretic and research tools,

but rather as potentially useful additions to them. Note particularly

222
that the present availability of a number of tools for dealing with
the concept of process in specific theory-research frameworks should
not be taken as an indication that those tools are complete. Indeed,
much the opposite is true, for the existing tools are in many cases
only starting points which require further development. The nature
of such development has been indicated with regard to theory, research,
and analysis (sections 3.u, n.3, u.u),2 and it is clear that scholars
in the discipline of communication will have to share in this develop—
ment (both in the challenges and in the headaches) if they wish a
continued ability to interpret the results of their studies in terms
of the concept of process (of. Krippendorff, 1969b, p. 35).

If scholars do make use of the existing tools and begin the
development of new tools, the body of research which develops will
not exist as an entity distinct from present research, but instead
will be an extension of it. That is, research which operationalizes
the concept of process can be seen as an extension of present research
in the same respect that a process design can be seen as an extension
of a difference design through the addition of another observation
point (see section 2.1.3). In this sense, research which operation—
alizes "process" is a refinement of current research, capable of
collecting new or additional information. It is also possible, of

course, to take the "opposite" point of view which sees current

 

21n slightly more specific terms, development is needed in ana—
logic (and symbolic) forms of theory, in research techniques involving
both design and measurement, and in analysis techniques for reducing
and testing research information. Any further indication of needed
development is beyond the scope of this chapter; see sections 3.1+, n.3,
and HA, for further information.

223
research as comprised of the special "point" and "difference" cases
of a more general "process" research framework. But the question of
whether or not this view is appropriate is one Which can only be
answered as the concept of process is increasingly operationalized in
communication research.

Increased operationalization of the concept of process will
present researchers in human communication with at least one problem
WhiCh they have not often confronted in the past (Sidman, 1960, p. 5”).
Specifically, they will have to consider in more depth the question of
Whether to fOcus their research on individual behavior'or on group
behavior. At first glance, such a decision may appear trivial, being
"merely" a matter of choosing whether to consider individual values or
some composite value such as a mean. However, the choice of research
focus is not this simple, particularly when one is involved in the
study events as processes, and especially if those events are complex.
That is, Sidman (1960, pp. us—su) has pointed out that a behavioral
process fOr a group may'have no counterpart in the same behavioral
process fOr an individual, and that in some cases, ". . . individual
and group curves simply cannot provide the same information, even if
their fOrms are identical (p. 53)."

A fOcus on individual behavior in research will thus be likely
to supply infOrmation which differs in important ways fromtthe infOr-
mation Obtained if'group behavior is the primary fOcus. The decision
on WhiCh of these two fOci to Choose is closely tied to the purpose
for Which the research is being conducted, but the matter is even more

complex. That is, the decision is affected, as well, both by

221+
considerations of generality and variability (Berlo, 1967, pp. 13—15;
Sidman, 1960, Chs. 5, 6), and by considerations regarding measurement
(see sections u.3.3, I51.4.1). The need to make a decision on the
individual or group focus in research has been encountered in the past
in studies of human interaction and of language development, and
promises to be a decision which is increasingly necessary as the con-
cept of process continues to be operationalized in research.

In view of the emphasis which has been placed on the concept
of process in the paper as a whole, it is important to note again that
this concept is not the only one of importance in research and theory
related to human communication. The concept has been purposely
narrowed from its broader usage , and the additional concepts which
were introduced in so doing must not be overlooked. In particular,
interaction and emergence are also important concepts in the study of
human communication,3 and in many respects need explication in a
manner similar to that employed here . The concept of process was
chosen from among these concepts and others entirely because the limits
of space dictated a choice of the single most important and useful con-
cept. Indeed, it is clear that a fuller understanding Of human com—
munication events will come only as these various concepts begin to be

considered together in research.

 

3This choice of additional concepts is not meant to disparage
the importance of the other concepts introduced in sections 1. 2 and
2 . 1. However, interaction and emergence are perhaps the most impor-
tant of the various concepts , with respect to the conduct of theory
building and research relevant to communication (see page l+0).

225

This recognition of the necessity of incorporating and Opera-
tionalizing a number of different concepts in the conduct of theory
building and researCh can be seen, too, in Krippendorff's recent and
important comments on communication research (1969b). An examination
of Krippendorff's requirements fOr~what he terms "elementary communica-
tion data" reveals that not only must the concepts of process, inter-
action, and emergence be incorporated in theory and operationalized in
research, but the concept of systemlmust be considered, as well (1969b,
pp. 23-33; cf. pages l6-l8).u Indeed, unless these concepts and others
are incorporated in theory and especially in research, Krippendorff'has
noted the resulting data will probably not produce evidence that is
capable of semantic interpretation in terms of the concept of commun—
ication (1969b, pp. 2-8, 3H-36).

Such a situation means that to be adequate, research on human
communication must become a considerably more sophisticated endeavor
than it is at present, not only in operationalzing the concept of
process as considered in this paper, but also in operationalizing the
related concepts Which cannot be considered here. At the same time,
theory too must become more sophisticated if it is to be useful in
adequately explaining, predicting, and/or controlling events (see page
6). Theory must incorporate the concept of process, together'with the

related concepts, and in this respect is likely to become someWhat

 

“Again, these four concepts are not the only ones Which must be
considered in dealing with communication events, as is clear in sections
1.2 and 2.1. HOwever, fOr the present discussion these four>are the
most important.

226

more "complex" than it is at present—-a potentiality Which has already
been recognized (Carroll, 1968, pp. 2-10; Jenkins, 1970). There is no
implicit assumption here that complex behavior, e.g. , human communica-
tion, requires complex theory, but it is interesting to note by way of
analogy that the theory of relativity, often held as a paragon of
economy in explanation and prediction, is actually an exceedingly
complex and intricate structure.

This potential complexity in theory and research, the new
problems to be encountered and considered in conducting studies, and
the difficulties in dealing with the concept of process itself, all
point to one end: there is no panacea in studies of human communica-
tion. This paper has attempted to isolate one fOrm of infOrmation
which appears vital, yet is lacking in such studies, i.e., infOrma-
tion on communication events as processes , and has sought to provide
a basis fOr theory and research which will provide that infOrmation.
In so doing it has uncovered many more problems than it has solved,
perhaps because the suggested approaches will change not only the
ways in which communication events are studied, but also the very
content of those studies.

The potential changes should prove highly beneficial to the
study of human.communication, however, both in improving explanation
and prediction of what happens over time when humans communicate, and
in opening a door to two areas of theory and research WhiCh scholars
have seen as important in communication, but also as unuseable until
now, namely, system theory and analysis (Carroll, 1968; Harrison,

1967a; of. Buckley, 1967, p. 39) and cybernetics (Krippendorff,

227
1969a, pp. 117—118, 129—132). In general, then, the concept of
process has an important place in human communication research, and
it is apparent why "MOst Soholars engaged in the study of human com-
munication will agree with Berlo (1960) in Choosing the phrase, 'the

process of communication,' to describe the fOcal point of their

 

discipline."

LIST OF REFERENCES

. I l i Ill.l’|0|l

LIST OF REFERENCES

Ackoff, R. L. Towards a behavioral theory of communication. Mgmt.
Sci., 1958, u, 218-23u.

Ackoff, R. L., Gupta, S. K., and Minas, J. S. Scientific method:
pptimizing applied research decisions. New York: WiIey, 1962.

 

Arundale, R. B. The diversity in systems approaches: implications
for the study of communication. Paper contributed to National
Society fOr the Study of Communication, New York, April, 1968.

Ashby, W. R. Ap_introduction Ep'cybernetics. New York: Wiley, 1963.

 

 

Ashby, W. R. Constraint analysis of many-dimensional relations.
General Systems, 1969, 9, 99-105.

 

Ashby, W. R. Principles of the self-organizing system. In W. Buckley
(Ed.), MOdern systems research fOr‘the behavioral scientist.
Chicago: Alaine, 1968. Pp.1083118.

 

 

Bales, R. Interaction ppocess analysis. Cambridge, Mass.: Addison-
Wesley,71950.

 

Barker, R. G. Ecological psychology: concepts and methods for
studying the environment pf human behavior. StanfOrd, Calif.:
StanfOrd UniVersity Press, 1968.

 

 

 

Barnlund, D. C. Toward a meaning centered philosophy of communica-
tion. g, Communication, 1962, 12, 197-211.

 

Barnlund, D. C. Interpersonal communication: survey and studies.
Boston: Houghton-Mifflin, 1968.

 

 

Bartlett, M. S. An introduction :p_stochastic processes. Cambridge,
Eng.: Cambridge University Press, 1966.

 

 

BeaCh, E. F. Economic models: ap_exposition. New York: Wiley, 1957.

 

 

Bellman, R. Functional equations. In R. D. Luce, R. R. Bush, and
E. Galanter (Eds.), Handbook pf_mathematical psychology.
Vol. 3. New York: Wiley, 1965. Pp. “87-513.

 

 

Bellugi, Ursula, and Brown, R. (Eds.) The acquisition pf_langpage.
Lafayette, Ind.: Society for Research in Child Development,
196”.

 

228

229

Berelson, B. The state of communication research. Publ. Opin.
Quart., 1959, 23, 1—17.

 

Bergland, G. D. A guided tour of the fast Fourier transform, IEEE
Spectrum, 6 (July, 1969), u1—52.

Berlo, D. K. The pgocess p£_communication. New York: Holt, Rinehart,
and Winston, 1960.

 

 

Berlo, D. K. Hypothesis testing and the goal of social science.
Unpublished manuscript, Department of Communication, Michigan
State University, 1967.

Berlo, D. K. Graduate education in communication: a call for a
community of scholars. Unpublished manuscript, Department of
Communication, Michigan State University, 1969.

Berlyne, D. E. Conflict, arousal, and curiosity. New York:
MCGraw—HiII, I960.

 

Berlyne, D. E. Structure and direction ip_thinking. New York:
Wiley, 1965. "‘""'

 

Bertalanffy, L. General system theory—-a critical review. General
Systems, 1962, 7, l—20.

Bertalanffy, L. General system theory. New York: Braziller, 1968.

 

Bettinghaus, E. P. Communication models. In J. Ball and F. C.
Byrnes (Eds.), ResearCh, principles, and practices in visual

communication. Washington, D. C.: National Education.
Association, 1960. Pp. 16—28.

 

 

 

Bock, R. D. Multivariate analysis of variance of repeated measure—
ments. In C. W. Harris (Ed.), Problems in measurin chan e.
Madison, Wis.: The University of Wisconsiﬁ'PEESs, 1967.

Pp. 85—121.

Bohrnstedt, G. W. Observations on the measurement of change. In
E. F. Borgotta (Ed.), Sociological methodology: 1969. San
Francisco: Jossey-Bass, 1968. Pp. 113-133.

 

Borden, G. A. Computer simulation of verbal communication. Paper
presented at International Communication Association,
Minneapolis, May, 1970.

Bracewell, R. The Fourier transtrmlapd ips applications. New York:
McGraw-Hill, 1965.

 

230

Brodbeck, May. Models, meanings, and theories. In L. Gross (Ed.),
S sium pp_sociological theory. Evanston, Ill., Row,
Peterson and CO., 1959. Pp. 373—903.

 

Brown, R., and Bellugi, Ursula. Three processes in the child's
acquisition of syntax. In E. H. Lenneberg (Ed.), New

directions ip_the stu of langpage. Cambridge, Maggi: NET
Press, 1964. P57. 3 — 61:-

 

Bryne, E. R. (Ed.) PrOposed definitions: systems science. Newsletter
of the Systems Science and Cybernetics Gro : Institute of’
Electrical and Electronics Engineers, AprrI, 1967. FE. 5:6.

 

 

 

 

Buckley, W. Sociolo and modern systems theory. Englewood Cliffs,
N.J.: Prentice—Hall, 1967}

Buckley, W. (Ed.) Modern systems research for the behavioral scientist.
Chicago: Aldine , T968 .

 

Campbell, D. T., and Stanley, J. C. E erimental and gpasieexperimental
desrgns for research. ChicagozRggd-MENally, 1963.

 

Campbell, J. The life cycle of a communication systemu a speculation,
frcmlSystems, models, and communication. Pa r presented to
National Society for the Study of Communication, Cleveland,
April, 1969.

Carroll, T. W. , and Farace, R. V. Systems analysis, computer simula-
tion, and survey research: applications to social research in
developing countries. Unpublished manuscript, Department of
Communication, Michigan State University, 1968. .Also in G. D.
Hursh (Ed.) Survey researCh in developing nations. New York:
Holt, Rinehart, and Winston,—ih press.

 

 

Cassirer, E. Ap_essay ppoan. New Haven, Conn.: Yale University
Press, 194”.

Chapple, E. D. The interaction chronograph: its evolution and present
application. Personnel, 19u8—u9, 25, 295—307.

Cherry, C. Qp_human communication. New York: Wiley, 1957.

 

Chorley, R. J., and Haggett, P. Trend—surface mapping in geographical
research. In B. J. L. Berry, and D. F. Marble (Eds.), Spatial
analysis. Englewood Cliffs, N.J.: Prentice-Hall, 1968.

o “2170

Chronkite, G. L. Out of the ivory palaces: a proposal fOr useful
research in communication and decision. In R.J. Kibler and
L. L. Barker (Eds.), Conceptual Frontiers in S eech—Communica—
tion. New York: Speech Association of ArneFfoa'J,LT—969". "W . 3435.

 

231

Clarke, D. L. Analytical archaeolog. London: Methuen, 1968.

 

Coombs, C. H. Theory and methods of social measurement. In L.
Festinger and D. Katz (Eds.), Research methods in the behavioral
sciences. New York: Holt, Rinehart, and Winstoh? I953.
Pp.'E7T3535.

 

Coombs, C. H. A theory pf data. New York: Wiley, 1961+.

 

Cox, D. R. Flaming 9}: experiments. New York: Wiley, 1958.

 

Cuénod, N., and Durling, A. A discrete time approach for systems
analysis. New York: Academic Press, 1969.

 

Deutsch, K. W. On communication models in the social sciences. Publ.
Opin. Quart., 1952, 16, 356-380.

 

Deutschmann, P. J. The sign—situation classification of human com-
munication. J. Communication, 1957, 7, 63-73.

 

Deutschmann, P. J. and Danielson, W. A. Diffusion of knowledge of the
major news story. Journ. Quart., 1960, 37, 3u5—355.

 

Dubin, R. TheorLbuilding. New York: The Free Press, 1969.

 

Ekman, P. A methodological discussion of nonverbal behavior. J.
Psychol., 1957, us, 1u1-1u9.

Ekman, P., Friesen, W. V., and Taussig, T. G. VID—R and SCAN: tools
and methods for the automated analysis of visual records. In
G. Gerbner, O. R. Holsti, K. Krippendorff, W. J. Paisley, and
P. J. Stone (Eds. ) , The analysis of communication content.
New York: Wiley, 1969. Pp. 297—31—2.

 

 

Erwin, Susan M. Imitation and structural change in children's
language. In E. H. Lenneberg (Ed.), New directions in the
stug pr: lanﬂge. Cambridge, Mass.:-WTTress, 1961?
Pp. 163-189.

 

Erwin, Susan M., and Miller, W. R. Language development. In H. W.
Stevenson (Ed.), Child psychology. Yearb. nat. Soc. Stud.
Educ., 1963, 62, Part I. Pp. 108—1u3.

 

 

Fararo, T. J. Stochastic Processes. In E. F. Borgotta (Ed.),
Sociological methodology: 1969. San Francisco: Jossey-Bass,
1968. Pp. 245-260.

 

Fearing, F. Toward a psychological theory of human communication.
g; Pers., 1953, 22, 71—88.

232

Ferguson, G. A. Nonparametric trend analysis. Montreal: McGill
University Press, 1965.

 

Flanders, N. A. Teacher influence in the classroom. In E. J. Amidon
and J. B. Hough (Eds.), Interaction analysis: theOry, research,
and application. Reading, Mass.: Addison—Wesley, 1967.

 

Frahm, J. H. Interaction analysis. Unpublished manuscript, Department
of Communication, Michigan State University, 1969.

Frahm, J. H. Verbal—nonverbal interaction analysis: exploring a new
methodology for quantifying dyadic communication systems .
Unpublished doctoral dissertation, Michigan State University,
1970.

Frahm, J. H. , and Harrison, R. P. Verbal-nonverbal interaction
analysis: examining dyadic communication. Paper presented at
Speech Association of America, New York, December, 1969.

Gabor, D. Theory of communication. J. 31: the Institution _o_f_
Electrical Engineers, 1946, 93, Part III. Pp. 1429-991.

 

 

Gaito, J. , and Wiley, D. E. Univariate analysis of variance pro-
cedures in the measurement of change. In C. W. Harris (Ed.),

Problems in measurin chan e, Madison, Wis.: The University of
Wisconsm_Press, . 5. 60-814.

Gelbaum, B. R., and March, J. G. Mathematics for the social and
behavioral sciences. PhiladeIphia: W. B. Saunders Co. , 1969.

 

 

Gerard, R. W. (Ed.) Concepts of biology. Behavioral Sci., 1958, 3,
92—209.

 

Gerard, R. W. Units and concepts of biology. In W. Buckley (Ed.),
Modern pystems research for the behavioral scientist. Chicago:
Kldine, 1968. Pp. 51—58.

 

Gibson, J. J. The problem of temporal order in stimulation and
perception. J. Psychol., 1966, 62, 1u1—1u9.

Goldberg, S. Introduction ’_c_o_ difference equations. Englewood Cliffs,
N.J.: Prentice—Hall, 1958.

 

 

Greenberg, H. J. , and Graham, J. T. Electroencephalographic changes
during learning of speech and non—speech stimuli. J. Verb. learn.
Verb. Beh., 1970, 9, 27u-281.

 

 

233

Gullahorn, J. T., and Gullahorn, J. E. The computer as a tool for

theory development. In D. Hymes (Ed.), The uses of computers
ip_anthropology. The Hague: Mouton, 1965. Pp. M25LEHB.

 

Hall, A. D., and Fagen, R. E. Definition of a system. In.W. Buckley
(Ed.), Modern systems researCh fOr the behavioral scientist.
Chicago: Aldine, 1968. Pp. 81—92.

 

Hamming, R. W. Numerical methods fOr scientists and engineers.
New York: McGraw-Hill, 1962.

 

Hannan, E. J. Time series analysis. London: Methuen, 1960.

 

Harris, C. W. (Ed.) Problems ip_measuringchange. Madison, Wis.:
The University of Wisconsin Press, 1967.

 

Harrison, R. P. Communication systems analysis: re-modeling man.
Paper presented to Central States SpeeCh Association,
Indianapolis, Ind., April, 1967. (a)

Harrison, R. P. Toward a theory of visual communication. Paper
presented to Association fOr Education in Journalism, Iowa
City, Iowa, August, 1967. (b)

Harrison, R. P. Communication theory. In R. V. Wiman and W. C.
Meierhenry (Eds.), Educational media: theo into ractice.
Columbus, Ohio: Charles E.’Merr111 C ., 1963. Pp. - . (a)

Harrison, R. P. Verbal-nonverbal interaction analysis: the sub—
structure of an interview. Paper presented to Association fOr
Education in Journalism, Berkeley, Calif., August, 1969. (b)

 

Heirich, M. The use of time in the study of social change. Amer.
Sociological Rev., l96u, 29, 386-397.

 

Holtzman, W. H. Statistical models fOr the study of change in the
single case. In C. W. Harris (Ed.), Problems in measuring

chan e. Madison, Wis.: The University of Wiscchsin Press,
I967IL'Pp. 199-211.

 

Hovland, C. I., Janis, I. L., and Kelley, H. H. Communication and
pgrsuasion. New Haven, Conn.: Yale University Press, 1953.

 

 

HOwes, D., and Osgood, C. E. On the combination of associative
probabilities in linguistic contexts. Amer. J, Psychol.,
195m, 67, 2u1—258. “""

Insko, C. A. Primacy vs. recency in persuasion as a function of the
timing of arguments and measures. J, abnorm. soc. Psychol.,
1969, 69. Pp. 381—391.

 

23H

Insko, C..A. Relation of cognitive and.memorial processes to
persuasion. Address to Seminar in Social Psychology,
Department of PsyChology, Michigan State University, May 19,
1970.

Jaffe, J., and Feldstein, S. Rhythms of dialogue. New York:
Academic Press, 1970. '__

Jenkins, J. Pilgrim's progress: models I have tried and found wanting.
Address to International Communication Association, Minneapolis,
Minn., May 9, 1970.

Jordan, N., Jensen, B. T., and Terelinski, S. J. The development of
cooperation among 3-man crews in a simulated man—machine infor-
mation processing system. J: soc. Psychol., 1963, 59, 175-18”.

 

MMLL.M(MJ %ﬁmnmmmﬁmm W$hgm,DCuTm®mn
Books, 1968.

 

Kaplan, A. The conduct of inquiry. San Francisco: Chandler Publishing
Co., I969. '__

Katzer, J. A.theoretical model of human language processing. Un-
published doctoral dissertation, Michigan State University,
1970.

Kelly, G. A. A theopy pf personality. New York: W. W. Norton, 1963.

 

Kibler, R. J., and Barker, L. L. (Eds.) Conceptual frontiers in
spgech-communication. NEW'YOFk: SpeeCh AssociatiOn of'America,
196 .

 

 

King, L. J. Statistical analysis in geography. Englewood Cliffs,
N.J.: Prentice:Hall, 1969. ——.

 

Kirk, R. E. Experimental design: Eocedures for the behavioral
sciences. *Belmont, Calif.: BroOks-COle PublishingICo., 1968.

 

 

 

Kivlin, J. E., Roy, P., Fliegel, F. C., and Sen, L. K. Communication
ip_India: experiments ip_introducipg change. Hyderbad,—India:
NatiOnal Institute Of Community Development, 1968.

 

 

 

 

Klir, J. Processing of general systems activity. General Systems,
1967, 12, 193-198.

 

Koenig, H. E. Mathematical models of socioeconomic systems: an
example. IEEE Trans. pp_Systems Science and Cybernetics,
1965, l, 91:95.

 

 

KOenig, H. E., Tbkad, Y., and Kesavan, H. K. Analysis pf discrete
physical systems. New YOrk: MCGraw-Hill, I967.

 

235

Krippendorff, K. Values, modes, and domains of inquiry into com-
munication. J. Communication, 1969, 19, 105-133. (a)

 

Krippendorff , K. On generating data in communication research .
Paper presented to National Society for the Study of Communica-
tion, Cleveland, Ohio, April, 1969, and revised November, 1969.
(b) Also in J. Communication, 1970, 20, 2ul—269. (Note that
adding the quantity 236 1 3 to the page references in this
paper will closely approximate the pagination of the J.
Communication version.)

 

 

Kuhn, T. S. The structure 9: scientific revolutions. Chicago:
University of Chicago Press, 1962.

 

 

Lasswell, H. D. The structure and function of communications in
society. In L. Bryson (Ed.), The communication of ideas.
New York: Harper and Brothers, I998? Pp. BfSlf—

 

 

Lazarsfeld, P. F. The use of panels in social research. Proc. 91: the
Amer. Philosophical Soc., 1998, 92, L105-1410.

 

 

Lazarsfeld, P. P., Berelson, 13., and Gaudet, H. The Eople's choice.
New York: Columbia University Press, 1991+.

 

Lazarsfeld, P. P., and Fiske, M. The "panel" as a new tool for
measuring opinion. Publ. Opin. Quart., 1938, 2, 596-612.

 

 

Lennard, H. L., and Bernstein, A. Patterns 12 human interaction.
San Francisco: Jossey-Bass, 1969.

 

Lenneberg, E. H. The natural history of language. In F. Smith and
G. A. Miller (Eds.), The enesis of langge. Cambridge,
Mass.: MIT Press, 1966. . -25'2.

Lindquist, F. E. Design and analysis of experiments 11:— psychology
and education. Bostonzi-Iougnton biffl'in, 1953.

 

 

Lindsley, O. R. Direct behavioral analysis of psychotherapy sessions
by conjugately programmed closed-circuit television.
Psychother.:'Iher., Res., and Pract., 1969, 6, 71—81.

 

Lord, F. M. Elementary models for measuring change. In C. W. Harris

(Ed.), Problems _1p measurin change. Madison, Wis.: The
University of Wisconsm Press, I967. Pp. 21-38.

Lowry, I. S. A short course in model design. In B. J. L. Berry and
D. F. Marble (Eds.), Spatial analysis. Englewood Cliffs, N.J.:
Prentice—Hall, 1968. Pp. 53-69.

 

Luce, R. D., Bush, R. R., and Galanter, E. HandboOk 9_f_ mathematical
psychology. New York: Wiley, 1963, 2 vols.; 1965, Ivol.

 

 

236

Mahl, G. F. The lexical and linguistic levels in the expression of
the emotions. In P. Knapp (Ed. ) , Expression of the emotions
in man. New York: International Universitwar'e-ss, 1963.
Ppl'77L105.

 

 

Mahl, G. F., and Schulze, G. Psychological research in the extra—
linguistic area. In T. A. Sebeok, A. S. Hayes, and M. C.
Bateson (Eds.), ApprOaches to semiotics. The Hague: Mouton,
196M. Pp. 51—129. """“""'

 

Maruyama, M. The second cybernetics: deviation—amplifying mutual
causal processes. Amer. Scientist, 1963, 51, 169-179.

 

Mason, C. J. W. Pattern recognition bibliography. In E. R. Bryne
(Ed.), Newsletter of the Systems Science and Cybernetics Group:
Institute pf Electricafand EIectronics Engineers , Ucﬁber,
19763‘ﬁeoember, 1970.

 

 

 

McNemar, Q. Psychological statistics. New York: Wiley, 1962.

 

Miller, G. A. Language and communication. New York: McGraw-Hill,
1951.

 

Miller, G. A. The magical number seven, plus or minus two: some
limits on our capacity for processing information. Psychol.
Rev., 1956, 63, 81-97.

Miller, G. R. Operationism in communication research. Unpublished
manuscript, Department of Communication, Michigan State
University, 1967.

Miller, G. R. Human information processing: some research guidelines.
In R. J. Kibler, and L. L. Barker (Eds.), Conceptual frontiers
in s eech—communication. New York: Speech Association of
A-m-erica, 969. Pp. 51-68.

 

 

Miller, J. G. Living systems: basic concepts. Behavioral Sci.,
1965, 10, 193-237. (a)

 

Miller, J. G. living systems: structure and process. Behavioral Sci.
1965, 10, 337-379. (b)

 

Miller, J. G. Living systems: cross-level hypotheses. Behavioral
Sci., 1965, 10, 380—911. (c)

 

Miller, N. , and Campbell, D. T. Recency and primacy in persuasion as
a function of the timing of speeches and measurement. J.
abnorm. soc. ngchol., 1959, 59, 1-9.

 

Milsurn, J. H. Biological control systems analysis. New York:
McGraw-Hill, 1966.

 

237

MOrris, C. D. J. InfOrmation as a construct in human communication
systems. Unpublished doctoral dissertation, Michigan State
university, 1968.

Oppenheim,.A. V. Speech spectrograms using the fast FOurier trans-
fOrmu IEEE Spectrum, 7 (August, 1970), 57-62.

Osgood, C. E., and Sebeok, T. A. (Eds.) Ps Cholin 'stics: §_survey
pf_theopy_and research problems, wnEE ﬁleBO%§, A. R., Jr.
surve of'psyCholingpistic research, 1959-1969. Bloomington,

.A
In .: ndia'na niversity ss,

 

 

Parzen, E. Stochastic pgocesses. San Francisco: Holden-Day, 1962.

 

Pierce, J. R. Symbols, signals, and noise. New York: Harper and
Row, 1961.

 

Platt, J., and Miller, J. G. Handling infOrmation overload.
Ekistics, 1969, 26, 295-296.

Quenouille, M. H. Associated measurements. London: Butterworths
Scientific PublicatiOns,’l952.

 

Quenouille, M. H. The anal sis of multiple time series. London:
Charles Griffiﬁ and 0., I957.

 

Rappoport, A. Mathematical aspects of general systems analysis.
General Systems, 1966, 11, 3-11.

 

Rappoport, A. Foreward. In W. Buckley (Ed.), MOdern s stems research
for the behavioral scientist. Chicago: Aldrne, i968.
Pp. xiii-xxii.

 

ReichenbaCh, H. The direction pf_time. Berkeley, Calif.: University
of CalifOrniE Press, 1956.

 

Reichenbach, H. The philospphy pflppace and time. New York: Dover,
1958.

 

 

Rogers, E. M. Mass communication and the diffusion of innovations:
conceptual convergence of two research traditions. Paper
presented to Association fOr Education in Journalism, Boulder,
Col., August, 1967.

Rogers, E. M. with Shoemaker, F. F. Communication _o_f_ innovations: g
cross-cultural approach. New Ycrk: Free Press of Glencoe,
fOrthcoming.

 

 

 

238

Rogers, E. M., and Stanfield, J. D. Adoption and diffusion of new
products: emerging generalizations and hypotheses . Paper
presented to Conference on the Application of the Sciences to
Marketing Management, Lafayette, Ind., July, 1966.

 

Rudner, R. S. Philoso hy of 'social science. Englewood Cliffs, N.J.:
easement—ime?"

Sarbaugh, L. E. Applying communication principles to international
communication. Paper presented to National Society for the
Study of Communication, Boulder, Col. , August, 1968.

Scheflen, A. E. Natural history method in psychotherapy: communication—
al research. In L. A. Gottschalk and A. H. Auerbach (Eds.),
Methods g research in_ Eychotherapy. New York: Appleton-
Century-Crofts, 1966. Pp. 263—289.

 

Scheflen, A. E. Human communication: behavioral programs and their
integration in interaction. Behavioral Sci. , 1968, 13, 99—55.

 

Shannon, C. E. A mathematical theory of communication. Bell System
Tech. J., 1998, 27, 3'47-923; 623-656.

 

Shannon, C. E., and McCarthy, J. (Eds.) Automata studies. Princeton,
N.J.: Princeton University Press, 1956.

 

Shannon, C. E. , and Weaver, W. The mathematical theory 9: communica—
tion. Urbana, I11.: The University of I111n01s Press, I963.

 

 

Sidman, M. Tactics 91: scientific research. New York: Basic Books,
1960.

 

Siegel, S. Non-parametric statistics for the behavioral sciences.
New York: McGraw-Hill, 1956.

 

Smith, F., and Miller, G. A. (Eds.) The genesis 31: language: 3 psy—
cholinguistic gpproach. Cambridge, Mass.: MITI Press, 1966.

 

 

Smith, S. Prominent american ghosts. Cleveland, Ohio: World
Publishing Co. , 1967.

 

Snell, J. L. Stochastic processes. In R. D. Luce, R. R. Bush, and
E. Galanter (Eds.), Handbook 9: mathematical psychology. Vol. 3.
New York: Wiley, 1965. F5. L#114485.

 

Spiegel, M. R. A lied differential ﬂuations. Englewood Cliffs,
N.J.: Prentice-Hall, 1958.

 

239

Stanfield, J. D., Lin, N., and Rogers, E. M. Simulation of innova—
tion diffusion. AID Diffusion Project Working Paper N'uTrber
'7, Department of Communication, Michigan State University,
1965.

 

Star, S. A., and Hughes, J. A. Report on an educational campaign:
the Cincinnati plan for the United Nations. Amer. J.
Sociology, 1950, 55, 389—900.

Starkweather, J. A. Measurement methods for vocal information. In
G. Gerbner, O. R. Holsti, K. Krippendorff, W. J. Paisley,
and P. J. Stone (Eds. ), The analysis of communication content.
New York: Wiley, 1969. P5.— 3I'373I7— _

 

Stevens, S. A definition of communication. J. acoust. Soc. Amer.,
1950, 22, 689-690.

 

Taylor, W. L. "Cloze procedure": a new tool for measuring readability.
Journ. Quart., 1953, 30, 1415-933.

 

Taylor, W. L. Recent develOpments in the use of cloze procedure.
Journ. Quart., 1956, 33, L+2448.

 

Templin, M. C. The study of articulation and language development
during the early school years. In F. Smith and G. A. Miller
(Eds.), The genesis of language: a psycholinguistic approach.
Cambridge, Mass.: MIT—Press, 1966-.- Pp. 173—180.

 

 

Thomas, G. B. Calculus and analytic geometry. Reading, Mass.:
Addison—Wesley, 1953.

Thompson, D' A. W. nggowth and form. New York: Macmillan, 191+2.

 

Troldahl, V. C. Using statistics. In G. D. Hursh (Ed.), Survey
research _ip developing nations. New York: Holt, Rinehart,
and Winston, in press.

Uhr, L. Pattern recognition. In L. Uhr (Ed.), Pattern recognition.
New York: Wiley, 1966. Pp. 365—381.

 

Van Wylen, G. J. Thermodynamics. New York: Wiley, 1959.

 

Wald, A. Sequential analysis. New York: Wiley, 19147.

 

Watanabe, 8. (Ed.) Methodologies g pattern recognition. New York:
Academic Press, 1969.

 

 

Watzlawick, P., Beavin, J. H., and Jackson, D. D. Pragmatics pf:
human communication. New York: Norton, 1967.

 

 

290

Webb, E. J., Campbell, D. T., Schwartz, R. D., and Sechrest, L.
Unobtrusive measures: nonreactive research 2 the social
sciences. Chicago: RandT'IcNally, 1966.

 

 

 

Whitehead, A. N. Science and the modern world. _ New York: New
American 1.15' rary, 1925.

 

Whitehead, A. N. ' Process and reality. New York: Harper and Row,
1929.

 

Wiens, A. N. , Saslow, G., and Matarazzo, J. D. Speech interruption
behavior during interviews . Psychother.: Ther. , Res . , Pract. ,
1966, 3, 153-158.

 

 

 

Wright, H. F. Recording and ylyzing child behavior. New York:
Harper and Row, 1967.

 

Zeisel, H. Say _i: with figures. New York: Harper and Brothers, 1957.

 

APPENDICES

291

APPENDIX A

Schematic Presentation of Experimental Designs

In their brief booklet on experimental design, Campbell and
Stanley (1963) present a series of "pre-," "true," and "quasi-" experi-
mental designs, each design.having a unique schematic or graphic repre-
sentation. Because these designs are referred to at severaltpoints in
this paper, they are presented here for convenience. Campbell and
Stanley explain their code as fOllows:

In presenting the experimental designs, a unifOrm code
and graphic presentation will be employed to epitomize most,
if not all, of their distinctive features. An X will repre-
sent the exposure of a group to an experimental variable or
event, the effects of which are to be measured; 0 will refer
to some process of observation or measurement; the X5 and Os
in a given row are applied to the same specific persons.

The left-to-right dimension indicates the temporal order,
and X5 and Os vertical to one another are simultaneous. To
make certain important distinctions, as between Designs 2
and 6, or between Designs 9 and 10, a symbol R, indicating
random assignment to separate treatment groups, is necessary.
This randomization is conceived to be a process occurring at
a specific time, and is the all-purpose procedure fOr
achieving pretreatment equality of groups, within known
statistical limits. .Along with this goes another graphic
convention, in that parallel rows unseparated by dashes
represent comparison groups equated by randomization, while
those separated by a dashed line represent comparison groups
not equated by random.assignment. A symbol fOr matching as
a process fOr the pretreatment equating of comparison groups
has not been used, because the value of this process has been
greatly oversold and it is more often a source of mistaken
inference than a help to valid inference. . .. A.symbol M
fOr:materials has been used in a specific way in Design 9

[to indicate equivalent materials] (1963, p. 6).

Each of the fOllowing presentations includes the number given to
the design in the Campbell and Stanley volume, a brief descriptive title

fOr the design, a page reference to Campbell and Stanley (1963), and the

 

 

 

292

schematic fOr the design.

(1) One-Shot Case Study (p. 6):
X 0
(2) One-Group Pretest-Posttest (p. 7):
O X 0 In

(3) Static Group Comparison (p. 12):

 

O

(9) Pretest-Posttest with Control Group (p. 13): i
R O X 0
R O O

(5) Solomon Four Group (p. 29):

R O X 0 -
R O O
R X 0
R O

(6) Posttest-Only with Control Group (p. 25):
R X 0
R O

(Hovland, §£_§l,) Pretest-Posttest with COntrol Group and Time Extension

(pp. 31—32):
R O X 0
R O O
R O X 0
R O O

293
(7) Time-Series (p. 37):
O O O O X 0 O O O
(8) Equivalent Time Samples (p. 93):

X10 X00 X10 X00

Note that X0 stands here fOr the absence of experimental stimuli.
(9) Equivalent Materials (p. 96):
maxlo beoo mgxlo de00 etc.
"The Ms indicate specific materials, the Ma, MC, etc., being in sampling

terms, equal to the sample Mb, Ma, etc. (p. 96)." NOte that here, as in

Design (8), a basic pattern of X10 X00 is repeated with the same group.

(10) Nonequivalent Cbntrol Group (p. 97):

O X 0

 

O O
(11) Counterbalanced (including Latin Square) (pp. 50-51):
1:1 t2 t3

Group A x10 x20 x30 xuo
Group B x20 Xuo x10 x30

Group D X 0 X30 X20 X10

(12) Separate—Sample Pretest-Posttest (p. 53):
R 0 (X)
R X 0
Note that the parentheses indicate an optional X. The fOllowing

variations in the basic design are also mentioned:

 

 

 

299

(a) Time Extension (p. 90, 53):

R O (X)

R X 0

'R """" 6 ' —(X)- ' - '
R X 0

(b) Time Extension (p. 90, 53):

R O (X)
R 0 (X)
R X 0

(c) Retested Group (p. 90, 59):
R O X 0

R X 0

(13) Separate-Sample Pretest—Posttest with Control Group (p. 55):

R 0 (X)

R X 0
_R. _ a .......
R O

(19) Multiple Time-Series (p. 55):

(15) Institutional Cycle--Patched-up (p. 56, 57):

 

E?“ l, in.
u
I

295
Note that the X, here, occurs in a cycle. The schematic given here is
a specific example of a patched-up design. See pp. 57—61 for fUrther

infOrmation.
(16) Regression-Discontinuity Analysis (p. 61):

Note that no sOhematic is presented by Campbell and Stanley fOr this

particular "design." The above is the present author's schematic, the

 

parentheses indicating an Optional or unrelated observation.

The fOllowing schematics are added to the above list for
completeness. They fOllow the notation used above, but are not
necessarily taken from.Campbell and Stanley (1963).

(A) TWO-wave Panel Design (Campbell, 1963, p. 67, 68):

(a) (b)

<2}<—*—2>—---<2><§H2>

(Unacceptable) (Better)

"Here the spanning parentheses indicate occurrence of the O or X on
the same interview; the question mark, ambiguity of classification into
X and nO-X groups (Campbell, 1963, p. 67)."

(B) "Multiple-wave" Panel Design (Lazarsfeld,1938, 1998):

{29% Wi-C-ZH HE}

296

The above schematic is the present author's representation of one
possible application of a "multiple wave" panel design. The Xa, etc.,
indicate a continuing event, such as an election campaign, while the
question marks indicate ambiguity in classification or lack of control
over exposure. Many other variations are possible fOr other types of
events, and fOr purposes of gaining precision (Lazarsfeld, 1938, 1999,

1998; Zeisel, 1957, pp. 215-259).

 

297

APPENDIX B

Derivation of Minimum Number of Measurements

 

As indicated in section 9.2, the problem considered here is

r LIZ-1

that of determining the minimum number of measurements required to
completely characterize a sinusoidal signal. As discussed, it will

be assumed that the frequency,f, of this signal is known, thereby

 

making the period, p, of the signal, p = l/f. The sinusoidal signal

‘w--

to be considered.will be the general case presented graphically in

Figure B-l, below:

 

 

 

‘At-a

 

 

 

 

 

 

t, t t tx t, t

Figure B—l. General Sinusoidal Signal

In this case, a value, vn, is obtained at a measurement point,
tn, where the value v is measured from an arbitrary reference, v0, and
time is measured from a reference point, to. Successive measurements

are separated by an interval of At, so that tn+1 = tn + At.

298

The concern here will be, of course, with intervals of At where
At ;=p/2, as considered in section 9.2. The general equation for

the value, vn, is (cf. Milsum, 1966, pp. 150-151):

 

v = K + A cos wtn (1)
where A is the amplitude of the sinusoidal signal measured from its
mean, and w is known, given that w = 2mf and the frequency, f, is 5‘
known. The term A coscmgv therefOre, describes the variation of the
sinusoidal signal about its mean, which in this case is a constant
value, K. i

Note that the quantity wtn in equation (1) is an angle expressed
in radian measure. Hence, if tn is initially zero, the quantity wtn
will be an angle of zero. If tn is p/2 or one4half of a cycle of the
sinusoidal signal, the quantity wtn will be an angle of m since wtn =
2mf-tn = 2mf7°p/2 = 2nf-1/2-l/f = m. Similarly, if tn is p or one
full cycle, then the quantity wtn will be an angle of 2m. In short,
it is possible to consider a time interval of At, where At = p/2, in
terms of an angle of m, and it will be convenient to do so at points
below.

Note, also, that in the derivation below, the quantities v and
t will always be known, since measurement of a signal or form of vari—
ation is considered here to involve both determination of the variable
value and determination of the time at which that measurement is taken.
Examination of equation (1) will show, then, that the only terms to be
determined are K and A, since all other terms are known. Determination
of these terms results in complete characterization of the sinusoidal
signal. The general procedure will be, then, to determine whether or

not equation (1) can be solved, and under what conditions, given one,

299

two, or three measurement points.1

One measurement point.-— Given a value v1 obtained at a time

 

t1, the result of substituting in equation (1) would be:

v1 = K +.A cos wtl .
Clearly, the above equation is one equation in two unknowns (K and A),
and cannot be solved under any conditions. Hence, one measurement
point will not completely characterize a sinusoidal signal.

Two measurement points.—- Given values v1 and v2 obtained at

 

times t1 and t2, it is possible to write two equations using equation

(1) above:

v1 K + A cos wtl , and

v2 = K_+ A cos wt2 . (2)

This pair of equations may be solved for K and A. Solving for A, first,

it is possible to rewrite the above as:

K = v1 - A cos wtl , and

K = v2 - A cos wt2 ,
hence:
v1 - A cos wtl 2 v2 - A cos wt?
and:
A cos wtl — A cos wt2 = y1 - v2
or:
A = V1 - V2 (3)

 

cos wtl - cos wtz

 

lAppreciation is expressed to John J. Forsyth for guidance
in this approach .

' E. 8 L711},

 

250
Solving fOr K, second, by substituting equation (3) into one of the

initial pair of equations:

V1 : K + vl - V2 cos wtl
cos wtl - cos wtz

 

01":

 

K = V1 - V1 ‘ V2 cos wtl (9)
cos wtl - cos wt2

Note that equations (3) and (9) are general forms applying to situa-
tions Where t2 = t1 + At and At ;=p/2, i.e., At ;:m. If At = m,
certain simplifications are possible, since cos wt2 = cos w(t1 + At) =
cos w(tl + m) = -cos wtl.

An examination of equations (3) and (9) shows, of course, that
all values on the right side of the equal sign are known values, as
would be expected in any solution of two equations in two unknowns.

It is the curious property of measurements on sinusoidal signals, how-
ever, that a pair of measurements taken at an interval whose midpoint
is tX or ty in Figure B-l will yield equal values for v1 and v2. Such
a situation may occur regardless of whether the interval is At = p/2
or At < p/2. Note the fOllowing crucial point: if the values v1 and
v2 are equal, equations (3) and (9) above break down, i.e., if v1 =
v2, it is not possible to solve the pair of equations (2) considered
above! Hence, two measurement points are sufficient to completely
characterize a sinusoidal signal, but only if the measured values are
not equal. If v1 = v2, it is necessary to obtain a third value, V3,
which is not equal to the other two. Such a value, v1 = v2 ¢ v3 will

permit a solution to the pair of equations above. A third value

251

requires, however, a third measurement point.

Three measurement points.-- The reason underlying the breakdown

 

of equations (3) and (9) when v1 = v2 can be seen by examining again
the initial pair of equations (2). That is, v1 may equal V2 only if
the points t1 and t2 are symmetrical around the a point like tX or ty
in Figure B-l. In such cases, the cosine terms in equations (2) are
identical, and since the values v1 and v2 are equal, the result is two
identical equations in two unknowns-—a situation which has no solution.
The need, then, is to obtain a third measurement, V3, at time t3 which
will not yield an equation identical to those in (2).

Note, however, that if v1 = v2 and t1 and t2 are separated by
an interval of At = p/2, all successive measurements spaced at At =

p/2 will yield the same value pf_y} That is, suCh a measurement scheme

 

would be "locked in" to measurement of the mean of the sinusoidal
signal, as can be seen by examining Figure B—l. Such a situation would
not provide information leading to the solution of equations (2), for
the reasons noted immediately above; hence, it would not be possible
to characterize the sinusoidal signal in this case, even with three
measurement points.

On the other hand, if v1 = v2 and t1 and t2 are separated by
an interval of At < p/2, then a third measurement at time t3 could pp:_
yield a value V3 identical to v1 and V2. That is, measurement at an
interval of At < p/2 could not be "locked in" to producing identical
values. Such a situation would produce at most two identical values
in any three measurements, or in this case, v1 = V2 i V3. The presence

of at least two distinct values of v would allow solution of equations

252

(2), and consequently would characterize the sinusoidal signal.

Reviewing the above discussion, then, it may be seen that,
given the frequency of a sinusoidal signal, three measurements of
that signal spaced at intervals of At < p/2 are sufficient in all
cases to completely characterize that signal.

Further considerations.—— As noted, the assumption above has

 

been that the frequency of the sinusoidal signal under investigation
is known. If this assumption is removed, then it is clear that three
measurement points is an absolute minimum number. That is, equation
(1) would be revised under these conditions to the following:

Vn = K + A cos 2nftn

where K, A, and f are unknown. SuCh a situation of three unknowns
clearly requires at least three measurements which provide three
different values of vh in order that a solution be found.

As in the case of three measurement points above, it may be
impossible, because of the measurement scheme and the interval At to
obtain three different values of Vn° Unlike the case above, however,
suCh situationg are numerous and are not as easily specified, so that
it is more difficult to determine, for example, the conditions under
which four (or more) measurements might be sufficient to completely
characterize a sinusoidal signal. Since such a determination moves
beyond the scope of the present discussion of the minimum number of
measurements required, given the frequency value, it will be more
expedient to turn to another type of sufficient condition. That is,

when the frequency of a sinusoidal signal is unknown, three measure-

ments of that signal are sufficient in all cases to completely

253

characterize that signal if v1 # v2 f v,.

"mmm