ATTRIBUTION OF RESPONSIBILITY AS A
FUNCTION OF THE STRUCTURE,
QUALITY, AND INTENSITY OF
THE EVENT









By
JEFFERSON LEWIS SULZER










A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA
June, 1964















ACKNOWLEDGMENTS


The writer wishes to express his appreciation for the guidance

and advice given him by the members of his supervisory committee: Dr.

Marvin E. Shaw, chairman, Dr. J. Mason Wright, Dr. Henry S. Pennypacker,

Dr. Sidney M. Jourard, and Dr. J. V. D. Saunders. He would also like

to express his gratitude to Mrs. Louanne Antrim for typing the.manu-

script and to the students who served as subjects in this study.















TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS . . . . . . . . . . . . ii

LIST OT TABLES . . . . . . . ... ... .. v

LIST OF FIGURES . . . . . . . . .. . . . vii

CHAPTER

I INTRODUCTION ... .............. .... 1

The Perception of Causality . . . . 3
Attribution of Responsibility . . . . 8
Experimental Studies of Attribution of
Responsibility . . . . . . .. 14
The Problem .. . . . . . . . 18
Theoretical Framework . . . . . .. 21
Hypotheses: Attribution of Responsibility
(AR) . . . . . . . . . . 24
Hypotheses: Assignment of Sanctions (AS) . 26

II STIMULUS MATERIALS .... .............. 29

Event Structure (Levels) .. . . . . 29
Outcome Selection . . . . . ... 31
Order of Presentation . . . . ... 34

III EXPERIMENT 1. ATTRIBUTION OF RESPONSIBILITY . . . 36

Method .. ... ..... ............ 36

Subjects ....... ............ 36
Stimulus Materials. . . . . . . 36
Administration . . . . . . ... 37

Results ... . . . . . ..... .. .. 38

Order Effects . . . . . . . .. 38
Experimental Variables . . . . . .. 41
Summary of Results of Experiment 1 . . .. 57

Discussion .................. 59









Page

IV EXPERIMENT 2. ASSIGNMENT OF SANCTIONS . . . .. 66

Method . . . . . . . . .. . 66

Subjects . . . .. . ..... .. 66
Stimulus Materials . . . . . . .. 66
Administration . . . . . . . 66

Results ... . . . . .. . . 67

Order Effects . ............ 68
Experimental Variables . . ....... 71
Summary of Results of Experiment 2 ..... 84

Discussion .................... 86

V SUMMARY AND CONCLUSIONS . . . . . . ... 90

Attribution of Responsibility . . . .. 91
Assignment of Sanctions . . . . .. 92
Attribution of Responsibility and Assignment
of Sanctions ............... 94
Conclusions ......... .. ..... 101

APPENDICES . . . ........ .... ...... .. 103

A. INSTRUCTIONS .................... 104
B. STIMULUS MATERIALS . .. . . . . . 110
C. RAW DATA .. . . .......... .... .. 121
D. SUPPLEMENTARY ANALYSIS . . . . . . .. . 140

REFERENCES ................. ....... 149

BIOGRAPHICAL SKETCH . . . . . . . . . . 153















LIST OF TABLES


Table Page

1. Mean Attribution of Responsibility Ratings for each
Stimulus Item in each of the Four Orders of
Presentation . . . . . . . .... . 39

2. Summary of the Analysis of Variance for Attribution
of Responsibility Ratings, Experiment 1 . . .. 40

3. Mean Attribution of Responsibility Rating and Variance
for Positive and Negative Outcomes at each of the
Five Levels . . . . . . . .... . 44

4. Mean Attribution of Responsibility Rating and Variance
for Low, Moderate and High Outcome Intensity
at each of the Five Levels . . . . . ... 46

5. Mean Attribution of Responsibility Rating and Variance
for each Outcome Quality and Intensity at each
of the Five Levels . . . . . . . ... 53

6. Mean Attribution of Responsibility Rating and Variance
for Positive and Negative Outcome Quality at Low,
Moderate and High Outcome Intensity . . ... 54

7. Summary of the Analysis of Variance for the Sanction
Ratings, Experiment 2 . . . . . .... .69

8. Mean Sanction Ratings for each Stimulus Item in each of
the Four Orders of Presentation . . ... ... . 70

9. Mean Sanction Rating and Variance for Positive or
Negative Outcomes at each of the Five Levels . . 74

10. Mean Sanction Rating and Variance for Low, Moderate
and High Outcome Intensity at each of the Five
Levels . . . . . . . . ... ..... 75

11. Mean Sanction Rating and Variance for Positive and
Negative Outcome Quality at Low, Moderate and
High Outcome Intensity . . . . . . . 81

12. Mean Sanction Rating and Variance for each Outcome Quality
and Intensity at each of the Five Levels . . .. 82









Table Page

13. Mean Ratings for Abstract Structure, Outcome, Respon-
sibility and Sanctions for each Stimulus Item . 96

14. Mean AR, Mean AS and Percent of Subjects who Attributed
Responsibility and Assigned Sanctions for each
Stimulus Item . . . . . . . .... .98















LIST OF FIGURES


Figure Page

1. Mean attribution of responsibility assigned to the
abstract stories at each Level . . . ... .32

2. Mean attribution of responsibility for the abstract
and realistic stories at each Level ..... .42

3. Mean attribution of responsibility rating for Positive
and Negative outcomes at each Level. . . .. 47

4. Mean attribution of responsibility for Low (L),
Moderate (M), and High (H) intensity outcomes
at each Level .................. 50

5. Mean attribution of responsibility for High (H-) and
Low (L-) intensity negative outcomes at each
Level . . . . . . . .... ..... .52

6. Mean attribution of responsibility for each stimulus
item at each Level . . . . . . . . 56

7. Mean attribution of responsibility and assignment
of sanctions at each Level . . . . ... .72

8. Mean assignment of sanctions for Positive and
Negative outcomes at each Level . . . ... .76

9. Mean assignment of sanctions for Low (L), Moderate (M),
and High (H) outcomes at each Level ..... .79

10. Mean assignment of sanctions for each stimulus item
at each Level. . . . . . . . . .. 83















CHAPTER I


INTRODUCTION



The present paper is concerned with the attribution of respon-

sibility and assignment of sanctions as functions of three aspects of

the action outcome: causal structure, quality,and intensity. It re-

presents an experimental application of Fritz Heider's (1944, 1958a)

"naive analysis of action" and attempts to provide empirical data upon

which a more definitive theoretical framework may be developed.

Attribution of responsibility and assignment of sanctions are

related behaviors which depend upon two interacting factors: personality

characteristics and response traits of the attributor and perceived

characteristics of the stimulus situation. This study is concerned

only with stimulus factors and is based upon the assumption that

communalities of experience in dealing with social situations produce

common ways of perceiving and reacting to interpersonal action outcomes.

It should not be inferred from this selective emphasis that characteris-

tics of the perceiver are considered to be merely troublesome sources

of variability. On the contrary, the writer believes that individual

differences in reacting to the relevant stimuli constitute an important

area for future research. However, the experiments reported here were

conducted with the additional belief that the study of such individual

differences cannot proceed effectively until the critical aspects of

the stimulus situations are identified and brought under control.









In agreement with others who have been concerned with the

topic, the writer considers the perception of social responsibility to

be a process which is in many ways analagous to the perception of

physical causality. It is assumed, therefore, that the primary deter-

minant of responsibility attribution is the causal structure of the

events as they are perceived by the attributor. However, since

responsibility attribution concerns social (or interpersonal) events

it is likely that the final judgments are not based solely upon perceived

causal relationships but are subject to the influence of perceived de-

gree of intention as well as favorableness and unfavorableness of the

outcome. Once the judgment is made that another person is responsible

for a given outcome he becomes open to sanction. At this point, the

attributor may or may not ply-jobeetively-appropriatesanctions and

it is assumed that the primary determinants in this final judgmental

process include: perceived characteristics of the agent, interpersonal

relations between the attributor and the agent, history and personality

characteristics of the attributor, aspects of the current social environ-

ment as well as perceived quality and intensity of the outcome for which

the agent is held responsible. In the experiments reported here the

effects of causal structure and outcome upon the sanction judgments

receive primary consideration under the explicit assumption that the

other factors are not operative or are under experimental control.

To summarize the position taken in this paper then, it is assumed

that responsibility attribution is based upon the perception of personal

causality, which then determines whether or not the identified agent is

open to sanction. Whether sanctions will actually be applied, as well

as sanction direction and intensity, is primarily determined by the








perceived quality and intensity of the action outcome. This basic set

of assumptions, as well as some alternative views, will be expanded in

the following section.


The Perception of Causality

Basic to the attribution of responsibility is the process of

perceiving causality. In fact, one theorist, Albert Pepitone (1958),

defines responsibility attribution as the process of identifying the

causal agent for a social action. Since the research reported in this

paper is based upon a somewhat different definition of responsibility

it may be helpful to examine the assumption that causal perception is

fundamentally the same process regardless of the nature of the events

concerned. In the following discussion it will be argued that the

identification of causal agents for social events requires cognitions

(or inferences) which are not typically involved in a causal interpre-

tation of physical phenomena.

Until quite recently most of the empirical and theoretical explora-

tions of causal perception have been carried on by Europeans. The

three major treatments of the topic, Piaget (1930, 1932), Michotte

(1963), and Heider (1944, 1958a),represent very different approaches

and it is no surprise that they have produced different ways of viewing

the phenomena involved. Michotte (1963) has conducted an ingenious

series of experiments aimed at specifying the stimulus conditions under

which causal conception occurs. In this research he has manipulated

various physical dimensions to produce the perception of "mechanical"

causality, which he maintains is the basic, innate process involved in

all causal conception. Piaget's work has, of course, largely concen-

trated on the development of causal conception and he has relied heavily









upon qualitative differences in children's explanations of, and other

reactions to, relatively commonplace physical and social events. Like

Michotte he has argued that similar principles are involved in per-

ceiving physical and social causality but Piaget maintains that this

homogeneity decreases as the child grows older. Heider is in general

agreement with Piaget in regard to the development of causal thinking

but his theoretical treatment of the cognitive processes involved is

based on a common-sense analysis rather than on the protocols of

experimental subjects. Actually, Heider's analysis of causal percep-

tion in the social realm (which he calls "the naive analysis of action")

seems to reflect a highly sophisticated application of Gestalt principles

to the data of naive perception and it is precisely because his

analysis is conducted within an established theoretical frame of refer-

ence that he has produced insights of considerable value to general

psychology.

All three of these investigators have published findings indi-

cating that purely physical events (interactions between physical objects)

are sometimes interpreted as if they involved humans. Piaget's (1930,

1955) young children, :ichotte's (1963) and Heider and Simmel's (1944)

adults frequently described discrete physical events in terms of inten-

tion, motive, and interpersonal relations. Additional empirical support

for the assumption of homogeneity of physical and social causal per-

ception is provided by the extensive program of research conducted by

Muuss (1959, 1960, 1961) and his co-workers (Ojemann, Levitt, Lyle,

and Whiteside, 1955). For example, Muuss (1961) was able to get

positive transfer from training in multicausal analysis of social inter-

actions to the analysis of physical causality. Although other studies









could be cited, these seem sufficient to substantiate the claim that

some of the same basic principles are operative in all types of causal

perception. Of greater relevance to responsibility attribution; how-

ever, are the differences between physical and social events from the

viewpoint of the relatively unsophisticated observer.

In personal or social causality, as in physical causality, the

minimum requirements are that there is a sequential occurrence of

events which appears to transfer energy from one object (A) to another

object (B) so that some discriminable change of state occurs in B.

Alfred North Whitehead (1933) has stated that causal perception involves

the fundamental cognition that energy has recognizable paths through

time and space and that this represents a conception of nature in terms

of continuity. In a sophisticated, "scientific" approach to physical

causality the continuity of events is considered to be infinite or

complete, and the very fact that such continuity obtains, prevents us

from identifying the primary or original cause. As we look back in

time each event is found to have one or more relevant predecessor, and

questions concerning the origin of a sequence thus remain largely un-

answerable. This is also true of social or interpersonal events if we

take a highly sophisticated point of view. However, It seems likely that

ordinary causal perception does not typically involve notions of infinite

regression, multi-causality, and rigorous application of logic, but

instead follows what Abelson and Rosenberg (1958) have called the rules

of "psycho-logic." Heider (1958b) maintains that there is a fundamental

psycho-logical tendency for the naive perceiver to identify persons as

the absolute causal origins of events, and that this serves to transform

irreversible changes into reversible ones by making it feasible (and









legitimate) to apply sanctions to the responsible person.

Heider (1958b, p. 4) quotes Fauconnet as follows: "Man is,

in a certain sense, a first cause, if not of the physical movements

which constitute his acts, at least of their moral quality. . From

this perfect causality originates his responsibility." In identifying

a person as the first cause of an event, his intentions are of primary

importance because they provide the basis for the psycho-logical cog-

nition that the event originated in him. It is precisely because the

individual can be identified as the origin of an event that his fellows

feel justified in punishing him for harmful acts. Nowhere is this

clearer than in the McNaghten Rule which is the basic criterion of

criminal responsibility. H. A. Davidson (1952, p. 3) quotes the original

formula as: ". . he [the defendant] was laboring under such a defect

of reason from disease of the mind as not to know the nature and quality

of the act; or, if he did know it, that he did not know he was doing

what was wrong." The obvious implication of this legal formula is

that if a person had knowledge of the harmfulness of an act he was

committing we are led to infer that he intended to harm his victim, or

more generally, that he foresaw and intended to produce the specific

outcome in question.

A concern with the agent's intention appears to constitute a

fundamental difference in the perception of physical and social causality,

at least as they occur in normal adult observers. From his developmental

studies, Piaget (1930, 1932, 1955) concludes that although the child

begins by imputing motive and intention to physical objects and events

he gradually becomes more sophisticated and restricts his imputations

of intention to personal agents. Although this conclusion seems to be









contradicted by Michotte's (1963) and Heider and Simmel's (1944)

studies in which adults interpreted apparent interactions between

physical objects in personal terms (i.e., intention, purpose and social

influence), it is likely that these findings represent as if inter-

pretations of artificially discrete events which were so constructed

that they resembled human interaction. That is, the mechanical move-

ments used as stimuli in these experiments were different from ordinary

physical occurrences in some important respects, including the appearance

of spontaneous self-propulsion. Such spontaneous, self-caused movement

is but one basis for inferring intention on the part of the agent. A

related basis for this inference is provided where the agent appears to

show flexible or adaptative behavior in reaching his presumed goal.

This is clearer, perhaps, in Heider's (1958a) distinction between

personal and physical causality. For Heider, the conditions of personal

causality are met when both "equifinality" and "local causality" (e.g.,

transfer of energy) appear. In the case of impersonal causality,

equifinality does not occur because a physical event may produce one or

more of a large number of effects depending upon the particular circum-

stances. To take an example from Heider, a falling stone may hit a

man, fall on the ground, or start an avalanche. In the case of personal

causality, on the other hand, equifinality does occur, in that the

personal agent with the intention of producing a given effect can choose

among a variety of means to reach the same goal, even though environmental

circumstances may be opposed to the production of the outcome.

If one accepts the validity of these differences between personal

and impersonal causality the implications are far-reaching. First,

where personal causality operates a source outside the situation cannot









as simply change the outcome by manipulating physical objects and con-

ditions. Second, the perception of personal causality is more complex

and requires different kinds of judgments especially in regard to

the intentions of the agent. Third, where the causal relations are

somewhat ambiguous and a person is held responsible it is very likely

that the attributor is inferring at least some intention on the agent's

part. (In fact, equifinality almost necessitates the cognition that

the events are guided by a goal-oriented agent.) Finally, the stronger

the opposing environmental conditions which were overcome to reach a

final outcome the clearer the indication that the agent intended to

produce it. (This "overcoming of obstacles" has been used by Warner

[1928) and others as an indicant of drive strength in rodents; it is

perhaps unnecessary to point out that we are concerned here with the

phenomenological implications of the behavior as well as with its

operational utility.)


Attribution of Responsibility

The literature on responsibility attribution is not extensive

and at present there is a great need for conceptual clarification of

the construct "responsibility" and of the attribution process itself.

There seems to be general agreement among those who use the term that

responsibility attribution involves the designation of one or more

persons as the primary origin of a specific event (X) which has occurred

in the interpersonal life space. For Pepitone (1958) this is all that

responsibility attribution entails. Wright (1960, 1963) agrees with

Pepitone but complicates the picture somewhat by suggesting that

willingness to sanction the personal agent (P) for the commission of X

may be utilized as an indication that responsibility has, in fact, been









attributed to him. A statement by Shaw and Sulzer (1964, 1963, p. 1)

seems to reflect this same line of thinking. "When one person attributes

responsibility for an event to another individual, he blames that per-

son if the outcome is negative and praises (gives credit to) him if the

outcome is positive." This statement obviously implies that respon-

sibility attribution and the application or assignment of sanctions are

either identical processes or that they are highly correlated. From

the point of view taken here both of these assumptions are unwarranted

until their validity is attested to empirically. In the interest of

conceptual clarity, attribution of responsibility will be discussed in

the following paragraphs as a process which does not necessarily imply

that sanctioning, or willingness to sanction, will occur.


Theoretical analysis of responsibility attribution. From the

writings of Jean Piaget (1932, 1948, 1955) it is apparent that he

believes that the moral value of an action outcome, as well as

responsibility for having produced it, may be decided in terms of the

amount of damage it represents (objective responsibility) or in terms

of the actor's intentions and motives (subjective responsibility).

According to Piaget, adults typically make these judgments on the basis

of motive, while most children under the age of nine or so show an

almost total reliance upon outcome intensity. Piaget draws these con-

clusions from an examination of children's responses to pairs of short

stories which were designed so that one member of each pair included a

well-intentioned act resulting in considerable material damage while the

other story depicted a selfish or malicious act which produced relatively

minor consequences. From Piaget's journalistic presentation of his

results it is rather difficult to make a quantitative evaluation of his









findings. Furthermore, a careful examination of his stimulus materials

reveals that other factors (notably causal structure) are confounded

with those he attempted to vary. Thus, his statements in regard to the

age at which subjective responsibility predominates are in need of further

experimental support. Bandura and McDonald (1963) and Boehm (1963)

have produced results which show that preschool children are capable

of verbalizing in terms of motivation and intention. What is needed

now, is experimental evidence on the age at which the effects of outcome

intensity are reduced or (possibly) disappear.

Piaget's developmental statements are so dramatic that they tend

to obscure the fact that he was aware of other variables which affect

the attribution process. More explicit statements of greater theore-

tical utility (for present purposes) have been made by Pepitone (1958)

and Heider (1958a). Pepitone considers responsibility as one of the

three "dimensions of social causality," conceptually separate from

"Intentionality" and "Justifiability." The Responsibility dimension is

basically concerned with the cognitive processes involved in identifying

the causal agent for a social (interpersonal) act. Intentionality con-

cerns the agent's intentions, .which may be positive or negative depending

upon whether they are to benefit or harm the perceiver. His third di-

mension, Justifiability, is based upon the cognition that the act is

in agreement (or disagreement) with various logical and social norms, or

other kinds of generally accepted standards. Pepitone's triadic treat-

ment of social causality has many interesting implications for the

attribution process and it is unfortunate that he has not developed it

more fully. (Most of the research reported by Pepitone is concerned

with the effects of these variables upon sociometric ratings and with

the effect of status upon the attribution of causality.) Since Heider's








(1944, 1958a) analysis is similar to, but more extensive than, Pepitone's,

it has been assumed as the theoretical framework for the present paper.

Heider (1958a) extended his naive analysis of causality to in-

clude personal responsibility, which he conceptualizes as a cognized

link between the person and the final outcome. Intention is the central

factor determining intimacy of the link. Generalizing from his distinc-

tion between personal and impersonal causality, Heider maintains that

responsibility for the outcome may be attributed to the person, to the

environment, or to both. The environment includes all impersonal factors

which might be perceived as facilitating or inhibiting production of a

given outcome, such as "luck," task difficulty, coercion, social in-

fluence and norms, or even fate or "Supreme being." Thus, responsibility

for a given outcome need not be attributed solely to a personal origin.

In Heider's conceptualization, "personal responsibility . varies

with the relative contribution of environmental factors to the action

outcome; in general, the more they are felt to influence the action, the

less the person is held responsible" (1958a, p. 113). An important

implication of this assertion (if it is accepted), is that it is legiti-

mate to ask questions about the degree of perceived responsibility.

Drawing on the writings of Stern (1923), Fauconnet (1928) and

Piaget (1932, 1948) Heider described five "levels" in attribution of

responsibility which represent a progression from relatively primitive

to relatively sophisticated cognitive processes. The levels have been

labelled and restated by Shaw and Sulzer (1964) as follows:


Level I: Global-Association: The person is held responsible for any

effect with which he is connected in any way. In Piaget's

(1955) terms, responsibility at this most primitive stage








is determined by syncretistic, pseudo-causal reasoning

rather than by consideration of objective causal connections.

Thus, a person may be blamed for harmful acts committed by

his friends.


Level II: Extended Commission: The person is held responsible for

any effect that he produced by his actions, even though he

definitely could not have foreseen the consequences of his

actions. As in Piaget's (1932) "objective responsibility"

the person is judged according to what he does but not

according to his motives.


Level III: Careless Commission: The person is held responsible for any

foreseeable effect that he produced by his actions even

though the effect was not a part of his goals or intentions.

He is held responsible for the lack of restraint that a

wider cognitive field would have produced.


Level IV: Purposive Commission: The person is held responsible for

any effect that he produced by his actions, foreseeing

the outcome and intending to produce the effect. This

corresponds roughly to Piaget's "subjective responsibility"

in which motives are the central issue.


Level V: Justified Commission: The person is held only partly

responsible for any effect that he intentionally produced

if the circumstances were such that most persons would have

felt and acted as he did. That is, responsibility for the

act is at least shared by the coercive environment.









Heider apparently intended these "levels" to be a description

of developmental stages, supplementing or replacing Piaget's three-

stage theory of the development of causal thinking. However, they may

also be viewed as descriptions of the information which is sufficient for

attribution of personal responsibility at each level of sophistication.

(It should be remembered that one of the components of personal respon-

sibility is intentionality; therefore, such attribution implies that P,

the primary person, intended to produce the outcome.) For the most

unsophisticated individual, the minimal information contained in Level

I should be a sufficient basis for attribution and the information con-

tained in the "higher" levels would have the effect of unnecessary re-

dundancy, For such an individual, attribution at the different levels

should be highly similar. On the other hand, a more sophisticated

individual would be less likely to attribute personal responsibility

given the information at Levels I and II, because sizable inferential

leaps would be required in regard to the actor's intention. Thus,

attribution of personal responsibility by the sophisticated person

should gradually increase up to a maximum at Level IV, where the cues

clearly indicate intention, and be somewhat less at Level V, where he

might attribute some responsibility to both the person and the environ-

ment because of the extenuating circumstances. Shaw and Sulzer (1964),

using stories designed to represent each of Heider's Levels in respon-

sibility attribution have obtained some empirical support for these

expectations, since they found that children attribute responsibility

to child actors in a relatively undifferentiated manner. When the

stories involved adults, however, no significant differences were found

as a function of the subjects' ages. The implication of these findings








to Piaget's stage theory is considerable. Of greater relevance to the

present paper, however, is the fact that Shaw and Sulzer (1964) demon-

strated that Heider's Levels have considerable utility as a framework

for studying the variables involved in responsibility attribution.


Experimental Studies of Attribution of Responsibility

In spite of the fact that the literature on the perception of

causality has a fairly long history, relatively little research has been

conducted which bears directly on the attribution of responsibility.

With the recent revival of interest in interpersonal perception,

however, some studies have been reinterpreted in terms of causal percep-

tion and responsibility attribution. These studies, when added to those

specifically designed to investigate responsibility, comprise the be-

ginnings of a body of literature which will probably show rapid growth

in the next few years.

The development of a generally accepted methodological approach

has not progressed very far and reflects the fact that the topic has

only recently entered the domain of experimental psychology. To the

writer's knowledge, the only instruments which have been specifically

developed to study responsibility attribution are the Social Interaction

Series used by Wright (1960, 1963) and the stories representing Heider's

"Levels in responsibility attribution" reported by Shaw and Sulzer (1964).

The Social Interaction Series includes a set of 36 line drawings which

depict a man and woman, boy and woman, boy and man, and two men in

successive stages of a positive and negative interaction. Measures of

direction and amount of attribution of responsibility (AR) are obtain-

able and the instrument may be used with individuals or groups of subjects.

Shaw and Sulzer's Levels in Responsibility Attribution Stories have been








used in several experiments (Sulzer, Nickols, Blum and Brant, 1963;

Sulzer and Shaw, 1963; Shaw and Sulzer, 1964) and, in spite of varia-

tions in content, they have produced highly predictable AR patterns

using adult subjects. The instrument has an additional advantage in

that the story technique has been used extensively in developmental

studies of moral judgment and causal perception, and thus lends itself

to a unified treatment of these related phenomena. Relevant findings

have been reported by Piaget (1932, 1948, 1955), Harrower (1934),Cuber

and Pell (1941), Seeman (1947), Levitt (1955), Diggory (1962), Johnson

(1962a), Bandura and McDonald (1963), and Boehm (1963).

Shaw and Sulzer's (1964) study partially supported their hypo-

thesis that children would show relatively undifferentiated attribution;

however, the children attributed very much as did adult subjects when

the stories involved adult actors and activities. This indicates that

children may actually perceive responsibility in peers and in adults

in a different manner. Other studies reported by Seeman (1947),

Thibaut and Riecken (1955), Pepitone and Kleiner (1957), Pepitone and

Sherberg (1957) and Wright (1960, 1963) support the conclusion that

social status is an important determinant of perceived responsibility.

Seeman, using a "moral evaluations questionnaire" designed by Cuber and

Pell (1941), found that Negroes were considered to be less responsible

or less "wrong" than whites when they were described as engaging in

identical amoral behavior. Thibaut and Riecken's findings clearly

show that persons are more likely to perceive the locus of causality

as "internal" (own force) for high-status persons and as "external"

(induced force or coercion) for low-status persons. The studies con-

ducted by Pepitone and his associates are somewhat less relevant because








of the expost facto tie-in between status and responsibility, but

Wright's two experiments unequivocally demonstrate that there is a

tendency to attribute more responsibility to authorities than t6 peers.

An interesting sidelight of his results is that males attributed more

responsibility to female than to male authorities. In view of this

finding it is possible, as Wright (1960) has suggested, that much of

what has been interpreted as status effects may merely reflect a ten-

dency to attribute differently to individuals who are similar and dis-

similar to the attributor. It is interesting to note that both status

(power) and similarity (cognitive balance) effects are predictable

from Heider's (1958a) theory.

Of considerable theoretical interest are the experiments which

investigated locus of causality because they provide support for the

contention that responsibility may be attributed to the person or the

environment. This idea is particularly important because it has impli-

cations for theories other than those which deal per se with the attri-,

bution of responsibility. For example, Festinger (1957) has suggested

that one way of avoiding or reducing post-decision dissonance is by

attributing responsibility for the decision to the environment rather

than to self. It is entirely possible that a tendency to attribute

more to one or the other may prove to be a response trait with high

relevance as a predictor of other behavior. (At present, Wright [1960]

has conducted the only investigation relating attribution patterns to

personality factors.) However, Bialer (1960) has produced a preliminary

scale to measure locus of control in mentally retarded and normal children.

This scale was used by McConnel(1963) in an attempt to show that ex-

ternally controlled subjects would be more suggestible. Although this









hypothesis was not supported, he did find that locus of control scores

are partially developmental in nature but show a significant curvilinear

relationship with chronological age. Considering the work of Piaget

on the development of causal thinking, it seems likely that high

attribution to the environment does not mean the same thing for young

children as it does for older ones. Thus, McConnell's curvilinear

relationship may simply reflect the fact that his scale could benefit

from a distinction between "syncretistic" and "sophisticated" attribution

of responsibility to the environment.

Related to external-internal attribution patterns is the concept

of diffusion of responsibility, or attribution to a number of agents or

sources. Although there is relatively little published research utilizing

responsibility diffusion as an independent variable, it appears to be

a promising approach to the study of group dynamics and organizational

behavior. For example, Wallach, Kogan and Bem (1964) produced results

which indicate that diffusion or spreading of responsibility for poten-

tial failure among all the members of a group leads to greater risk-

taking. It remains to be seen whether diffusion of responsibility also

lowers individual achievement motive and/or has other undesirable effects

upon group performance. In regard to the effects of focal attribution,

Worthy, Wright, and Shaw (1963) found that an individual who has respon-

sibility attributed to him for his team's losses in a noncompetitive

game became less willing to enter into future interactions. This finding

is certainly relevant to potential group effectiveness, but is based on

an extreme case where a single member is rather consistently identified

as the cause of his group's poor performance. His reluctance to enter

into future interactions may simply reflect the negative reinforcement









he received for his past efforts as well as a lowered estimate of his

value to the group. -Although this study is a significant one, it

cannot be taken as indicative of reactions to having responsibility

attributed to an individual in all situations since it dealt only with

responsibility for a negative event. It seems likely that being held

responsible for both positive and negative events, or for positive

only, would have different effects upon the accused.

Most of the experimentation on responsibility attribution and

other moral behavior has been concerned only with negative outcomes.

Whether or not this reflects a bias on the part of those conducting

the research, it has led to a somewhat one-sided theoretical development

of the phenomena involved, and possibly restricts the general applica-

bility of the conclusions which have been reached. Experiments by

Wright (1960, 1963) and Shaw and Sulzer (1964) have reported some sig-

nificant differences in attribution of responsibility for positive and

negative outcomes; however, the outcome variable is in need of further

exploration. This deficiency provides a major raison d'tre for the

present investigation.


The Problem

The present investigation had as its primary concern the evalua-

tion of the effects of outcome characteristics and causal structure upon

the attribution of responsibility. In many respects, it extends the

work done by Shaw and Sulzer (1964) but was conducted with the explicit

expectation that more conclusive results would be obtained. The findings

of this investigation could be of critical importance to the develop-

ment of a more comprehensive theory of responsibility attribution and

havesome important implications for the more general topical areas








which have been labelled "interpersonal perception" and "moral judg-

ment." By including both positive and negative outcomes it was hoped

that results would be obtained which would encourage a reappraisal

of theories and models which have overemphasized negatively valued

behavior.

Existing evidence on the role of outcome in responsibility attri-

bution is meager and somewhat equivocal. Piaget (1932, 1948), in

his assessment of the development of moral judgment, varied outcome

intensity (amount of damage), but dealt only with negative outcomes.

The only studies explicitly designed to evaluate the role of outcome

in responsibility attribution were conducted by Wright (1960, 1963)

and Shaw and Sulzer (1964). In all of these experiments outcome

quality (positive and negative) was investigated rather than outcome

intensity. In his first experiment Wright (1960) obtained significant

outcome effects only in a second order interaction but in a later

replication (1963) (with a different experimenter and subject popula-

tion) he produced a significant main effect attributable to outcome

quality. Shaw and Sulzer (1964), utilizing Heider's Levels in Respon-

sibility Attribution, reported significant outcome effects in an ex-

periment concerned with children's activities, but a replication using

adult activities failed to confirm these results. There are no ready

explanations for the instability of these findings except for the

obvious possibility that the experimental manipulations were too weak.

The stimulus materials used by Wright and by Shaw and Sulzer contained

relatively mild outcomes and it was anticipated that this weakness would

be overcome in the present investigation in which both quality and inten-

sity were varied.








A secondary purpose of this investigation was to study the

effects of causal structure, outcome quality, and outcome intensity

upon the assignment of sanctions. Hypotheses were tested in regard

to the effects of these variables on judgments concerning the amount

of sanctioning considered to be appropriate. An additional set of

expectations concerning the relationship between attribution of

responsibility and assignment of sanctions were also evaluated.

Empirical studies relating these variables are almost as rare as those

concerned with responsibility attribution. Piaget is generally con-

sidered to be the first, and almost the only, researcher to study moral

judgment; however, Johnson (1962b)has turned up studies by Schallen-

berger (1894) and Barnes (1894, 1902) which anticipated Piaget's

methodology and many of his findings. From his research, Piaget (1932,

1948) concluded that the amount of punishment considered appropriate

for a given act is determined by the actor's motives rather than by

outcome intensity in children of age ten or older. Kohn (1959), in a

study comparing working-class and middle-class families, concluded

that middle-class parents were far more likely to decide on an appropriate

punishment on the basis of judged intent behind a misdeed. This suggests

that population differences may account for the discrepancy between the

results of Piaget (1932, 1948), Boehm (1963) and Bandura and McDonald

(1963). However, there is still no basis for concluding that outcome

characteristics do not exert a powerful influence on adults' sanctioning

behavior. Piaget's position on this point is rather puzzling. For

example, Piaget himself assumes that outcome intensity is a potent

determinant of sanctioning behavior in adults when he makes an attempt

to explain the origin of objective responsibility. Quoting Piaget:









The hypothesis may therefore be advanced that evaluations
based on material damage alone are the result of adult
constraint refracted through childish respect far rather
than a spontaneous manifestation of the child mind.
Generally speaking, adults deal very harshly with clumsi-
ness. In so far as parents fail to grasp the situation
and lose their tempers in proportion to the amount of
damage done, in so far will the child begin by adopting
this way of looking at things and apply literally the
rules thus imposed, even if they were only implicit
(1948, pp. 126-127).

Although this statement clearly indicates that he believes

sanctioning by adults is often based upon outcome intensity, Piaget

concludes that subjective responsibility (judgment based on intention

and motive) completely replaces objective responsibility as a child

matures.

Thus these answers [his subject's responses] present us with
two distinct moral attitudes one that judges actions
according to their material consequences, and one that only
takes intentions in account. These two attitudes may co-
exist at the same age and even in the same child, but
broadly speaking, they do not synchronize. Objective
responsibility diminishes on the average as the child grows
older, and subjective responsibility gains correlatively
in importance. We have therefore two processes partially
overlapping, but of which the second gradually succeeds in
dominating the first (1948, p. 129).

In the present paper no attempt was made to unravel the apparent

contradictions in Piaget's views of moral judgment. Nevertheless, it

was expected that the research reported in the following pages would

clearly show that outcome characteristics have a decided and predic-

table effect upon judgments of appropriate punishment and reward.


Theoretical Framework

The theoretical framework for the investigations reported here

was derived from Fritz Heider's (1958a)"naive analysis of action," in

general, and his analysis of attribution of responsibility, in particular.

These analyses indicate that judgments in regard to responsibility for








a given outcome are based upon the perception of personal causality

which is, in turn, based upon an evaluation of the relative contribu-

tion of environmental forces and personal forces (power, intention

and motive) to the final outcome. The experimental approach utilized

in these investigations was based upon the assumption that Heider's

Levels in Responsibility Attribution, as they have been restated by

Shaw and Sulzer (1964), provide an accurate and comprehensive repre-

sentation of the causal structure upon which most responsibility

attribution occurs.

In regard to the relationship of the independent variables

(causal structure, outcome quality and outcome intensity) to the de-

pendent variables (attribution of responsibility and assignment of

sanctions), additional propositions were stated which were not derived

from Heider, although they appear to be congruent with his views.

These propositions may be generally stated as follows:

1. Responsibility attribution is primarily determined by the variables

represented in the structure of personal causality.

2. Outcome characteristics (i.e., quality and intensity) are a

secondary source of variability.

3. Attribution of responsibility determines whether an individual

is open to sanctioning but does not dictate the amount of sanctioning

(punishment or reward) which will be considered appropriate.

4. Judgments with respect to appropriate sanctions are primarily

determined by the perceived quality and intensity of the outcome for

which the individual is held responsible.

5. Causal structure partly determines the degree to which sanctioning

judgments are susceptible to the influence of outcome characteristics.








Before stating the specific hypotheses, it might be helpful to

examine an operational definition of each of the major variables.


Causal Structure. This complex variable is represented by

stories constructed in conformity with Heider's Levels in Responsibility

Attribution. Five different "Levels" or structures were represented,

each of which portrays a specifiable relationship between a primary

person (P) and the final outcome (X) for which his responsibility is

to be assessed. To facilitate communication these Levels have been

labelled as follows: Level I, Global Association; Level II, Extended

Commission; Level III, Careless Commission; Level IV, Purposive

Commission; Level V, Justified Commission.


Outcome Quality (Q). Favorableness (Positive Q) or unfavor-

ableness (Negative Q) of the final outcome as judged by a sample of

subjects drawn from the target population. In this paper Outcome

Quality is concerned with benefit or harm to persons rather than to

physical objects. All outcomes were evaluated as occurring to hypo-

thetical others, rather than to the self.


Outcome Intensity (I). Degree of favorability or unfavorability

of the outcome as judged by the same sample of subjects. The final set

of outcomes represents three intensities (Low, Moderate and High) for

each outcome quality. An attempt has been made to include a broad

range of outcomes, from those which are almost neutral to those which

represent an extreme degree of harm or benefit (e.g., loss of life or

rescue from death).


Attribution of Responsibility (AR). The degree to which P is









perceived as responsible for a given outcome is directly assessed by

thesubject's self-recorded ratings after having read each story re-

presenting a specific structure-outcome combination.


Assignment of Sanctions (AS). The degree as well as the direc-

tion (punishment or reward) of the sanctions considered appropriate

in each case is directly assessed by the subject's self-recorded ratings.


Hypotheses: Attribution of Responsibility (AR)

In addition to the propositions stated above the following

hypotheses are based upon two general assumptions. First, in regard

to Outcome Quality, it was assumed that there are cultural forces or

norms which are differentially directed toward the production of posi-

tive and negative outcomes. Heider (195e)refers to these forces as

"Ought" and "Ought Not." Assuming that these forces are operating,

AR for positive outcomes, which are in keeping with the Ought forces,

should be less than AR for negative outcomes, which are in opposition

to these "supra individual norms." In other words, it was assumed that

greater intention to produce an outcome would be inferred when such

production required P to overcome opposing cultural norms. Secondly,

it was assumed that the perceived degree of compliance with, or

opposition to, the Ought forces is a positive function of perceived

Outcome Intensity. Thus, the more intense the outcome, the more the

primary actor, P, will be perceived as responsible for negative outcomes.

Before listing the specific hypotheses, however, one additional

assumption should be mentioned. On the basis of the second assumption

P should be perceived as less responsible for High intensity positive

outcomes. It seems likely, however, that extremely positive outcomes









may be seen as so far beyond the call of duty that P may be seen as

more responsible if the positive outcome exceeds cultural expectations.


Hypothesis 1. Mean AR for abstract representations of Heider's

Levels increases more or less linearly up to Level IV, followed by a

drop at Level V.


Hypothesis 2. Mean AR for realistic representations of Heider's

Levels follows the same general trend as the abstract representations

except for minor variations introduced by content.


Hypothesis 3. Overall AR is significantly greater for Negative

outcomes than for Positive outcomes.


Hypothesis 4. Mean AR increases significantly as a positive

function of Outcome Intensity.


Hypothesis 5. A significant Levels by Quality interaction is

expected, primarily due to differential effects of Outcome Quality at

Levels III, IV and V. That is, the effect of Outcome Quality should

be less at Level IV (where P's intention is clear) than at Levels III

and V. The greatest effect of Quality is expected to occur at Level

V, where the amount of justification is assumed to be highly sensitive

to cultural standards.


Hypothesis 6. A significant Levels by Outcome Intensity inter-

action is expected, primarily due to relatively large differences at

Levels III and V as compared with Levels I, II and IV. Level V is

expected to show the greatest effects of Outcome Intensity because of

the postulated involvement of Intensity in determining the amount of

perceived justification.









Hypothesis 7. A weak, but significant, interaction is predicted

between Outcome Intensity and Outcome Quality. Negative outcomes

should produce greater mean AR at all Intensities, however the relative

size of this difference is expected to increase as Intensity increases.


Hypothesis 8. A significant Levels by Intensity by Quality

interaction is expected, primarily due to greater differences among

Outcome Quality-Intensity combinations at Levels III and V than at the

other levels; however, AR is expected to be greater for Negative out-

comes at all Levels and Intensities. The smallest differences among

the Quality-Intensity combinations are expected to occur at Levels I

and IV.


Hypotheses Assignment of Sanctions (AS)

In most respects the hypotheses regarding the degree to which the

subject is willing to sanction P are based upon the propositions and

assumptions which were stated earlier. Causal structure (Levels) is

expected to contribute to the Sanction ratings to some extent but the

influence of outcome characteristics is expected to be greater in the

case of AS than in AR. Outcome Quality is assumed to determine the

direction of sanction (i.e., reward or punishment) while Outcome Inten-

sity is expected to be the major determinant of the amount of sanctioning

considered appropriate. Nevertheless, AS is expected to be greater

for Negative outcomes than for Positive outcomes, due to the fact that

the former are produced in opposition to cultural norms and thus may be

perceived as necessitating punishment in order to prevent their reoccur-

rence. On the other hand, it seems likely that Positive outcomes will

produce relatively low AS unless they are judged to be "beyond the call of

duty."









Hypothesis 1. Overall AS is significantly different over the

five Levels, generally following the trend of AR.


Hypothesis 2. AS is significantly greater for Negative than

for Positive outcomes.


Hypothesis 3. AS increases as a positive function of Outcome

Intensity.


Hypothesis 4. A significant interaction is expected between

Levels and Outcome Quality primarily due to greater differences between

AS for Positive and Negative outcomes at Levels III and V than at

the other Levels.


Hypothesis 5. A significant interaction is expected between

Levels and Outcome Intensity due to greater Outcome Intensity differences

at Levels III, IV and V.


Hypothesis 6. A significant interaction is expected between

Outcome Quality and Outcome Intensity partly due to a ceiling effect

for Negative outcomes. AS for Positive outcomes is expected to be a

positively accelerated, increasing function of Outcome Intensity; for

Negative outcomes a negatively accelerated, increasing function is

expected.


Hypothesis 7. A significant interaction is expected between

Levels, Outcome Quality and Outcome Intensity due to greater effects

of Outcome Quality and Outcome Intensity at Levels III, IV and V.


In addition to these formally stated hypotheses certain expec-

tations in regard to the relationship between AR and AS were also





28


evaluated. In general, these expectations involved the assumption

that AR is primarily determined by causal structure (Levels), while

AS is primarily determined by outcome characteristics. Evidence in

support of this assumption should show that the effects of Outcome

Quality and Intensity are greater for AS than for AR. Obviously, such

findings would also be related to the more general statement that AR

determines whether P is open to legitimate sanctioning.















CHAPTER II


STIMULUS MATERIALS



The stimulus materials for both experiments consisted of 30

short stories representing all possible combinations of the three

independent variables: Outcome Quality (Q), Outcome Intensity (I),

and event structure or Levels (L). The specific outcomes and struc-

tures which were used in the experimental phases of the study were

selected in a preliminary investigation in which an attempt was made

to maximize control over these variables. The following sections

describe this preliminary investigation and present and discuss some

of the relevant findings.


Event Structure (Levels)

In earlier studies using Heider's Levels in Responsibility

(Shaw and Sulzer, 1964) the stimulus materials were constructed on

the basis of Heider's (1958a) somewhat loose descriptions. The stories

for the different levels were composed around the basic characteristics

of each level: P's association with the agent or instrument (Level I),

unintended and unforeseeable consequences (Level II), unintended-but

foreseeable consequences (Level III), intended and foreseeable conse-

quences (Level IV), and justified consequences which are both intended

and foreseeable (Level V). The main criterion for success in controlling

the Levels variable was complete agreement among two or more judges








who classified the finished stories. Although this procedure is open

to criticism, its success is attested to by the predictability and

stability of ratings obtained from different sets of stories produced

in this manner. The main disadvantage of the procedure is that it

provides only a general definition of structure and thus runs the risk

of including differently perceived structures under the same Levels

designation and thereby increasing response variance. Shaw and Sulzer

(1964) suggested that one might overcome this disadvantage by having

more than one item represent each cell. In the present case this was

not feasible since the design required a minimum of 30 stories with

only one representative per cell. With two items per cell the task

would have been cumbersome and lengthy, and would have introduced

confounding factors such as fatigue and boredom. The decision to use

one item per cell, however, greatly increased the need to evaluate

various possible structures so that the stories representing each

Level might be equated.

The basic procedure followed in evaluating the structures was

to present 23 subjects (17 males and 6 females) with a number of abstract

representations of the Levels variables and have them indicate the

degree to which person P was responsible for the outcome in each case.

The structures were "abstract" in that they contained only the barest

statement of the Levels variables with a minimum of irrelevant content.

For example, a card which was intended to represent a Level I structure,

association by friendship, read as follows:

Person P is person O's friend.
0 harms X.
Is P responsible for harming X?

Thirty-six stories of this sort were typed on cards and projected on a








screen one at a time to the entire group of subjects. To insure

attentiveness and to pace the task, the experimenter read each card

aloud while it was being projected. The subjects rated each story on

a zero-to-ten-point scale to indicate degree of P's responsibility for

each outcome. Responsibility was defined for them (as in earlier studies

in which the author participated) as "the basis upon which we might

assign praise or blame to a person for a given event." The complete

set of instructions is entered in Appendix A-i.

Mean ratings and variance were obtained for each of the stories.

From the initial set, one to three were selected from each level,

primarily on the basis of equivalence of means. These stories and

their means are presented in Appendix B-l. Figure 1 shows the mean

rating for each of the selected stories at each Level as well as the

combined mean (dotted line) for all of the stories at each level.

From the plotted points it is obvious that variability at each level

was not great. Furthermore, the overall mean closely resembles the

findings from the studies in which more realistic content has been in-

cluded. One exception is the dropoff at Level V. This is in line with

the expectations derived from Heider but has not been produced in

every case in the earlier studies. Very likely these failures were due

to the difficulty of achieving the desired balance between perceived

outcome intensity and amount of coercion. With the relatively abstract

and content-free stories it seems likely that the judgment is made with

greater emphasis upon the nature of the coercion.


Outcome Selection

Previous research on the effect of outcome on the attribution of

responsibility has been primarily confined to outcome quality, i.e.,









/ 8
/ a



- I
O- 4-

2I

S 2 o *'


l I II
I 1L Im : 1
LEVELS

Fig. 1. Mean attribution of responsibility assigned to the
abstract stories at each Level.








favorableness or unfavorableness of outcome. The degree, or intensity,

of outcome has not been the subject of direct systematic investigation,

although Wright (1960) and Shaw and Sulzer (1964) were aware of the

potential effects of this variable. Because of the equivocal nature

of the few relevant findings, it was assumed that clear-cut effects

attributable to outcome characteristics would not be obtained unless

the outcomes themselves were carefully selected. Although considerable

inter-subject agreement was expected in the designation of events as

positive or negative, it seemed likely that a greater amount of variance

might be obtained in regard to outcome intensity. In view of these

considerations, outcomes were selected on the basis of intensity ratings

assigned by subjects drawn from the target population.

From a large r*.o61 of items, 114 were selected and presented

to 40 subjects (24 males and 16 females) who designated each outcome

as favorable or unfavorable and rated it on a zero-to-ten-point scale

to indicate perceived degree of favorability or unfavorability. Each

item included an object, e.g., "a student," "a farm worker," or "a

busload of school children," and an event which affected the object,

e.g., "gets kicked out of school," "is made foreman on a ranch," and

"are seriously injured in a forest fire." Prior to obtaining the

ratings, the criteria for selection of items were established as

follows: Low intensity items should have a mean rating of .5 to 1.0

and a range of 0 to 2.0; Moderate intensity items should have a mean

of 5.0 and a range of 4.0 to 6.0; and High intensity items should have

a mean of 9.5 to 10.0 and a range of 8.0 to 10.0. The experimental

design required five outcomes of each quality-intensity combination.

The criteria, as stated above, apparently proved too rigorous. Among








the positive items only three Low, two Moderate, and nine High met

criterion, while among the negative items, three Low, no Moderate, and

nine High were acceptable.

Using the same outcomes, a second set of ratings was obtained

in which a different group of 25 subjects (23 males and 12 females)

were required to assign only a low, high and "in between" value.

(The instructions for this final rating are entered in Appendix AT2.)

From these results the final set of outcomes was selected and is-

shown in Appendix B-2, classified by Intensity and Quality. For both

Low and High outcomes, inter-subject agreement was better than 90

per cent. The problem of excessive variability in judging the moderate

intensity items still remained. Therefore, Moderate outcomes were

accepted if they elicited better than 60 per cent agreement.

Once a sufficient number of outcomes at the three levels were

available, there remained only the problem of combining them with the

appropriate event structure (i.e., Levels) to complete the stimulus

materials. Each of the acceptable outcomes within each Quality-Intensity

combination were randomly assigned to one of the abstract structures

already selected. Finally, the structures and outcomes were combined

in relatively realistic short stories and were thoroughly examined to

insure their conformity to experimental requirements. The complete set

of stimulus items classified by Level, Outcome Quality, and Outcome

Intensity is entered in Appendix B-3.


Order of Presentation

Since the early days of psychophysical experiments, the order

in which objects are presented has been regarded as a potential source

of variance. Waver and Zener (1928), Helson (1947), and others have






35


demonstrated that physical stimuli appearing early in a series may

exert an important effect by providing a subjective standard against

which later stimuli are judged. McGarvey's (1943) excellent study

provides evidence of controlled anchoring effects in the judgment of

verbal materials (moral acts and occupational status). In order to

evaluate potential order effects and increase generalizability of

the findings, four independent random orders were used in the present

experiments. The location of each stimulus item in each of the four

Orders of presentation is shown in Appendix B-4.















CHAPTER III


EXPERIMENT 1. ATTRIBUTION OF RESPONSIBILITY



Method



Subjects

This experiment utilized data from 96 subjects, 64 males and 34

females. The subjects were drawn on a volunteer basis from two large

courses of introductory psychology and were awarded research credit

for their participation.


Stimulus Materials

The stimulus materials were described in detail in the preceding

chapter. Briefly, they consisted of 30 short stories which represented

combinations of the three independent variables: structure or Levels

(L), Outcome Quality (Q), and Outcome Intensity (I). Six stories were

composed on structures representing each of the five Levels in Attribu-

tion described by Shaw and Sulzer (1964). Three of the stories at each

Level contained Positive (favorable) Outcome Quality and three contained

Negative (unfavorable) Outcome Quality. Each of these three represented

a different degree of Outcome Intensity, Low, Moderate and High, as

determined by pre-experimental ratings. The complete set of stories

was reproduced on a six-page response sheet in four different random

orders which were labelled A, B, C and D for identification.









Twenty-four subjects received each order. Beneath each story there

was a blank and a rating scale ranging from zero to ten. (An example

of the stimulus materials is entered in Appendix B-3.)


Administration

The subjects were assembled separately in three groups of 43,

32 and 24 individuals in the early evening within a period of one week.

They remained seated in a large lecture room for the 45 minutes required

for everyone to complete the assigned task. The doors were closed to

prevent interruptions; the response sheets and pencils were distri-

buted and then the instructions were read by the experimenter. The

subjects were told to read each story and decide whether or not P was

responsible for the specified event. If they thought P was not respon-

sible they were to write "No" in the blank under the story and circle

the zero end of the scale. If they thought P was responsible they

were to circle a number from one to ten to indicate the amount of P's

responsibility for the event. Thus, the ratings assigned to each story

could vary from zero (no responsibility) to ten (maximum AR). As in

earlier studies using this type of material, responsibility was defined

for the subjects by telling them that: "If a person is responsible for

something that means that we might praise or blame him for it." (The

complete instructions are entered in Appendix A-3.)

The instructions were apparently adequate for task execution since

few questions were asked and no discernible errors were made in using



The reader has probably noticed that this definition encouraged
the subjects to relate responsibility attribution to sanctioning (i.e.,
praise and blame). This may constitute a methodological weakness but it
was made necessary by a desire to follow the procedure employed by Shaw
and Sulzer (1964).









the scale. The subjects worked quietly and steadily for an average

of approximately 30 minutes. In no case did an entire session run

longer than 45 minutes, including the time required to distribute the

materials and read the instructions. At the end of each session the

subjects were asked to refrain from discussing the content with their

classmates. Whether or not they complied with this request, no

differences were found among the results from the different sessions.



Results


Table 1 shows mean AR for each stimulus item obtained under

each order of presentation, as well as mean AR and variance for all

orders combined. An analysis of variance, conforming to a Lindquist

(1953) Type VI, was conducted on the AR ratings and the results of

this analysis are summarized in Table 2. The effects which were tested

over each error term are listed under the column headed "M.S. tested."


Order Effects

The different Orders of presentation were utilized in an attempt

to increase the generalizability of the findings. Therefore, Order is a

control, rather than an experimental, variable and its effects will be

considered separately. Table 2 indicates that the main effect and all

interactions involving Order produced F ratios with associated probabili-

ties in excess of the chosen level of significance (.05). The only inter-

actions which closely approached the required level were Quality by

Order (Q x 0) and Intensity by Quality by Order (I x Q x 0). Since the

Order effects did not attain significance, there is no reason to discuss

them in more detail; the results of the analysis of variance indicate









Table 1


Mean Attribution of Responsibility Ratings for dach Stimulus Item
in each of the Four Orders of Presentation





Orders of Presentation Total

Item A B C D 7 0


I +L .62
+M 1.79
+H 1.08
-L .96
-M 1.58
-H .96

II +L 2.50
+M 1.04
+H 4.08
-L .38
-M 1.92
-H 2.38

III +L 4.79
+M 5.17
+H 2.92
-L 8.92
-M 8.92
-H 9.54

IV +L 9.13
+M 8.83
+H 9.33
-L 9.29
-M 6.92
-H 9.42

V +L 2.00
+M 6.67
+H 4.63
-L 5.96
-M 6.08
-H 9.46

Grand Mean 4.92
o2 17.31


1.41 1.50 2.75
.87 1.42 .96
1.00 .88 1.92
.67 .12 .42
1.13 1.50 1.38
1.29 1.33 1.58

3.25 4.21 4.38
1.58 1.54 1.96
6.96 6.67 6.12
.92 .92 1.38
3.38 2.79 2.92
2.67 2.88 2.42

6.33 5.92 5.83
4.21 4.08 3.00
5.17 5.17 3.46
9.25 8.00 9.21
8.33 8.54 8.58
9.63 9.12 9.54

8.67 9.29 9.33
9.17 7.96 8.79
9.00 8.92 9.67
8.79 9.25 9.13
6.92 7.17 7.79
9.25 8.83 9.13

2.21 2.75 2.21
6.92 6.75 6.33
5.54 5.87 5.33
5.25 5.29 5.00
5.25 6.62 6.08
9.04 8.00 9.29

5.13 5.11 5.20
16.41 15.45 15.99


1.57 9.59
1.26 7.63
1.22 8.00
.54 3.54
1.40 6.30
1.29 4.88

3.58 8.71
1.53 3.43
5.96 11.94
.90 2.47
2.75 9.77
2.58 7.82

5.72 10.69
4.11 10.19
4.18 9.81
8.84 3.42
8.59 4.50
9.46 1.20

9.10 3.38
8.69 4.05
9.23 2.62
9.11 3.85
7.20 5.68
9.16 2.79

2.29 7.96
6.67 8.22
5.34 9.66
5.37 12.05
6.01 9.67
8.95 3.57

5.09 16.28


* The letters L, M and H refer to Low, Moderate, and High Outcome Intensity.
The plus and minus signs refer to Positive and Negative Outcome Ouality.









Table 2


Summary of the Analysis of Variance for Attribution of
Responsibility Ratings, Experiment 1


Source df Mean Square F p


Total Between Ss 95

Order (0) 3 11.17 .55 N.S.
Error 92 20.34

Total Within 2784

Levels (L) 4 5,288.90 546.44 *
Intensity (I) 2 306.82 52.78
Quality (Q) 1 437.89 62.29 *
L x I 8 170.59 30.85 *
L x Q 4 806.36 135.49 *
I x Q 2 22.08 2.62 N.S.
L x I x Q 8 160.91 37.82 *
L x 0 12 13.28 1.37 N.S.
Ix 0 6 7.52 1.29 N.S.
Q x 0 3 17.01 2.42 N.S.
L x I x 0 24 5.77 1.04 N.S.
L x Q x O 12 6.54 1.10 N.S.
I x Q x O 6 17.18 2.04 N.S.
L x I x Q x O 24 3.81 .90 N.S.

Total Error Within 2668 M.S. Tested

error wl 368 9.68 L, L x O
error w2 184 5.81 I, I x 0
error w3 92 7.03 Q, Q x 0
error w4 736 5.53 Lx I, Lx I x Q
error w5 368 5.95 L x Q, Lx Q x O
error w6 184 8.43 I x Q, I x Q x 0
error w7 736 4.25 L x I x Q
LxIxQx0
Total for Experiment 2879


*p .05









that the findings of this experiment were not unduly influenced by

anchoring effects.


Experimental Variables

An examination of Table 2 reveals that virtually all of the

experimental variables produced significant effects upon the AR ratings.

The main effects of Levels (L), Outcome Intensity (I), and Outcome

Quality (Q), as well as most of the interactions of these variables,

produced highly significant F-ratios. The only combination of experi-

mental variables which failed to achieve the required .05 level was

the interaction between Outcome Intensity and Quality (I x Q). These

effects will be considered in more detail and related to the hypotheses

in the following paragraphs.


Hypothesis 1. Mean AR for abstract representations of Heider's

Levels increases more or less linearly up to Level IV, followed by a

drop at Level V.


Hypothesis 2. Mean AR for realistic representations of Heider's

Levels follows the same general trend as the abstract representations

except for minor variations introduced by content.


The results of the analysis of variance revealed that the Levels

variable had a highly significant effect on the AR ratings. Further

evidence in support of Hypotheses 1 and 2 is provided by Figure 2,

which compares mean AR obtained for the abstract stories (dashed lines)

with mean AR for the realistic stories (solid lines). The solid line

in Figure 2 shows overall mean AR for all six stories at each of the

five Levels and clearly reveals a more or less ogival increasing function







10
/I;\
8 /\
z / \o



6 ./

z2 I
< "ABSTRACT"

S I -

I IETII EZ: 2
LEVELS

Fig. 2. Mean attribution of r eponsibility for the abstract
and realistic stories at each Level.








up to Level IV (Purposive Commission) and a rather sharp drop at

Level V (Justified Commission). Comparing this curve with the dashed

lines representing the ratings obtained with the Abstract structures,

the similarity is striking. The major difference is a consistent

tendency for the realistic ratings to more nearly approximate the over-

all mean (5.09). This could represent regression effects, but may also

reflect a basic difference in the rating of content-free and content-

rich material. At any rate, the similarity between the two curves is

sufficient to demonstrate that the translation from the abstract

structures was reasonably successful. From the results of the analysis

of variance and the means shown in Figure 2 we may conclude that Hypo-

theses 1 and 2 were strongly supported. However, there are indications

that the statement in regard to the linearity of the relationship be-

tween AR and Levels I through IV may be an oversimplification.


Hypothesis 3. Overall AR is significantly greater for Negative

outcomes than for Positive outcomes.


The results of the analysis of variance summarized in Table 2

indicate that the main effect of Outcome Quality was highly reliable

and thus lend support to the hypothesis. Table 3 shows the overall means

for Positive and Negative outcomes as 4.70 and 5.48, respectively, and

indicates that the difference was in the predicted direction.


Hypothesis 4. Mean AR increases significantly as a positive

function of Outcome Intensity.


This hypothesis is given general support by the significant main

effect for Outcome Intensity which is shown in the analysis of variance















Table 3


Mean Attribution of Responsibility Rating and Variance for
Positive and Negative Outcomes at
each of the Five Levels


Outcome Quality

Level Positive Negative Combined


I 7 1.35 1.08 1.21

C2 8.37 5.02 6.70

II X 3.69 2.08 2.88

o2 11.25 7.35 9.94

III 7 4.67 8.97 6.82

o2 10.71 3.15 11.54

IV 7 9.01 8.49 8.75

o2 3.38 4.91 4.21

V 7 4.77 6.77 5.77

a2 11.92 10.80 12.35


Total 7 4.70 5.48 5.09

G2 15.26 17.01 16.28









table. More specific evidence is provided by Table 4 which shows the

grand means for Low, Moderate and High Intensity outcomes as 4.70,

4.82 and 5.74, respectively. The insignificant increase from L6w to

Moderate Intensity (t = .46) contrasts sharply with the jump of almost

a full scale value from Moderate to High (t = 7.83, 1918 df). From

this comparison it is obvious that the significant Intensity effect

is largely the result of greater AR for the High Intensity outcomes.


Hypothesis 5. A significant Levels by Quality interaction is

'expected, primarily due to differential effects of Outcome Quality at

Levels III, IV and V. That is, the effect of Outcome Quality should

be less at Level IV (where P's intention is clear) than at Levels III

and V. The greatest effect of Quality is expected to occur at Level

V, where the amount of justification is assumed to be highly sensitive

to cultural standards.


This hypothesis is generally supported by the significant L x Q

effect obtained in the analysis of variance. Differences between each

pair of means were evaluated by Duncan's (1955) New Multiple Range Test,

as described by Edwards (1960). The only means which were not signifi-

cantly different (p > .05) were: Level I Positive versus Level I Nega-

tive, Level III Positive versus Level V Positive, and Level III Nega-

tive versus Level IV Positive. All other differences exceeded the

shortest significant ranges of approximately .40. (Summaries of this

and other Multiple Range Tests are entered in Appendix D.)

The means and variance for Positive and Negative outcomes at

each of the five Levels are entered in Table 3; however, Figure 3

shows the interaction in somewhat clearer form. From Figure 3 it is














Table 4


Mean Attribution of Responsibility Rating and Variance
for Low, Moderate and High Outcome Intensity
at each of the Five Levels


Outcome Intensity

Level Low Moderate High Combined


I X 1.06 1.33 1.26 1.21

C2 6.80 6.93 6.41 6.70

II X 2.24 2.14 4.27 2.88

02 7.38 6.94 12.69 9.94

III 7 7.28 6.35 6.82 6.82

02 9.47 12.35 12.48 11.54

IV X 9.11 7.94 9.19 8.75

02 3.60 5.39 2.69 4.21

V Y 3.83 6.34 7.15 5.77

U2 12.34 9.01 9.84 12.35


Total X 4.70 4.82 5.74 5.09

02 17.13 14.85 16.26 16.28








10

X-. 8, 0
z NEGATIVE I

61/ x







r -

I I IV I
S4 EV POSITIVE






Fig. 3. Nean attribution of reaponalbility rating for Positive
and Negative outcaue at each Level.

Z 2 ,x
LLJ -
20 -



LEVELS

Fig. 3. Mean attribution of responsibility rating for Positive
and Negative outcomes at each Level.









apparent that there were strikingly different Quality effects at the

different Levels, but some of the effects were not in agreement with

the predictions. Although larger differences were obtained at Levels

III and V than at Level IV, the largest difference occurs not at Level

V (Justified Commission) but at Level III (Careless Commission).

Also unanticipated was the large difference in favor of Positive out-

comes at Level II and the small difference in favor of Positive out-

comes at Level IV. An examination of mean AR for each stimulus item

shown in Table 1 reveals that the story including the Moderate outcome

(IV-M) is responsible for depressing the negative mean at Level IV.

(The failure of this item to pull its expected rating is attributable

to ambiguity in regard to whether P's actions were justified. This

point will be discussed in more detail later.) The difference in

favor of Positive Outcome Quality at Level II reflects greater AR for

both High and Low Intensity items and is, therefore, a more serious

blow to the hypothesis. From the means displayed in Table 3 and Figure

3 it appears that the detailed predictions stated in Hypothesis 5

were not unequivocally supported. Important unexpected results were:

the largest difference was obtained at Level III; the highest AR for

Negative outcomes occurred at Level III; and Positive AR was greater

than Negative at Levels II and IV.


Hypothesis 6. A significant Levels by Outcome Intensity inter-

action is expected, primarily due to relatively large differences at

Levels III and V as compared with Levels I, II and IV. Level V is

expected to show the greatest effects of Outcome Intensity because of

the postulated involvement of Intensity in determining the amount of

perceived justification.









The significant F ratio obtained for L x I provides strong

general support for the hypothesis. The means and variances for Lov,

Moderate and High Outcomes, which are entered in Table 4, show a some-

what complex relationship in that mean AR for High Intensity Outcomes

is not consistently greater than for Low and Moderate Intensity. Figure

4 provides a comparison of the means in graphical form but does little

to clarify the relationship. A comparison of the means at each Level

by Duncan's Multiple Range Test revealed the following: Level I, no

significant differences; Level II, High was significantly greater than

Low and Moderate; Level III, Low was significantly greater than

Moderate; Level IV, Low and High were significantly greater than

Moderate; Level V, High was significantly greater than Moderate and

Low, and Moderate was significantly greater than Low.

Obviously, the hypothesized relationships were not obtained,

except at Level V where Outcome Intensity showed a fairly clearcut

positive relationship with AR. It seems quite likely that the signifi-

cant L x I interaction was produced by the relatively large differences

in favor of High Intensity at Level V and Level II, as compared with

the other Levels. The elevated mean for High Intensity at Level II

primarily reflects the unusually large rating pulled by the High Inten-

sity, Positive item. In view of the inexplicable basis for this large

mean, it would be unwise to take it as evidence of an Outcome Inten-

sity effect. Furthermore, although the means at Level V showed the

expected relationship, the means for Moderate and High Intensity dis-

played very little interaction over the five Levels.

In summary, the results shown in Table 4 and Figure 4 provide

only partial support for Hypothesis 6 and (possibly) indicate a failure






10
100

Z H
46- / L
/ x \
- //4 \

2 0 L
z 0

I I III I
LEVELS

Fig. 4. Mean attribution of responsibility for Low (L),
Moderate (M), and High (H) intensity outcomes
at each Level.








to achieve the necessary control over the Intensity variables, par-

ticularly in combination with Positive Outcomes. Figure 5, which shows

mean AR at each Level for only the High and Low Negative Outcomes.

reinforces this conclusion. Although the means displayed in Figure 5

are not in complete agreement with the predictions made in Hypothesis

6, they clearly reveal that P was perceived as more responsible for

High Intensity Negative Outcomes, regardless of the structure (Levels)

with which they were combined. A final point of considerable theoretical

importance is that the ratings at Level V did not drop significantly

for High Negative Outcomes as they did for High Positive Outcomes.

Table 5 shows the means for these items as 8.95 and 5.34, respectively.


Hypothesis 7. A weak, but significant, interaction is predicted

between Outcome Intensity and Outcome Quality. Negative outcomes

should produce greater mean AR at all Intensities, however the relative

size of this difference is expected to increase as Intensity increases.


The results of the analysis of variance failed to support this

hypothesis since the F ratio for I x Q failed to attain the required

level of significance. Table 6 shows the means and variances for this

interaction as well as the differences between the means for Positive

and Negative outcomes at each Intensity. The increasing difference is

in line with the prediction made in Hypothesis 7 and it is possible

that a design permitting multivariate analysis would be more sensitive

to this interaction.


Hypothesis 8. A significant Levels by Intensity by Quality

interaction is expected, primarily due to greater differences among

Outcome Quality-Intensity combinations at Levels III and V than at the







10


Z I


<- --



z 2
l -...,-0
------ -- --
2 0

I L. n n [
LEVELS

Fig. 5. Mean attribution of responsibility for High (H-)
and Low (L-) intensity Negative outcomes at each
Level.












STable 5


Mean Attribution of Responsibility Rating and Variance
for dach Outcome Quality and Intensity
at each of the Five Levels


Levels in Attribution

Outcomes I II III IV V


Positive

Low X 1.57 3.58 5.72 9.10 2.29
02 9.59 8.71 10.69 3.38 7.96

Moderate X 1.26 1.53 4.11 8.69 6.67
02 7.63 3.43 10.19 4.05 8.22

High X 1.22 5.96 4.18 9.23 5.34
02 8.00 11.94 9.81 2.62 9.66

Negative

Low x .54 .90 8.84 9.11 5.37
02 3.54 2.47 3.42 3.85 12.05

Moderate X 1.40 2.75 8.59 7.20 6.01
o2 6.30 9.77 4.50 5.68 9.67

High 7 1.29 2.58 9.46 9.16 8.95
02 4.88 7.82 1.20 2.79 3.57


Grand Mean X 1.21 2.88 6.82 8.75 5.77

02 6.70 9.94 11.54 4.21 12.35
















Table 6


Mean Attribution of Responsibility Rating and Variance for
Positive and Negative Outcome Quality at
Low; Moderate and High Outcome Intensity


Outcome Quality

Outcome Intensity Positive Negative Difference



Low X 4.45 4.95 .50

o2 15.40 18.76


Moderate X 4.45 5.19 .74

o2 15.00 14.47


High X 5.19 6.29 1.10

o2 15.09 16.85




Total 7 4.70 5.48 5.09

C2 15.26 17.01 16.28








other Levels; however, AR is expected to be greater for Negative

outcomes at all Levels and Intensities. The smallest differences among

the Quality-Intensity combinations are expected to occur at Levels I

and IV.


General support for this hypothesis was provided by the analysis

of variance which produced a significant F ratio for the L x I x Q

interaction. The means entered in Table 5, however, indicate that not

all of the expected relationships were obtained. The differences

among the means at each Level were evaluated by Duncan's Multiple Range

Tests. Mean AR for each combination of the variables is shown in Figure

6 and an examination of this figure clearly reveals the major points of

congruence with and departure from the expectations. As was predicted,

the greatest differences occurred at Levels III and V and the smallest

differences occurred at Levels I and IV. In disagreement with expectn-

tions, AR was not significantly greater for Negative Outcomes at all

Levels and Intensities; nevertheless, only a few dramatic reversals

are evident.

In Figure 5 the means at Level I are clustered together around

the minimal AR rating and partly reflect the fact that many subjects did

not hold P responsible at this Global Association level. The dispersion

at Level II, which reveals significantly greater AR for High and Low

Positive outcomes than for Negative outcomes, was not anticipated and

is in opposition to the hypothesized relationships. At Level III

much greater attribution was drawn by Negative outcomes but the expected

Outcome Intensity effects were not obtained. As predicted, the Level

IV means are clustered at near maximum AR, except for the ModerPte

Intensity, Negative item. The significant difference between the mean







iO POSITIVE
e-- H
Sx-x M M .. --
S o0-- L
z 8 NEGATIVE /
*- -- H eH Hx.
Q< x---x M / \
r: 6 o---oL *




| 2l 0

0-
I I I


LEVELS

Fig. 6. Mean attribution of responsibility for each stimulus
item at each Level.









for this item and the others at Level IV strongly supports the conten-

tion that its structure was differently perceived.2

At Level V, only the means for Moderate outcomes show lack of

agreement with the hypothesis. Mean AR for High and Low Intensity

outcomes, on the other hand, dramatically illustrate the expected

effects of Quality and Intensity. For both Positive and Negative out-

comes AR was significantly greater for High than for Low Intensity, but

Negative outcomes appeared to be less influenced by the structure of

Justified Commission. The High Intensity Negative mean did not drop

significantly from its value at Level IV (and III) while the means for

Low Negative and High Positive were equally depressed by the presence

of Outcome-facilitating circumstances. Attribution for the Low Positive

outcome at Level V strikingly demonstrates the influence of justification,

since it fell almost to its Global Association value.


Summary of Results of "xpcrirent 1

Order Effects. The order in which the stimulus materials was

presented produced no significant effects.




2This item reads as follows:

P sent in a very bad report on an employee so that he
would lose his job. The employee was fired. Is P respon-
sible for the employee losing his job?

It seems likely that the content of this item caused it to be
perceived, at least partly, in the framework of Justified Commission
(Level V). Several considerations may have led to this, including the
right of a superior to fire an employee and/or the inference that the
employee deserved to lose his job because of incompetence, etc. Un-
fortunately, this ambiguity was not discovered in the pre-test period;
if it had been, the item would have been reworded to conform more
closely to the structure of personal causality. This could be accom-
plished by modifying "employee" with the adjectives "good" or "effi-
cient" or by including a purely personal motive for P's actions.









Structure Effects. The causal structure (Levels) of the actions

described in the stories had a highly significant effect upon AR. As

predicted, AR increased up to the Purposive Commission Level and dropped

off at the Justified Commission Level. The stories which contained

realistic content produced the same general results as did the abstract

structures upon which they were composed; however, the ratings for

the realistic stories were somewhat closer to the grand mean.


Outcome Quality. The overall difference between Positive and

Negative (favorable and unfavorable) outcomes proved to be statistically

reliable and was in the predicted direction, i.e., AR was greater for

Negative than for Positive outcomes.


Outcome Intensity. The intensity (degree of favorableness or

unfavorableness) of the action outcome produced a significant effect

upon AR. As predicted, a positive relationship between Outcome Inten-

sity and AR was obtained; however, there was only a small difference

between Low and Moderate outcomes. Apparently, the significant effect

was produced by the relatively large mean for High Intensity outcomes.


Levels by Outcome Quality (L x Q). The interaction between

Levels and Quality was statistically significant and generally in con-

formity with expectations. Outcome Quality had little effect on attri-

bution at Levels I and IV but mean AR was greater for Negative outcomes

at Levels III and V. A difference in favor of Positive outcomes at

Level II occurred in all four Orders and indicates the operation of

some undetected factors.


Levels y Outcome Intensity (L x I). A significant L x I effect








was obtained but the differences among the means were not wholly in

agreement with expectations. As predicted, the largest differences

occurred at Level V and showed a positive relationship between AR and

Intensity. Relatively small differences were obtained at the other

Levels, except at Level II where the mean for High Intensity outcomes

reflected the elevated ratings assigned to the Positive items.


Outcome Intensity y Outcome Quality (I x Q). The interaction

between Outcome Intensity and Quality was the only experimental effect

which failed to achieve significance. Nevertheless, the trend was as

predicted, i.e., increasing differences between Positive and Negative

outcomes as Intensity increased.


Levels by Intensity Quality (L x I x Q). In support of the

relevant hypothesis, L x I x Q proved to be statistically reliable. Also

in keeping with expectations, large differences favoring Negative out-

comes occurred at Levels III and V which revealed a positive relation-

ship between AR and Outcome Intensity. Unanticipated findings were:

larger means for Positive outcomes at Level II, nearly equal means for

Negative outcomes at Levels III and IV, and reversals of the expected

relationship between Positive and Negative Moderate Intensity outcomes

at Levels IV and V.



Discussion


The results indicate that causal structure, as represented in

Heider's Levels, is the main determinant of attribution of responsibility,

but they also show that such attribution is influenced by the perceived

quality and intensity of the outcome for which responsibility is








assessed. These findings argue against accepting Piaget's (1932)

conclusion that mature persons attribute responsibility solely on the

basis of the actor's motive and intention. It seems likely that Piaget

was led to this erroneous conclusion because his methodology (a forced

choice between pairs) provided no way to measure the interaction or rela-

tive degree of influence of causal structure and outcome characteristics.

Generally, the shape of the curve relating AR to Levels was as

predicted and closely resembled the curves found by Shaw and Sulzer

(1964) and Sulzer, Nickols, Blum,and Brant (1963) who used the same

framework but stories with different content. The predictability and

stability of these relationships is a strong argument in favor of the

utility of the Levels approach in experimental investigations of

responsibility attribution. The similarity of the overall ratings

assigned to the abstract and realistic stories indicates that the under-

lying judgmental process was relatively rational or objective but the

significant outcome effects show that the judgments can be influenced

by subjective factors. Probably, attribution was more objective for

the abstract structures, since these materials were almost completely

devoid of social content.

In discussing studies of perceptual distortion several writers

(e.g., Thrasher, 1954; Asch, 1956; Blake, Helson, and Mouton, 1956;

Sherif and Sherif, 1956; and Kelley and Lamb, 1957) have concluded

that the effects of the experimental manipulations are greater as the

stimulus materials increase in ambiguity and lack of structure.

Ambiguity or lack of structure, in itself, doesn't necessarily produce

perceptual distortion. It is merely a condition which favors such

distortion by bringing about a greater reliance on less central stimuli.








Judgments based upon these stimuli need not be less accurate than those

based upon central ones. It is just that many perceptual experiments

(including the present one) are generally designed so that these cues,

if utilized, will lead to judgments that are not in complete agreement

with some ostensibly objective standard. It was expected, therefore,

that attribution for those causal structures which were most ambiguous

in regard to P's linkage with the final outcome would be most influenced

by non-causal factors, i.e., characteristics of the outcome itself.

In the present experiment, ambiguity per se was not varied;

however, the structure at Levels II and III, and to some extent at

Level V, was more ambiguous in regard to personal causality than was

the structure at Levels I and IV. For example, at Level I, P had no

stated part in harming or benefitting the object person (0); he was

merely "associated" with the active agent. Few individuals would infer

that P caused or intended to cause the outcome under these circumstances.

The Level IV stories also lacked ambiguity since the causal link and P's

intentions were quite clear. At Levels II and III, on the other hand,

P clearly caused the outcome but information about his intentions was

somewhat ambiguous. Presumably, this is what made these Levels more

susceptible to outcome influences. At the Extended Commission Level

(II) the nature of this influence was not in line with expectations,

but AR for Positive and Negative outcomes at the Careless Commission

Level (III) was strikingly different in the predicted direction.

Attribution for the Positive items at Level III was only slightly

greater than at Level II, while AR for the Negative items was nearly as

great as at the Purposive Commission Level (IV). Since the Level III

stories involved harmful effects which were carelessly, rather than








deliberately, caused, it would appear that: (a) the subjects felt

that P did not intentionally hamn 0 but should be held totally respon-

sible anyway because he was careless (or ruthless); or (b) they inferred

that P did intend to cause the outcome. The first alternative is based

upon an assumption that the subjects were, in this case, making judg-

ments of P:'b "legal responsibility" (Cf. Piaget, 1948, p. 123ff.). The

second alternative assumes that the negative quality of the outcomes

influenced the subjects to make erroneous inferences. Theavailable

evidence provides no basis for choosing between these alternatives, but

one rational consideration tends to favor the legal responsibility

interpretation. As Piaget (1932, 1948) and Shaw and Sulzer (1964) have.

suggested, parents tend to react strongly to carelessly caused damage,

and the installation of thoughtfulness or foresight is heavily emphasized

in child training. Thus, it may be that the differences obtained at /

Level III merely reflect an internalization of the standards which are

implicit in parental behavior. If this is the case, differences between

the sanction ratings for Negative and Positive outcomes at Level III

should be even greater than were the observed differences in AR.

The effects of outcome characteristics at Level V require a

different explanatory framework because the stories at thii level in-

volved justifiability, in addition to local causality and intention.

The stories were composed, in fact, to represent P as intentionally

causing the final outcomes.but doing so under extenuating circumstances

(externally-derived motives or coercion) which were disposed toward its

production. An attempt was made to hold the strength of these coercive

forces relatively constant for all stories (as indicated by ratings

assigned to the abstract Level V items), so that an difference in AR








would reflect only the influence of outcome characteristics. Whether

or not this attempt was successful, attribution of responsibility at

Level V was apparently quite susceptible to these influences. The

results show that the High Intensity negative outcome was perceived as

far less justified than were the other outcomes, that the Low Negative

and High Positive outcomes were seen as equally justified, and that the

Low positive outcome was perceived as so strongly justified (or incon-

sequential) that hardly any responsibility was attributed to P. These

findings are in line with the assumption that justification is largely

determined by whether, and to what degree, the action outcome is con-

gruent with, or in opposition to, culturally derived "Ought" forces or

norms. This is equivalent, perhaps,.to Heider's (1944, 1958a)statement

that justification is determined by a consideration of the motive under-

lying the act. The value of Heider's distinction between intention

(what P did) and motive (why he did it) derives from its theoretical

utility in these cases. The former is one of the conditions of personal

causality; the latter is the main basis for justification. To the degree

that P's motive is perceived as a culturally acceptable basis for his

actions, the attributor will feel that he would have felt and acted in

the same way under those circumstances, i.e., that P's behavior was

justified to some extent. Consequently, he will attribute less

responsibility to P than he would it P's motives were less acceptable.

The influence of Outcome Intensity on AR at Level V can be ex-

plained be extending this line of reasoning. As was mentioned above,

an attempt was made to hold the coercive force (externally-derived

motive) constant in all of the Level V stories. It seems obvious,

however, that a coercive force of given strength might be perceived as








sufficient justification for producing a positive or mildly negative

outcome but not for a more extreme harmful act. The results showed

that this is, in fact, true. Apparently the coercive force was per-

ceived as sufficiently strong to justify almost all of the outcomes

to some degree. Only the extremely negative outcome, which involved

murder based on a revenge motive, failed to produce AR that was lower

than at the Purposive Commission Level. Apparently, a socially accep-

table motive of comparatively great strength is required to justify,

and lower attribution for, such an outcome.

Most of this would be expected on the basis of a common sense

analysis. Nevertheless, congruency with common sense does not detract

from the theoretical value of the findings. The obtained results suggest,

for example, that it is feasible to develop a rather precise quantita-

tive statement of the relationship between Outcome Quality and Intensity

on the one hand, and degree of perceived justification or responsibility

on the other. With this information, it should be possible to construct

stimulus materials in such a way that coercive forces are matched with

outcome characteristics so that any desired degree of perceived justifi-

cation (and AR) could be obtained. With perfect matching, the effects

of Outcome Quality and Intensity on AR could be completely deleted or

greatly magnified. The data obtained in the present study suggest that

motives for negative outcomes would have to appear approximately twice

as strong as for positive outcomes of equivalent intensity to produce

an equal amount of justification. This is only a tentative statement

of the relationship; a more comprehensive empirical base is needed for

a more exact specification. Hopefully, such information will be forth-

coming. It would be a significant contribution to the development of









a quantitative theory of responsibility attribution.

In the interest of completeness, a final interpretation of out-

come effects should be mentioned which is more parsimonious, and less

dramatic, than those offered above. It is possible that outcome effects

were greatest at Levels II, III and V precisely because there was the

greatest "room" for these effects to occur at these Levels. At the

Global Association Level (I) the fact that little attribution occurred

at all provided virtually no base for outcome effects to work upon.

At the Intentional Commission Level (IV), on the other hand, P was so

clearly responsible that virtually all subjects rated him as maximally

responsible. Thus a ceiling effect and a floor effect may account for

the fact that the influence of outcome characteristics was confined

to Levels for which the abstract structures pulled AR ratings near the

middle of the scale. However, even if this partially accounts for the

interactions between Levels and Outcomes, it does not explain the direc-

tion of the differences which were obtained. The results seem to warrant

the conclusion that attribution of responsibility is greater for negative

than for positive outcomes and greater for extreme outcomes than for

mild ones.















CHAPTER IV


EXPERIMENT 2. ASSIGNMENT OF SANCTIONS



Method


Subjects

The 120 subjects (80 males and 40 females) who provided the

data for this experiment were drawn from the same population as were

those who were utilized in Experiment 1. A total of 128 subjects

actually participated, but prior to recording the data the response

sheets of eight subjects were randomly discarded to equalize the number

in the four orders of presentation.


Stimulus Materials

The stimulus materials were identical to those used in Experi-

ment 1 except that the line reading, "Is P responsible for . ?"

was deleted.


Administration

The administrative procedures, except for the instructions, con-

formed to those in Experiment 1. The subjects were assembled in three

separate groups of 40 in the same large lecture room and at the same

time of the evening within one week of the final administration of the

first experiment. When all of the subjects were seated the stimulus

materials were distributed and the experimenter read the instructions.









The subjects were told to read each story and decide whether P should

be rewarded or punished for the final event described in the story. If

they thought P should be neither rewarded or punished they were to enter

a zero on the response sheet; reward or punishment was to be indicated

by a plus or minus sign. The amount of sanctioning was indicated by

circling a number on a ten-point scale. Before they began work at

the task, the subjects were instructed to form some idea of what they

might consider to be maximum and minimum punishment and reward. This

suggestion, which resembles Kilpatrick and Cantril's (1960) "self-

anchoring scaling" in some respects, was employed to orient the subjects

to task requirements and to establish individually meaningful scale

limits before they began the actual process of rating. (The complete

instructions are entered in Appendix A-4.)



Results


The scale values obtained in this experiment ranged from +10

(maximum reward) through zero (neither punishment nor reward) to -10

(maximum punishment). Because a few ratings of objectively inappropriate

sign were assigned (e.g., a rating of -2 to a Positive outcome or a

,rating of +2 to a Negative outcome), the following procedure was adopted:

1. A constant of -10 was added to all ratings assigned to negative

outcomes and a constant of +10 was added to all ratings assigned

to positive outcomes.

2. Statistical analyses were carried out on the absolute values of

the resulting quantities.

Thus, in the tables which follow, a mean value of 20 indicates that the

maximum appropriate sanction (either reward or punishment) was assigned,









a value of zero indicates that no sanctions were assigned, and values

less than 10 indicate that inappropriate sanctions were assigned. The

degree of "inappropriateness" can easily be determined by btracting

such values from 10.

As in Experiment 1, a Lindquist Type VI analysis of variance was

conducted on the ratings. A summary of this analysis appears in Table

7. The mean sanction ratings (AS) assigned to each stimulus item in

each of the four orders of presentation are displayed in Table 8.


Order Effects

Table 7 reveals that the order of presentation had a significant

and pervasive effect on the sanction ratings. The main effect, and

virtually all of the interactions of Order with the other variables

produced F ratios with associated probabilities less than the selected

.05 level of significance. The interaction of Order with the Levels

variable (L x 0) was the only exception. These findings were unexpected

and, because of their scope and the complexity of the interactions in-

volved, they defy meaningful psychological interpretation. An examina-

tion of Table 8 does little to clarify the situation. The differences

among the grand means were considerably greater than were obtained in

Experiment 1 (.99 versus .28), as were the differences among the means

shown for each stimulus item. The only reasonable conclusion which can

be drawn is that sanction judgments are far more sensitive to the order

of presentation of these materials than are judgments of responsibility.

In view of this, it is fortunate that the design provided a way of

extracting this source of variance. In the following presentation the

effects of Order are not considered. This decision was based on the

arbitrary definition of Order as a control, rather than an experimental,

variable.









Table 7


Summary of the Analysis of Variance for the
Sanction Ratings, Experiment 2


Source df Mean Square F p


Total Between Ss 119

Order 3 155.37 8.73 *
Error 116 17.79

Total Within 3480

Levels (L) 4 2,039.41 284.04 *
Intensity (I) 1 2,776.84 538.15 *
Quality (Q) 1 3,775.64 474.92 *
L x I 8 273.02 56.88 *
L x Q : 4 1405.32 259.28 *
I x Q 2 15.25 2.13 N.S.
L x I x Q 8 355.92 79.62 *
L x O 12 3.70 .72 N.S.
I x 0 6 53.33 6.71 *
Q x 0 3 113.75 18.99 *
L x I x 0 24 10.28 2.14 *
L x Q x 0 12 16.93 3.12 *
I x Q x 0 6 19.71 4.41 *
L x I x Q x 0 24 15.70 2.62

Total Error Within 3364 M. S. Tested

error wl 464 7.18 L, L x 0
error w2 232 5.16 I, I x 0
error w3 116 7.95 Q, Q x 0
error w4 928 4.80 L x I, L x I x 0
error w5 464 5.42 L x Q, L x Q x 0
error w6 232 4.47 I x Q, I x Q x O
error w7 928 5.99 L x I x Q
LxIxQxO
Total for Experiment 3599


* p < .05










Table 8


Mean Sanction Ratings for Mach Stimulus Item in each
of the Four Orders of Presentation


Orders

Item* A B C D Total


I +L 10.03 10.67 10.09 10.23 10.26
+M 9.90 10.33 10.63 9.90 10.19
+H 11.00 10.10 11.90 11.67 11.17
-L 9.83 9.53 10.00 9.90 9.82
-M 10.17 10.23 10.00 10.13 10.13
-H 11.90 13.73 10.97 11.27 11.97

II +L 11.27 12.03 10.50 10.73 11.13
+M 9.97 10.03 10.17 9.83 10.00
+H 11.03 11.40 11.40 11.23 11.27
-L 10.03 10.13 10.07 10.00 10.06
-M 10.17 11.23 11.13 10.87 10.85
-H 12.37 13.60 12.13 10.53 12.16

III +L 9.80 9.70 9.60 9.63 9.68
+M 9.20 8.67 9.10 8.63 8.90
+H 8.43 10.93 9.80 10.37 9.88
-L 12.67 15.33 12.67 11.83 13.12
-M 16.37 17.17 15.17 14.53 15.96
-H 17.90 19.07 18.13 17.37 18.12

IV +L 12.70 13.93 12.53 13.23 13.10
+M 12.80 14.63 11.83 11.63 12.72
+H 17.43 19.03 19.27 17.90 18.41
-L 13.07 14.50 14.67 11.23 13.37
-M 12.83 13.93 12.60 12.90 13.07
-H 17.60 18.27 18.57 17.20 17.91

V +L 8.93 9.33 9.33 9.87 9.37
+M 12.27 13.03 11.07 11.80 12.04
+H 11.60 11.83 11.23 12.47 11.78
-L 12.13 11.97 12.17 11.63 11.97
-M 14.27 14.37 14.57 12.80 14.00
-H 18.90 18.47 17.90 16.93 18.05

Grand Mean 12.21 12.93 12.31 11.94 12.35
o2 13.84 13.12 12.36 12.79

* The letters L. M and H refer to Low. Moderate, and High Outcome Intensil


The plus and minus signs refer to Positive and Negative


ty.


Outcome Quality.









Experimental Variables

The analysis of variance summarized in Table 7 shows that the

experimental variables produced highly significant effects on the sanc-

tion ratings. Only the Intensity by Quality interaction failed to

produce a significant F ratio. Incidentally, it is interesting to note

that the experimental variables accounted for more than 95 per cent

of the total variance, while Order, interactions involving Order, and

error accounted for the remainder. In the following pages, the nature

of the experimental effects will be examined and related to the hypotheses.


Hypothesis 1. Overall AS is significantly different over the

five Levels, generally following the trend of AR.


The analysis of variance revealed that the Levels variable had

a highly significant effect on the sanction ratings. Mean AS and AR

for each Level is shown in Figure 7 and an examination of this figure

indicates that the sanction ratings and responsibility ratings followed

the same general trend, as was expected. It is also apparent that

mean AS was lower than AR at every Level. The size of this difference

increased progressively from Levels I through IV and diminished at

Level V. These results have rich theoretical implications; however,

caution must be exercised in interpreting them in view of the fact

that the scale used in the second experiment differed in some respects

from the scale used in Experiment 1.


Hypothesis 2. AS is significantly greater for Negative than for

Positive outcomes.


The significant F ratio produced by Outcome Quality provides






20

0
/
S16- /
/ \

14


S2 AS
e---
S10

Z o
I It mI I
LEVELS


Fig. 7. Mean attribution of responsibility and assignment
of sanctions at each Level.









general support for this hypothesis while the means shown at the bottom

of Table 9 reveal that the difference was in the expected direction.

The relationship between Outcome Quality and Levels will be considered

under Hypothesis 4.


Hypothesis 3. AS increases as a positive function of Outcome

Intensity.


The analysis of variance showed that Outcome Intensity had a

significant effect on the sanction ratings. The means entered at the

bottom of Table 10 reveal that the expected positive relationship be-

tween AS and Outcome Intensity was obtained. Mean AS for the Low,

Moderate and High Intensity outcomes, respectively, were 11.19, 11.79

and 14.07. Comparisons of these means by t tests yielded significant

values of 3.53 for the difference between Low and Moderate and 10.00

for the difference between Moderate and High (df = 2348, p < .05 in

both cases). As was the case with the AR ratings, the magnitude of

the difference increased as Outcome Intensity increased.


Hypothesis 4. A significant interaction is expected between

Levels and Outcome Quality primarily due to greater differences between

.AS for Positive and Negative outcomes at Levels III and V than at the

other Levels.


The analysis of variance revealed that the Levels by Quality

interaction was highly significant. Figure 8, which shows mean AS for

Positive and Negative Outcome Quality,"indicates that the interaction

was produced in the predicted manner. The results of Duncan's Multiple

Range Test provide additional statistical support for the hypothesis.














Table 9


Mean Sanction Rating and Variance for Positive and'Negative
Outcomes at each of the Five Levels




Outcome Quality

Level Positive Negative Combined


I X 10.54 10.64 10.59

02 4.12 4.04 4.07

II X 10.80 11.02 10.91

02 4.11 5.02 4.57

III X 9.49 15.73 12.61

02 7.60 11.29 19.19

IV X 14.74 14.78 14.76

02 15.92 14.95 15.04

V 7 10.94 14.67 12.87

2 8.28 14.43 14.60


Total T 11.33 13.37 12.35

02 11.19 14.23 13.75














Table 10


Mean Sanction Rating and Variance for Low, Moderate
and High Outcome Intensity at
each of the Five Levels


Outcome Intensity

Level Low Moderate High


I X 10.04 10.16 11.57

O2 1.12 1.44 8.24

II X 10.60 10.42 11.71

02 3.27 2.30 7.19

III X 11.04 12.43 14.00

02 7.26 21.11 25.93

IV X 13.23 12.90 18.16

02 7.65 14.13 6.04

V X 10.67 13.02 14.92

02 7.21 9.13 18.50


Total 7 11.19 11.79 14.07

o2 6.52 11.12 18.99






20


(D 18
Z
16 -

w NEGATIVE / A---- x
14 14
/
Z 12 -
< 0

io -
tPOSITIVE
8 -


I 11 11S IZ I2
LEVELS

Fig. 8. Mean assignment of sanctions for Positive and Negative
outcomes at each Level.









Mean AS for Positive and Negative outcomes was not significantly

different at the Global Association (I), Extended Commission (II), and

Purposive Commission (IV) Levels, but was significantly greater for

Negative outcomes at the Careless Commission (III) and Justified Commis-

sion (V) Levels. A striking aspect of this relationship is that mean

AS for Positive outcomes at Level III was less than 10. This indicates

that the subjects felt that P should be mildly punished rather than

rewarded for producing a good outcome carelessly. Unexpectedly,

mean AS for Negative outcomes reached its peak at Level III and was

almost a full scale value larger than at Level IV. At Level V the mean

for Positive outcomes dropped far below its maximum value at Level IV,

while the Negative mean showed only a slight (insignificant) drop. The

exact mean values and variances for the relationships shown in Figure 8

are entered in Table 9.

Since AS for the High Negative outcome at Level IV was less than

was anticipated, it seems likely that the item itself was at fault. All

of the High Intensity Negative outcomes received maximum Intensity

ratings in the pretest, but this particular item differed from the

others in an important respect. Whereas all of the other High Negative

items included loss of life or severe mutilation, the item at Level IV

depicted P as intentionally causing the ouster of a college student.

Apparently, this outcome was not perceived as sufficiently extreme to

require maximum punishment.


Hypothesis 5. A significant interaction is expected between

Levels and Outcome Intensity due to greater Outcome Intensity differences

at Levels III, IV and V.









The means and variances for Low (L), Moderate (M) and High (H)

Intensity outcomes at each of the five Levels are entered in Table 10.

The analysis of variance revealed that the Levels by Intensity inter-

action was highly significant. The means were compared by Duncan's

Multiple Range Test. Figure 9 shows the relationship in graphic form.

It is apparent that the hypothesis was generally confirmed since the

greatest differences among the means were obtained at Levels III, IV

and V, as was predicted. An examination of the means for Low and HIgh

outcomes clearly reveals that a significant positive relationship between

AS and Intensity was obtained at all Levels and that the strength of

the relationship generally increased from Levels I through IV, and

decreased slightly at Level V. The means for Moderate outcomes, how-

ever, were located as anticipated (i.e., between the means for L and

H) only at Levels III and V. It seems likely that the Moderate mean

would have fallen between the means for Low and High at Level IV as

well, if item IV-M had been properly controlled. This conclusion is

reinforced by the fact that mean AS for the Moderate Positive item was

14.00.


Hypothesis 6. A significant interaction is expected between

Outcome Quality and Outcome Intensity partly due to a ceiling effect

for Negative outcomes. AS for Positive outcomes is- expected to be a

positively accelerated, increasing function of Outcome Intensity; for

Negative outcomes a negatively accelerated, increasing function is;

expected.


The analysis of variance revealed that the Quality by Intensity

interaction failed to achieve the required level of significance.






20


18


16


14

12


10

0


I.


Fig. 9. Mean assignment of sanctions for Low (L), Moderate
(M), and High (H) outcomes at each Level.


z



Cr.

z

w
2


*-- H
x-x M
- 0---o L


-


I II E
LEVELS









Table 11 shows mean AS and variance for the Positive and Negative out-

comes at each Outcome Intensity. An examination of the means entered

in this table indicates that AS for both Positive and Negative outcomes

tended to be positively accelerated functions of Outcome Intensity.


Hypothesis 7. A significant interaction is expected between

Levels, Outcome Quality and Outcome Intensity due to greater effects of

Outcome Quality and Outcome Intensity at Levels III, IV and V.


The results of the analysis of variance indicated that the L x

Q x I interaction was statistically significant, thus providing general

support for the hypothesis. Mean AS and variance for each combination

of the experimental variables are entered in Table 12. Separate

Multiple Range Tests were conducted in order to compare the means at

each Level. Figure 10 shows the mean sanction ratings for each item in

graphical form. From the relationships shown in this figure and the

results of the multiple range comparisons it is apparent that the inter-

action was produced in the predicted manner. Generally, the differences

among the Quality-Intensity combinations were much greater at the three

higher Levels.

Since Figure 10 presents the findings in much more detail than

did the other figures it reveals several relationships that are worthy

of consideration. In general, it shows that assignment of sanctions

was affected at least as much by outcome characteristics as by causal

structure, particularly at the higher Levels. At Level I hardly any

sanctions were assigned at all except for the High Intensity outcomes.

This was only slightly less true at Level II. The means shown at Level

III, however, reveal a dramatic contrast between AS for Negative and
















Table 11


Mean Sanction Rating and Variance for Positive and Negative
Outcome Quality at Low, Moderate and
High Outcome Intensity




Outcome Quality

Outcome Intensity Positive Negative Combined



Low X 10.71 11.67 11.19

2 6.01 6.58 6.52


Moderate 7 10.77 12.80 11.79

02 8.13 12.07 11.12


High N 12.50 15.64 14.07

o2 17.40 15.69 18.99




Total 7 11.33 13.37 12.35

02 11.19 14.23 13.75












Table 12


Mean Sanction Rating and Variance for each Outcome Quality
and Intensity at each of the Five Levels


Levels in Attribution

Outcomes I II III IV V


Positive

Low 7 10.25 11.13 9.68 13.10 9.37
02 1.13 5.90 1.70 8.11 3.68

Moderate X 10.19 10.00 8.90 12.72 12.04
02 2.04 0.49 8.14 12.52 7.80

High X 11.17 11.27 9.88 18.41 11.78
02 8.64 5.04 11.83 6.60 9.10

Negative

Low X 9.82 10.06 13.12 13.37 11.97
02 1.02 0.09 6.21 7.23 7.37

Moderate 7 10.13 10.85 15.96 13.07 14.00
02 0.86 3.78 9.13 15.79 8.61

High 7 11.97 12.16 18.12 17.91 18.05
02 7.58 9.01 6.07 4.99 8.27


Grand Mean X 10.59 10.91 12.61 14.76 12.87

02 4.07 4.57 19.19 15.04 14.60






20


POSITIVE
e--* H
x--x M
- o--o L
NEGATIVE
*---- H
- x----x M
o----o L


z
J



LEVELS
LEVELS


Fig. 10. Mean assignment of sanctions for each stimulus
item at each Level.


I
/
I


(D
Z








<:
Ll


V


8

O "


r
IF. :


II









Positive outcomes. High and Low Intensity Positive outcomes received

objectively inappropriate sanction ratings of approximately the same

amount, while AS for the Negative outcomes was much greater and'showed

a highly significant positive relationship with Outcome Intensity. At

Level IV this relationship was disturbed by the failure of the Moderate

Negative item to pull its expected rating, but it appeared quite clearly

again at Level V. These differences among the means at the three

higher Levels provide considerable support for the prediction that the

sanction ratings would be less influenced by causal structure than by

outcome characteristics.


Summary of Results of Experiment 2


Order Effects. The analysis of variance revealed that the order

of presentation of the stories had a pervasive influence on the sanction

ratings; however, these effects could not be meaningfully interpreted

because of their complexity.


Structure Effects. As was expected, causal structure (Levels)

produced mean AS ratings which followed the same trend as AR, but

structure effects were less pronounced in the case of the sanction

ratings.


Outcome Quality. In line with expectations, the mean amount of

sanctions assigned to Negative outcomes was significantly greater than

mean AS for Positive outcomes.


Outcome Intensity. Outcome Intensity had a significant effect

upon the sanction ratings. The results showed that AS was a positively

accelerated, increasing function of Outcome Intensity as was hypothesized.









Levels bOutcome Quality. As was expected, a significant

interaction was obtained for Levels by Quality. Mean AS was almost iden-

tical for Positive and Negative outcomes at Levels I, II, and IV, but

was reliably greater for Negative outcomes at Level III (Careless

Commission) and Level V (Justified Commission). Inappropriate sanctions

were assigned to Positive outcomes at Level III, whereas AS for Negative

outcomes reached its maximum at this Level.


Levels b Outcome Intensity. As predicted, a significant inter-

action was obtained which indicated that AS was greater for High than

for Low outcomes at all five Levels and that the difference was greatest

at the Careless, Purposive and Justified Commission Levels. Mean AS

for Moderate outcomes was less clearly in line with expectations,

mainly because of low ratings assigned to the Moderate Negative item

at Level IV.


Outcome Quality by Outcome Intensity. The analysis of variance

revealed that the expected interaction between Quality and Intensity

was not significant. This was the only hypothesis which failed to

obtain substantial support.


Levels y Intensity by Quality. A significant second-order

interaction was obtained. In keeping with the relevant hypothesis,

differences among the ratings for the Quality-Intensity combinations

were found at all Levels but were greatest at the three highest Levels.

A strong positive relationship between AS and Intensity was apparent

for Negative outcomes at Levels III and V. Mean AS for Positive out-

comes showed only mild Intensity effects, reflecting the fact that they

received relatively weak sanction ratings at all Levels, except at









Level IV, where the High Positive item produced slightly greater AS

than did the High Negative item.



Discussion


The results suggest that if sanctions are assigned, outcome

characteristics exert a stronger influence on the amount of sanctioning

considered to be appropriate than does the causal structure under which

the outcomes are produced. The major contribution of causal structure

seems to be in determining whether or not sanctions will be assigned.

Outcome Quality had relatively little effect on AS at the Global

Association (I) and Extended Commission (II) Levels, but exerted a

strong influence on the ratings at Levels III and V. The greatest

effect of Outcome Quality occurred at Level III (Careless Commission),

where the amount of sanction assigned for Negative outcomes was much

greater than for Positive outcomes and revealed a strong positive rela-

tionship with Outcome Intensity. The subjects showed a consistent

reluctance to "reward" P for carelessly-produced Positive outcomes and,

in fact, a slight tendency to "punish" him for these effects. Negative

sanctions at this Level were expected to be rather susceptible to the

,influence of Intensity. The obtained relationships seem to confirm

Piaget's observation that adults react harshly to carelessly caused

damage and tend to "lose their tempers in proportion to the amount of

damage done . ." (1948, p. 127). The observed tendency to punish for

carelessness that produces benefit to others was not expected, however,

and it has less apparent basis in common sense. At first glance, it is

difficult to see why rewards (albeit in small amounts) were assigned

to P when he could not possibly have known that the benefit would occur









(i.e., at Level II), and yet not to reward him when he might have

foreseen the benefit to 0. Perhaps the explanation resides in the fact

that benefit to 0 was foreseeable but, nevertheless, was produced

carelessly (i.e., without thought of 0). An examination of the three

stories involved seems to support this interpretation. (See Appendix

B for the complete stories.)

In the Low Ihtensity story, a little girl got an extra serving

of ice cream and cake because P left the table to play just after the

refreshments had been served. If P's behavior was perceived as im-

polite, or even thoughtless, this would explain the fact that the

subjects felt he should be punished. In the Moderate Intensity story,

P caused another salesman to get a bonus because he did not feel like

driving to the next town to close a sale. It seems likely that punish-

ment was assigned to P in this case because he was perceived as lazy or

careless about his occupational commitments. In the High Intensity

item, P's blind neighbors were saved from their burning house by the

firemen P had called in concern about damage to his own house. In this

case, he was almost certainly punished for his ruthless lack of concern

for his neighbor's welfare and his selfish interest in his own property.

The results obtained seem to warrant the conclusion that carelessness,

in itself, is negatively valued in our culture and is therefore con-

sidered to require some degree of negative sanctioning regardless of

the quality of the outcome which is produced.

At the Purposive Commission Level (IV) mean AS was closely

clustered around a relatively low value, except for the two High Intensity

items. Due to the operation of uncontrolled variables, two of the Nega-

tive items failed to pull their expected ratings; nevertheless, the









results clearly indicated that Outcome Intensity exerts a strong

influence on sanction judgments even when P is depicted as intentionally

causing the event. Outcome Quality, on the other hand, failed 1o show

any striking effects upon AS at this Level. It must be tentatively

concluded, therefore, that individuals are as willing to reward as to

punish when the event is obviously produced intentionally.

At Level V (Justified Commission) the results revealed that both

Outcome Quality and Intensity had powerful effects on the amount of

sanction assigned. The differences among the mean ratings rather

unequivocally dictate the conclusion that: (a) when the amount of

coercive force is held relatively constant, P is sanctioned less for

producing beneficial effects than for producing harmful effects; and

(b) the amount of negative sanction considered appropriate increases

with perceived seriousness (Intensity) of the outcome.

The results of this experiment lend themselves to the inter-

pretation that individuals in our culture are expected to benefit one

another. There is, therefore, no reason to reward them for good effects

unless the benefit is so great, and produced under such clear-cut con-

ditions of personal causality, as to appear extraordinary or even heroic.

The High Intensity Positive story at Level IV, which was the only High

Intensity Positive item to produce AS equal to the equivalent Negative

item, seems to conform to what most people would regard as heroic. In

this story, P left his position of safety and risked his life in order

to save a number of children from a forest fire. On the other hand,

bad effects seem to require punishment even when they are not intended

or are produced under some form of coercion.

Recently, Solomon (1964) discussed the perennial controversy over









the relative merits of using punishment and reward in the control of

behavior. He suggested that experimentation on the effects of punish-

ment has been discouraged, in large part, by the "legend" that punish-

ment inevitably produces disabling side effects such as rigidity and

impaired cognitive functioning. Some writers (e.g., Skinner, 1948,

1961) doggedly maintain that positive reinforcement is a more effective

means of control. This belief is in keeping with a humanitarian point

of view, but, whether or not it is typical of psychologists, it is

apparently not shared by most laymen. The results of the present study

indicate that the tendency to punish is considerably stronger than is

the tendency to reward. Whether one prefers to interpret this in terms

of retributive counter-aggression or behavior control, one conclusion

is unequivocally dictated. For the population represented here large

amounts of reward are considered to be appropriate only for extraordi-

narily positive action outcomes. It remains to be seen whether this

finding is restricted to hypothetical situations such as were used in

the present experiments.

The findings reported here argue against Piaget's (1932, 1948)

somewhat idealistic conclusion that moral judgments of this sort are

made in accordance with a rather sophisticated consideration of the

actor's intentions and motives. They do indicate, however, that moral

judgments are strongly influenced by culturally derived expectations,

and that the primary utilization of sanctions is in punishing indivi-

duals who fail to act in accordance with these expectations. The major

evidence of rational considerations appears in deciding whether or not

P is open to sanctioning and (if he is) in the tendency to make "the

punishment fit the crime."















CHAPTER V


SUMMARY AND CONCLUSIONS



The purpose of the research reported in these pages was to

evaluate the effects of causal structure and outcome characteristics

upon attribution of responsibility and sanction assignment. Causal

structure was represented by stories which were modelled upon Heider's

(1958a) Levels of Responsibility Attribution. The two outcome charac-

teristics, quality and intensity, were based upon ratings made by

subjects drawn from the target population. It was generally hypothesized

that the amount of responsibility attributed (AR) to an individual (P)

for an action outcome would be principally determined by the causal

structure within which the outcome was produced, and that outcome charac-

teristics would exert a detectable, but less powerful, influence. The

results from the first experiment provided unequivocal support for these

general expectations, although all of the more detailed hypotheses were

not supported. The second major set of hypotheses concerned the effects

of these same variables upon the assignment of sanctions (AS). In this

case, it was hypothesized that outcome characteristics were the primary

determinants and that causal structure would play a somewhat secondary

role. The results from the second experiment partially confirmed these

expectations but showed that the general trend of AS over the different

Levels was fairly similar to AR. These findings and the conclusions

drawn from them will be briefly summarized in the following sections.









Attribution of Responsibility

In Experiment 1, significant effects were found for Levels (L),

Outcome Quality (Q) and Outcome Intensity (I) and for all of thd inter-

actions among these variables, except for the interaction between Out-

come Quality and Intensity. As expected, relatively small amounts of

responsibility were attributed at the Global Association and Extended

Commission Levels, where the link between P and the final outcome was

weak and involved no possibility that he intended to harm or benefit

the object person (0). At the Careless and Purposive Commission Levels,

however, P was held highly responsible for producing the outcome. An

unexpected finding was that P was held equally responsible for harmful

acts at these two Levels. Two possible explanations were offered. The

first assumes that ambiguity in the Level III (Careless Commission)

structure allowed the subjects to infer that P did intend to harm 0,

in which case the Level III and Level IV (Purposive Commission) struc-

tures would have been perceived as identical. The second alternative

assumes that carelessness is negatively valued in our culture and that

P was seen as "legally" responsible because of his failure to consider

the probable consequences of his behavior.

Although these alternatives are not mutually exclusive, it was

pointed out that the second alternative would be favored if the sanction

ratings for Negative and Positive outcomes at Level III showed even

greater differences. (Large differences at this Level were obtained in

Experiment 2 )

At Level V, where P was depicted as causing the final outcome

under coercive circumstances, AR was much less than at Levels III and

IV. This was interpreted as indicating that P's actions were perceived









as justified to some extent. The amount of responsibility attributed

at this Justified Commission Level was highly related to outcome

characteristics. Generally, much less responsibility was attributed

for Positive than for Negative outcomes, and AR showed a strong posi-

tive relationship with Outcome Intensity. These findings suggest that

the amount of perceived justification is jointly determined by the

strength of the coercive force and the quality and intensity of the out-

come.

The results confirmed the expectation that causal structure (as

represented by Heider's Levels) is the primary determinant of respon-

sibility attribution. Apparently, however, attribution is highly sen-

sitive to the perceived quality and intensity of the effects for which

responsibility is assessed. The influence of outcome quality on AR was

interpreted as indicating that less responsibility is attributed for

actions which are congruent with culturally derived "Ought" forces than

are actions which occur in opposition to these normative standards. It

seems likely that greater intentionality (and less justification) is

inferred when the individual must overcome these forces to produce a

given outcome.

Outcome intensity has its greatest effects when the structure

encourages a consideration of the motives underlying P's actions. With

motive and outcome quality held constant, more personal responsibility

is attributed as outcome intensity increases. Apparently, the degree

of benefit or harm which results from P's actions is one of the major

determinants of perceived justification.


Assignment of Sanctions

In Experiment 2, as in Experiment 1, significant effects were









found for Levels, Outcome Quality, Outcome Intensity and all interactions

among these variables except for the interaction between Q and I.

(Although significant interactions were also obtained between these

variables and Order of presentation, these findings are excluded from

the present summary.) Virtually all of these effects were produced in

the manner set forth in the detailed hypotheses. The effects of causal

structure were most clearly demonstrated at the Global Association and

Extended Commission Levels, where fewer subjects assigned sanctions than

at the higher Levels. This had the effect of reducing the apparent in-

fluence of outcome characteristics at Levels I and II and supports the

contention that sanctions are not assigned in cases where P is not con-

sidered responsible. At Levels III (Careless Commission), IV (Purposive

Commission) and V (Justified Commission), the influence of causal struc-

ture was revealed in the differences among the outcome effects .

At Level III, negative sanctions (punishment) were assigned for

both harmful and beneficial effects; however, AS was much greater for

harmful outcomes and showed a significant positive relationship with

Outcome Intensity. These relationships suggest that carelessness is

negatively valued in our culture (as was suggested earlier) and,,there-

fore, that such behavior is seen as deserving punishment regardless of

the quality of the event produced. At the Purposive Commission Level,

virtually all subjects assigned sanctions but there were no apparent

differences attributable to Outcome Quality. (This was partly due to

the fact that the Moderate-Negative item included uncontrolled variables.)

Relatively weak sanctions were assigned to all but the two High Intensity

outcomes. Mean AS for the High Intensity Positive item at this Level

was the only positive mean to exceed the Grand Mean for the experiment.