University of Southern California Dissertations and Theses

Conflicting Motives In The Prisoner'S Dilemma Game

(USC Thesis Other)

Conflicting Motives In The Prisoner'S Dilemma Game

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS
This material was produced from a microfilm copy of the original document. While
the most advanced technological means to photograph and reproduce this document
have been used, the quality is heavily dependent upon the quality of the original
submitted.
The following explanation of techniques is provided to help you understand
markings or patterns which may appear on this reproduction.
1. The sign or "target" for pages apparently lacking from the document
photographed is "Missing Page(s)". If it was possible to obtain the missing
page(s) or section, they are spliced into the film along with adjacent pages.
This may have necessitated cutting thru an image and duplicating adjacent
pages to insure you complete continuity.
2. When an image on the film is obliterated with a large round black mark, it
is an indication that the photographer suspected that the copy may have
moved during exposure and thus cause a blurred image. You will find a
good image of the page in the adjacent frame.
3. When a map, drawing or chart, etc., was part of the material being
photographed the photographer followed a definite method in
"sectioning" the material. It is customary to begin photoing at the upper
left hand corner of a large sheet and to continue photoing from left to
right in equal sections with a small overlap. If necessary, sectioning is
continued again — beginning below the first row and continuing on until
complete.
4. The majority of users indicate that the textual content is of greatest value,
however, a somewhat higher quality reproduction could be made from
"photographs" if essential to the understanding of the dissertation. Silver
prints of "photographs" may be ordered at additional charge by writing
the Order Department, giving the catalog number, title, author and
specific pages you wish reproduced.
5. PLEASE NOTE: Some pages may have indistinct print. Filmed as
received.
Xerox University Microfilms
300 North Zeeb Road
Ann Arbor, Michigan 48106
74-28,417
BEALE, Darryl Kurtz, 1944-
CONFLICTING MOTIVES IN THE PRISONER’S DILEMMA
GAME.
University of Southern California, Ph.D., 1974
Psychology, experimental
i
\
i
1 University Microfilms, A X ERO X Company, Ann Arbor, Michigan
THIS DISSERTATION HAS BEEN MICROFILMED EXACTLY AS RECEIVED.
CONFLICTING MOTIVES IN THE PRISONER'S DILEMMA GAME
by
Darryl Kurtz Beale
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
August 1974
UNIVERSITY O F SOUTHERN CALIFORNIA
THE GRADUATE SCHOO L
UNIVERSITY PARK
LOS ANGELES, CALI FORNIA 9 0 0 0 7
This dissertation, written by
.............D£rr^l__Kurtz__Beale............
under the direction of h%?..... Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by The Graduate
School, in partial fulfillment of requirements of
the degree of
D O C T O R OF P H IL O S O P H Y
DISSERTATION COMMITTEE
Chairman
ACKNOWLEDGMENT
During my graduate experience I have become acutely
aware of the important influence which various teachers and
colleagues have exercised upon my thinking. It would be a
formidable task to provide a comprehensive list of all of
those individuals, so I hereby acknowledge their influence
collectively. Special thanks, however, are extended to
Dr. James P. Kahan for his many helpful suggestions during
my graduate career.
I must acknowledge the forbearance and assistance of
my wife Barbara during the last few months of this project.
Her encouragement and the absence of complaints while she
was temporarily widowed for this cause were greatly
appreciated.
Finally, I wish to extend my thanks to my parents,
Mr. and Mrs. David Beale, without whose help over the years
energies by necessity would have been directed to less
creative endeavors.
TABLE OF CONTENTS
LIST OF TABLES
iv
LIST OF FIGURES
V
Chapter
I. INTRODUCTION 1
Experimental Games in Psychology
Prisoner's Dilemma
Behavior in PD Games
Motives in PD Games
Manipulation of Motives
Levels of Cooperation and
Motivational Orientations
Effects of Selecting Motives
Motivational Inclinations in
Iterated Play
II. THE PROBLEM 22
III. METHOD 29
Subjects
Apparatus
Procedure
IV. RESULTS 35
V. DISCUSSION 55
BIBLIOGRAPHY 66
APPENDICES 72
iii
LIST OF TABLES
Table _ Page
1. Contingency Rates of Cooperation Based on
motive and outcome on previous trial .... 36
2. Proportion of Cooperative Response and
Proportion of Time Players Indicated
Joint Motive Across Trials ................. 39
3. One Stage Transition M a t r i x ............. 40
4. One Stage Transition Matrix for Motives . . . 42
5. Conditional Probabilities of Motive and
Behavior Changes ............................ 44
6. Absolute Mean Differences Between Level of
Cooperation and Proportion of Joint
Motive Between Dyad Members Across Trials . 45
7. Contingency Rates of Cooperation Based on
Prediction of Other Player's Motive .... 47
8. Contingency Rates of Cooperation for
Player Motivated by Non-Joint Gain Based
on Prediction of Other Player's Motive . . . 48
9. Contingency Rates of Cooperation for
Player Motivated by Joint Gain Based on
Prediction of Other Player's Motive .... 49
10. Contingency Rates of Joint Motive Choice
Based on Prediction of Other Player's
Motive....................................... 51
11. Prediction of the Other Player's Motive
Based on Outcome of Previous Trial ..... 52
12. Proportion of Cooperative Responses Based
on Player's Motive, Prediction of
Opponent's Play, and Prediction of
Opponent's Motive for Each Outcome on
the Previous Trial.......................... 54
iv
LIST OF FIGURES
Figure Page
1. Prisoner's Dilemma Game Matrix as Defined
by Rapoport and Chammah (1965) ............. 9
2. Normal Payoff Matrix (a) Transformed Into a
Matrix of Joint Gain (b), Own Gain (c), and
Relative Gain (d) for Player A ............. 16
3. Normal Payoff Matrix (a) Transformed Into a
Matrix of Joint Gain (b), Own Gain (c), and
Relative Gain (d) for Player A ............. 18
4. The Payoff Apparatus As It Appears to
Player B (Column Player).................... 31
5. Master Console and Example Data Card......... 32
v
CHAPTER I
INTRODUCTION
Many behavioral phenomena are difficult to study
because of their complexity. Some of these phenomena can
be explored in the laboratory by transforming their basic
elements into a simple game structure. One such group of
phenomena that lends itself to this type of transformation
includes situations which involve the necessity of making
decisions in circumstances when the resultant outcomes are
affected by the decisions of at least one other party.
Of particular interest are those decision-making
situations which present some degree of conflict of
interest among the decision-makers. Thus, problems of
political science have been subjected to rigorous game
theoretical analyses (Shapley and Shubik, 1954; Luce and
Rogow, 1956) to provide a model of how power is distributed
in a body of men with partially the same and partially
opposite interests. Results of such analyses generally are
in agreement with the findings of political scientists
about the distribution of power and additionally call
attention to other aspects of power often under-emphasized
by traditional analyses (Luce and Rogow, 1956).
As indicated above, phenomena that lend themselves
to game theoretical analyses are situations which involve
the necessity of making decisions when the outcomes that
result will be affected by at least one other decision
maker. The decisions may be effected under conditions of
certainty, risk, or uncertainty (Luce and Raiffa, 1957).
A decision is made under conditions of certainty if each
action is known to lead to a specific outcome. A risky
decision is one which leads to one of a set of possible
outcomes, each occurring with a known probability. If the
decision leads to a set of possible outcomes whose
probabilities are unknown then the decision-maker is
making his decision under conditions of uncertainty.
Of late, there has been an increasing trend in
using game situations to investigate decision-making under
conditions of uncertainty because experimental games afford
an opportunity to create and examine interpersonal conflict
in these situations under controlled conditions. The
results of these studies have relevance to intergroup and
international conflict as well (Deutsch, 1969). Luce and
Raiffa (1957) point out that the problem of conflict is,
in their opinion, really a problem of individual decision
making under a mixture of risk and uncertainty, the
uncertainty aspect of the decision arising from one
decision-maker's ignorance as to what the other decision
maker will do.
Experimental Games in Psychology
Game theory is primarily a product of mathematicians
and not of investigators from the empirical fields. As
Rapoport (1966) argues, it is concerned with the formal
aspect of rational decision (decision processes devoid of
content). Classifying a situation as a game according to
game theory, then, should not be based on content-related
criterion, but rather on
. . . whether certain choices of actions and certain
outcomes can be unambiguously defined, whether the
consequences of joint choices can be precisely
specified, and whether the choosers [decision-makers]
have distinct preferences among the outcomes.
(Rapoport, 1966:17)
One general area of research to which psychologists
have applied concepts from game theory is the extent to
which human decision-makers behave according to the model
of rational behavior developed in the theory of games.
The meaning of "rational" is not clearly specified, but
generally it seems to include an assumption one makes
about the players maximizing something (utility), and an
assumption about complete knowledge (of the outcomes) on
the part of the player in a very complex situation (Luce
and Raiffa, 1957:5). Given that these assumptions are
true, the psychologist's task is to find the utility
functions in order to explain the behavior of the players
as rational.
One of the simplest families of experimental games
is that which restricts both the number of players and
3
the number of alternative choices for each player to two.
In these games the outcome is determined by the patterns of
responses made by the two players. There are only four
outcomes since each of the two players is allowed only two
alternatives. When an outcome occurs, each player receives
the prescribed payoff for that outcome. Therefore, there
are two payoffs associated with each outcome, one payoff
for each player, or eight possibly payoffs that can be
assigned to any game. The two-person, two-strategy games
are defined by mathematical relationships between the
eight possible payoffs. Rapoport and Guyer (1966) have
described the 78 possible different games that are defined
on the basis of ordinal relationships among these payoffs.
That is, if each player's payoff in a 2 x 2 game is ranked
ordinally, if each player can be regarded as equivalent,
and if games which can be obtained from each other by
interchange of rows and/or columns can be regarded as
equivalent, there are exactly 78 different games. One
basis upon which their classification of these games is
made rests upon the number and type of equilibrium out
comes. An outcome is said to be in equilibrium if neither
player can improve his payoff by a unilateral shift in
strategy. If both players are satisfied (i.e., they each
receive their maximum payoff) the equilibrium is called
absolutely stable. An equilibrium is strongly stable if
conditions exist such that if either player deviates from
equilibrium the other becomes satisfied. A strongly stable
equilibrium may be deficient if some other outcome is
better for both players.
A second basis upon which games are classified
according to Rapoport and Guyer (1966) deals with whether
either or both of the players have dominant strategies. A
game contains a dominant strategy if, and only if, for at
least one of the players, one of the alternatives is
preferable; that is, the payoff for the actor's one
alternative is always greater than the payoff for his
choosing any other alternative, no matter which course of
action the other player chooses.
A third characteristic often used to categorize
games is whether the pairs of scores for each outcome sum
to a constant number. In constant sum games, on any
trial, the gain of one player plus the gain of the other
player always equals a constant figure. When that figure
is zero, one player's gains are by definition the other
player's losses, and the game is said to be zero-sum.
These games are also known as strictly competitive (Luce
and Raiffa, 1957:64), since the players are strict
adversaries of each other and have strictly opposing
preference patterns for the outcomes of the game. Non-
zero-sum games, or non-strictly competitive games, are
those in which the outcomes do not sum to a constant number
and hence the players do not necessarily have opposing
preference patterns. These games provide the players with
an opportunity to cooperate as well as compete. As Luce
and Raiffa (1957) point out, most economic, political, and
military conflicts can be realistically transformed into
game form only if their non-strictly competitive nature is
acknowledged (e.g., war is probably not strictly competi
tive since both parties presumably prefer a draw to mutual
annihilation).
Prisoner's Dilemma
Two-person, two-strategy, non-zero-sum games have
been the ones most often used in laboratory settings (see
Oskamp, 1971 for a review). In some of these games it is
difficult to specify what is optimal behavior, but these
games on a simple level seem to be representative of actual
social interaction situations. By far the most popular of
these games has been the Prisoner's Dilemma (PD) game, a
two-person, two-strategy, non-zero-sum game with each
player having a dominant strategy and whose equilibrium
outcome is strongly stable and deficient. Its popularity
has grown due to its unique structure which contains the
essence of many real-life conflict situations in which
achieving group goals is in conflict with achieving
individual goals. Each player in the game is tempted to
seek his own personal gain by competing, but if both
players compete, they gain less than what they could gain
6
if both are willing to trust each other and cooperate.
According to Luce and Raiffa (1957) the name for this game
is attributed to A. W. Tucker. Davis (1970) interprets
PD using the following example: Two men suspected of
committing a crime together are arrested and placed in
separate cells by the police. Each suspect may either
confess or remain silent, and each one knows the possible
consequences of his action. If one suspect confesses and
his partner does not, the one who confessed turns state's
evidence and goes free and the other one goes to jail for
twenty years. If both suspects confess, they both go to
jail for five years. If both suspects remain silent, they
both go to jail for a year for carrying concealed
weapons— a lesser charge. Assuming there is no "honor
among thieves" and each suspect is concerned with his own
well-being, the suspects must decide whether to confess or
remain silent. Since each must make his decision without
knowing what his partner will do, he must consider each of
his partner's alternatives and anticipate the effect of
each of them on himself. If his partner confesses, he must
either remain silent and go to jail for twenty years, or
confess and go to jail for five years. On the other hand,
if his partner remains silent, he can serve a year by
being silent also, or win his freedom by confessing. In
either case, he is better off confessing— this is his dom
inant strategy.
Rapoport and Chammah (1965) define PD much more
concisely. In their conceptualization of the game, each
player has two alternatives with respect to the other
player: cooperation (C) and defection (D). They also
label the outcomes that may occur for each player. If
both choose C, they each receive the reward payoff (r).
If they both choose D, each receives the punishment payoff
(p). If one chooses D, and the other chooses C, the
player who chose C receives the sucker's payoff (s) and
the player who chose D receives the temptation payoff (t)
(see Figure 1). The required constraints upon the payoffs
for the PD game when the payoffs are symmetric have been
formally expressed by the following two inequalities:
t > r y p > s, and (1)
2r > s + t (2)
The dominant strategy for the PD game for both
players is D since by inequality (1) t is greater than rf
and p is greater than s. The players will always be ahead
on any single play by choosing D. A second implication by
inequality (1) is that mutual cooperation is preferable to
mutual defection, because r is greater than p. Inequality
(2) insures that a maximal gain resulting from alternating
cooperation and defection in iterated play is not possible.
8
PLAYER B
C D
PLAYER A
D
Fig. 1.— Prisoner's dilemma game matrix as defined
by Rapoport and Chammah (1965).*
*First entry in each cell is Player A's payoff.
t, s
9
Behavior in PD Games
The many experiments conducted using the PD game,
seem to have as their goal determining under what
conditions players cooperate and under what conditions
they play rationally (in a game-theoretic sense). Among
the important variables that determine how a player
behaves are the manipulations of the entries in the
payoff matrix, the way the other person plays, and
whether communication between the players is allowed.
In general, performance in an iterated PD game is
more competitive than cooperative (Rapoport and Chammah,
1965). Scodel et al. (1959), Scodel and Minas (1960),
and Minas et al. (1960), varied the payoff matrix by
increasing the difference between the temptation payoff
and the sucker payoff. The effect of enlarging this
"competitive" index was to produce a larger percentage of
competitive play, as expected. In other games, they
relaxed some of the restrictions of PD in an effort to
produce cooperation, so that the worst payoff was given
for joint defection and a smaller than normal reward was
assigned to the temptation payoff. VThile these variations
did produce more cooperation, the number of competitive
choices still exceeded the number of cooperative choices
and tended to increase over trials. Rapoport and Chammah
(1965) have been able to achieve levels of cooperation up
to 77 percent by increasing the payoff for joint
10
cooperation and decreasing the competitive index. Oskamp
and Kleinke (1970), in a series of experiments, examined
behavior in PD as a function of magnitude of reward.
Their first study used five different reward levels ranging
from minus five cents to three dollars for the r payoff
and results showed no differences in the amount of
cooperative behavior (the five matrices all had the same
structure in terms of the interrelationships of the payoff
values in the four cells). In their second experiment no
significant difference in cooperation was found between
conditions of no reward (playing for points) or monetary
reward. Their results are consistent with the results of
a majority of PD studies (e.g., Evans, 1964; Wrightsman,
1966; Knox and Douglas, 1968). Contradictory results,
however, have been reported (Oskamp and Perlman, 1965;
Radlow et al., 1968) which indicate that subjects do
behave differently in the PD game when motivated by
meaningful incentives versus low incentives. Some of these
subjects who receive meaningful incentives become more
cooperative and others less cooperative but when incentives
are low there is less variability among subjects (Knox and
Douglas, 1971).
The game playing behavior produced by the strategy
of the other player has been investigated by using a
confederate and/or a pre-planned program of moves. Of
all the variables affecting game playing behavior, this
11
variable has been one of the most extensively studied
(Oskamp, 1971). Early investigations of this variable
indicated that there was no effect of the "other's"
strategy on the choice behavior of subjects (Bixenstine
et al., 1963; McClintock et al., 1963). Later investiga
tions, however, have found that the strategy of the other
player does affect cooperation. Specifically, a 100
percent cooperative strategy has been found to produce
substantially greater cooperation than a 0 percent
cooperative strategy, particularly in early trials of
iterated play (Lave, 1965; Sermat, 1967). Relatively high
levels of cooperation (80 percent) on the part of the other
player also produce more cooperation (Knapp and Podell,
1968). A contingent strategy, such as tit-for-tat matching
(a strategy which, on every trial, matches the subject's
previous response) produces significantly higher coopera
tion than a non-contingent strategy having the same level
of cooperation (Solomon, 1960). Several studies that have
examined the effects of changing strategies during game
play have shown that cooperation levels change depending
on the change in the "other's" strategy. Rapoport and
Mowshowitz (1966), for example, found that raising the
probability of a cooperative response on the part of. the
other player following jointly cooperative responses
produced substantially more cooperation on the part of the
subjects (see also Komorita, 1965). Bixenstine and Wilson
(1963) systematically changed the cooperation level
quadratically so that subjects initially were presented
with low levels of cooperation, then high levels, then low
levels again over the course of 200 trials. Their results
showed that although this condition produced more
cooperation than the reverse, subjects tended to exploit
the highly cooperative trials.
Communication has been shown to affect game playing
behavior. Loomis (1959) and Deutsch (1958) reported that
when communication was allowed, cooperative responses
averaged well over 50 percent. Others have indicated that
as the amount of communication increases there is an
increase in the amount of cooperation (Scodel et al.,
1959; Bixenstine et al., 1966; Wichmann, 1970; and
Voissem and Sistrunk, 1971). Kahan (1973) argues that in
the two-person PD game, the rule of no communication,
typically adhered to, is tacitly violated by the repeti
tion of play since repeated plays may be used as a vehicle
for the communication of a cooperative strategy. In his
three-person game he found far less cooperation than
comparable two-person studies and suggests this difference
may be.due to the inability of subjects to communicate in
this way in the multi-person game, when their anonymity
remains relatively stable.
13
Motives in PD Gaines
The selection of a strategy for the player is
somewhat different in PD than it is in the zero-sum game
where the dominant motive in the game is to obtain the
maximum payoff. The non-zero-sum games, including PD, by
definition allow for potential cooperation in terms of
increasing the gain of both players, as well as for
competition.
Recent investigations have focused on determining
the variables that influence these choices. One variable
that has received attention is the motivational orienta
tion of subjects playing experimental games (Messick and
McClintock, 1968; Pruitt, 1970). Studies using several
types of games have indicated that subjects' behavior in
game play depends upon three motives: (1) rivalistic— the
player's choice reduces his own payoff but reduces his
opponent's payoff by a larger margin; (2) cooperative— the
player's choice maximized both players' joint payoff; and
(3) simple maximization— the player's choice maximized
his own payoff (Schelling, 1958; Fouraker and Siegel,
1963).
More recently, these motives have been examined
using two-person, two-alternative, non-zero-sum games
(McClintock and McNeel, 1966a, 1966b, 1966c, 1967;
Messick and Thorngate, 1967; and Messick and McClintock,
1968). In these studies, one of the motives is tested
14
against at least one of the others. All of the studies
have included games in which subjects must choose between
maximizing their own absolute gain and maximizing their
gain relative to the other player. In these experiments
the motives are placed in conflict by various manipulations
of the amount of payoff associated with the four outcomes
(payoff matrix). In the PD game, for example, the strategy
for a high joint score (cooperative behavior) can be
placed in conflict with the strategy for a high score
(simple maximization) and the strategy for a high relative
score (rivalistic behavior); the latter strategies are not
in conflict with each other. Consider the payoff matrix
shown in Figure 2a. The first entry in each cell is
Player A's (row player) payoff. If Player A is motivated
to maximize joint gain the matrix may be transformed into
a matrix of joint payoffs (Figure 2b). Clearly, Player
A's choice should be C. If, however, Player A is motivated
for a high absolute gain the matrix may be transformed as
is indicated in Figure 2c. Here, alternative D is his
obvious preference. Finally, if Player A is motivated for
a high relative gain his preference is again for strategy
D (Figure 2d). Games such as these in which either of the
two alternatives can be dominated depending on the motive
have been called double-dominance games (Messick and
McClintock, 1968) .
In the example just presented, one motive is placed
15
PLAYER B
C D
PLAYER B
C D
C
PLAYER A
D
6, 6
00
00
1
8 , -8
a
1
to
1
to
12 0
0
b
-4
C
PLAYER A
D
6
00
1
o
00
-2
0 -16
16
d
0
Fig. 2.— Normal payoff matrix (a) transformed into
a matrix of joint gain (b), own gain (c), and relative
gain (d) for Player A.
16
in conflict with the other two. There are, however,
various manipulations which isolate each motive by placing
two in conflict and holding the third constant. An
example is shown in Figure 3. The original game matrix
is indicated in Figure 3a. The player motivated by joint
gain should prefer strategy C (Figure 3b). A player
motivated for own absolute gain should have no preference
for C or D (Figure 3c). If the player is motivated for
high relative gain then he should prefer strategy D
(Figure 3d). Thus, this game could be used to examine the
difference between the two motivational orientations of
joint gain and relative gain holding the motive for
absolute gain constant.
Manipulation of Motives
Several studies have manipulated the saliency of
the other person's score by including it in the feedback.
There have been two experimental conditions of feedback
in these experiments: (1) the subject's own cumulative
score, and (2) both the subject's cumulative score and the
cumulative score of the other player. McClintock and
McNeel (1966b, 1966c) report that subjects who were given
both their own score and that of the other player displayed
significantly more rivalistic behavior than subjects who
were given only their own score. Messick and Thorngate
(1967) reported similar results using a different type of
17
PLAYER B PLAYER B
C D
C
PLAYER A
D
C D
C
PLAYER A
D
-3
Fig. 3.— Normal payoff matrix (a) transformed into
a matrix of joint gain (b), own gain (c), and relative
gain (d) for Player A.
two-person, two-alternative, non-zero-sum game. In their
games much greater payoffs were given to both the players
when they chose the same alternative. In the notation of
Figure 1 and Rapoport and Chammah (1965), r > p >t > s.
Thus, C would yield the greatest payoff if it was mutual
and the lowest if it was unilateral; and D was the only
alternative for which one could receive a greater payoff
that one's partner.
Another type of manipulation of the form of the
feedback was reported by Messick and McClintock (1968) .
The three conditions of the display of cumulative feedback
were: the player's own score, the sum of the player's
score and that of the partner (joint score), and the
difference between the player's score and that of his
opponent (difference score). In this study, several
different payoff matrices were employed causing many
different motivationally conflicting situations. Each
situation placed one motive in conflict with either one or
both of the other motives. When the motive for a high
joint score is in conflict with the motive for a high
score, the subjects who received cumulative feedback in
terms of joint score chose the alternative favoring a
high joint score more than did the other groups. The
effect was even more pronounced for the motivational
conflict between own gain and maximum gain over the other
player. In this case, the group which received cumulative
19
feedback, in terms of the difference between their score
and that of their opponent, chose the relative gain
alternative much more often than did the members of either
of the other feedback groups.
Messick and Thorngate (1967) manipulated the feed
back of the subjects in a very different way. One of the
members of each group was given one payoff matrix, and the
other member of each pair was given another payoff matrix.
In every condition the conflict of motives was between own
score and score relative to the other player. Some of the
matrices were designed so that both subjects believed that
they could not get a higher score than the other person,
but they could get a lower score. Other groups were led
to believe the reverse. The groups which thought they
could only do as well as the other players exhibited more
rivalistic behavior than did the other group. Apparently,
fear of obtaining a score lower than the other player is
a stronger motive for rivalistic behavior than is the
incentive to obtain a much larger score than that of the
other player.
McClintock and McNeel (1967) manipulated the motive
for gain relative to the other player by varying the
interpersonal experience prior to the experiment. They
created three experimental conditions: (1) no prior
experience; (2) friendly prior experience; and (3) hostile
prior experience. The subjects in the latter two groups
20
were given pre-game experience consisting of ten trials
in which each player was informed that the other player
had given him a high positive or a low score on each of
the trials. The game used in the experiment was a maxi
mizing difference game which places own score in conflict
with relative score. The results were that those in the
friendly condition showed less rivalistic behavior than
did those in either of the other groups. While these
results indicate differences in behavior due to motive,
sampling problems may exist in that the conclusions may
hold only for persons given their pre-game experiences.
In summary, the research involving motivational
questions in two-person, two-response, non-zero-sum games
can be interpreted in the general context of two basic
problems. The first of these is the measurement of the
relative strengths of underlying motives effecting the
responses made in such games. The ratios of the payoff
matrix were manipulated to vary the relative strengths of
the conflicting motives. The studies indicate that
increasing the reward associated with a motive results in
a corresponding increase in responses attributed to the
motive. The other basic problem is that of investigating
the effect of variables assumed to act on one motive
independent of other motives.
21
CHAPTER II
THE PROBLEM
The investigation of motives in two-person, two-
alternative, non-zero-sum games has concentrated on the
player's responses as the dependent variable. The present
study additionally obtained measures of subjects' motives
directly as an adjunctive dependent variable. This
technique allows the investigation of motivational aspects
of any particular game without changing its motivational
properties. The PD game situation was selected because of
the nature of the motivational conflict inherent in that
particular game.
The main problem that was considered in this paper
bore on the relationship between cooperation and the
motives. One method of measuring cooperation in PD is to
simply record the proportion of C responses. However, PD
is a social situation which is influenced by the perform
ance of the other member of the dyad.
One way to standardize the experimental situation
is to compare trials which have the same outcome on
previous trials. Rapoport and Chammah (1965) accordingly
indicate that the dependent variables with respect to PD
22
are the conditional probabilities of cooperation given
which of the four outcomes (CC, CD, DC, DD) occurred on the
previous trial. To use the terminology of Rapoport and
Chammah (1965), the selection of the index of cooperation
on any trial depends upon the "state" of the trial which
is determined by the outcome of the previous trial.
Three questions were considered with respect to the
relationship between the motives and the four indices of
cooperation. First, was there any relationship between
levels of cooperation and the motivational orientations of
the subjects. More specifically, are there any significant
differences in rates of cooperation based on the motive
operating at the time? The second question was how are
the seven measures (the three motives and the four indices
of cooperation) interrelated? The interrelationships were
of interest as an indication of basic underlying traits.
The last question concerned whether the procedure of
having subjects indicate their motives during game play
alters the game so that the results of this study are not
comparable to other PD experiments.
Levels of Cooperation and
Motivational Orientations
The question undertaken here was the primary one
and the essence of the paper. The basic assumptions for
the research were that three specific motives account for
some differences in performance. The most important
23
question of the paper was, therefore, are the cooperation
levels based on one motive significantly different from
cooperation levels based on another motive.
On logical grounds, when players are motivated to
obtain a high joint score, they should demonstrate more
cooperation than when motivated for an absolute or relative
gain. This prediction was based on the fact that coopera
tion by one player leads to a higher joint score regardless
of the response of the other member of the dyad. This
condition does not exist for either of the other motives.
Hypothesis 1 was that when players chose the motive of a
high joint score, they would demonstrate more cooperation
than when indicating one of the other motives.
The second prediction was that when persons attemp
ted to maximize their own gain, they would have higher
rates of cooperation than when attempting to maximize
their score relative to the other player's score. The
rationale was that mutual cooperation acts as an incentive
when persons desire a high score but not for those persons
when desiring a high relative score. In Deutsch's (1958)
terms, mutual trust can occur even under circumstances
where the people involved are overtly unconcerned with
each other's welfare provided that the characteristic of
the situation is such as to lead one to expect one's trust
to be fulfilled. One characteristic which he believes to
facilitate this "trusting" response is the opportunity and
ability to communicate fully a system for cooperation
which defines mutual responsibilities and also specifies
a procedure for handling violations and returning to a
state of equilibrium with minimum disadvantage if a
violation occurs. With respect to two-person PD, the
iterated play perhaps provides the vehicle for such
communication and a violation of responsibility on the part
of the other player to cooperate would simply result in a
return to strategy D. Therefore, Hypothesis 2 was that
when players chose the motive of high score, they would
demonstrate more cooperation on all four state-related
indices than when choosing the motive for high relative
score, and when players are motivated for high relative
gain, they would demonstrate the lowest amount of
cooperation since they would have had no reason to play C
except to use that choice to induce a C response by the
other player and then take advantage of that response.
Effects of Selecting Motives
Earlier in the paper, the procedure of having
subjects indicate their motives was advocated on the
grounds that other methods of measuring motives changed
the payoff matrix resulting in a different two-person,
two-alternative, non-zero-sum game. If the indicating of
motives changes the perception of PD, players in a sense
would also be playing a different game. A control group
25
that played PD without indicating their motives was
compared with those who play PD and do indicate their
motives in order to detect any significant differences due
to this procedure. Hypothesis 3 was that there would be
no significant differences in levels of cooperation
between those subjects who were given the opportunity to
indicate their motives and those subjects who were not.
Motivational Inclinations in
Iterated Play
The last type of problem that was investigated was
the effect that iterated playing of PD had on the motiva
tional orientations of the subjects. Three different
effects were examined.
One type of effect is the change of motivational
preferences for the group as a whole. Previously, it was
assumed that one motive dominates the other two over the
course of the experiment? however there may be motivational
dimensions that are important in PD. Even given that this
assumption is true, there may be a relationship between the
seven measures. Interrelationships involving both the
motivational inclinations and performance measures were of
interest because Rapoport and Chammah (1965:78) have
already established the interrelations between the game
play measures and because one concern of this paper was to
determine the relationship between these two types of
measures. Rapoport and Chammah (1965:96-97) reported a
26
general fluctuation in the rate of cooperation which
would be consistent with such a motivational shift. They
reported an initial decrease in cooperation followed by
an increase. Kelley and Stahelski (1970) investigated
changes in game play behavior as a function of the behavior
of the opponent and found that when cooperative and
competitive persons interacted the cooperative ones tended
to become like the competitive ones; that is, they played
competitively. They also found that because of this
behavioral change on the part of the cooperators,
competitors misjudged the intentions of the cooperators as
being competitive. Although their experiments indicate
that players change their behavior over trials as a
function of their partner's play, the question of whether
their motives change was still left unanswered. Coopera
tive players may be forced to play competitively without
changing their initial motivational orientation to the
game. Kelley and Stahelski (1970) suggest this may be
true. The player approaches the game
. . . not in an ad hoc or arbitrary way, but with the
orientation that he intends to adopt in all such
situations . . . the subjects general orientation in
these relationships does not necessarily imply
corresponding behavior, (p. 86)
In the present study, the motivations of the subjects were
made on every trial by each subject, in this way, not only
could behavioral changes be noted, but motivational
changes as well.
27
Another type of motivational change is that the
members of each dyad tend to develop motivational patterns
similar to that of their partner. The rationale for such
a change is that the PD situation is very ambiguous to
naive players and that the other player's responses change
the perception of the game. Rapoport and Chammah (1965:
58-60) reported high correlations between the cooperation
levels of the members of the same dyads. The greatest
cause of these correlations was the result of the "lock-in"
effect. The lock-in effect is the high proportion of
either mutual cooperation or defection. The players
either engage in tacit collusion or are both trapped in
mutual defection where neither player feels justified to
cooperate with a defecting opponent. The presence of the
lock-in effect, however, does not guarantee that subtle
motivational changes do occur. A final aim of this study,
then, was to find out whether in fact these motivational
changes do occur that correspond to the greater similarity
between dyad members.
28
CHAPTER III
METHOD
Subjects
Sixty male subjects were used in the study. Sixteen
of these were used as control subjects who played PD
without indicating their motives. The remaining 44
subjects were required to specify their motive on every
trial of play. They were also required to specify their
prediction of how they thought the opponent would play and
what motive they thought the opponent would indicate. Each
subject was paid one dollar to participate in the experi
ment and their gains or losses were added to or subtracted
from the initial dollar at the exchange rate of one cent
for every six points accumulated. All subjects were
volunteers from Cerritos College.
Apparatus
Each subject was seated in front of a box displaying
the payoff matrix, and containing four response switches.
Two response switches were for the subject to indicate his
response and his motive. The other two response switches
were for subjects to indicate their prediction of how the
other player would play and what the other player would
indicate as his motive. Behind each cell of the matrix
was a light bulb such that any one of the four cells could
be lit individually. The payoff matrix apparatus is shown
in Figure 4. The boxes were connected to a master console
(Figure 5) on which subjects' responses were displayed and
recorded. The master console was set up such that the
appropriate cell of the matrix would light up depending
upon the choices made by the subjects. For example, if
Player A (row player) chose alternative 1 and Player B
(column player) chose alternative 2, the experimenter would
light the cell indicating the payoffs to Player A and B
or "-12" and "12" points respectively.
Procedure
The players were seated with a partition between
them and each was presented with the appropriate payoff
apparatus. The experimental instructions (Appendix A),
which had been pre-recorded, were then played to the
subjects. The instructions explained that the subjects'
payoffs were determined not only by what one did, but also
by what one's opponent did. They were told that both
players could act in their own interest or in the interest
of the pair. The instructions also indicated that choos
ing alternative "2" always obtained a greater payoff for
the individual than alternative "1," while choosing
alternative 1 always meant a greater payoff for his oppon
ent and the dyad as a whole. The strategy choice of the
PLAYER B
MY MOTIVE
X
JOINT
ABSOLUTE
RELATIVE
PREDICTION OF HIS PLAY
Y ^ HOW 1
ROW 2
PREDICTION OF HIS MOTIVE
JOINT
Z ^ ABSOLUTE
RELATIVE
W
you
get
6
he
gets
6
you
get
12
he
gets
-12
you he you he
get gets get gets
-12 12 -6 -6
READY
Fig. 4.— The payoff apparatus as it appears to Player B (column player)
Player A's responses
o o o o o
Card « * ,
for
trial)
Response
Indicator
Lights
o o o o o
Player B's responses
Power On
Matrix On
Indicator Matrix
Light Cell On
o o
Reset
O
1 2 J A R 1 2 J A R
my
play
my
motive
predict
his play
predict
his motive
TRIAL # DYAD# DATE
EXPERIMENTAL CONDITION COMMENTS:
my
play
my
motive
predict
his play
predict
his motive
1 2 J A R 1 2 J A R
Fig. 5.— Master console and example data card.*
*The top of the data card represents row player's responses and the
bottom represents column player's responses.
opponent could be ascertained simply by noting which of the
four lights was illuminated; that is, based on the outcome
the subject could deduce what the opponent had played.
The subjects were informed that they had one dollar
in credit prior to the beginning of the experiment. Any
point gained or lost would be added to or subtracted from
their dollar at a conversion rate to be told them later.
The instructions also emphasized the importance of the
lack of overt communication between the subjects during the
experiment. In addition to selecting a strategy, the
subjects in the experimental group were instructed that
they were to indicate their motive on every trial and to
indicate their prediction of their opponent's strategy and
motive on every trial.
After presenting the instructions, the experimenter
lit up one of the cells of the matrix to determine whether
or not each subject could determine what alternative his
opponent chose and what payoffs each had received if that
outcome had resulted during game play.
All subjects played PD for a total of 50 trials.
The experimenter recorded the outcome and motive for each
player on each trial as well as the prediction of the
other player's motive and strategy on each trial on
response cards that fit into the master console as shown
in Figure 5.
After finishing the last trial all subjects were
33
paid and the nature of the experiment was explained to
them.
34
CHAPTER IV
RESULTS
The main purpose of this investigation was to
examine the relationship between levels of cooperation in
Prisoner's Dilemma and the motive of the player. There
were two hypotheses concerned with the existence of such a
relationship. The first prediction was that subjects when
motivated by a high joint gain would demonstrate more
cooperation than when motivated by one of the other
motives. The second hypothesis was that subjects would
demonstrate more cooperation when motivated by an absolute
gain than when motivated by a relative gain. Table 1
shows the proportion of cooperative responses based on the
motive of the player for each state of the game, as well
as the proportion of cooperative responses for the sub
jects who played without indicating their motives (control
group) . As is indicated in the bottom row of the table,
subjects when motivated by a high joint gain played
cooperatively 59.1 percent of the time. When motivated by
a high absolute gain or a high relative gain the levels
of cooperation dropped considerably (18.1 and 14.6 percent
cooperation levels, respectively). For each dyad the
35
TABLE 1.— Contingency rates of cooperation based on motive and outcome on previous
trial
Probability C Given Motive
Over
Motive Control
Player
Played
Opponent
Played Joint Absolute Relative
C C .833 .225 .250 .500 .391
C D .414 .197 .097 .230 .217
D C .652 .259 .232 .342 .385
D D .561 .140 .123 .235 .226
.591 .181 .146 .280 .267
GJ
< y »
motives were ranked based on level of cooperation, so
that the motive which resulted in the highest level of
cooperation was given the rank 1, the motive which
produced the next highest level of cooperation was given
the rank 2, and the motive which resulted in the least
amount of cooperation was ranked 3. A Friedman Analysis
of Variance by ranks (Siegel, 1956:166) was then conducted
on these data which indicated that the rank was dependent
upon the motive ( = 11.82, df_= 2, £ ^ .01). These data
appear to be supportive of hypothesis 1 and non-supportive
of hypothesis 2. There appears to be little if any
distinction between game play behavior under motivations
of absolute or relative gain. For this reason, the
remainder of the data analyses were conducted by grouping
the responses based on absolute and relative gain into a
category called "non-joint" gain. The third prediction
dealt with the difference in cooperation levels between
subjects indicating motives and subjects not given the
opportunity to indicate motives. The difference between
the cooperation levels of these two groups is also shown
in Table 1, and indicates little, if any, difference
between subjects in these two conditions. Apparently,
having subjects indicate their motives does not, in itself,
change the play of those subjects . This finding is
evidence in support of hypothesis 3.
Another purpose of the present investigation was
37
to examine the changes of motive and level of cooperation
that take place as the game progresses. For this purpose,
the level of cooperation and the proportion of trials in
which subjects indicated joint motive were calculated for
each block of ten trials. These results appear in Table 2.
On the first ten trials subjects seemed to cooperate
relatively more often than on later trials. Furthermore,
both the level of cooperation and the proportion of trials
in which the motive was high joint gain seem relatively
constant after those initial ten trials.
Table 3 presents the one-stage transition matrix
for the proportion of cooperative responses from trial to
trial. The most frequently repeated outcome was mutual
defection, which occurred about 59.7 percent of the time
following a trial in which mutual defection occurred. The
least frequent change that occurred was the change from
mutual defection to mutual cooperation. This transition
occurred only 6.3 percent of the time. On trials where one
player cooperated and the other defected there was over a
50 percent likelihood that the next trial would result in
mutual defection? the player who played cooperatively
obviously changed his play (The probability of CD to CD
or from DC to DC was only .120.). In fact, over 75 percent
of the time, the cooperator on trial X would play non-
cooperatively on trial X + 1 if his opponent played non-
cooperatively on trial X.
38
TABLE 2.— Proportion of cooperative response and proportion of time players indicated
joint motive across trials*
Block 1 Block 2 Block 3 Block 4 Block 5
Proportion
Cooperation
Proportion
Joint Motive
.377
.282
.264
.259
.245
.264
.277
.273
.255
.277
*Each block represents ten trials.
UJ
VO
TABLE 3.— One stage transition matrix
# of C
Choices
Outcome
2C
on Trial X
1C
+ 1
OC
Outcome 2C .278 .444 .278
on 1C .107 .365 .528
Trial X OC .063 .340 .597
over outcomes .100 .360 .540
40
A transition matrix for motives (joint versus non
joint) is presented in Table 4. Note here that if players
are mutually playing a given trial motivated by a non-joint
(NJ) gain there is a 71 percent likelihood that they will
be similarly motivated on the next trial. This table
indicates a few rather interesting findings. First, since
level of cooperation and motive are related as shown
previously and displayed in Table 1, it was expected that
the transition matrix for motives would appear similar to
the transition matrix for game play with respect to the
transition probabilities. For the most part this is true.
There are, however, several notable exceptions. As was
shown in Table 3, the likelihood of the transition from a
single cooperative response to a single cooperative
response was .365. However, the likelihood of the motiva
tional transition from one joint motive to one joint
motive was .495. Obviously, when a player changes his
game play from cooperative to competitive after having
been exploited (Rapoport, 1967), there is not necessarily
a corresponding change in his motivation (The probability
of J - NJ or NJ - J repeat was .414.). In fact, just
as it was pointed out that the cooperator on trial X
had over a 75 percent likelihood of competing on trial
X + 1 if his opponent competed on trial X, it should also
be noted that the subject motivated by a joint gain on
trial X will only change his motive to NJ on trial X + 1
TABLE 4.— One stage transition matrix for motives
# of Joint
Motive Choices
Outcome
2J
on Trial X
1J
+ 1
OJ
Outcome 2J .540 .340 .120
on 1J .078 .495 .427
Trial X OJ .027 .263 .710
over motive! outcome .093 .353 .555
42
given his opponent had a non-joint motive on trial X
about half of the time. The conditional probabilities of
behavior change given motive change and motive change
given behavior change are further clarified in Table 5.
The evidence presented in Table 5 seems to clearly
support the notion that if a player changes his motive
(from joint to non-joint or vice versa) he usually changes
his corresponding behavior except when the previous out
come was mutual defection. On the other hand, if a player
changes his behavior he is just as likely to retain the
same motive as change it. The reason for the exception as
noted for the DD situation could well be due to a motive
change from an absolute gain motive to a relative gain
motive.
Table 6 presents data concerning the similarity
between dyad members across blocks of trials. The differ
ence in cooperation level was averaged across dyads so
that the numbers given represent measures of dyad simi
larity (average difference in level of cooperation). In
addition, the same measures were computed for difference
in proportion of trials in which joint motive was
indicated. Contrary to the findings by Rapoport and
Chammah (1965), but consistent with those of Kelley and
Stahelski (1970), it appears that dyad members seem not to
become more similar as trials progress. The same may be
said of similarities in motivational inclinations. The
TABLE 5.— Conditional probabilities of motive and behavior
changes
Probability of
Motive Change
Given Behavior
Change on Trial
X + 1
Probability of
Behavior Change
Given Motive
Change on Trial
X + 1
Outcome CC .538 .913
on
CD .500 .971
DC .461 .849
Trial X DD .555 .584
over outcome .512 .750
44
TABLE 6.— Absolute mean differences between level of cooperation and proportion of
joint motive between dyad members across trials*
Block 1 Block 2 Block 3 Block 4 Block 5
Difference in
Cooperation
Level
.173 .109 .182 .209 .200
Difference in
Proportion of
Joint Motive
.345 .173 .273 .273 .300
*Each block represents ten trials.
<_n
level of difference between dyad members for both of
these measures seems to stay relatively constant. Notice
that the difference in proportion of joint motive parallels
the difference in level of cooperation across trials. One
reason that has been given to explain the supposed increase
in dyad similarity across trials previously reported has
been the lock-in effect mentioned earlier. In the present
investigation, however, it appears that the overall
competitive nature of the game and the fact that only 50
trials were conducted probably prevented the competitive
lock-in which is characteristic of situations of high
dyadic similarity.
Intuitively, it seems that there should be a
relationship between a player's choice and his prediction
of the other player's motive. If the player feels that
the other player is motivated by joint gain he may play
differently than when he feels the other player is
motivated by relative or absolute gain. Tables 7 through
9 show this relationship. Notice that if a player
predicts his opponent's motive as a high joint gain
(Table 7) he is twice as likely to play cooperatively than
when he predicts a non-joint motive for his opponent.
Tables 8 and 9, however, show that this difference can be
accounted for by the player1 s own motive. When the
player's motive is a non-joint gain (Table 8) there appears
to be little, if any, difference in his cooperation level
46
TABLE 7.— Contingency rates of cooperation based on prediction of other player's motive
Probability
C Given
Prediction of the Other
Player's Motive
Player
Played
Opponent
Played Joint Non-Joint Total
C C .675 .397 .500
C D .471 .179 .230
D C .538 .293 .342
D D .402 .186 .235
over previous outcome .479 .222 .280
TABLE 8.— Contingency rates of cooperation for player motivated by non-joint gain based
on prediction of other player's motive
Probability
C Given
Prediction of the Other
Player's Motive
Player
Played
Opponent
Played Joint Non-Joint Total
C C .222 .235 .233
C D .214 .145 .152
D C .250 .246 .247
D D .129 .128 .128
over previous outcome .171 .162 .163
TABLE 9.— Contingency rates of cooperation for player motivated by joint gain based on
prediction of other player1s motive
Probability
C Given
Prediction of the Other
Player1s Motive
Player
Played
Opponent
Played Joint Non-Joint Total
C C .806 .882 .833
C D .650 .289 .414
D C .842 .519 .652
D D .656 .474 .557
over previous outcome .716 .481 .589
due to his prediction of his opponent's motive, or for that
matter the state of the game (outcome on the previous
trial). The slight difference that is apparent shows a
lower level of cooperation when the opponent previously
played competitively. On the other hand, when the
player's motive is for a high joint gain (Table 9) there is
a large difference in cooperation level due to prediction
of the other player's motive. Indeed, not only are there
differences due to the prediction of the opponent's motive
but also differences due to the outcome on the previous
trial. Evidently, the player motivated by a high joint
gain is generally more responsive to changes on the part
of the opponent.
Table 10 shows the likelihood of indicating a joint
motive depending on how the player predicts the opponent's
motive. The results seem consistent with the results shown
in Tables 7 through 9, thus providing further evidence for
the relation between level of cooperation and motivational
inclination.
Table 11 shows the prediction of the other player's
motive based on the outcome on the previous trial.
Notice that the only difference in prediction of the other
player's motive seems to be when the previous outcome was
mutual cooperation (which increases the likelihood of
joint motive prediction for the opponent) versus the other
states.
50
TABLE 10.— Contingency rates of joint motive choice based on prediction of other player's
motive
Probability
Joint Motive
Given
Prediction of the Other
Player1s Motive
Player
Played
Opponent
Played Joint Non-Joint Total
C C .775 .250 .444
C D .588 .235 .296
D C .487 .172 .235
D D .504 .166 .240 ‘
over previous outcome .558 .187 .270
U1
TABLE 11.— Prediction of the other player’s motive based on
outcome of previous trial
Outcome
Previous
on
Trial
Player Opponent
Probability Player
Predicts Joint
Played Played Motive for Opponent
C C .370
C D .199
D C .173
D D .220
52
A summary of all the variables affecting level of
cooperation that were investigated in this study is pre
sented in Table 12. These results are in accord with other
investigations. The greatest level of cooperation occurred
on trials that immediately followed mutual cooperation.
Examining the marginal column, it is apparent that the
highest level of cooperation takes place on those trials
when the subject is motivated by a high joint gain and
when he predicts his opponent's play to be cooperative and
his opponent's motive to be high joint gain. The player's
motive (ignoring the outcome of the previous trial) seems
to be the most important determiner of cooperation. The
next important determiner seems to be the player's
prediction of the opponent's play (Tables 7, 8, 9, and 12).
53
TABLE 12.— Proportion of cooperative responses based on player's motive, prediction of opponent's play,
and prediction of opponent's motive for each outcome on the previous trial
Player* s
Motive
Prediction of
Opponent's Play
Prediction of
Opponent's Motive
Outcome on
CC CD
Previous
DC
Trial
DD Total
joint C joint .880
.917
.938 .761 .838
joint C non-joint 1.000 .400 .545 .560 .607
joint D joint .500 .250 .333 .389 .371
joint D non-joint .714 .250 .500 .431 .412
non-joint C joint .333 .375 .364 .146 .227
non-joint C non-joint .143 .224 .286 .242 .240
non-joint D joint .000 .000 .111 .095 .077
non-joint D non-joint .300 .093 .216 .071 .116
.500 .230 .342 .235 .280
u i
CHAPTER V
DISCUSSION
The results of the present investigation strongly
support the notion that self-report motivational responses
are related to behavior in the prisoner's dilemma
situation. The relationship that stands out most clearly
is between the motive of high joint gain and the four
indices of cooperation taken as a whole. When selecting
the high joint gain motive, subjects consistently
demonstrated more cooperation than when indicating one of
the other motives, no matter what the outcome of the
previous trial.
One of the assumptions of the present investigation
was that there are three motives operating in prisoner's
dilemma (high joint gain, high relative gain, and high
absolute gain). Several of the findings, however, suggest
that the motivational conflict inherent in prisoner's
dilemma cannot be differentiated on the basis of the
behavior of the players. In particular, the distinction
between maximizing one's own gain or receiving more
relative to the other player was far less important than
whether or not a player was motivated to obtain a high
payoff for the dyad as a whole. This outcome would
strongly suggest that the emphasis given the cooperative-
competitive dimension in game research has not been
misplaced (Kelley and Stahelski, 1970). Previous studies
examining the three-motive model generally have been ones
in which motives have been created for the players by any
one of a number of manipulations of the payoff matrices,
the form of feedback of the score, the prior experiences
of the subjects, et cetera. In these situations it is
possible that the player's play is affected by the
manipulations but their motives are not. Since no direct
measure of motivation was taken in those studies it is
difficult to interpret the motivation of subjects. For
instance, in the Messick and McClintock (1968) experiment
cited earlier, it was found that when the motive of high
absolute gain was placed in conflict with the motive of
high relative gain, subjects who received feedback in
terms of the difference between their score and their
opponent's score chose the alternative assumed to reflect
the relative gain motive more often than other subjects.
In that experiment it is quite likely that subjects could
have interpreted and performed their task as having a
certain goal implied by the form of the feedback. A
difference in game play behavior attributed to a difference
in absolute or relative gain may only become apparent when
subjects are told to maximize difference or absolute gain
56
either explicitly or implicitly as in the example just
cited. In an extensive review of the gaming literature,
Wrightsman et al. (1972) indicate,
. . . the structure of the conflict situation must be
considered as a major factor in any general theory of
conflict . . . this means careful attention to (a) the
motivational structure . . . and (b) the demand
characteristics established by the experimental
procedure . . . . (p. 186)
In the present experiment, subjects' motives were not
manipulated and results showed that game play behavior is
not different for subjects indicating these two motives of
absolute and relative gain.
There are at least two explanations for this find
ing. First, there exists the possibility that £s could
not clearly differentiate the two motives of absolute and
relative gain. Post experimental debriefing, however,
indicated that subjects did know the difference. A second
explanation, and the one that seems most reasonable is
that subjects may have no opportunity to differentiate
between those motives at least on the basis of game play in
PD. In order to test this hypothesis behavior in different
games should be examined and compared to behavior in PD.
These results indicate a simplication involving the
motivational indications by the subjects. Since the
distinction between relative gain and absolute gain was
much less important than the distinction between these two
motives and joint gain the three category system can be
57
reduced to one bipolar system in studying the PD game.
As was previously pointed out, one of the major
limitations of the other methods of measuring motives in
non-zero-sum games was that the measurement process
involved changes the very nature of the game. The results
show that this is not the case when subjects are allowed
to indicate their motives directly. Those who were given
the opportunity to indicate motives showed no important
differences in cooperation levels from the control group
which did not have this opportunity.
Since it was found that having subjects indicate
their motives does not affect game play behavior and since
a relationship exists between subjects' motives and their
behavior given the previous state (e.g., the motive of
joint gain was particularly potent when the outcome of the
preceding trial was mutual cooperation) the feasibility of
having subjects indicate their motives in conjunction with
playing prisoner's dilemma, or any other game, is sug
gested. This can clarify many problems that to now have
appeared unresolvable (e.g., the effects of monetary
payoffs on cooperation).
The overall game play behavior was extremely
competitive. The overall level of cooperation was only
.280. This level of cooperation is barely within the
range found by Rapoport and Chammah (1965) which varied
between .27 and .77. However, as Scodel et al. (1959),
58
Scodel and Minas (1960), Minas et al. (1960), and Terhune
(1968) found, increasing the difference between the
temptation payoff and the sucker payoff increased
competitive play and in the present experiment the
difference between these two indices was 24 points, a
relatively large difference. In Rapoport and Chammah's
(1965) terms, the cooperation index (r - p)/(t - s) was
equal to .50.
Both the proportion of cooperative response and the
proportion of trials in which joint motive was indicated
remained relatively consistent across trials. Contrary
to the findings of Rapoport and Chammah (1965), who
reported initial decreases in cooperation followed by an
increase, in the present study cooperation levels on the
first ten trials were greatest and dropped about 10 per
cent for the remaining 40 trials. It seems that the
assumption that one motive dominates a player's behavior
over the course of the experiment is supported, and
correspondingly the player1s behavior in the game remains
relatively constant, at least for the first 50 trials.
The fact that there were only 50 trials, of course, could
be one possible explanation for this behavior. The changes
in behavior that might have been expected due to a lock-in
could not have happened since for a lock-in to take place
more than 50 trials are usually required.
As indicated by the one stage transition matrix
59
(Table 3), however, players do change the behavior
depending upon the outcome of the previous trial; their
behavior is not absolutely constant. One of the important
changes that takes place is the change that occurs when
one player plays cooperatively and the other player plays
competitively. This outcome repeats itself very rarely
(probability of CD or DC repeat equals .120). Obviously,
if a player had just been exploited (he lost 12 points and
his opponent gained 12 points), he is not likely to be
exploited again. Instead, he changes his behavior. This
change brings about a mutually competitive outcome (DD)
since the exploiter has no reason to change his play.
Thus, the change to outcome DD is more likely than any
other change following a CD or DC outcome on the previous
trial. This change is not surprising and compares
favorably with most of the findings from prisoner's
dilemma studies. In previous studies, however, the ques
tion of whether the player's motive changes as well as his
behavior in these circumstances has been left, for the
most part, unexamined. Rapoport and Chammah (1965), as
was mentioned earlier, reported a change in rate of
cooperation which they believed was consistent with a
motivational shift. Kelley and Stahelski (1970) , however,
pointed out that the players may have an orientation that
they intend to adopt in all such situations. The one stage
transition matrix for motive (Table 4) and the conditional
60
probabilities of motive and behavior changes (Table 5)
get at the heart of this question. The change in motives
for the players is different from the change in their
behavior. This is especially evident in the change from
the joint to non-joint and non-joint to joint transitions.
These transitions repeat themselves with probability .414.
In other words, even though the player who has been
exploited changes his behavior, he may not necessarily
change his motive. He may feel, for example, that he
would like to play for a high joint gain yet feel forced
to act competitively as a self defense to prevent being
exploited any further. The point proposed by Kelley and
Stahelski (1970) seems to be supported by these data; the
likelihood of repetition from trial to trial of a particu
lar set of motives for the players is high when compared
to the analogous transitions for cooperative and competi
tive responses.
The data shown in Table 5 indicate the strongest
support for this argument. The likelihood of a motive
change given a behavior change is only .512 and this is
fairly representative across all outcomes on the previous
trial. On the other hand, if a player changes his motive
he changes his behavior 75 percent of the time. This
figure, however, is somewhat deflated because of the
anomalous DD previous outcome state. There is over a 90
percent likelihood of a behavior change given a motive
61
change if the previous outcome of DD is eliminated from
consideration. Perhaps the reason that the DD state
decreases the likelihood of a behavior change given a
motive change may be that the motive change under these
circumstances may be one from absolute gain to relative
gain or vice versa. Here, the motive would change but the
game play would not.
With respect to the predictions of increased dyad
similarity, it appears that 50 trials simply are not
enough to show this similarity. The data indicated that
the average difference in cooperation level between dyad
members stayed relatively the same across trials as did
the difference in the proportion of trials in which dyad
members indicated joint motive. One of the reasons
generally given for the increase in dyad similarity is the
lock-in effect and 50 trials, as previously mentioned,
simply are not enough to permit a lock-in to take place.
Although the overall competitive nature of the game seems
high, with an increased number of trials there would
probably have been an even greater competitive tone due to
the possibility of this competitive lock-in. This seems
especially evident considering that Kelley and Stahelski
(1970) found that when a cooperative and competitive
person interact the cooperative one tends to play like the
competitive one. Because of this change on the part of
the cooperator, the competitor misjudges the intentions of
62
the cooperator as competitive. Based upon the present
research and Kelley and Stahelski, it seems to be even
more evident that the proportion of competitive play should
increase since predictions of the other players motive
affects both a player's behavior and motive.
Kelley and Stahelski (1970) have indicated that a
person's action at any given point is likely to depend on
what he thinks the other person will do, and this in turn
is likely to reflect what he thinks the other person is
trying to accomplish in the relationship, particularly
whether he is trying to satisfy their joint interests or
merely his own interests. They assume (a) that there are
two types of individuals— cooperative and competitive;
(b) that they have different views of the world; and
(c) these views are due to experience in social situations.
These assumptions may be interpreted as reflecting a
reciprocity of interactions with the other player,
including the norms which prescribe both what to expect
from the other and how to act toward the other. Their
results lead to the deduction that cooperators and
competitors will have different beliefs about what other
persons are like with respect to cooperativeness and
competitiveness. Specifically, they infer that coopera
tors will believe others are heterogeneous as to their
cooperativeness versus competitiveness, whereas competitors
will believe other persons to be uniformly competitive.
63
Further,
Being expectant of variability in the types of persons
they meet, cooperators should show a greater desire
than competitors to obtain information about potential
partners (or opponents) before interacting with them.
Similarly, their view of heterogeneity should heighten
the cooperators' motivation to make discriminations
among other persons and to become capable at doing so.
The data shown in Tables 7 through 9 seem to be consistent
with those ideas. Competitors did not play differentially
based on their prediction of the other player's motive
but the cooperators did. It is possible that although
competitors indicated a prediction of the other player's
motive it really did not mean anything to the player since
he was playing competitive and the other player's motive
would not change his play. A cooperator, however, would
pay careful attention to his prediction of the other
player's motive since the player's behavior would change
due to this prediction.
In summary, the present investigation has demon
strated three major findings. First, the motives with
respect to prisoner's dilemma may be reduced to a bipolar
scale of joint versus non-joint motive. There seems to be
no clear cut distinction in the game play behavior between
the motives of absolute gain or relative gain. Secondly,
players of prisoner's dilemma may change their behavior
without a corresponding change in their motive. The
opposite, however, is not true, generally speaking. This
phenomenon seems to take place most often just following
64
the exploitation of one player and seems due to the
cooperator playing competitively as a self defense to
prevent further exploitation. The third finding is support
for Kelley and Stahelski's (1970) idea that cooperators
are more sensitive to the type of opponent and change
their behavior depending on their perception of him
(Tables 1, 8, and 9).
65
BIBLIOGRAPHY
BIBLIOGRAPHY
Bixenstine, V. E.; Levitt, C.; and Wilson, K.
"Collaboration Among Six Persons in a Prisoner's
Dilemma Game." Journal of Conflict Resolution, X
(1966), 488-496.
________ ; Potash, H. M.; and Wilson, K. "Effects of Level
of Cooperative Choice by the Other Player on Choices
in a Prisoner's Dilemma Game." Journal of Abnormal
and Social Psychology, LXVI (1963), 308-313.
______ , and Wilson, K. "Effects of Level of Cooperative
Choice by the Other Player on Choices in a Prisoner's
Dilemma Game." Journal of Abnormal and Social
Psychology, LXVII (1963), 139-147.
Davis, M. D. Game Theory. New York: Basic Books, 1970.
Deutsch, M. "Trust and Suspicion." Journal of Conflict
Resolution, II (1958), 265-279.
________ . "Socially Relevant Science: Reflections on Some
Studies of Interpersonal Conflict." American
Psychologist, XXIV (1969), 1076-1092.
Evans, G. "Effect of Unilateral Promise and Value of
Rewards Upon Cooperation and Trust." Journal of
Abnormal and Social Psychology, LXIX (1964), £87-590.
Fouraker, L. E., and Seigel, S. Bargaining Behavior. New
York: McGraw Hill, 1963.
Kahan, J. P. "Noninteraction in a Three-Person Prisoner's
Dilemma Game." Behavioral Science, XVIII (1973),
124-127.
Kelley, H., and Stahelski, A. "Social Interaction Basis
of Cooperators' and Competitors' Beliefs About
Others." Journal of Personality and Social
Psychology, XVI (1870), 66-91.
67
Knapp, W. M., and Podell, J. E. "Mental Patients,
Prisoners, and Students with Simulated Partners in a
Mixed Motive Game." Journal of Conflict Resolution,
XII (1968), 235-241.
Knox, R. E., and Douglas, R. "Low Payoffs and Marginal
Comprehension, Two Possible Constraints Upon Behavior
in the Prisoner's Dilemma." Paper presented at
meeting of Western Psychological Association, San
Diego, March, 1968.
_________, . "Trivial Incentives, Marginal
Comprehension, and Dubious Generalizations from
Prisoner's Dilemma Studies." Journal of Personality
and Social Psychology, XX (1971), 160-165.
Komorita, S. S. "Cooperative Choice in a Prisoner's
Dilemma Game." Journal of Personality and Social
Psychology, II (1965), 741-745.
Lave, L. B. "Factors Affecting Cooperation in the
Prisoner's Dilemma." Behavioral Science, X (1965),
26-38.
Loomis, J. L. "Communication, the Development of Trust
and Cooperative Behavior." Human Relations, XII
(1959), 305-315.
Luce, R. D., and Raiffa, H. Games and Decisions. New
York: Wiley, 1957.
______ , and Rogow, A. A. "A Game Theoretic Analysis
ofCongressional Power Distributions for a Stable
Two-Party System." Behavioral Science, I (1956),
85-95.
McClintock, C. G.? Harrison, A.; Strand, S.; and Gallo, P.
"Internationalism-Isolationism, Strategy of the Other
Player and Two-Person Game Behavior." Journal of
Abnormal and Social Psychology, LXVII (1963), (T3T-
635.
________ , and McNeel, S. P. "Reward and Score Feedback
as Determinants of Cooperative and Competitive Game
Behavior." Journal of Personality and Social
Psychology, XV (1966a), 606-613.
______ , ' "Reward Level and Game Playing
Behavior." Journal of Conflict Resolution, X (1966b),
98-KJ2.
68
________ r . "Cross-Cultural Comparisons of
Interpersonal Motives." Sociometry, XXIX (1966c)‘ ,
406-427.
_________> ' "Prior Dyadic Experience and Monetary
Reward as Determinants of Cooperative and Competitive
Behavior." Journal of Personality and Social
Psychology, V (1967), 282-294.
Messick, D. M., and McClintock, C. G. "Motivational Bases
of Choice in Experimental Games." Journal of
Experimental Social Psychology, IV (1968), 1-25.
______, and Thorngate, W. B. "Relative Gain Maximization
in Experimental Games." Journal of Experimental
Social Psychology, III (1967), 85-101.
Minas, J. S.; Scodel, A.; Marlowe, D.; and Rawson, H.
"Some Descriptive Aspects of Two-Person Non-Zero-Sum
Games." Journal of Conflict Resolution, IV (1960),
193-197.
Oskamp, S. "Effects of Programmed Strategies on Coopera
tion in the Prisoner1s Dilemma and Other Mixed Motive
Games." Journal of Conflict Resolution, XV (1971),
225-259. ---------------------------------
_____ , and Kleinke, C. "Amount of Reward as a Variable
in the Prisoner's Dilemma Game." Journal of
Personality and Social Psychology, XVI (1970), 133-
140.
_____ , and Perlman, D. "Factors Affecting Cooperation
in a Prisoner's Dilemma Game." Journal of Conflict
Resolution, IX (1965), 359-374.
Pruitt, D. G. "Motivational Processes in the Decomposed
Prisoner's Dilemma Game." Journal of Personality and
Social Psychology, XIV (1970), 227-238.
Radlow, R.; Weidner, M.; and Hurst, P. "The Effect of
Incentive Magnitude and 'Motivational Orientation'
Upon Choice Behavior in a Two-Person Non-Zero Sum
Game." Journal of Social Psychology, LXXIV (1968),
199-208.
Rapoport, Amnon, and Mowshowitz, A. "Experimental Studies
of Stochastic Models for the Prisoner's Dilemma."
Behavioral Science, XI (1966), 444-458.
69
Rapoport, Anato1. Two Person Game Theory: The Essential
Ideas. Ann Arbor: The University of Michigan Press,
1966.
________ • "Exploiter, Leader, Hero, and Martyr. The Four
Archetypes of the 2 X 2 Game." Behavioral Science,
XII (1967), 81-84. -------------------
________ / and Chammah, A. M. Prisoner1s Dilemma: A Study
in Conflict and Cooperation. Ann Arbor: The
University of Michigan Press, 1965.
________ , and Guyer, M. "A Taxonomy of 2 X 2 Games."
General Systems, XI (1966), 203-214.
Schelling, T. C. "The Strategy of Conflict: Prospectives
for a Reorientation of Game Theory." Journal of
Conflict Resolution, II (1958) , 203-26Tl
Scodel, A., and Minas, J. S. "The Behavior of Prisoners
in a 'Prisoner's Dilemma' Game." Journal of
Psychology, L (1960), 133-138.
________ } ; Ratoosh, P,; and Lepitz, M. "Some
Descriptive Aspects of Two-Person, Non-Zero-Sum
Games." Journal of Conflict Resolution, III (1959),
114-119.
Sermat, V. "The Effect of An Initial Cooperative or
Competitive Treatment Upon a Subject's Response to
Conditional Cooperation." Behavioral Science, XII
(1967), 301-313.
Shapley, L. S., and Shubik, M. "A Method for Evaluating
the Distribution of Power in a Committee System."
American Political Science Review, XLVIII (1954),
787-792.
Siegel, S. Nonparametric Statistics for the Behavioral
ScienceiT New York: McGraw Hill, 1956.
Solomon, L. "The Influence of Some Types of Power
Relationships and Game Strategies Upon the Development
of Interpersonal Trust." Journal of Abnormal and
Social Psychology, LXI (I960)", 223-230.
Terhune, K. W. "Motives, Situation, and Interpersonal
Conflict Within the Prisoner's Dilemma." Journal of
Personality and Social Psychology, VIII (1968), 1-24.
70
Voissem, N., and Sistrunk, F. "Communication Schedule and
Cooperative Game Behavior." Journal of Personality
and Social Psychology, XIX (1971)', 160-167.
Wichmann, H. "Effects of Isolation and Communication on
Cooperation in Two Person Games." Journal of
Personality and Social Psychology, XVI (1970), 114-
12157
Wrightsman, L. S. "Personality and Attitudinal Correlates
of Trusting and Trustworthy Behavior in a Two-Person
Game." Journal of Personality and Social Psychology,
IV (1966), 328-332.
________ ; O'Connor, J.; and Baker, N. J. (eds.).
Cooperation and Competition: Readings on Mixed
Motive Games. Belmont: Brooks/Cole, 1972.
71
APPENDICES
APPENDIX A
INSTRUCTIONS
In this experiment you will be playing a game for
money in which the outcome depends not only on what you do,
but also upon what the other player does. The other
player's choices affect the amount of money you receive,
and your choices affect what the other player receives.
There are four characteristics of the game that you
should understand in order to play the game intelligently:
First, each of you have two choices, these choices are
indicated by the "1" and "2" on switch "W"; second, you
should understand that your choice "2" will always result
in more points for you than choice "1." It should also
be understood that your choice "1" will always result in
more points for the other player than your choice "2";
third, after each trial, one of the four cells in the
matrix will light up. On each trial you will receive the
number of points indicated in the cell which is lit and
these points will be converted to money after the game.
Thus, if you make choice "2" and the other player makes
choice "1," you would receive 12 points and the other
player would lose 12 points. If however, you make choice
"2" and the other player makes choice "2" you each would
lose 6 points; fourth, you should be able to detect which
choice each of you has made. This task is not difficult.
Player A controls which row or horizontal set will be
selected and Player B controls which column or vertical
set will be selected. Notice that any time you make choice
"2" you receive more points on that trial than if you had
made choice "1." Look at the display in front of you (E
lights cell "1,1"). What response would you have made if
this were the outcome? What response would your opponent
have made? If this were the outcome that had resulted,
Player A must have made choice "1," and Player B must have
made choice "1." In this case Player A and Player B both
would have received 6 points each.
The following instructions are inserted for the
experimental group of subjects only:
On the left hand side of the machine you will
notice that there are three additional switches, they are
marked switch "X," switch "Y," and switch "Z" respectively.
Switch "X" is used to indicate the selection of your
motive. Switch "Y" will indicate your prediction of what
you think the other player1s choice will be, and finally
switch "Z" will indicate your prediction of what you
believe the other player's motive will be. A description
of the three motives are listed next to switches "X" and
"Z" which are as follows: JOINT) which is a motive for a
high combined score for both yourself and the other player;
ABSOLUTE) which is a motive for a high score for yourself
only regardless of what the other player gets; and
75
RELATIVE) which is the desire for a score higher than the
other player. For your convenience, notice that these
definitions are posted above each machine.
On each trial you are to select one of the two
choices on switch "W" and indicate one of the three motives
on switch "X." In addition, you are to predict which
choice you believe the other player will select, which is
controlled by switch "Y" and, finally which motive you
believe the other player will indicate by using switch
"Z." It is important that you realize that only switch
1 1 W" controls which cell will light up in the matrix.
The following instructions are again for all sub
jects:
When you have finished making your responses on each
trial indicate that you are ready to see the outcome by
depressing the "ready" button located below the matrix.
It is of the utmost importance that you do not
communicate with each other in any form whatsoever. This
includes sighing, laughing, or any other form of communica
tion which might indicate how you feel about given out
comes, or how you would like the other player to behave.
After both of you have pressed the ready button I
will light up the appropriate cell. When the light is
turned off you should immediately begin making choices for
the next trial. Do not change your responses after you
have pressed the ready button.
76
Are there any questions before we begin game play?
If not, let us begin. Begin making your choices for trial
one.
77
APPENDIX B
MOTIVE DEFINITIONS
1. JOINT*****THE DESIRE FOR A HIGH COMBINED SCORE FOR
BOTH YOURSELF AND THE OTHER PLAYER
2. ABSOLUTE*****THE DESIRE FOR A HIGH SCORE FOR YOURSELF
ONLY
3. RELATIVE*****THE DESIRE FOR A SCORE HIGHER THAN THE
OTHER PLAYER
79

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A Factor Analysis Of The Semantic-Evaluation Abilities

PDF

A Temporal Approach-Avoidance Conflict In An Academic Test Situation

PDF

Death Anxiety In Leukemic Children

PDF

Intellectual And Cognitive Factors In The Production Of Psychological Stress Reactions

PDF

Non-Specific Treatment Factors And Deconditioning In Fear Reduction

PDF

Protein synthesis in the hippocampus of rats during learning assessed by radioautography

PDF

Recognition Accuracy In The Method Of Single Stimuli: A Test Of An Operational Definition Of The Distinctiveness Of Stimuli

PDF

Decision Making: An Individually Parameterized Deterministic Model

PDF

Age Differences In Serial Reaction Time As A Function Of Stimulus Complexity Under Conditions Of Noise And Muscular Tension

PDF

Psychomotor Performance And Change In Cardiac Rate In Subjects Behaviorally Predisposed To Coronary Heart Disease

PDF

Aggression As A Function Of Arousal And Friendship Ties

PDF

The Effect Of Irrelevant Environmental Stimulation On Vigilance Performance

PDF

Human Performance As A Function Of The Joint Effects Of Drive And Incentive Motivation

PDF

The Effect Of Discriminability On The Partial Reinforcement Effect In Human Gsr Conditioning

PDF

Defense Choice And Identification

PDF

Figural And Symbolic Divergent-Production Abilities In Adults And Adolescents

PDF

The effects of an RNA polymerase in the improvement of verbal learning and memory

PDF

The Subjective Utility Function As An Estimator Of Optimal Test Weights For Personnel Selection

PDF

Differential effects of epinephrine and propranolol on shuttle box avoidance learning in rats of different ages

PDF

Habituation Of The Multiple Unit Discharge Response To White Noise Stimulation In The Unanesthetized Rabbit

Asset Metadata

Creator Beale, Darryl Kurtz (author)

Core Title Conflicting Motives In The Prisoner'S Dilemma Game

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, experimental

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Kahan, James P. (committee chair), Cliff, Norman (committee member), Smith, Robert A. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-727825

Unique identifier UC11356602

Identifier 7428417.pdf (filename),usctheses-c18-727825 (legacy record id)

Legacy Identifier 7428417.pdf

Dmrecord 727825

Document Type Dissertation

Rights Beale, Darryl Kurtz

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA