Full download ebook at ebookgate.

com

Game Theory and its Applications in the

Social and Biological Sciences 2nd Edition
Andrew M. Colman

https://ebookgate.com/product/game-theory-and-its-
applications-in-the-social-and-biological-
sciences-2nd-edition-andrew-m-colman/

Download more ebook from https://ebookgate.com

More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Matrices and Society Matrix Algebra and Its

Applications in the Social Sciences Ian Bradley

https://ebookgate.com/product/matrices-and-society-matrix-
algebra-and-its-applications-in-the-social-sciences-ian-bradley/

Fuzzy Set Theory Applications in the Social Sciences

1st Edition Michael Smithson

https://ebookgate.com/product/fuzzy-set-theory-applications-in-
the-social-sciences-1st-edition-michael-smithson/

Biocalorimetry 2 Applications of Calorimetry in the

Biological Sciences 2nd Edition John E. Ladbury

https://ebookgate.com/product/biocalorimetry-2-applications-of-
calorimetry-in-the-biological-sciences-2nd-edition-john-e-
ladbury/

Microcontroller Theory and Applications with the PIC18F

2nd Edition M. Rafiquzzaman

https://ebookgate.com/product/microcontroller-theory-and-
applications-with-the-pic18f-2nd-edition-m-rafiquzzaman/
Computational Finance And Its Applications II M.
Costantino

https://ebookgate.com/product/computational-finance-and-its-
applications-ii-m-costantino/

Bionanotechnology Biological Self Assembly and Its

Applications 1st Edition Bernd H. A. Rehm

https://ebookgate.com/product/bionanotechnology-biological-self-
assembly-and-its-applications-1st-edition-bernd-h-a-rehm/

Multidimensional Liquid Chromatography Theory and

Applications in Industrial Chemistry and the Life
Sciences Steven A. Cohen

https://ebookgate.com/product/multidimensional-liquid-
chromatography-theory-and-applications-in-industrial-chemistry-
and-the-life-sciences-steven-a-cohen/

Recent advances in operator theory and its applications

the Israel Gohberg anniversary volume International
Workshop on Operator Theory and its Applications IWOTA
2003 Cagliari Italy 1st Edition Marinus A. Kaashoek
https://ebookgate.com/product/recent-advances-in-operator-theory-
and-its-applications-the-israel-gohberg-anniversary-volume-
international-workshop-on-operator-theory-and-its-applications-
iwota-2003-cagliari-italy-1st-edition-marin/

Quantum Measurement Theory and its Applications 1st

Edition Kurt Jacobs

https://ebookgate.com/product/quantum-measurement-theory-and-its-
applications-1st-edition-kurt-jacobs/
INTERNATIONAL SERIES IN SOCIAL PSYCHOLOGY

Game Theory &

its Applications
IN THE SOCIAL AND BIOLOGICAL SCIENCES

r *“■
Ifc

ANDREW M. COLMAN
Game Theory and its Applications
INTERNATIONAL SERIES IN SOCIAL PSYCHOLOGY

Series Editor: VV. Peter Robinson, University of Bristol

Assertion and its Social Context

Children's Social Competence in Context
Emotion and Social Judgements
Genius and Eminence
Making Sense of Television
Psychology of Gambling
Social Dilemmas: Theoretical Issues and Research Findings
The Theory of Reasoned Action
Game Theory and its Applications
in the Social and Biological
Sciences
Second Edition

Andrew M. Colman

Routledge
Taylor & Francis G r o u p

LONDON A N D NEW YORK

 First published 1982 as Game Theory and Experimental Games

Published 1995 by Butterworth-Heinemann Ltd

Published 1999 and 2003

by Routledge
27 Church Road, Hove, East Sussex BN3 2FA

Simultaneously published in the USA and Canada

by Routledge
270 Madison Avenue, New York N Y 10016

Transferred to Digital Printing 2008

Routledge is an imprint of the Taylor & Francis group, an Informa business

© 1982, 1995, 1999 Andrew M. Colman

The right of Andrew M. Colman to be identified as the author of this work

has been asserted by him in accordance with the Copyright, Designs and
Patents Act 1988

A l l rights reserved. No part of this book may be reprinted or

reproduced or utilized in any form or by any electronic,
mechanical, or other means, now known or hereafter
invented, including photocopying and recording, or in any
information storage or retrieval system, without permission i n
writing from the publishers.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloging in Publication Data

A catalogue record for this book is available from the Library of Congress

ISBN 978-0-7506-2369-8

Printed and bound in the UK by TJ1 Digital, Padstow, Cornwall

This publication has been produced with paper manufactured to strict environmental
standards and with pulp derived from sustainable forests.
 Contents

Preface to the First Edition ix

Preface to the Second Edition xi

Part I Background

1 Introduction 3
1.1 Intuitive background 3
1.1.1 Head On 4
1.1.2 Price War 5
1.1.3 Angelo and Isabella 5
1.2 Abstract models: basic terminology 6
1.3 Skill, chance, and strategy 10
1.4 Historical background 12
1.5 Summary 14

2 One-person games 15
2.1 Games against Nature 15
2.2 Certainty 15
2.3 Risk 17
2.4 Expected utility theory 19
2.5 Uncertainty 23
2.5.1 Insufficient reason 25
2.5.2 Maximax 26
2.5.3 Maximin 26
2.5.4 Minimax regret 28
2.6 Summary 32

3 Coordination games and the minimal social situation 33

3.1 Strategic collaboration 33
3.2 Coordination games 33
3.3 The minimal social situation 40
3.4 The multi-person minimal social situation 48
3.5 Summary 50
vi Contents

Part II Theory and empirical evidence

4 Two-person zero-sum games 53

4.1 Strictly competitive games 53
4.2 Extensive and normal forms 54
4.3 Games with saddle points: Nash equilibria 57
4.4 Games without saddle points 62
4.5 Dominance and admissibility 69
4.6 Methods for finding solutions 71
4.7 Ordinal payoffs and incomplete information 78
4.8 Summary 83

5 Experiments with strictly competitive games 85

5.1 Ideas behind experimental games 85
5.2 Empirical research on non-saddle-point games 87
5.3 Empirical research on saddle-point games 91
5.4 Framing effects 94
5.5 Critique of experimental gaming 97
5.6 Summary 99

6 Two-person mixed-motive games: informal game theory 100

6.1 Mixed-motive games 100
6.2 Subgame perfect and trembling-hand equilibria 101
6.3 Classification of 2 X 2 mixed-motive games 107
6.4 Leader 108
6.5 Battle of the Sexes 110
6.6 Chicken 111
6.7 Prisoner's Dilemma 115
6.8 Comparison of archetypal 2 X 2 games 118
6.9 Theory of metagames 121
6.10 Two-person cooperative games: bargaining solutions 126
6.10.1 Maximin bargaining solution 129
6.10.2 Nash bargaining solution 130
6.10.3 Raiffa-Kalai-Smorodinsky bargaining solution 131
6.11 Summary 132

7 Experiments with Prisoner's Dilemma and related games 134

7.1 The experimental gaming literature 134
7.2 Strategic structure 135
7.3 Payoffs and incentives 139
7.4 Communication effects 141
7.5 Programmed strategies 142
7.6 Axelrod's computer tournaments 144
7.7 Sex differences and cross-cultural studies 149
 Contents vii

7.8 Attribution effects

7.9 Framing effects
7.10 Summary

8 Multi-person cooperative games: coalition formation

8.1 N-person cooperative games
8.2 Characteristic function and imputation
8.3 Core and stable set
8.4 Harold Pinter's The Caretaker
8.5 Shapley value
8.6 Kernel, nucleolus, and least core
8.7 Coalition-predicting theories
8.7.1 Equal excess theory
8.7.2 Caplow's theory
8.7.3 Minimal winning coalition theory
8.7.4 Minimum resource theory
8.8 Experiments on coalition formation
8.9 Summary

9 Multi-person non-cooperative games and social dilemmas

9.1 N-person non-cooperative games: Nash equilibria
9.2 The Chain-store paradox and backward induction
9.3 Auction games and psychological traps
9.4 Social dilemmas: intuitive background
9.4.1 The "invisible hand" and voluntary wage restraint
9.4.2 Conservation of natural resources
9.4.3 The tragedy of the commons
9.5 Formalization of social dilemmas
9.6 Theory of compound games
9.7 Empirical research on social dilemmas
9.7.1 Group size effects
9.7.2 Communication effects
9.7.3 Individual differences and attribution effects
9.7.4 Payoff and incentive effects
9.7.5 Framing effects
9.8 Summary

Part III Applications

10 Social choice and strategic voting

10.1 Background
10.2 Alternatives, voters, preferences
10.3 Voting procedures
viii Contents

10.4 Voting paradoxes 237

10.5 Arrow's impossibility theorem 244
10.6 Proportional representation: single transferable vote 246
10.7 Strategic (tactical) voting 250
10.8 Sophisticated voting 258
10.9 Empirical evidence of strategic voting 267
10.10 Summary 270

11 Theory of evolution: strategic aspects 272

11.1 Historical background 272
11.2 Strategic evolution and behavioural ecology 272
11.3 Animal conflicts and evolutionarily stable strategies 276
11.4 An improved multi-person game model 281
11.5 Empirical evidence 288
11.6 Summary 293

12 Game theory and philosophy 294

12.1 Relevance of game theory to practical problems 294
12.2 Rationality in games 297
12.2.1 Coordination games 297
12.2.2 Prisoner's Dilemma games 300
12.3 Newcomb's problem 304
12.4 Kant's categorical imperative 308
12.5 Plato, Hobbes, Rousseau: social contract theories 310
12.6 Evolution and stability of moral principles 314
12.7 Summary 316

Appendix A simple proof of the minimax theorem 317

A.l Introductory remarks 317
A.2 Preliminary formalization 317
A.3 The minimax theorem 318
A.4 Proof 319
A.5 Remark 324

References 325

Index 363
 Preface to the First Edition

The primary aim of this book is to provide a critical survey of the essential
ideas of game theory and the findings of experimental research on strategic
interaction. In addition, I have reported some new experiments using lifelike
simulations of familiar kinds of strategic interactions, and included
discussions of recent applications of game theory to the study of voting, the
theory of evolution, and moral philosophy. The time has (alas) long since
passed when a single person could reasonably hope to be an expert on all
branches of game theory or on all of its applications, and I have not
attempted to achieve the impossible. But I thought it worthwhile, none the
less, to aim for a fairly comprehensive coverage of important topics, with
particular emphasis on those that seemed to be most relevant to naturally
occurring strategic interactions.
Game theory and the experimental gaming tradition have grown up in
relative isolation from each other. Game theorists, in general, remain largely
oblivious of the empirical studies that have been inspired by the theory, and
experimental investigators have tended to assume that the nuts and bolts of
the theory do not concern them. Both parties are the losers from this divorce,
and I have therefore tried to contribute towards a reconciliation by
examining in detail, for the first time in a single volume, both sides of the
story.
My goal has been to introduce and evaluate the fundamental theoretical
ideas, empirical findings, and applications as clearly as possible without
over-simplifying or side-stepping important difficulties. In so far as I have
succeeded, this is the kind of book that I should have liked to have read
when I first became interested in game theory and experimental games, or
better still, before I developed any interest in these matters. Wherever
possible, I have attributed seminal contributions to their originators and
cited the original sources: ideas are almost invariably expressed more clearly
and forcefully by their inventors or discoverers than by subsequent
commentators. But I have also cited many useful review articles, which will
be of assistance to readers wishing to pursue particular topics in depth.
The most important chapters [the numbering of chapters has been
amended here to correspond with the second edition] for social psycholo
gists and others whose primary interest is in such strategic phenomena as
cooperation, competition, collective equilibria, self-defeating effects of
x Preface to the First Edition

individual rationality, coalition formation, threats, altruism, spite, escala

tion, social entrapment, and so forth, are chapters 1, 2, 3, 6, 7, 8, and 9.
Mathematically inclined readers should pay special attention to chapters 3,
4, 6, 8, 10, and to the Appendix in which the minimax theorem is rigorously
proved. The chapters most relevant to sociology, economics, and politics are
chapters 1, 2, 3, 6, 8, 9, and 10. Biological applications are discussed in
chapter 11, but chapters 1, 2, 6, 7, and 9 provide a necessary background.
Philosophical applications are dealt with primarily in chapter 12, to which
chapters 1, 2, 3, 6, 9, and 11 provide the necessary background. Most of the
translations from original French and German sources in chapter 12 and
elsewhere are my own.
I am indebted to a number of people who contributed to this book in
various indirect ways. In particular, I am grateful to Michael Argyle, Alan
Baker, Barbara Barrett, Dorothy Brydges, Roy Davies, Julia Gibbs, Gabriele
Griffin, John Lazarus, Nicholas Measor, Richard Niemi, Ian Pountney, Albert
W. Tucker, Diane Williams, Bill Williamson, and the Research Board of the
University of Leicester. I should be delighted to receive comments from
readers, indicating their reactions to the final product.

Andrew M. Colman
 Preface to the Second Edition

The first edition of this book was entitled Game Theory and Experimental
Games: The Study of Strategic Interaction and was published by Pergamon
Press in 1982. It aimed to bridge the gap between game theory and its
applications by providing an introduction to the theory and a reasonably
comprehensive survey of some of its major applications and associated
experimental research. Its more specific objectives were to explain the
fundamental ideas of mathematical game theory from first principles and to
provide an introductory survey of experimental games and other applica
tions of the theory in social psychology, decision theory, evolutionary
biology, economics, politics, sociology, operational research, and
philosophy.
The first edition was favourably received and generously reviewed on
both sides of the Atlantic and adopted as a text for a number of specialist
courses. Demand for the book, though modest, remained remarkably steady
for many years, but theoretical developments and new empirical findings
accumulated over the years, making the need for a revision of the original
text increasingly difficult to ignore. This second edition is so radically
revised as to be hardly the same book, and I believe it to be a significant
improvement on the first. The principal changes that I have introduced are
as follows. I have modified the title, at the suggestion of the publisher, to
indicate the scope of the book more accurately. I have corrected the errors
and plugged the gaps that have been pointed out to me, and I have
introduced numerous amendments and improvements to every chapter. I
have thoroughly updated the contents of the book to include significant or
interesting theoretical developments and empirical research findings related
to coordination games, social dilemmas, strategic aspects of evolutionary
biology, framing effects, strategic voting, and many other areas of research.
In the light of comments from readers and reviewers I have introduced a
little more formal mathematics where omitting it seems to have created
more problems than it solved.
In addition, I have improved the book's coverage by incorporating into
this second edition a number of important topics that have developed
recently or were missing from the first edition for other reasons. Chapter 3
xii Preface to the Second Edition

now indudes a great deal more theoretical and empirical work on

coordination games and the minimal sodal situation. Chapter 6 includes a
section on subgame perfect and trembling-hand equilibria and a further
section on bargaining solutions for two-person cooperative games; chapter
7 indudes a discussion of Axelrod's tournaments of Prisoner's Dilemma
game computer programs and a brief review of cross-cultural comparisons
of cooperativeness; chapter 8 is renamed and largely rewritten to include a
detailed review of the major theories of coalition formation; chapter 9 is
renamed and restructured and indudes a discussion of the Chain-store
paradox and backward induction; chapter 10 is renamed and restructured
and incorporates material on strategic voting, which was in a separate
chapter of its own in the first edition, and a discussion (requested by a
number of readers) of proportional representation voting procedures,
chapter 11 is renamed and includes a refutation of the notion that the
strongest always survive in evolutionary games, and chapter 12 is renamed
and restructured and includes discussions of philosophical problems related
to coordination games and Newcomb's problem.
I no longer believe that "ideas are almost invariably expressed more
clearly and forcefully by their inventors or discoverers than by subsequent
commentators", as I said in the preface to the first edition. I now believe, on
the contrary, that innovators sometimes struggle to understand their own
inventions or discoveries and that their successors often understand and
explain them better. I have therefore been more liberal with my citations of
secondary sources in this edition.
I remain indebted to the people who helped me with the first edition and
were acknowledged in its preface. Preparation of this edition was facilitated
by Grant No. L122251002 from the Economic and Social Research Council as
part of the research programme on Economic Beliefs and Behaviour.
Numerous thoughtful readers in Britain, the United States, the Netherlands,
and elsewhere have made helpful suggestions that have been incorporated
into this second edition, and I am grateful to them all. They are too
numerous to list exhaustively, but special thanks are due to Jerome
Chertkoff, whose thoughtful review in Contemporary Psychology included
constructive suggestions for improvement that I have implemented fully, to
Werner Tack and Manfred Vorwerg, who also offered important practical
advice that I have followed, to Michael Bacharach of Oxford University and
other colleagues involved in the Framing, Salience and Product Images
research project for inspiration and advice, to Brian Parkinson and David
Stretch, who provided useful comments, to Roy Davies, who helped me to
improve the mathematical appendix and the general presentation of this
edition, to Zhu Zhifang of Wuhan University in the People's Republic of
China, who discovered an important though deeply hidden error while
working on a Chinese translation of the book, and to Kathy Smith for
preparing many of the payoff matrices. But in spite of everyone's help and
 Preface to the Second Edition xiii

my own best efforts, this book is not free of errors. It cannot be error-free,
because if it were, then the sentence immediately preceding this one would
be an error, which would mean that it was not error-free after all. I should
be grateful to readers who spot any more serious defects or who have any
other comments to make.

Andrew M. Colman
- Part I -
Background
 Introduction

1.1 Intuitive background

Game theory is a branch of mathematics devoted to the logic of decision

making in social interactions. It is applicable to any social interaction with
the following three properties:
(a) there are two or more decision makers, called players;
(b) each player has a choice of two or more ways of acting, called strategies,
such that the outcome of the interaction depends on the strategy choices
of all the players;
(c) the players have well-defined preferences among the possible outcomes,
so that numerical payoffs reflecting these preferences can be assigned to
all players for all outcomes.
Any social interaction with these three properties is a game in the
terminology of game theory - or to be more precise could be modelled by a
game, which is really an abstract mathematical invention.
An essential feature of these social interactions is that each decision maker
has only partied control over the outcome. It is immediately obvious that
games like chess and poker are games in the technical sense, but that other
activities such as patience (solitaire), doll play, hopscotch, and solitary
computer games such as Nintendo or Sega, in spite of being games in the
popular sense of the word, are not. More importantly, many economic,
political, military, and interpersonal conflicts that are seldom if ever referred
to as games in everyday speech nevertheless have the three properties that
define them as games in the terminology of game theory. In spite of its name,
game theory is not specifically concerned with recreations and pastimes, and
a less misleading name for it would have been the theory of interdependent
decision making, but it is too late to rename game theory without risking even
worse confusion.
The principal objective of mathematical game theory is to determine,
through formal reasoning alone, what strategies the players ought to choose
in order to pursue their own interests rationally and what outcomes will
result if they do so. Formal game theory is therefore normative rather than
positive or descriptive inasmuch as it seeks to discover how players ought to
4 Background

behave in order to pursue their own interests most effectively but makes no
predictions about how they will behave in any actual interaction, and for
that reason it cannot be tested by experimental methods. But in non-
mathematical applications in the social and biological sciences, game theory
has provided a useful framework for explaining and predicting behaviour in
a wide range of situations involving interdependent decision making, and
predictions derived from informal game theory have been tested through
empirical research. Even in cases in which satisfactory formal solutions
cannot be found, informal game theory has often provided deep and
illuminating insights. Certain important features of individual and col
lective rationality, cooperation and competition, trust and suspicion, threats
and commitments cannot even be clearly described, let alone explained,
without the framework of game theory.
Traditional theories in social psychology and related fields lack the
necessary concepts with which to deal rigorously with interdependent
decision making. They have tended to emulate the theoretical models that
are used in classical physics, in which the behaviour of objects is explained
in terms of responses to external forces. But inanimate objects do not make
deliberate choices, and their behaviour is not governed by any assumptions
about how other objects will behave. One-way causal models may be
appropriate for explaining certain involuntary human responses, such as
kicking the air in response to a doctor's tap on the patellar tendon or
blinking the eye in response to the intrusion of a speck of dust. But a person
may kick the air voluntarily as an act of symbolic aggression in the hope of
gaining some strategic advantage during a competitive interaction, or may
deliberately wink an eye in an attempt to increase the intimacy of a personal
relationship. These are deliberate actions whose outcomes depend on the
decisions of other people as well, and one-way causal models of the kind
that have proved so successful in the physical sciences are unlikely to
explain them adequately. Psychologically interesting and significant forms
of social behaviour almost all involve deliberate decisions, and they are
sometimes amenable to game theory analysis. A few simple examples will
help to provide an intuitive understanding of the kinds of social interactions
involving interdependent decision making that fall within the compass of
game theory.

1.1.1 Head On

Two people are walking briskly towards each other along a narrow corridor.
They are heading for a collision, which they would both prefer to avoid. To
simplify the analysis, let us assume that the walkers have to make their
decisions immediately, and therefore simultaneously, and that each has just
three strategies from which to choose: swerve left, swerve right, or keep
 Introduction 5

going straight ahead. If both keep going straight ahead, or if both swerve to
the same side of the corridor, then they will collide; all other strategy
combinations lead to non-collision outcomes, which they both prefer. The
outcome of the game depends on the decisions of both walkers, whose
interests coincide exactly inasmuch as their preferences among the outcomes
are identical.

1.1.2 Price War

Three retail companies are each trying to carve out a larger slice of a market
for which they all compete. Each has to decide, in ignorance of the decisions
of the others, whether or not to advertise its product. If all three companies
advertise simultaneously, none will gain or lose market share, but all will
lose the cost of their advertisements. The status quo market shares will also
be preserved, slightly more cheaply, if all three decide not to advertise.
Other strategy combinations result in increased market share for one or two
of the companies at the expense of the other(s). In this example, the players'
interests are mutually opposed, because one company's gain necessitates a
loss for at least one of the others.

1. 1.3 Angelo and Isabella

This poignant example is taken from Shakespeare's Measure for Measure (II.
iv). Angelo is holding a prisoner whom he intends to execute. The prisoner's
sister, Isabella, comes to plead for her brother's life. Angelo is sexually
attracted to her and at length proposes the following ungentlemanly
bargain: "You must lay down the* treasures of your body" to save your
brother. Isabella initially declines the offer: "More than our brother is our
chastity". Angelo then complicates the game by threatening not merely to
kill the prisoner, as he had originally intended, but to subject him to a
lingering death unless Isabella submits. At this point Isabella faces a choice
between submitting and refusing, and whichever option she chooses there
are three courses open to Angelo: to spare the prisoner's life, to execute him
humanely, or to subject him to a lingering death. The outcome depends on
the choices of both Angelo and Isabella, and all but one of the possible
outcomes lead to unattractive payoffs for Isabella (the exception being
Isabella's refusal coupled with a reprieve from Angelo), although some are
clearly worse than others. Both protagonists would prefer to see the prisoner
executed humanely rather than cruelly, other things being equal, so the
players' interests are partly opposed and partly coincident (Schelling, 1960,
p. 140).
6 Background

1 .2 Abstract models: basic terminology

Game theory deals with social interactions like the ones described above by
abstracting their formal, logical properties. A game is a purely imaginary
idealization of a social interaction. A real social interaction is invariably too
complex and transitory to be clearly perceived and perfectly understood, so
it is replaced by a deliberately simplified abstract structure whose rules and
basic elements - players, strategies, and payoffs - are explicitly defined and
from which other properties can be deduced by formal reasoning. These
deductions apply not to the social interaction itself but to the game that
models it, and they are true, provided only that the reasoning does not
contain errors, whether or not the game corresponds accurately to the
original social interaction. But if the game does not correspond to social
reality in important particulars, then its practical value is at best limited; and
if it does not yield insights that transcend a common-sense understanding of
the social interaction, then it serves no useful purpose at all.
The formal models devised by game theorists are sometimes criticised
for their lack of concreteness and their failure to capture the full
complexity of the social realities that they are designed to represent. This
criticism stems from a profound misunderstanding of the purpose of a
formal model, which is to reduce reality to its bare essentials by
deliberately excluding non-essential details. In some fields of investigation
formal models have proved so successful that many people forget that
they are merely abstractions and confuse them with reality itself. Euclidian
geometry is a striking (and indeed a classical) example, which was
eventually shown to correspond imperfectly to reality, although it remains
adequate for most ordinary purposes.
The examples in the previous section illustrate the fact that the decision
makers or players (to use the technical term) may be individual human
beings or corporate decision-making bodies, and there are even applications
discussed elsewhere in this book in which they represent non-human
animals. The essential attribute of a player is the capacity to choose or, in
other words, to make decisions. These decisions are called moves or strategies
- the distinction between these two concepts will be explained shortly.
In some games, the outcome depends partly on the invisible hand of
chance, and consequently the moves or actions of the (other) players do not,
on their own, fully determine the outcomes that follow. In order to handle
games like these, a fictional player called Nature is brought into the game,
and it is assumed that Nature chooses her moves according to the laws of
probability. Poker is a typical example: Nature makes the first move by
arranging the deck in a particular (random) order, and from that point on
she plays no part in the game until the next hand is to be dealt. In many
economic, political, military, and interpersonal games of everyday life,
Nature also plays an important part.
 Introduction 7

A game must involve at least two players, otherwise there could be no

interdependence of choice, but in the widest interpretation of game theory
this includes interactions between one ordinary player and Nature.
Furthermore, each player must have at least two courses of action from
which to choose, because an agent with only one way of acting would have
no effect on the outcome of the game and could therefore be ignored. In
defining the set of strategies facing a player, as in everyday life, it is
necessary to remind oneself that doing nothing is often one of the available
options.
The rules of the game specify what moves or actions are available to each
player, how the moves are made, including what the players are permitted
to know when they choose their moves, and what outcome is associated with
each possible combination of decisions by all the players. The rules of chess,
for example, specify three possible outcomes: White wins, Black wins, or a
draw. The rules governing the games outlined in the previous section were
specified informally in their descriptions. The outcome of Head On is either
a collision or a non-colUsion. In Price War, the outcomes are various
distributions of market share among the three companies. Angelo and
Isabella is a slightly more complicated case, which will help to clarify the
subtle though important distinction between moves (or actions) and
strategies.
One possible outcome of Angelo and Isabella is this: Isabella surrenders
her chastity to Angelo and her brother is subjected to a lingering death.
From Isabella's point of view, this is the worst possible outcome. It is
assumed in general that players have consistent preferences across the set of
possible outcomes. At the very least, each player must be able, in principle,
to arrange the outcomes in order of preference from best to worst with equal
ranks assigned to those between which the player is indifferent. It is then
possible to assign numerical payoffs to the outcomes, indicating the players'
relative preferences among them. In Angelo and Isabella, it is possible to
assign Isabella's lowest payoff, say 1, to her least preferred outcome, 2 to her
next-to-worst outcome - presumably a surrender of her chastity to Angelo
coupled with a humane execution of her brother - and so on up to her
favourite outcome, namely a preservation of her chastity and a reprieve for
her brother. Angelo's lowest payoff may (reasonably) be assigned to the
outcome in which Isabella withholds her sexual favours and her brother is
reprieved, and his highest to the one in which Isabella submits and her
brother is humanely executed as originally planned. Ordinal payoffs like
these are all that are needed for solving some games, but in others the
payoffs must be measured on interval scales. In other words, it is sometimes
necessary to know not merely which outcomes a player prefers to which
others, but also how strong the preferences are relative to one another.
In either event, a payoff function is defined for every player in the game.
Mathematically speaking, an individual player's payoff function is a
8 Background

relation that associates each possible outcome of the game with a unique
(ordinal or interval) payoff to that player; in other words it specifies what
payoff the player receives for every conceivable outcome of the game. A
complete payoff function for all of the players is simply an amalgamation of
the individual players' payoff functions; it specifies the payoffs to all players
for every conceivable outcome of the game.
A strategy is a complete plan of action, specifying in advance what moves
a player will make in every possible situation that might arise in the course
of the game. If each player has only one move to make, and if the moves are
made simultaneously or (what amounts to the same thing) in ignorance of
the moves chosen by the other player(s), then the available strategies are
simply the moves, and there is no distinction between the two concepts. In
Head On, for example, each player has only one move to make, namely
swerve left, swerve right, or keep going straight ahead, and the moves are
made simultaneously, so they are also the strategies available to the players.
Similarly in Price War, the players' strategies or moves are simply to
advertise or not to advertise. But in other cases, the players' strategies are
more complicated than the individual moves of the game. In Angelo and
Isabella, Isabella has just two moves, namely submit (S) or don't submit
(DS), and these are also her strategies. But the rules of this game stipulate
that Angelo makes his move after - and in full knowledge of - Isabella's. He
has just three available moves, namely reprieve (R), execute humanely (E),
or torture (T), but his strategy choice must specify in advance which move
he will choose in response to each of Isabella's possible moves. Although he
has only three moves or actions, it turns out that he has nine possible
strategies, as follows:
(1) If S, choose R; if DS, choose R.
(2) If S, choose R; if DS, choose E.
(3) If S, choose R; if DS, choose T.
(4) If S, choose E; if DS, choose R.
(5) If S, choose E; if DS, choose E.
(6) If S, choose E; if DS, choose T.
(7) If S, choose T; if DS, choose R.
(8) If S, choose T; if DS, choose E.
(9) If S, choose T; if DS, choose T.
Angelo's promise (or threat, depending on one's point of view) is to choose
his third strategy, but he is not bound by the rules of the game to honour it
if his bluff is called. In Shakespeare's play, Angelo's untrustworthiness
becomes tragically clear as the story unfolds and it transpires that he has in
fact chosen his fifth strategy in spite of his promise.
By choosing a strategy that specifies in advance what particular moves a
player will make in all possible contingencies, a player does not sacrifice
any freedom of action. All moves that are available to a player in the
 Introduction 9

extensive form of the game, which sets out the full sequence of moves in the
order in which they are made, are preserved in the normal form, which
collapses each player's sequence of moves into just one strategy choice. In
the normal form, a game involving a sequence of moves can be
represented statically, as though the outcome were determined by a single
choice on the part of each player, and this does not affect the strategic
properties inherent in the situation. The strategies chosen by the players
determine a unique sequence of moves and a definite outcome. On the
basis of their strategy choices, a referee could in principle make all the
moves in accordance with the players' plans and discover the outcome.
But this is a practical impossibility except in very simple games. In chess,
some 160000 different positions can arise after only two moves by each
player, and all the paper or all the computer disks in the world would not
have the storage capacity to record a player's strategy for an entire game.
Only much simpler games can be represented in normal form, and even in
these cases the way players actually think is often more usefully displayed
in extensive form, which is usually represented by a branching tree. The
normal form, though logically equivalent, is often less useful for practical
analysis, and it impedes the identification of subgame perfect equilibria,
which will be explained in section 6.2 (Myerson, 1991, chap. 5; Rasmusen,
1989, chap. 2).
Two technical terms, which need to be carefully distinguished, are used to
describe what the players know during the course of the game. In a game of
complete information, the players know not only the rules of the game and
their own payoff functions but also the payoff functions of the other
player(s). In other words, they know the available strategies and preferences
of all of the players. There is also an important assumption that each player
knows that the other(s) have complete information, and that they know that
the other player(s) know this, and so on; this is the assumption that the
description of the game is common knowledge (Heal, 1978; Lewis, 1969; see
also Bicchieri, 1993, passim; Sugden, 1991, pp. 764-765). Most games that
have been studied by game theorists, and most that have been used in
applications in the social and biological sciences, have been games of
complete information, but some important work has also been done on
games of incomplete information.
A game of perfect information is something rather different from one of
complete information. Informally, it is a game in which the players move
one at a time, rather than simultaneously, and a player choosing a move
always has full knowledge of all moves that have preceded it. Chess is a
game of perfect information, and so is Angelo and Isabella, but Head On
and Price War are not, because in these last two games moves are not made
one at a time and with the benefit of knowing what moves the other players
have chosen. All of these games are, however, games of complete
information.
10 Background

1 .3 Skill, chance, and strategy

It is possible to classify games according to certain family resemblances. This

is useful because the method used to analyse a particular game can often be
applied without modification of its underlying ideas to other games
belonging to the same general class; only a relatively few need therefore to
be studied in order to understand a much larger number. Perhaps the most
fundamental criterion of classification concerns the factors that influence the
outcomes of the games. It is accordingly useful to distinguish between
games of skill, games of chance, and games of strategy.
Games of skill and chance are species of one-person games, whereas
games of strategy involve two or more decision makers in addition to
Nature. Games of skill are often referred to as individual decision making under
certainty. Games of chance are called individual decision making under risk or
uncertainty or one-person games against Nature. These important distinctions
deserve the less cumbersome terminology of skill, chance, and strategy.
The defining property of a game of skill is the involvement of just one
player who has complete control over the outcomes so that each of the
player's strategy choices leads to a single certain outcome. Responding to an
IQ test might possibly be modelled by a game of skill, provided that there
is no significant element of luck: every possible combination of answers that
a testee might give to the questions constitutes a strategy, each strategy
choice leads to a definite outcome in the form of an IQ score, and it is
reasonable to assume that the player prefers some outcomes to some others.
In view of the fact that other players have no effect on the outcomes of
games of skill - not even Nature plays any part - these games constitute a
degenerate class from which the element of interdependence of choice is
lacking, and they do not, strictly speaking, qualify as games in the technical
sense outlined in section 1.2. But for the sake of completeness, and also to
place other one-person games in theoretical perspective, a brief discussion of
games of skill is included in section 2.2.
Games of chance are models of one-person decisions whose outcomes are
influenced by Nature. In a game of chance, the decision maker does not
control the outcomes completely, and strategy choices do not lead to
outcomes that can be guaranteed. The outcomes depend partly on the
choices of the decision maker - the agent to whom the adjective "one-
person" refers - and partly on the choices of the fictitious player, Nature.
Although these are one-person games, they involve two players,and they can
therefore be included within the domain of game theory.
Depending on how Nature's strategies are interpreted, games of chance
can be further subdivided into those involving risk and those involving
uncertainty. (This distinction was made explicit by Keynes, 1937; see also
Bicchieri, 1993, pp. 25-27; Lucas, 1981, p. 15.) In risky decision making, the
decision makers know the probabilities associated with each of Nature's
 Introduction 11

strategies and therefore of the outcomes associated with each of their own
strategy choices. Russian roulette is a good example of a game of chance
involving risk. For any number of live bullets that one chooses to load into
a revolver before pointing it at one's temple and pulling the trigger, the
probabilities associated with the two possible outcomes - life and death -
are obvious, but unless all of the bullet chambers are filled or all are empty,
the outcome cannot be predicted with certainty. The theory of games of
chance is essentially the theory of probability, which arose out of a
mathematical analysis of gambling. All games of this type can be
satisfactorily solved, as will be shown in chapter 2, provided that the
player's preferences among the outcomes are consistent.
There are other games of chance in which meaningful probabilities cannot
be assigned to Nature's strategies, and these are called decisions under
uncertainty. A doctor needing to choose a treatment for a patient with a rare
complaint may, for example, have no basis on which to judge the relevant
probabilities. Nature obviously has a hand in this game, because the
outcomes of the doctor's decisions are not certain, but in this case even the
probabilities are unknown. Ingenious solutions have been devised for
games of this type, but it will be argued in chapter 2 that none of them is
entirely sound.
Games of strategy model social interactions in which the outcomes
depend on the choices of two or more decision makers, each of whom has
partial control of the outcome. The players cannot meaningfully assign
probabilities to one another's choices, and therefore the decisions are said to
be made under uncertainty rather than risk. The bulk of the literature on
game theory is devoted to games of strategy, and it is fair to say that the
theory came into its own only when methods were discovered for solving
these games.
Games of strategy are especially useful for explaining social interactions,
and most of the chapters that follow are concerned with them. They can be
subdivided into two-person and multi-person games according to the
number of players (excluding Nature) involved. Both two-person and multi
person games can be further subdivided according to their structural
properties. One important structural property is the way in which the
players' payoff functions are related to one another. The players' preferences
may coincide perfectly as in Head On, they may be mutually opposed as in
Price War, or they may be partly coincident and partly opposed as in Angelo
and Isabella. Games in which the players' preferences coincide are called co-
ordination games, and research into such interactions is discussed in section
3.2. Games in which the players' preferences are mutually opposed are
called zero-sum games; they are dealt with in chapters 4, 5, and (part of) 8.
Most recreational games between two players or two teams are zero-sum in
the sense that what one player or team wins, the other must lose, and the
payoffs to the two players therefore add up to zero in each outcome, but
12 Background

most social interactions of everyday life lack this property. The rest of this
book is concerned mainly with two-person and multi-person games in
which the preferences of the players are neither perfectly coincident nor
entirely opposed. They are known technically as mixed-motive games.
Two-person zero-sum games are also called strictly competitive games. They
can be convincingly solved, provided that each player has only a finite
number of strategies, and empirical research has thrown some light on the
way in which people behave in strictly competitive interactions. Other types
of games, including multi-person zero-sum games and both two-person and
multi-person mixed-motive games, are not generally soluble in the formal
sense, but they provide important insights into some aspects of social
interaction, and some of them are more relevant to the social and biological
sciences than strictly competitive games.

1.4 Historical background

The forerunner of the theory of games of strategy was the theory of games
of chance. This earlier theory originated in the seventeenth century from
attempts to solve practical problems of gambling raised by members of the
dissolute French nobility before the revolution. From those frivolous origins
sprang the theory of probability, which provided a foundation for the later
development of statistics, quantum physics, and population genetics.
The theory of games of chance - or the theory of probability - is traced by
most historians to 1654, when an exchange of letters took place between
Blaise Pascal and Pierre de Fermat concerning the misfortunes of the French
nobleman, the Chevalier de Mere. De Mere had written to Pascal explaining
that he had won a considerable sum of money over a period of time by
betting even odds that at least one six would come up in four throws of a
die, only to lose it all by betting that at least two sixes would come up in 24
throws. According to de Mere's faulty calculations, his second bet ought to
have been at least as favourable as his first. The correct probabilities,
however, are in fact .5177 and .4914 respectively, that is, the first is
favourable but the second is not. De Mere must have been an extraordinarily
industrious and deep-pocketed gambler to have discovered by dint of sheer
bitter experience that the second bet was unfavourable, in other words to
have distinguished empirically between probabilities of .4914 and .5000. The
fundamental principles of probability theory were developed in order to
solve De Mdre's and other similar problems. The story is entertainingly told,
with extracts from the Pascal-Fermat letters, by Florence N. David (1962).
The first important contribution to the theory of games of strategy was
made by the German mathematician, Ernst Zermelo, in 1912. Zermelo
managed to prove that every strictly competitive game of perfect informa
tion that has a finite number of moves is strictly determined, that is, it
 Introduction 13

possesses either a guaranteed winning strategy for one of the players or

guaranteed drawing strategies for both. This result applies to games such as
noughts and crosses (ticktacktoe) and chess in which each player knows
what moves have been made previously, but it provides no method for
finding the winning or drawing strategies. Noughts and crosses has
drawing strategies for both players - if it is played rationally it must end in
a draw - but in the case of chess it is still not known whether White (or
conceivably Black) has a winning strategy or whether both players have
drawing strategies, although the latter is most likely.
Much of the groundwork of the theory of games of strategy was laid in a
series of papers by the French mathematician Emile Borel between 1921 and
1927 (see Frechet, 1953; Leonard, 1992), although Borel was anticipated more
than two centuries earlier by the Englishman James Waldegrave, who later
became the Earl Waldegrave (Dimand and Dimand, 1992). But Borel was
unable to prove the minimax theorem (see chapter 4 and the appendix) that
lies at the heart of formal game theory; in fact, he rashly conjectured that the
no such theorem could be proved. The minimax theorem was promptly
proved by the Austro-Hungarian mathematician John von Neumann in
1928. In 1934, independently of von Neumann, the British mathematician
Ronald Aylmer Fisher, who is remembered chiefly for his colossal
contribution to experimental design and statistics, proved the minimax
theorem for the special case in which each player has just two strategies.
Game theory became more widely known after the publication in the United
States of von Neumann and Morgenstern's classic Theory of Games and
Economic Behavior in 1944. This book stimulated a great deal of interest
among mathematicians and mathematically sophisticated economists, but it
was a later text entitled Games and Decisions: Introduction and Critical Survey
by Duncan Luce and Howard Raiffa (1957) that made the theory accessible
to a wide range of social scientists and psychologists.
Only a handful of empirical investigations of interdependent decision
making appeared before 1957, but the publication of Luce and Raiffa's book
was followed by a steady growth of experimental gaming as a field of
research (Smith, 1992). The first comprehensive review of experimental
games was written by Rapoport and Orwant in 1962, by which time 30
experiments had been published. By 1965, experimental games had become
sufficiently popular for the Journal of Conflict Resolution to begin devoting a
separate section of each issue to it, and by 1972 more than 1000 empirical
studies had appeared in that journal and elsewhere (Guyer and Perkel, 1972;
Wrightsman, O'Connor, and Baker, 1972). Towards the mid-1970s, many
researchers began to express dissatisfaction with experimental gaming
research, which was then dominated by extremely abstract versions of one
particular two-person game called the Prisoner's Dilemma game (see
chapters 6 and 7). This crisis of confidence in experimental games was
highlighted by an acrimonious debate in the European Journal of Social
14 Background

Psychology initiated by Pion (1974). The flow of experimental gaming

research nevertheless continued to increase, but less emphasis was placed
on the Prisoner's Dilemma game and more on newly developed multi
person games. By the mid-1990s some 2000 experimental gaming studies
had been published and empirical research continues to flourish.
Theoretical work on games of strategy also accelerated from the 1960s
onwards. The International Journal of Game Theory was founded in 1971 and
the journal Games and Economic Behavior in 1989. Russian scholars began to
make significant contributions to game theory in the late 1960s (see
Robinson, 1970; Vbrob'ev, 1977). In 1972 the Nobel Prize for Economics was
shared by Kenneth J. Arrow, whose work is closely related to game theory
and will be discussed in section 10.5, and in 1994 the prize was awarded to
the game theorists John Nash, John Harsanyi, and Reinhard Selten, whose
contributions are discussed at various points in the chapters that follow.
Numerous applications of the theory to economics were published from the
1950s, but applications to politics, biology, and philosophy, some of which
will be discussed in later chapters of this book, are more recent and are
developing rapidly.

1.5 Summary

Section 1.1 opened with an informal definition of game theory and an

outline of the types of social interactions to which it applies. An essential
property of these interactions is interdependent decision making, and it was
argued that game theory is more appropriate than the conventional one-way
causal models of social psychology and related fields for understanding
such interactions. Three hypothetical examples were given of social
interactions to which game theory could be applied. In section 1.2, the
important properties of games as abstract models, and basic technical
terminology, were explained with the help of the examples given earlier.
Section 1.3 focused on the various categories into which games can be
grouped. Games of skill, which constitute a degenerate class without true
interdependence, games of chance, which involve either risky gambles or
uncertainty, and games of strategy were distinguished. Games of strategy
were further subdivided into coordination, zero-sum, and mixed-motive
varieties. Section 1.4 was devoted to a synopsis of the historical develop
ment of game theory and its principal applications from the mid
seventeenth century to the present.
 ------- 2 -------
One-person games

2.1 Games against Nature

This chapter introduces the theory of one-person games. These are decision
making problems that are sometimes called one-person games against
Nature, and because of their non-social properties some theorists do not
consider them to be genuine games. The dilemmas discussed in section 2.2,
in which Nature plays no part, are certainly not games in the strict sense,
because they do not involve interdependent decisions, but they represent a
limiting case that serves as a convenient and logical point of departure. In
contrast, risky decisions, which will be discussed in sections 2.3 and 2.4, are
interdependent in the formal sense, although they involve no social
interaction, because Nature functions as an additional player, and some of the
techniques that are used to analyse them will reappear in more sophisticated
forms in the solution of two-person and multi-person games in later chapters.
In particular, the fundamental ideas underlying probability theory and
expected utility theory provide essential tools for the solution of more
complex games. Section 2.5 will deal with individual decision making under
uncertainty rather than risk, and the ideas developed in that section will also
reappear in later chapters in which two-person and multi-person games,
which invariably involve decisions under uncertainty, are introduced.

2.2 Certainty

Individual decisions under certainty are games of skill in which the solitary
player knows exactly what the outcome of any strategy choice will be. The
player has complete control of the outcomes, which are not affected by the
actions of any other interested parties or Nature.
It is worth pointing out that sports such as golf and archery are not games
of skill in this strict sense, because an element of chance limits the players'
control over the outcomes, which are never certain. To put it another way,
Nature influences the outcomes to a degree that depends to some extent on
the player's level of skill. Solving a crossword puzzle, on the other hand, is
a game of skill, and so is writing a computer program to perform certain
16 Background

logical operations most economically, and so too is shopping for various

foods in order to minimize the overall cost while ensuring a supply of
certain basic nutrients - this is a classic game of skill called the diet
problem.
Mathematically, games of skill are simply optimization problems, and
from a game theory point of view they are uninteresting, because the
solution in all cases is simply to choose the strategy that maximizes the
payoff, that is, the option that leads to the best outcome. But in spite of this,
a branch of mathematics called linear programming and large tracts of
operational research, management science, welfare economics, and behav
ioural science are devoted to these non-games. The reason is that optimal
strategies are often difficult and sometimes impossible to find. The decision
maker may wish to maximize the payoff but be unable to determine which
strategy to choose in order to do so.
The following infamous problem of the travelling salesman illustrates the
unexpected difficulties that can arise in decision making under certainty. A
travelling salesman needs to visit a number of specified cities exactly once,
using the shortest possible route, before returning to base. The options or
strategies are the various orders in which the cities may be visited during
the tour, and the distance between any pair of cities can be looked up on a
map. Optimizing the payoff amounts to minimizing the distance, and the
salesman's deceptively simple objective is to choose the shortest tour. In the
language of linear programming, the salesman needs to minimize the
objective function that defines the length of the tour.
The problem is that the number of options rises astronomically with the
number of cities. For 10 cities, there are 3628800 different tours from which
to choose. This is easy to prove. There is obviously a choice of ten cities for
the first visit, and for each of these there are nine possibilities left for the
second. There are therefore 10 X 9 = 90 different ways in which the salesman
can begin the tour by visiting two cities. For each of these 90 opening pairs,
there are eight possibilities left for the third visit, and so on. The total
number of options is therefore 1 0 x 9 x 8 x 7 x 6 x 5 x 4 x 3 x 2 x 1 =
3 628 800. For 20 cities, the number of options is 2 432 902 008 176 640 000, and
for 30 cities it exceeds 265 thousand billion billion billion. It would test the
patience of the most industrious travelling salesman to compare the length
of each tour with that of each of the others, and even the world's fastest
supercomputer would take many billions of years to solve the travelling
salesman problem for a tour of the 52 county towns of England and Wales
or the 50 state capitals of the United States. But no satisfactory alternative
method has been found, and this simple game of skill turns out to be an
intractable problem. Mathematicians have come very close to proving that
efficient solutions to problems of this type are impossible in principle (see
Papadimitriou and Steiglitz, 1982, for a full account of this intriguing branch
of mathematics).
 One-person games 17

Although games of skill, or decisions under certainty, constitute a vast

and thriving area of research, nothing more will be said about them here
because they lack the interdependence of choice that is characteristic of
game theory in the strict sense.

2.3 Risk

A risky decision is a game of chance in which the solitary decision maker is

pitted against the fictitious player, Nature. The decision maker does not
know with certainty what moves Nature will make but can nevertheless
assign meaningful probabilities to them and therefore to the various
possible outcomes. It was pointed out in chapter 1 that this is precisely the
problem facing a player of Russian roulette.
On the face of it, a decision involving only risk presents few problems. It
can apparently be solved by working out, according to the principles of
probability theory, the expected value of each available strategy choice and
then choosing the one that is best according to this criterion. The following
imaginary game of chance will illustrate the basic ideas. A competitor in a
television quiz game, after answering several trivia questions correctly, is
offered a choice between two alternatives:
(1) A coin will be tossed; if it lands heads, the competitor will receive a prize
of £1000, but if it lands tails, no prize will be given.
(2) The competitor must select one of three envelopes, which are known to
contain prizes of £900, £300, and £150 respectively, although there is no
way of knowing which prizes are in which envelopes.
What is the rational strategy in this game? Intuitively, and also according to
elementary principles of probability theory, the expected value of (1) is
1000/2 + 0 / 2 = £500, and the expected value of (2) is 900/3 + 300/3 4-1 5 0 / 3
= £450. To maximize the expected value, the competitor should therefore
choose (1): it pays better than (2) "on average", in the sense that if the game
were repeated many times, a player would tend to win more money by
choosing (1) than (2). It therefore seems obvious to many people that the
principle of maximizing expected value can be generalized to all purely
risky decisions.
There are two serious objections to this straightforward type of solution,
however. First, it is well known that rational human beings do not always
maximize expected value. Roulette (the French, not the Russian variety) is a
classic game of chance in which Nature's moves and their associated
probabilities are transparently obvious, yet millions of people play this
game in the full knowledge that the expected value is always negative,
otherwise the house would not make a profit. The same applies to pools and
lotteries, and to one-armed bandits, fruit machines, and other gambling
18 Background

devices found in amusement arcades. In all such games, the expected

monetary value of not playing at all, which is zero, is higher than that of
playing for any stakes. One might respond by classifying gambling as an
irrational form of behaviour in which neglect of the principle of expected
value maximization is merely evidence of mental pathology, but in that case
what is one to make of insurance policies, which sane and prudent citizens
buy regularly? The expected monetary value of insurance is negative for
essentially the same reason: insurance companies are out to make profits,
and they calculate their premiums according to probability estimates so that
their clients are bound to lose out "on average". To call the purchase of
insurance policies irrational would be to warp the notion of rationality
unreasonably.
A second objection arises from situations in which the principle of
expected value maximization appears to break down by generating
obviously absurd solutions. Perhaps the most famous example of this is the
St Petersburg paradox which the Swiss mathematician Daniel Bernoulli first
presented to the St Petersburg Academy in the early eighteenth century. The
rules of the St Petersburg game are as follows. A coin is tossed, and if it falls
heads, then the player is paid one rouble and the game ends. If it falls tails,
then it is tossed again, and this time if it falls heads the player is paid two
roubles and the game ends, otherwise it is tossed again. This process
continues for as long as necessary, with the payoff doubling each time, until
the player wins something, and then it stops. Assuming that the house has
unlimited funds, what is the expected value of playing this game? Or to put
the question another way, how much should a rational person be willing to
pay for the privilege of playing it?
The expected value can be calculated very simply. The player wins one
rouble with probability 1 / 2 , two roubles with probability 1 / 4 , four roubles
with probability 1 / 8 , and so on. The expected value of playing this game is
therefore
(1/2)(1) + (l/4)(2) + (l/8)(4) + ..
and it is easy to see that the sum of this endless series is infinite, because
each of its terms is equal to 1 /2. According to the principle of expected value
maximization, it would therefore be rational to offer one's entire fortune,
such as it is, to play the St Petersburg game once, because "on average" the
payoff is worth more than any fortune. But this conclusion is self-evidently
absurd, because there is a high probability of losing everything: to begin
with, there is a 50 per cent chance of reducing one's entire fortune to one
rouble on the very first toss. According to the expected value principle, this
is more than counterbalanced by the very small probability of winning a
vastly greater fortune if there is a long series of tails. But the St Petersburg
game is fatally damaging to the principle of expected value maximization,
as Bernoulli was the first to realize, because it would clearly be irrational to
 One-person games 19

stake one's fortune on it. The question that arises naturally from this is
whether a better principle can be found for solving games of chance
involving pure risk

2.4 Expected utility theory

The risky choices of rational decision makers are guided by their preferences
among the possible outcomes, but the examples of section 2.3 show clearly
that the superficially plausible principle of expected value maximization
cannot satisfactorily distinguish common sense from foolishness in risky
decision making.
Daniel Bernoulli's great contribution was the suggestion, which seems
obvious with hindsight, that the values of things to a person are not simply
equivalent to their monetary values. A small sum of money could be
precious to a pauper but almost worthless to a millionaire, even if the
pauper and the millionaire were one and the same person, for example the
day before and the day after winning a fortune in a lottery. People's
decisions are guided more by what the outcomes are worth to them than by
the cash value of the outcomes. Bernoulli (1738) called the subjective
desirability of an outcome its "moral worth", and this hypothetical quantity
later became known as utility.
Utilities must obviously be closely related to monetary values in cases in
which the outcomes of decisions are merely varying amounts of money or
can be straightforwardly translated into money. A large amount of money
cannot have less utility for a rational person than a small amount, because
it is easy to convert a large sum into a smaller sum by disposing of some of
it. Utility must therefore be a monotonically increasing function, but it may not
be a linear function of monetary value, which means that an increase in x
units of money may not increase its recipient's utility by the same amount
irrespective of the starting point. Bernoulli suggested a "law of diminishing
marginal utility", as it was later called, according to which the greater a
person's starting capital, the smaller the utility (or disutility) of any fixed
monetary gain (or loss). Bernoulli went further and claimed that the
functional relation between money and utility is logarithmic, which implies
that equal increases in utility correspond to equal proportional increases in
money, just as logarithms increase by equal steps as their corresponding
numbers increase by equal proportions. Historians of psychology (e.g.
Boring, 1950) seldom if ever point out that Bernoulli anticipated by more
than a century the famous psychophysical law of Gustav Theodor Fechner,
which propounds the same logarithmic relation between sensation and
stimulus intensity in general.
Bernoulli's theory of utility fell into disrepute for two main reasons. The
first was its arbitrariness: there seemed to be no logical or empirical reason
20 Background

why the relation should be logarithmic or why it should be the same for all
people in all circumstances. The second major limitation of Bernoulli's
theory was that it provided no method for assigning utilities to non
monetary outcomes. How, for example, can the logarithmic function be used
to assign utilities to the outcomes of a game of Russian roulette? The answer
is that it cannot. Similarly, recalling the examples presented in chapter 1,
Bernoulli's theory provides no way of assigning interval-scale utilities to
preserving one's chastity and saving one's brother's life, increasing a market
share, or even to avoiding a collision in a corridor. But in order to solve such
games we often need utilities that are measured quantitatively on an
interval scale, and we shall also be needing interval-scale utilities for solving
certain classes of two-person and multi-person games. Fortunately, a
brilliant solution is at hand.
John von Neumann and Oskar Morgenstern (1947) rehabilitated the
concept of utility by proposing a completely new theory that suffered from
neither of the defects of the old one. The essential ideas are simple, and the
theory is remarkably flexible (for a review of later theoretical developments,
see Fishbum, 1988). Von Neumann and Morgenstern's utility theory is based
on the assumption that a player can express a preference (or indifference),
not only between any pair of outcomes, but also between one outcome and
a lottery between a different pair of outcomes. To give a concrete
interpretation of these ideas, consider once again the imaginary television
quiz game outlined in section 2.3. For von Neumann and Morgenstern's
utility theory to work, a player must be able to express a preference (or
indifference) between any pair of outcomes - prizes of £1000, £900, £300,
£150, and nothing - and also between such pairs as the following:
(1) £300 for certain; or
(2) a 50:50 lottery between £1000 and nothing.
If the player turns out to prefer (2) to (1), then a new choice can be offered
between a fixed amount of money larger than £300 and (2). It is reasonable
to assume that there is some amount that the player will consider no less and
no more desirable than (2). It is then possible to convert the monetary
outcomes of the game into utilities, provided only that the player's
preferences are consistent. If the amounts involved are not too large, then it
is very likely that the player's utilities will be directly proportional to the
monetary values, because most people value £4 twice as much as £2, and £3
three times as much as £1, and this linear relation will be reflected in their
preferences between fixed amounts and lotteries. When large amounts are
involved, however, this linear relationship between utility and cash value is
likely to break down or at least bend, because £100 million does not seem
twice as desirable as £50 million to most people.
Von Neumann and Morgenstern succeeded in proving two things about
utilities that are determined in this way. First, it is always possible in
 One-person games 21

principle to convert a player's consistent preferences into utilities. Second, if

players apply the principle of maximization outlined in section 2.3 to
expected utilities rather than to expected monetary values, then they are in
fact choosing according to their preferences. This method of decision
making is called the principle of expected utility maximization.
Expected utilities, which are simply weighted average utilities (weighted
according to the corresponding probabilities), are purely psychological
quantities, and in the light of this principle it is not necessarily irrational to
gamble or to buy insurance policies. The expected utility of a lottery
involving the probable loss of a small stake and the improbable gain of a
huge prize may be positive even if the expected monetary value is negative;
and for most people the expected utility of a lottery involving the probable
loss of an insurance premium and the improbable repayment of the value of
one's property after a burglary, for example, may be considerably greater
than that of the premium, even if the expected monetary value of the lottery
is worth much less than the premium. It is equally clear that the paradoxical
quality of the St Petersburg game dissolves in this new light. There is no
reason why a person should prefer to stake everything on gambling with a
high probability of losing it all just because the expected monetary value of
the game is theoretically infinite.
Expected utility (EU) theory has another important consequence. It
enables numerical utilities, corresponding to degrees of preference, to be
assigned to games whose outcomes are inherently qualitative. The method
of quantifying preferences on an interval scale of utilities can be explained
with the help of the game of Angelo and Isabella, which was outlined in
chapter 1. We begin by assigning arbitrary utilities, say 1 and 10, to each
player's least and most preferred outcomes respectively. (In more familiar
interval scales such as the centigrade or Celsius scales of temperature, two
values corresponding, for example, to the freezing and boiling points of
water, are also assigned entirely arbitrarily.) It is then possible to determine
the utilities of the other outcomes by finding lotteries involving the two
extreme outcomes that the player considers equally preferable to each of the
other outcomes. Let us consider one possible outcome from Isabella's point
of view. It may turn out that she is indifferent between the outcome labelled
(1) below, and the lottery involving her least and most preferred outcomes
labelled (2):
(1) She preserves her chastity and her brother is executed humanely; or
(2) A die will be rolled. If it comes up six she will surrender her chastity to
Angelo and her brother will be tortured to death; if any other number
comes up she will preserve her chastity and her brother will be
reprieved.
Isabella is indifferent between one of the intermediate outcomes and a
1 / 6 : 5 / 6 lottery involving her least and most preferred outcomes. If her
22 Background

least preferred outcome is arbitrarily assigned a utility of 1 and her

favourite outcome a utility of 10, then the utility of the outcome under
consideration is
(1/6)(1) + (5/6)(10) = 8.5.
This utility is much closer to 10 than to 1, which reflects the fact that the
outcome is a relatively attractive one for Isabella - compared to the worst
and the best that can happen. Every possible outcome of the game can be
assigned a utility in this way, and the utilities represent Isabella's true
preferences. The same can be done from Angelo's point of view, and thus the
payoffs of this game can be expressed numerically. This method can be used
in any other game, provided that the players can express consistent
preferences between outcomes and hypothetical lotteries involving out
comes. Utilities derived in this way are measured on interval scales. This
means that the zero points and units of measurement are arbitrary - they are
fixed by the numbers chosen for the least and most preferred outcomes - so
that if they are all multiplied by a positive constant, or if a constant is added
to each of them, the information that they contain is unaffected. The
information concerns relative preferences. It is important to bear in mind that
a player's utilities, thus defined, may be influenced by any factors that have
a psychological bearing on the player's satisfaction or dissatisfaction with
the outcomes, including spiteful or altruistic feelings towards the other
players, religious beliefs, idiosyncratic tastes, phobias, masochistic tenden
cies, and so on. Expected utility (EU) theory is neutral with regard to the
roots of a player's preferences.
Von Neumann and Morgenstern's utility theory has not solved all
problems concerning the assignment of numerical payoffs to the outcomes
of games, and some of the residual problems will surface in later chapters.
The theory does, on the other hand, provide a convincing method for
solving games of chance involving pure risk. It follows almost tautologically
from the theory that in any game of this type a rational decision maker will
choose the strategy that maximizes expected utility. When the outcomes are
relatively small amounts of money, or can be readily interpreted as such, the
principle of maximizing expected monetary value outlined in section 2.3
will generally yield solutions that closely approximate those generated by
the principle of maximizing expected utility, because utility in such cases
will usually be a linear function of monetary value.
Utility functions can be constructed quite successfully for a wide range of
risky choices - for critical reviews of the relevant theory and empirical
research, see Bell, Raiffa, and Tversky (1988, chaps 4-9, 15-21), Dawes
(1988), and Rapoport (1989). When these functions are based on subjective
rather than objective probabilities, they are called subjective expected utility
(SEU) functions. Investigations of the behaviour of people in games of
chance involving pure risk have produced rather surprising results. In some
 One-person games 23

very simple games of this type, experimental subjects have tended to choose
strategies that are manifestly non-optimal. Humphreys's (1939) light
guessing experiment is a classic in the field, and many later experiments
were modelled on it (e.g. Goodnow, 1955; Ofshe and Ofshe, 1970; Siegel and
Goldstein, 1959; Siegel, Siegel, and Andrews, 1964). In these experiments,
the subjects were typically seated in front of two light bulbs and asked to try
to predict which of the two would light up on each trial. The bulbs were, in
fact, lit up in a random pattern according to probabilities fixed in advance,
for example, 80 per cent for the left-hand bulb and 20 per cent for the right
hand bulb. Subjects typically began by distributing their guesses roughly
equally between the two options. After a number of trials they usually began
to increase the frequency of choosing the bulb that was being lit up more
often. After very many trials, by which time the subjects had enough
information to judge the probabilities reasonably accurately and were
therefore playing a game of chance involving pure risk, most subjects settled
into a non-optimal matching strategy. In other words, they tended to choose
strategies with probabilities approximately equal to the probabilities with
which the bulbs lit up, for example 80 per cent left and 20 per cent right in
the case mention above. This is clearly non-optimal because the probability
of a correct guess is maximized by choosing the more frequently illuminated
bulb on every single trial. This optimal strategy obviously yields 80 per cent
correct guesses, whereas the matching strategy yields a probability of only
(.8)(.8) + (.2)(.2) of being correct, that is, 68 per cent correct guesses. These
and other similar experiments provide vivid illustrations of irrational
human behaviour in games of chance involving pure risk. A thought
provoking critical review of expected utility (EU) theory has been provided
by Frisch and Clemen (1994); see also the reviews of behavioural decision
research by Einhorn and Hogarth (1988), Kahneman and Tversky (1982),
Rapoport (1989), Slovic, Fischhoff, and Lichtenstein (1977) and Poulton
(1994).

2.5 Uncertainty

Games of chance involving uncertainty rather than risk crop up frequently

in everyday life. In these cases the player does not know for certain the
outcomes of the available strategy choices and cannot even assign
meaningful probabilities to them. Uncertainty and risk both imply a degree
of ignorance about the future, but uncertainty is a profounder type of
ignorance than risk.
The following example of a journalist's dilemma will illustrate the general
ideas. A freelance journalist wishes to send an article through the post to a
magazine in a foreign country. If the article is delivered safely, she will
receive a fee of £200. It is possible, however, that the article will be lost in the
24 Background

post. The cost of normal postage is negligible, but for an additional £5 she
can register the parcel; if it is then lost, the Post Office will reimburse the
£200 that she would have been paid for the article.
If the journalist chooses to register the parcel, then she loses £5 in postal
registration charges but receives £200 either from the Post Office (if the
parcel is lost) or from the magazine (if it is delivered safely). If she chooses
not to register it, she receives nothing if it is lost and £200 with no postal
registration charge to pay if it is delivered safely. The journalist's monetary
payoffs in each of the four possible outcomes, determined by her own and
Nature's strategy choices, are most neatly displayed in what is called a payoff
matrix (Matrix 2.1).
It is possible, of course, that the figures in the payoff matrix do not
accurately reflect the journalist's utilities, because there may be other

Matrix 2.1
Journalist's Dilemma
Nature
Lose Deliver

Register 195 195

Journalist
Don't register 0 200

psychologically significant aspects of the outcomes apart from their

monetary values that affect her preferences. Her reputation for reliability,
for example, or her satisfaction at seeing her article in print, may be worth
more to her than the figures in the payoff matrix suggest. But the
discussion in section 2.4 showed that considerations like these present no
problems in principle, and we may therefore assume that the payoff
matrix accurately reflects the journalist's utilities, or rather that they have
been adjusted to do so.
What principles of rational choice can be offered in a case like this? If the
journalist knows the probability of the parcel being lost, then the game is a
straightforward risky decision, and the principle of expected utility
maximization explained in section 2.4 is indicated. If, for example, she
knows that the chances are one in ten that it will be lost, then the expected
utility of registering it is
 (l/10)(195) + (9/10)(195) = 195,
and the expected utility of not registering it is
(l/10)(0) + (9/10)(200) = 180,
and if she were to choose rationally according to her preferences she would
accordingly register the parcel. But suppose that she has no knowledge of
Nature's probabilities; in other words, she has not the faintest idea of how
likely it is that the parcel will be lost. The country for which it is destined,
let us say, is going through a period of industrial unrest and no one knows
whether or not the postal services are affected, in fact no one is even willing
to hazard a guess. In that case, we have on our hands a decision involving
uncertainty rather than risk, and a number of principles of rational choice
have been suggested for solving them.

2.5.1 Insufficient reason

In practice, a decision involving uncertainty can always be transformed into

a risky one by assigning probabilities to Nature's strategies arbitrarily. The
principle of insufficient reason, advanced by the English clergyman and
mathematician Thomas Bayes in the middle of the eighteenth century, and
later championed by Laplace and many other influential figures, is based on
this idea. According to the principle of insufficient reason, one is supposed
to be justified in assigning equal probabilities to events in the absence of any
sufficient reason to believe that they are unequal.
Using this principle, the journalist in the example could cut the Gordian
knot by simply assuming that the probability of the parcel being lost is 1 / 2 ,
and having made that assumption she could apply the principle of expected
utility maximization straightforwardly. This method of solving - or, I should
prefer to say, side-stepping - the problem of uncertainty is, in my opinion,
illusory. If the probabilities are unknown, then they are simply unknown,
and there is no logical justification for calling them equal. "Subjective
probabilities", or degrees of belief, which are popular with many modem
Bayesian statisticians, have not provided a firm foundation for the principle
of insufficient reason. In spite of the useful distinction between risk and
uncertainty made long ago by John Maynard Keynes (1937), there are
nevertheless still those who believe that " a player always has some idea of
what the payoffs are, so we can always assign him a subjective probability
for each possible payoff" (Rasmusen, 1989, p. 66) and that "all decision
problems may be treated, subjectively at least, as risky choice" (Shepsle,
1972, p. 278, italics in original).
The principle of insufficient reason can be shown to lead to contradictions
as follows. Suppose the journalist in the example constructs a slightly
26 Background

different abstract model of her dilemma, by incorporating three possible

states of Nature, namely loss of the parcel, safe and timely delivery of the
parcel, and delayed delivery of the parcel. She would presumably be
entitled to assign equal probabilities of 1 / 3 to each of these events. But we
have already seen that she is entitled to assign probabilities of 1 / 2 to two of
them. By modelling the game in various other logically unassailable ways,
she could assign virtually whatever probabilities she chose to Nature's
strategies. The principle of expected utility maximization, if applied to these
fanciful probabilities, would yield different "solutions" depending on the
models that happen to have been chosen. Treating uncertainty as though it
were merely risk is an illusory solution to the problem, although some game
theorists disagree with this view.

2.5.2 Maximax

This principle of choice can hardly be considered rational, but it is worth

discussing, partly as a basis for discussions of more sophisticated principles
based on similar ideas, and partly because it is often used in practice by
naive decision makers such as children and by voters in elections (see
chapter 10). The maximax principle counsels the decision maker to choose
the strategy that yields the best of all possible payoffs. In the case of the
journalist's dilemma, the highest possible payoff if the parcel is registered is
195, and the highest possible payoff if it is not registered is 200. These two
figures are the two row maxima of the payoff matrix. The maximum of the
two maxima is 200, and it corresponds to the journalist's decision not to
register her parcel. If she follows the maximax principle, she will therefore
neglect to register it.
The maximax principle amounts to angling for the best possible outcome
of the game without any consideration of the less favourable outcomes that
are possible. It is an ultra-optimistic approach to decisions under uncer
tainty, because it implicitly assumes that the less favourable outcomes will
not arise. Although it possesses a certain innocent charm, and although it is
widely adopted in certain classes of situations (see section 10.9), it is
transparently silly and cannot be said to characterize the behaviour of
rational decision makers.

2.5.3 Maximin

Inspired by von Neumann and Morgenstern's (1944) classic text on game

theory, the statistician Abraham Wald (1945) suggested this principle. The
decision maker is advised to begin by identifying the lowest payoff in each
row of the payoff matrix, in other words to look for the worst possible payoff
 One-person games 27

that can result from each possible strategy choice. The final step involves
choosing the strategy associated with the best of these worst possible
outcomes, thereby maximizing the minimum payoff, hence the name. In the
journalist's dilemma, the maximin principle obviously leads her to choose the
strategy of registering the parcel, because the minimum payoff in that row
is 195, which is better than the minimum of 0 if she does not register it.
Whereas the maximax principle is ultra-optimistic, the maximin principle
could be said to be ultra-pessimistic, because it amounts to choosing the best
of the worst possible outcomes based on the assumption that whatever
strategy is chosen, the worst will happen. It is essentially a conservative
method of maximizing security rather than of trying to get the most out of
the game by taking chances. This can be seen most clearly in a payoff matrix
such as Matrix 2.2.

Matrix 2.2
Nature
X Y

0 1000

1 1

According to the maximin principle, the player should in this case choose
the strategy corresponding to the second row, labelled B, which guarantees
a minimum payoff of 1 unit, rather than A, which risks a zero payoff but
offers the possibility of a huge reward. In fact, no matter how large the top
right-hand matrix element might be, a player guided by the maximin
principle would never choose A,
The pessimistic policy of anticipating the worst may be quite sensible
when a great deal is at stake, or where the player's prime consideration is to
avoid the worst outcomes at any cost. It is the principle governing statistical
decision-making, where the primary consideration is to avoid a Type I error
(concluding that an effect is real when it may be due to chance) rather than
a Type II error (concluding that it is due to chance when it is in fact real). But
it does not seem intuitively reasonable to adopt this policy for all decisions
under uncertainty, as we shall now see.
Consider the maxim "Nothing ventured, nothing gained". An inter
pretation of this maxim in game theory terms is provided by a consideration
of what would happen if it were applied in industry to decide when to
devote funds to research and product development (see Moore, 1980, pp.
28 Background

284-286). The possible strategies are to support a research and development

(R&D) project or to neglect it. Nature's moves determine whether or not the
R&D will succeed it is supported. If the cost of the R&D project is
represented by c and the potential return if it is successful by r, then the
decision is represented by Matrix 2.3.

Matrix 2.3 Research and

Development
Nature
R&D Succeeds R&D Fails

Support R&D r- c -c
Player
Neglect R&D 0 0

The minimum payoff if the research and development is supported (row

1) is clearly -c, and the minimum if it is neglected (row 2) is zero. The
maximum of these two minima is zero no matter what the values of r and
c might be. Thus, by adopting the ultra-pessimistic maximin principle,
industry would never support research and development that might fail, no
matter how small the cost and how profitable the potential rewards. This
would be an absurdly unenterprising approach to business management,
and it highlights the limitations of the maximin principle. From a logical
point of view, the principle is unassailable in so far as it attains the clearly
specified goal of ensuring the best of the worst possible outcomes in any
contingency that might arise. But, whereas this is a reasonable goal in some
circumstances, it does not correspond to intuitive notions of rationality in
others.

2.5.4 Minimax regret

This principle was first advanced by Savage (1951). Many decision theorists
consider it to be the soundest method of solving problems involving
decisions under uncertainty, because it is neither too optimistic nor too
pessimistic, and it combines the best aspects of other approaches, but it
contains a subtle though fatal flaw.
Savage's ingenious idea was to transform the original payoff matrix into
a regret matrix (sometimes called a loss matrix). Each element of the regret
 One-person games 29

matrix represents the positive difference between the corresponding payoff

in the payoff matrix and the highest payoff that could occur under the same
state of Nature. These new numbers reflect the decision maker's "opportu
nity loss" or degree of regret, if Nature's strategy were to be revealed, for not
having chosen differently in those circumstances. Each element of the payoff
matrix is converted into an element of the regret matrix by asking: Could the
player have obtained a higher payoff by choosing differently if Nature's
strategy had been known in advance? If not, then the corresponding regret
is zero. If the player could have done better, then the regret is equal to the
amount by which the payoff could have been improved by choosing the
most profitable strategy given advance knowledge of Nature's intentions.
To illustrate these ideas, consider once again the journalist's dilemma
introduced earlier, and suppose that she chooses to register the parcel. If it
is delivered safely, her regret for having wasted money registering it
amounts to £5, because in those circumstances she could have achieved a
payoff £5 higher by not registering it. If the parcel is lost, on the other hand,
then she has no cause for regret because, given that state of Nature, she
could not have done better by not registering it. Arithmetically, the regret
matrix is calculated by subtracting each element of the payoff matrix from
the largest element in its column (Matrices 2.4).

Matrices 2.4 Journalist's Dilemma

Nature Nature
Lose Deliver Lose Deliver

Register 195 195 0 5

Journalist
Unregistered 0 200 195 0

Payoff matrix Regret matrix

Notice that every element of the regret matrix is either positive or zero.
This is the case even if some or all of the original payoffs are negative - there
is no such thing as negative regret. The minimax regret principle for making
decisions under uncertainty is as follows. Choose the strategy that
minimizes the maximum possible regret. If the journalist registers the parcel,
then the maximum regret (the largest element in the top row of the regret
matrix) is £5, and if she does not register it, then the maximum regret (in the
bottom row) is £195. According to the minimax regret principle, she should
30 Background

choose the strategy corresponding to the minimum of these two maxima, in

other words she should register the parcel.
It is worth examining the minimax regret solution to the game of research
and development mentioned earlier. The payoff matrix is reproduced in
Matrices 2.5, together with its regret matrix, which has been calculated in the
usual way by subtracting each matrix element from the largest in its
column.

Matrices 2.5 Research and Development

Nature Nature
R&D R&D R&D R&D
succeeds fails succeeds fails

Support R&D r-c -c 0 c

Player
Neglect R&D 0 0 'r-c 0

Payoff matrix Regret matrix

According to the minimax regret principle, the player should examine the
maximum in each row of the regret matrix and select the row with the least
of these maxima. The maximum in the top row of the regret matrix is c,
because the cost must be greater than zero, and the maximum in the bottom
row is r - c on the reasonable assumption that the cost of R&D (c) is less
than the return if it succeeds (r) so that r - c is greater than zero. The player
should therefore support R&D if c is less than r-c, which is equivalent to c
being less than r/2, and neglect R&D otherwise. In other words, industry
should support research and development if and only if the cost of it is less
than half the potential return. This rule of thumb is, in fact, quite commonly
applied in research management (Moore, 1980, p. 286), although it is
usually justified by experience and intuition rather than game theory
analysis.
Unfortunately, the minimax regret principle suffers from a fatal flaw,
which was exposed by Chemoff (1954). In some cases it violates a condition
of rationality called the independence of irrelevant alternatives. (This condition
reappears in a slightly modified guise in section 6.10 in connection with
bargaining solutions to two-person cooperative games and in section 10.5 in
connection with social choice theory.) The following example is based on
Luce and Raiffa (1957, p. 288). Suppose a customer in a restaurant consults
 One-person games 31

the menu and finally decides to order poached salmon. The waiter then
points out helpfully that, although it is not on the menu, roast chicken is also
available. "In that case", says the customer, "Til have the lamb chops". This
behaviour seems irrational, because the addition to the list of the option of
roast chicken seems irrelevant to the customer's preference for poached
salmon over lamb chops. But this strange phenomenon can arise from the
application of the minimax regret principle, as we can see by returning for
the last time to the journalist's dilemma.
The minimax regret principle suggests that the journalist will register the
parcel if she is rational. But suppose that she contemplates a third strategy,
namely to register the parcel and simultaneously to bet a colleague £200 at
even odds that it will be delivered safely. If the parcel is lost, she will then
lose the £5 postal registration charge and will also have to pay her colleague
£200, but the Post Office will reimburse her £200, so she will lose £5 overall.
If the parcel is delivered safely, she will lose the £5 registration charge but
will receive £200 from the magazine and will win another £200 from her
colleague, so that her net gain will be £395. The addition of this new strategy
enlarges the game without affecting the payoffs associated with the original
strategies; the new payoff matrix is the same as the old one with an
additional row joined on to it. The new regret matrix can be calculated in the
usual way, but all the regret values turn out to be quite different. The new
payoff matrix and its regret matrix are shown in Matrices 2.6.
The maximum in the first row of the regret matrix is 200, the maximum in
the second row is 195, and the maximum in the third row is 200. The
minimum of these maxima is 195, and it corresponds to the strategy "Don't
register"! Something very curious indeed has happened, because in the
original formulation of the problem, the journalist's preferred strategy
according to the minimax regret principle was to register the parcel. Once an

Matrices 2.6 Journalist's Dilemma

Nature Nature
Lose Deliver Lose Deliver

Register 195 195 0 200

Journalist Don't register 0 200 195 195

Register & bet -5 395 200 0

Payoff matrix Regret matrix

32 Background

additional strategy was added to the list, the strategy that was rejected in the
original analysis emerged as the rational choice.
At an abstract level, this is similar to what happened to the customer in
the restaurant who ordered poached salmon rather than lamb chops but
changed the order to lamb chops on being told that roast chicken was also
available. If this reversal by the restaurant customer seems irrational on the
grounds that the availability of the third alternative is irrelevant to the
customer's relative preferences between the other two, then exactly the
same can be said of the journalist. The minimax regret principle has fallen
foul of the independence of irrelevant alternatives, and it is therefore
considered to be suspect by many decision theorists.
Games of chance involving uncertainty are evidently more intractable
than those involving mere risk. When probabilities cannot be assigned to the
outcomes, the principle of expected utility maximization, which provides
compelling solutions to risky decisions, cannot be applied meaningfully.
Four decision rules designed to generate rational decisions under uncer
tainty have been examined, and they have all been found wanting. Others
have been advanced from time to time, but none is entirely watertight. In
fact it was proved many years ago (Milnor, 1954) that no decision rule can
satisfy all of the criteria, of which the independence of irrelevant alternatives
is one, that one would require of a completely convincing solution to the
problem.

2.6 Summary
Section 2.1 drew attention to the existence of the class of one-person games
in which a solitary decision maker is pitted against Nature, and it was
pointed out that analyses of these games are based on ideas that crop up in
the solution of more complicated games involving two or more decision
makers. Section 2.2 discussed decisions under certainty, which are called
games of skill, and section 2.3 decisions under risk. The superficially
plausible principle of expected value maximization was shown to lead to
difficulties, and in section 2.4 it was shown how the worst of these
difficulties can be overcome by the application of von Neumann and
Morgenstern's expected utility (EU) theory. A method was outlined for
assigning numerical utilities on an interval scale of measurement to games
whose outcomes are essentially qualitative in character. Section 2.5 was
devoted to one-person decisions involving uncertainty rather than risk.
Various decision rules - the principle of insufficient reason, maximax,
maximin, and minimax regret - were examined critically, and all were found
to violate intuitive notions of common sense, although the last two were
found to have some merits. It was pointed out that no entirely satisfactory
decision rule is possible.
 3
Coordination games and the minimal
social situation

3.1 Strategic collaboration

This chapter is devoted to two classes of games involving players whose
interests are best served through strategic collaboration. The first is the class
of coordination games, which are characterized by coincidence of the
players' interests, and the second is the minimal social situation, which has
intrigued game theorists and experimental investigators for some time. The
minimal social situation is a class of games involving incomplete informa
tion; in its strict form, the players are ignorant not only of the nature but
even of the fact of their strategic interdependence.
In a coordination game, it is in every player's interest to try to anticipate
the others' choices in order to obtain a mutually beneficial outcome, and
they all know that the other players are similarly motivated. If the social
situation makes communication between the players impossible, then subtle
and interesting psychological problems arise, and the outcome often
depends on the level of strategic intuition of the players. In the minimal
social situation it is once again in the players' mutual interest to collaborate.
But the problems confronting the players are quite different in this case,
because they arise from a lack of information about the rules of the game.
The central question in the minimal social situation is whether the players
can learn, through playing the game a number of times, to cooperate
without conscious awareness of the need for cooperation and without even
being aware of each other's existence.
In the following section the general strategic properties of coordination
games will be examined. Theoretical and empirical work on the minimal
social situation will be discussed in section 3.3, a multi-person general
ization of the minimal social situation will be outlined in section 3.4, and a
brief summary of the chapter will be given in section 3.5.

3.2 Coordination games

The defining property of a coordination game is agreement among the players
as to their preferences among the possible outcomes; in particular, an
34 Background

outcome that is considered best by one player is considered best by the

others. A pure coordination game is one in which all of the players' preferences
are identical. In a coordination game there is no conflict of interest between
the players: their sole objective is to coordinate their strategies in such a way
as to obtain an outcome that they all favour. The game Head On, introduced
in section 1.1.1, is an example of a coordination game, because the outcome
is either a collision or the avoidance of a collision and both players prefer the
avoidance of a collision. In Head On, two people are walking briskly
towards each other along a narrow corridor. Each player chooses one of
three strategies - swerve left, swerve right, or continue straight ahead - in
an effort to avoid colliding with the other player, and an outcome that is
good (or bad) for one is similarly good (or bad) for the other.
Coordination games are clearly games of strategy according to the
definition given in chapter 1. In treating them as such, I am following
Schelling's (1960) "reorientation of game theory", although many author
ities do not regard them as genuinely strategic games. In their classic
textbook, Luce and Raiffa (1957), for example, asserted that any group of
decision makers "which can be thought of as having a unitary interest
motivating its decisions can be treated as an individual in the theory" (p. 13),
and furthermore that if the players have the same preference pattern over
the outcomes "then everything is trivial" (p. 59); "certainly in the extreme
case where there is perfect agreement the analysis is trivial" (p. 88). There
are, none the less, at least three good reasons for taking coordination games
seriously. First, although they may appear trivial from the point of view of
formal game theory, informal analysis and empirical evidence discussed in
this chapter shows that the problems confronting decision makers in these
games are far from inconsequential (see also Bacharach, 1994; Crawford and
Haller, 1990; Kandori, Mailath, and Rob, 1993; Young, 1993). Second, there is
a growing body of experimental evidence that coordination failures often
occur in these games, even in some cases in which communication is
allowed (e.g., Cooper, De Jong, Forsythe, and Ross, 1992a, 1992b; Van
Huyck, Battalio, and Beil, 1990). Third, it is aesthetically appealing to treat
pure coordination games as one extreme of a classification in which strictly
competitive structures (two-person zero-sum games) are at the opposite
extreme and mixed-motive games occupy the middle ground.
The theoretical symmetry between zero-sum and coordination games can
be illustrated by a pair of simple examples that share certain superficial
features in common. Two-person hide-and-seek is a classic example of a
zero-sum game: one player chooses a hiding place from among a number of
locations, and the other chooses a location to search. Each outcome -
discovery or non-discovery - represents a victory for one player and a defeat
for the other, so the game is strictly competitive. By way of contrast, if two
people are accidentally separated in a crowded shopping centre, they
usually end up playing a game that I shall call Rendezvous. Once again they
Another random document with
no related content on Scribd:
teitä, milloin vain tarve vaatii.

Sen sanottuaan herra Wolodyjowski alkoi nyhtää pystyjä,

vahattuja viiksiään, sillä hän oli tyytyväinen itseensä, kun oli heti
mennyt in medias res [itse asiaan. Suom. huom.] ja paljastanut
tunteensa neidille. Aleksandra-neiti istui hämillään ja vaiti, ollen
kaunis kuin kevätpäivä. Kalvas puna nousi hänen poskilleen, ja
silmät lymysivät pitkien ripsien alle, jotka loivat varjoa kukkeille
poskipäille.

— Tuo hämi on hyvän merkki, — ajatteli herra Wolodyjowski.

Hän yskähti ja jatkoi:

— Te tiedätte, armollinen neiti, että minä olen ottanut isoisänne

jälkeen laudalaisten johdon käsiini?

— Tiedän kyllä, — vastasi Oleńka. — Ukkovainaja ei voinut itse

ottaa osaa viimeiseen sotaretkeen, mutta hän oli sangen iloinen
saadessaan kuulla kenelle Vilnon vojevoda oli luovuttanut
laudalaisten sotalipun. Hän sanoi kuulleensa kehuttavan teitä
erinomaiseksi soturiksi.

— Sanoiko hän todellakin niin?

— Kuulin itse, kuinka hän ylisti teitä taivaaseen saakka, ja taistelun

jälkeen tekivät laudalaiset samoin.

— Minä olen tavallinen sotilas enkä ansaitse sellaista ylistystä,

yhtä vähän kuin muutkaan. Iloitsen kuitenkin siitä, etten ole neidille
tuiki tuntematon, sillä nyt ette voi luulla, että joku tuntematon ja
epäluotettava vieras on pudonnut eteenne kuin pilvistä. On hyvä
aina tietää kenen kanssa on tekemisissä, sillä on paljon sellaisia
ihmisiä, jotka kuljeskelevat ympäri sanoen olevansa familiantteja
[tarkoittaa ylhäissukuista puolalaista aatelista. Suom. huom.], mutta
jotka, Jumala paratkoon, usein eivät ole edes tavallistakaan
aatelissukua.

Herra Wolodyjowski oli tahallaan johtanut keskustelun niin, että

sai tilaisuuden kertoa, mitä miehiä hän oli. Oleńka myönsi heti:

— Teitä, herra eversti, kukaan ei epäile, sillä täällä Liettuassa asuu

samanniminen aatelissuku.

— Nämä ovat Ossorya-sukua, mutta minä olen Korczakin

Wolodyjowskeja, jotka ovat kotoisin Unkarista ja erään Attilan ritarin
jälkeläisiä. Tämä ritari oli vihollisen ahdistaessa tehnyt Pyhälle
Neitsyelle lupauksen, että kääntyy katoliseen uskoon, jos jää eloon.
Hän piti sanansa, kun oli pelastunut pääsemällä kolmen virran yli, ja
nuo kolme virtaa ovat juuri kuvattuina meidän vaakunassamme.

— Te ette olekaan siis kotoisin näiltä seuduilta?

— En ole, armollinen neiti. Ukraina on minun syntymäseutuni, ja

siellä omistan vielä tänä päivänä kylän, joka äskettäin on joutunut
vihollisen käsiin. Olen pienestä pitäen palvellut armeijassa ja
harrastanut enemmän yleistä hyvää kuin omia yksityisiä asioitani.
Ensin olin vähävenäläisen vojevodan, ikuisesti kaivatun ruhtinas
Jeremin palveluksessa, ottaen osaa kaikkiin hänen sotaretkiinsä.
Jumala voi todistaa, armollinen neiti, etten ole tullut tänne
ylistämään itseäni. Tahtoisin vain saattaa tietoonne, etten ole mikään
tyhjänkerskaaja, joka toitottaa suuria sanoja, mutta pelkää
henkeään. Minä olen uskollisesti palvellut isänmaatani halki koko
elämäni ja niittänyt vähäisen kunniaakin tahraamatta omaatuntoani.
Niin on asian laita, kautta Jumalani, ja sen voivat luotettavat miehet
todistaa.

— Jospa kaikki olisivatkin teidän kaltaisianne, herra eversti! —

huoahti Oleńka.

— Armollinen neiti ajattelee varmaankin sitä väkivallantekijää, joka

rohkeni nostaa kätensä teitä vastaan?

Aleksandra-neiti loi katseensa maahan eikä vastannut sanaakaan.

— Hän on saanut minkä ansaitsi, — jatkoi herra Wolodyjowski. —

Sanotaan hänen toipuvan, mutta tuomiotaan hän ei kuitenkaan voi
välttää. Kaikki kunnon ihmisetkin tuomitsevat häntä sanoen hänen
liittyneen vihollisiimme ja saaneen heiltä apua, mutta se ei ole totta.
Ne kasakat, joiden kanssa hän hyökkäsi kimppuumme, eivät olleet
viholliselta, vaan maantieltä koottuja.

— Mistä te sen tiedätte? — kysyi Oleńka vilkkaasti kohdistaen

sinisten silmäinsä katseen herra Wolodyjowskiin.

— Kasakoilta itseltään. Kyllä se Kmicic on kummallinen mies! Kun

minä ennen kaksintaistelua syytin häntä petturiksi, niin hän ei
väittänyt vastaan, vaikka olin väärässä. Se mies on ylpeä kuin itse
paholainen.

— Oletteko kertonut tästä kaikille?

— En vielä, kun en ole sitä tiennyt, mutta tästä puolin tulen

kertomaan. Pahinta vihollistakaan ei pidä syyttää moisesta
tapahtumattomasta rikoksesta.
 Aleksandra-neidin katse kohdistui taas pienikasvuiseen ritariin
ilmaisten myötätuntoa ja kiitollisuutta.

— Te olette yhtä kunnioitettava kuin harvinainen ritari!…

Herra Wolodyjowski kiersi viiksiään tyytyväisesti.

— Asiaan, Michal! — ajatteli hän. Sitten hän lausui:

— Sanonpa armolliselle neidille vielä jotakin!… Minä moitin herra

Kmicicin menettelyä, mutta en ihmettele, että hän sillä tavalla koetti
saada haltuunsa teidät, jolle itse Venus sopisi kamarineidiksi. Se oli
hänen epätoivonsa, joka johti hänet moiseen kehnoon tekoon ja
varmasti vielä johtaa, jos vain tilaisuus tarjoutuu. Mitenkä te, joka
olette niin tavattoman kaunis, voisitte olla yksin ja ilman suojelijaa?
Useita on maailmassa sellaisia Kmicicejä, useita sydämiä te vielä
sytytätte, yhä useammat vaarat ovat edessänne. Jumala soi minulle
armon pelastaa teidät tällä kertaa, mutta Marsin torvet soivat jo…
Kuka on teitä vastedes suojeleva? Armollinen neiti, sotureita
syytetään huikentelevaisuudesta, mutta väärin. Minunkaan sydämeni
ei ole kivestä eikä ole voinut kestää teidän verratonta sulouttanne…

Herra Wolodyjowski lankesi polvilleen Oleńkan eteen.

— Armollinen neitiseni, — jatkoi hän, — olen perinyt sotalipun

teidän isoisältänne, sallikaa minun periä hänen poikansatytärkin!
Jättäkää minun huostaani teidän suojelemisenne, niin voitte olla
rauhassa ja huoletta minun sodassa ollessanikin, sillä pelkkä nimeni
riittäisi teitä suojelemaan.

Oleńka oli hypähtänyt pystyyn tuolilta ja kuunteli hämmästyneenä

herra
Wolodyjowskin puhetta. Tämä jatkoi:

— Olen varaton sotilas, mutta olen aatelismies ja rehellinen mies

ja vannon, ettei aateliskilvessäni enemmän kuin
omallatunnollanikaan ole pienintäkään tahraa. Minä teen ehkä väärin
pitäessäni tällaista kiirettä, mutta se teidän pitää ymmärtää, sillä
isänmaa kutsuu minua, enkä minä voi sitä jättää edes teidän
tähtenne… Ettekö anna minulle toiveita? Ettekö sano hyvää sanaa?

— Te pyydätte minulta mahdottomia, hyvä mies… Kautta Jumalan,

minä en voi! — vastasi Oleńka pelokkaasti.

— Se riippuu kokonaan teistä…

— Juuri siksi vastaan teille suoraan: ei! Oleńka rypisti

kulmakarvojaan ja jatkoi:

— Armollinen herra eversti! Olen teille suuressa

kiitollisuudenvelassa, en kiellä sitä. Pyytäkää minulta mitä hyvänsä,
olen valmis antamaan kaikki paitsi kättäni.

Herra Wolodyjowski nousi.

— Te ette siis välitä minusta? Niinkö?

— En voi!

— Onko tuo viimeinen sananne?

— Viimeinen ja peruuttamaton. — Ehkä tämä kiireellisyys tuntuu

teistä vastenmieliseltä? Antakaa minulle toiveita!

— En voi, en voi…
 — Täällä minulla siis ei ole onnea enemmän kuin sitä on ollut
missään muuallakaan! Armollinen neiti ei siis palkinnut minun
palveluksiani, vaikk'en tullut palkintoa hakemaan, kun tulin
pyytämään kättänne, sillä olen tehnyt sen vapaasta tahdosta. Jos te
olisitte vastannut minulle, että annatte sen velvollisuudesta, olisin
minä siitä kieltäytynyt. Missä ei ole hyvää tahtoa, siinä ei ole
onneakaan! Te olette ylenkatsonut minut… mutta Jumala suokoon,
ettette saisi vielä huonompaa kuin minä! Lähden talosta niinkuin
tulinkin enkä ikänä enää palaja. Täällä minua ei pidetä minään.
Sanottu ja tehty! Eläkää onnellisena vaikkapa juuri tuon Kmicicin
kanssa. Ehkä olette minulle vihoissanne juuri siksi, että erotin teidät
miekallani. Koska pidätte häntä minua parempana, niin ette sovi
minulle.

Oleńka tarttui kaksin käsin päähänsä ja toisti muutaman kerran:

— Hyvä Jumala! Hyvä Jumala!…

Mutta Oleńkan epätoivo ei vaikuttanut enää herra Wolodyjowskiin,

joka kumarrettuaan poistui vihaisena. Ritari nousi heti hevosensa
selkään ja ajoi pois.

— Ikänä en jalallani sinne astu! — sanoi hän. Sotilaspalvelija

Syruc, joka ajoi jäljessä, joudutti hevosensa everstin viereen ja
kysyi:

— Mitä herra eversti sanoo?

— Hölmö! — vastasi herra Wolodyjowski.

— Teidän armonne sanoi sen jo äsken tänne ajaessamme.

Seurasi hiljaisuus. Sitten herra Michal alkoi taas jupista:

 — Kiittämättömyyttä minulle siellä syötettiin… Ylenkatseella
palkittiin rakkauteni… Tulen sotaväessä naimattomana palvelemaan
kuolemaani saakka. Niin näkyy olevan tähtiin kirjoitettu… Hitto
vieköön! Moinen kohtalo!… Missä yritätkin, aina rukkaset… Tässä
maailmassa ei ole oikeudenmukaisuutta!… Mitä lie hänellä sitten ollut
minua vastaan?

Herra Wolodyjowski rypisti kulmakarvojaan ja pinnisti ajatuksiaan.

Yht'äkkiä hän löi kämmenellään reiteensä.

— Nyt tiedän! — huudahti hän. — Hän rakastaa vielä sitä toista…

Toisin ei voi olla!

Mutta tämä havainto ei karkoittanut pilviä hänen kasvoiltaan.

— Sitä pahempi minulle, — ajatteli hän hetken kuluttua, — sillä jos

neiti Billewicz vielä kaiken tämän jälkeen rakastaa häntä, niin hän ei
enää lakkaa rakastamasta häntä… Kmicic on tehnyt niin paljon
pahaa kuin on voinut. Nyt hän lähtee sotaan, niittää kunniaa ja
pelastaa maineensa… Eikä häntä voi tästä estää, vaan pikemminkin
auttaa, koska isänmaan etu on kysymyksessä… Ja hän on oiva
sotilas… Mutta millä ihmeen tavalla hän on voinut voittaa tytön
rakkauden?… Niin, kukapa sen arvaa?… Toisilla on sellainen onni,
että sen kuin katsovat vain tyttöön, niin tämä on valmis vaikka
tuleen… Jospa saisi tietää miten he menettelevät tai jos oppisi jonkin
taikatempun, niin ehkä toiste paremmin vetelisi!… Ansiot eivät
merkitse mitään, kun on kysymys naisista. Herra Zagloba oli
oikeassa sanoessaan, että nainen ja repolainen ovat kaksi kavalinta
otusta maan päällä… Surkeata, että pitikin käydä näin huonosti! Niin
hiton sievä tyttö ja niin kunnollinen, sanotaan. Mutta ylpeä kuin itse
paholainen!… Ties tokkohan tuo Kmiciciä kuitenkaan huolinee,
vaikka rakastaakin, sillä onhan Kmicic julmasti loukannut ja häväissyt
häntä… Olisihan hän voinut rauhallisesti tulla tyttöä noutamaan,
mutta rupesi mieluummin ryöstäjäksi… Tyttö lienee valmis
luopumaan naimisiinmenon haaveista… Minun on vaikea ollakseni,
mutta tyttö-paran kai vieläkin vaikeampi…

Ja herra Wolodyjowski tuli liikutetuksi ajatellessaan Oleńkan

kohtaloa.
Hän alkoi huojuttaa päätään, huoahdella, kunnes vihdoin virkkoi:

— Jumala häntä auttakoon! En ole hänelle vihoissani. Eivät ne

olleet ensimmäiset rukkaseni, mutta tytölle se oli ensimmäinen suru.
Tyttöparka on ihan suunniltaan epätoivosta, kun olen sitä Kmiciciä
niin pahanpäiväisesti kolhinut ja päälle päätteeksi vielä ärsyttänyt…
Sitä ei olisi pitänyt tehdä… Mutta täytyy hyvittää. Ansaitsisinpa
selkääni moisesta menettelystä. Kirjoitan tytölle
anteeksipyyntökirjeen ja lupaan auttaa häntä minkä voin.

Syruc, joka taas ajoi everstin viereen, keskeytti hänen

päivittelynsä sanoen:

— Anteeksi, herra eversti, mutta tuolla mäellä ajaa herra Charlamp

muutamien muitten kanssa.

— Missä?

— Tuolla noin!

— Siellä näkyy olevan kaksi ratsumiestä… mutta herra Charlampin

pitäisi olla vojevodan luona Vilnossa. Kuinka sinä tuntisit hänet näin
kaukaa?

— Keltaisesta hevosesta tunnen. Senhän tuntee joka mies koko

armeijassa.
 — Niin todellakin… keltainen näkyy olevan. Mutta se voi olla joku
toinen.

— Ei ole, sillä minä tunnen sen myöskin käynnistä… Se on herra

Charlamp aivan varmaan.

Molemmat kannustivat hevosiaan, vastaantulevat tekivät samoin,

ja pian herra Wolodyjowski tunsi, että se todellakin oli herra
Charlamp.

Tämä oli Wolodyjowskin vanhoja tuttuja, kokenut, oivallinen

sotilas ja luutnantti Liettuan armeijan ratsuosastossa. Ennen hän ja
herra Wolodyjowski olivat tuon tuostakin otelleet keskenään, mutta
sitten heistä oli tullut hyvät ystävät yhteisillä sotaretkillä saman lipun
alla. Herra Wolodyjowski karahdutti hänen luokseen ja huusi
levittäen käsivartensa:

— Hei! Terve, sinä komeanenäinen mies! Mitä kuuluu? Mistä tulet?

Luutnantti, jolla totisesti oli jättimäinen nenä, heittäytyi everstin

käsivarsille, ja he syleilivät toisiaan iloisesti. Lauhduttuaan herra
Charlamp sanoi:

— Sinun luoksesi tulen, tuoden käskykirjeen ja rahaa.

— Käskykirjeen ja rahaa?… Keneltä?

— Vilnon vojevodalta, meidän hetmaniltamme. Hän lähettää

sinulle kirjelmän pyytäen, että heti ryhtyisit pestaamaan väkeä, ja
toisen herra Kmicicille, jonka myös pitäisi olla näillä seuduilla.

— Herra Kmicicille?… Mitenkä kaksi voi pestata samalla

paikkakunnalla?
 — Hänen pitäisi lähteä Trokyyn ja sinun jäädä tänne.

— Mistä tiesit tulla tänne minua etsimään?

— Hetmani itse on kovasti sinua tiedustellut, ja vihdoin hän saikin

muutamilta laudalaisilta tietää olinpaikastasi, ja minä ajoin heti
tänne… Sinä olet siellä yhä suuressa arvossa!… Kuulin ruhtinaan,
meidän hetmanimme, itse sanovan, ettei hän ollut ikänä odottanut
saavansa periä mitään vähävenäläisen vojevodan jälkeen, mutta
saikin periä hänen kaikkein suurimman soturinsa.

— Suokoon Jumala hänen saavan periä myöskin sotaonnea…

Minulle on suuri kunnia saada pestata sotaväkeä, ja minä ryhdyn
toimeen heti… Sotakelpoista väkeä täällä on riittävästi, kun vain on
millä heidät varustaa. Onko sinulla paljonkin rahaa mukanasi?

— Kun tulemme Pacuneleen, saat laskea.

— Sinä olet jo ollut Pacunelessa? Pidä varasi, sillä siellä on paljon

kauniita tyttäriä!

— Taidatpa samasta syystä viihtyä siellä niin hyvin… Mutta

odotapas, minulla on myöskin yksityinen kirje sinulle hetmanilta.

— Anna tänne!

Herra Charlamp otti esille kirjeen, jossa oli Radziwillien sinetti.

Herra Wolodyjowski avasi sen ja alkoi lukea:

»Kunnioitettu herra eversti Wolodyjowski!

Tietäen teidän hartaan pyrkimyksenne palvella isänmaata lähetän

teille samalla määräyksen ryhtyä pestaamaan sotaväkeä, mutta ei
tavalliseen tapaan, vaan suurella innolla, sillä periculum in mora
[viivyttelemisessä on vaara. Suom. huom.]. Jos tahdotte ilahduttaa
meitä, niin laittakaa lippukuntanne kuntoon heinäkuun loppuun
mennessä tai viimeistään syyskuun puoliväliin. Meitä huolestuttaa
se, mistä saisitte hyviä hevosia, etenkin kun voimme lähettää
niukasti rahoja, sillä emme ole meille vanhastaan nurjamieliseltä
valtiovarainhoitajalta saaneet; enempää. Antakaa puolet näistä
rahoista herra Kmicicille, jolle herra Charlamp myöskin tuo
määräyskirjelmän. Toivomme hänen ahkerasti toimivan
hyväksemme. Kuitenkin, kun olemme kuulleet huhuja mainitun
Kmicicin väkivallasta Upitassa, niin lienee paras, että otatte
haltuunne hänelle osoitetun kirjelmän ja ratkaisette, voiko sitä
antaa hänelle vai eikö. Jos hän mielestänne on syypää liian moniin
rikoksiin, niin älkää sitä hänelle antako. Me pelkäämme nimittäin,
että vihamiehemme, kuten valtiovarainhoitaja ja Vitebskin
vojevoda, alkavat huutaa, että me uskomme tärkeitä tehtäviä
kelvottomille henkilöille. Jos hän kuitenkaan ei ole syypää
suurempiin rikoksiin, niin antakaa kirjelmä, ja koettakoon herra
Kmicic palveluksessamme ahkeruudella hyvittää hairahduksensa,
älköönkä mikään tuomioistuin siinä tapauksessa vaatiko häntä
vastaamaan teoistaan, koska hän silloin on meidän oikeutemme
alainen. Osoittakoon tämä tehtävä samalla, että luotamme teihin
täydellisesti.

Janusz Radziwill,
Birźen ja Dubinkin vojevoda.»

— Hetmani on kovasti huolissaan siitä, mistä saat hevosia, —

sanoi herra Charlamp, kun lyhytkasvuinen ritari oli lopettanut
lukemisen.
 — Vaikeata niitä on saada, — vastasi herra Wolodyjowski. —
Täkäläinen alempi aatelisto tulee varmasti miehissä ensimmäiseen
kutsuntaan, mutta heillä on samogitialaiset hevosensa, jotka eivät
kelpaa sotapalvelukseen. Heidän pitäisi täydellä syyllä saada toiset.

— Ovathan ne kestäviä ja ravakoita hevosia.

— Mutta pienikasvuisia. Ja miehet täällä ovat päinvastoin

isokasvuisia. Kun he ajavat noilla hevosillaan, näyttää siltä, kuin
lippukunta ratsastaisi koirilla. Kyllä siitä hommaa tulee!… Mutta minä
ryhdyn heti asiaan. Annapas minulle Kmicicin määräyskirjelmä. Minä
annan sen itse hänelle, kuten hetmani käskee. Tämä tuli hänelle
otolliseen aikaan!

— Kuinka niin?

— Sen tähden että hän on tataarilaiseen tapaan ruvennut

ryöstämään neitosia. Hänellä on yhtä paljon prosesseja edessä kuin
hiuksia päässä. Siitä ei ole vielä viikkoa kulunut, kun minä hänen
kanssaan miekkailin.

— Oho! — huudahti Charlamp. — Jos sinä hänen kanssaan ottelit,

niin nyt hän kai makaa.

— Hän on jo toipumaan päin ja viikon, parin kuluttua terve. Mitä

uutta de publicis?

— Pahaa, vanhaan tapaan… Valtiovarainhoitaja Gosiewski on yhä

riidassa meidän ruhtinaamme kanssa, ja kun hetmanit ovat
eripuraisia, niin asiat eivät mene oikeita latuja. Vähän parempi on
asemamme kuitenkin, ja luulen, että jos yksimielisiä olisimme, niin
kyllä me vihollisen kanssa tulisimme aikoihin. Jumala suokoon, että
saamme niskasta pitäen karkoittaa heidät sinne, mistä ovat
tulleetkin. Valtiovarainhoitaja on syynä kaikkeen!

— Mutta toiset sanovat, että suurhetmani itse.

— Ne ovat pettureita. Vitebskin vojevoda niin väittää, sillä hän on

jo aikoja sitten vehkeillyt salaa yhdessä valtiovarainhoitajan kanssa.

— Vitebskin vojevoda on kelpo kansalainen.

— Oletko sinäkin Sapiehojen puolella Radziwilleja vastaan?

— Minä olen isänmaan puolella; se on jokaisen velvollisuus.

Siinäpä se pahin onkin, että sotilaatkin jakautuvat puolueisiin sen
sijaan että kävisivät sotaa vihollista vastaan. Mutta Sapieha on
kunnon mies, ja sen minä sanoisin itse ruhtinaallekin, vaikka olenkin
hänen palveluksessaan.

— Kunnon kansalaiset ovat koettaneet sovittaa heidät, mutta

turhaan! — sanoi Charlamp. — Kovasti nykyään lentää lähettejä
kuninkaalta meidän ruhtinaamme luo… Huhutaan siellä suuressa
maailmassa olevan taas jotakin tekeillä. Olemme odottaneet yleistä
kutsuntaa kuninkaan johdolla, mutta siitä ei ole tullut mitään.
Sanotaan, että jossakin muualla se voisi olla tarpeellisempaa.

— Ehkä Ukrainassa?

— Mistä minä sen tietäisin! Mutta luutnantti Brochwicz on kertonut

minulle, mitä hän omin korvin on kuullut. Tiesenhausen saapui
kuninkaan lähettinä meidän ruhtinaamme luo, ja he sulkeutuivat
huoneeseen ja puhelivat kauan keskenään. Mistä he puhuivat, sitä
Brochwicz ei voinut kuulla, mutta kun he tulivat ulos, kuuli hän
hetmanin itse sanovan näin: »Tästä voi syttyä vielä uusi sota.» Kyllä
me olemme julmasti ajatelleet ja arvailleet, mitä nuo sanat voisivat
merkitä!

— Hän kuuli varmasti väärin! Ketä vastaan nyt lähdettäisiin

sotimaan?
Keisari on pikemmin meidän kuin vihollisemme puolella, Ruotsin
kanssa
kestää aselepoa vielä kuusi vuotta, ja tataarit auttavat meitä
Ukrainassa, mitä he eivät suinkaan tekisi ilman Turkin suostumusta.

— Me emme myöskään keksineet mitään.

— Kun ei mitään ollut. Mutta Jumalan kiitos, että sain taas työtä.
Olen jo ikävöinyt taistelutanterille.

— Sinä viet siis itse kirjelmän Kmicicille.

— Kuten sanoin: hetmani käskee minua niin tekemään. Minun

täytyy sitäpaitsi muutenkin käydä tervehtimässä häntä, kuten
ritarillinen tapa vaatii, ja kun minulla on kirjelmä, on syytäkin. Mutta
annanko sen hänelle, se on toinen juttu. Täytyy harkita, koska se on
jätetty riippumaan minusta.

— Se on hyvä, sillä minun täytyy joutua taas matkaan. Minulla on

kolmas kirje herra Stankiewiczille. Sitten täytyy lähteä Kiejdanyyn
toimittamaan tykkejä ja sitten Birźeen katsomaan, onko
linnoituksessa kaikki valmiina puolustukseen.

— Vai Birźeen?

— Niin.
 — Se on minusta kummallista. Vihollinen ei ole saanut
ainoatakaan uutta voittoa ja on vielä kaukana Birźestä, joka sijaitsee
Kuurinmaan rajalla. Ja jos, kuten näkyy, uusia rykmenttejä
muodostetaan, pitäisi riittää väkeä niillekin seuduille, jotka jo ovat
vihollisen hallussa. Kuurinmaalaiset eivät ajattele sotaa meitä
vastaan. He ovat kyllä hyviä sotilaita, mutta heitä on niin vähäisen,
että Radziwill itse voisi kuristaa heidät yhdellä kädellä.

— Kummallista se on minustakin, — vastasi Charlamp, — etenkin

kun minua on käsketty joutumaan ja, jos kaikki ei ole kunnossa,
ilmoittamaan ruhtinas Boguslawille, jonka pitäisi lähettää insinööri
Peterson sinne.

— Mitä tämä merkinnee?! Kunhan tästä ei vain syttyisi sisällinen

sota! Jumala meitä siitä varjelkoon! Niin pian kuin ruhtinas Boguslaw
on mukana, on pirulla syytä iloita.

— Älä sano! Hän on rohkea soturi.

— Sitä en kiellä, mutta hän on pikemmin saksalainen tai

ranskalainen kuin puolalainen… Valtakunnasta hän vähät välittää,
mutta Radziwilleista kyllä. Se on juuri hän, joka lietsoo meidän
hetmanissamme, Vilnon vojevodassa, ylpeyden henkeä, vaikka sitä
muutenkin olisi hetmanissa tarpeeksi. Myöskin Sapiehojen ja
Gosiewskin eripuraisuus on hänen työtään.

— Sinähän olet suuri poliitikko, näemmä. Riennäpäs vain naimisiin,

Michal, jottei sinun älysi joutuisi hukkaan.

Wolodyjowski katsoi toveriinsa tutkivasti.

— Naimisiin?…
 — Niin juuri! Taidatpa ollakin kosiomatkalla, koska olet noin
paraatipuvussa?

— Jätä minut jo rauhaan!

— Tunnusta pois…

— Pitäköön kukin rukkasensa hyvänään… On niitä sinullakin.

Mutta nyt ei ole aikaa ajatella sellaisia, kun pää on täynnä sotaisia
asioita.

— Oletko valmis heinäkuussa?

— Heinäkuun lopulla olen, vaikka pitäisi nuo hevoset maasta

kaivaa.
Jumalan kiitos, että sain työtä, muuten olisi suru minut syönyt.

Tiedot hetmanilta ja ajatukset edessäolevasta ankarasta työstä

karkoittivat herra Wolodyjowskin huonon tuulen, ja ennenkuin he
olivat saapuneet Pacuneleen, hän ei enää muistanutkaan äskeistä
vastoinkäymistä. Tieto käskykirjeestä levisi nopeasti ympäri kylän.
Aateliset saapuivat heti kysymään, oliko se totta, ja kun herra
Wolodyjowski vakuutti sen olevan, teki se syvän vaikutuksen. Kaikki
olivat valmiita lähtemään sotaan, mutta muutamat valittivat sitä, että
se sattui juuri heinäkuun lopulle, kun elonkorjuu on ovella. Herra
Wolodyjowski lähetti heti pikalähettejä toisiin kyliin, Upitaan ja
suurimpiin kartanoihin. Illalla saapui muutamia Butrymeja,
Stakianeja ja Domaszewiczeja.

Vihollista haukuttiin kovasti ja ennustettiin loistavia voittoja.

Butrymit vain vaikenivat, mutta sitä ei pantu pahaksi, koska tiedettiin
heidänkin olevan yhtä miestä. Seuraavana päivänä jo kuhisi kylissä
kuin mehiläispesissä. Ei puhuttu enää herra Kmicicistä ja Aleksandra-
neidistä, sillä sotaretki oli yleisenä keskustelun aiheena. Herra
Wolodyjowski oli suonut Oleńkalle anteeksi rukkaset lohduttaen
itseään sillä, ettei se ollut ensimmäinen eikä viimeinen kerta. Hän ei
ollut vielä selvillä siitä, mitä tehdä Kmicicin käskykirjelmällä.
YHDEKSÄS LUKU.

Herra Wolodyjowskille alkoi nyt ankaran työn, kirjelmäin ja

matkustusten aika. Seuraavalla viikolla hän muutti Upitaan, jossa
pestaaminen alkoi. Aatelia, sekä ylempää että alempaa, tuli
tulvimalla hänen luokseen maineen houkuttelemana. Runsaimmin
saapui kuitenkin laudalaisia, joille oli hankittava hevosia. Tämä tuotti
herra Wolodyjowskille paljon puuhaa, mutta neuvokas kun oli se
onnistui hänelle kuitenkin. Samaan aikaan hän kävi tervehtimässä
herra Kmiciciä Lubiczissa. Tämä oli jo parantunut aika lailla, vaikk'ei
ollutkaan vielä jättänyt vuodetta. Jos herra Wolodyjowskin miekka oli
raskas, niin oli hänen kätensä kevyt.

Herra Kmicic tunsi hänet heti ja kalpeni hiukan nähdessään tulijan.

Hän vei vaistomaisesti kätensä vuoteen yläpuolella riippuvaa
miekkaa kohti, mutta huomattuaan hymyn vieraan kasvoilla ojensikin
laihtuneen kätensä hänelle sanoen:

— Kiitän, että tulitte tervehtimään. Se oli ritarillista.

— Tulin kysymään, kannatteko kaunaa minulle? — sanoi Michal-

herra.
 — En, sillä minut voitti miekkailumestari. Mutta huonosti minun oli
käydä.

— Mitenkä jaksatte nyt?

— Te taidatte ihmetellä, että pääsin hengissä käsistänne? Ainakin

minä sitä ihmettelen.

Herra Kmicic hymyili ja lisäsi:

— Mutta eihän asia ole silti menetetty. Voittehan tehdä minusta

lopun milloin haluatte.

— Enhän minä tullut ensinkään siinä tarkoituksessa…

— Te olette joko itse paholainen tai osaatte noitua, — keskeytti

Kmicic. — Minun ei luonnollisesti tee mieleni kehua itseäni nyt, kun
melkein palaan toisesta maailmasta, mutta ennenkuin olin tavannut
teidät, ajattelin aina: jollen ole ensimmäinen miekkailija
valtakunnassa, niin ainakin toinen. Sanokaa, missä olette oppinut
miekkailutaitonne?

— Osaksi on minulla ollut synnynnäisiä taipumuksia, — vastasi

Michal-herra, — ja sitäpaitsi on isäni opettanut minua pienestä
pitäen. Hän tapasi sanoa: »Jumala on antanut sinulle vaatimattoman
ulkomuodon, ja jolleivät ihmiset tule sinua pelkäämään, niin he
nauravat.» Sitten opin myöskin ollessani vojevodan palveluksessa
Vähä-Venäjällä. Siellä oli muutamia miehiä, jotka löivät minut.

— Onko mahdollista, että oli sellaisia?

— Oli niitä. Esimerkiksi herra Podbipienta, liettualainen ja

familiantti, joka kaatui Zbarazin luona. Rauha hänen sielulleen! Siinä
oli sitten niin julman väkevä mies, ettei uskaltanut ajatellakaan häntä
vastustaa, muuten hän olisi voinut yhdellä iskulla halkaista sekä
varustukset että vastustajan. Sitten oli siellä vielä ystäväni herra
Skrzetuski, josta varmaankin olette kuullut.

— Kuinka! Hänkö, joka hyökkäsi Zbarazista murtautuen

kasakkajoukkojen läpi! Kukapa ei olisi kuullut hänestä puhuttavan?
Niitä miehiäkö te olettekin? Ja zbarazilainen? Teen kunniaa! Siinä
tapauksessa olen varmasti kuullut teistä puhuttavan Vilnon
vojevodan luona. Onko teidän etunimenne Michal?

— On kyllä. Oikeastaan olen Jerzy ( = Yrjö) Michal, mutta koska

Pyhä Mikael johtaa koko taivaan sotajoukkoja ja on saavuttanut jo
niin monta voittoa helvetin lippukunnista, niin pidän mieluummin
hänet suojeluspyhimyksenäni.

— Niin, eihän Pyhä Yrjänä sentään vedä vertoja Pyhälle Mikaelille.

Te olette siis sama Wolodyjowski, joka kaatoi Bohunin.

— Sama.

— No, eipä ole häpeä saada sellaiselta mieheltä vasten kuonoa.

Jumala suokoon, että meistä tulisi ystävät. Te sanoitte minua tosin
petturiksi, mutta siinä te erehdyitte.

Sen sanottuaan Kmicic rypisti kulmakarvojaan, ikäänkuin hän olisi

tuntenut kipua haavassa.

— Minä myönnän erehtyneeni, — vastasi herra Wolodyjowski, —

mutta minä sain tietää sen jo teidän miehiltänne ennen kuin nyt
teiltä. Ja huomatkaa: muuten minä en olisikaan tullut tänne.
 — Kyllä minua on näillä seuduilla haukuttu! — selitti Kmicic
katkerasti. — Oli miten hyvänsä. Tunnustan, että olen syypää yhteen
ja toiseen, mutta kyllä paikkakunnan asukkaat ovatkin ottaneet
minut hävyttömästi vastaan.

— Eniten te olette vahingoittanut itseänne Wolmontowiczen palolla

ja viimeisellä ryöstöllä.

— Kyllä minua nyt uhataankin prosesseilla. Olen saanut jo koko

joukon haasteita. Eivät anna sairaankaan olla rauhassa. Olen
polttanut Wolmontowiczen, se on totta, ja ihmisiä olen tappanut,
mutta rangaiskoon Jumala minua, jos tein sen pahuuttani. Yöllä
ennen paloa olin tehnyt pyhän lupauksen elää kaikkien kanssa
sovussa, vieläpä maksaa upitalaisille vahingonkorvausta, sillä olin,
totta puhuen, tehnyt siellä pahaa jälkeä. Mutta kun minä sinä yönä
palasin kotiin, niin mitä näin? Minun toverini oli teurastettu kuin
härät! Kun sain tietää, että Butrymit olivat sen tehneet, silloin heräsi
itse saatana minussa… ja minä kostin kauheasti… Ja voitteko arvata
miksi heidät oli surmattu? Sain sittemmin tietää sen eräältä
Butrymilta, jonka kimppuun hyökkäsin metsässä: siksi, että he
olisivat tahtoneet tanssia Butrymien naikkosten kanssa eräässä
kapakassa… Kukapa ei olisi kostanut?

— Se on totta, — vastasi Wolodyjowski, — että teidän

tovereitanne oli kohdeltu liian julmasti, mutta tokkohan laudalaiset
heitä tuhosivat? Ei suinkaan; syynä heidän tuhoonsa oli heidän oma
maineensa, jonka he olivat tuoneet mukanaan, sillä jos jotkut
sotilaat olisivat kohteliaasti tyttöjä tanssittaneet, niin ei heitä
senvuoksi olisi surmattu.

— Miesparat! — jatkoi Kmicic seuraten omaa ajatustaan. — Kun

olen nyt maannut täällä kuumeessa, ovat he harva se ilta tulleet
luokseni tuon oven kautta… Olen nähnyt heidän seisovan vuoteeni
ympärillä sinisinä, raastettuina ja valittavan: »Jendrus, anna lukea
meille messuja, sillä me kidumme kovasti…» Hiukseni ihan ovat
nousseet pystyyn, ja tulikiven hajua on tuntunut huoneessa heidän
jälkeensä… Messuja olen kyllä luetuttanut; kunpa se sitten olisi
auttanut!

Seurasi hetken hiljaisuus.

— Ja mitä ryöstöön tulee, — jatkoi Kmicic, — niin te ette voi

tietää, että neiti Billewicz on pelastanut henkeni, kun laudalaiset
ajoivat minua takaa, mutta käskenyt sitten minua menemään
matkoihini. Mitäpä muuta olisin voinut tehdä?!

— Se oli kuitenkin aito tataarilainen teko.

— Te ette mahtane tietää, mitä on rakkaus ja kuinka

epätoivoiseksi ihminen voi tulla menettäessään sen, jota eniten
rakastaa.

— Enkö minä tietäisi mitä rakkaus on? — huudahti herra

Wolodyjowski harmistuneesti. — Aina siitä hetkestä saakka, jolloin
aloin kantaa miekkaa, olen alinomaa ollut rakastunut… Totta kyllä,
minä olen rakastanut useampaa kuin yhtä, mutta se johtuu siitä,
etten ole koskaan saanut vastarakkautta. Muussa tapauksessa
kukaan ei olisi ollut uskollisempi minua.

— Mitä rakkautta se sellainen on, että rakastaa milloin yhtä,

milloin toista, — arveli Kmicic.

— Kerronpa teille mitä olen omin silmin nähnyt. Chmielnickin

kapinan alkuaikoina Bohun, sama, joka nykyään Chmielnickin jälkeen
nauttii suurinta kunniaa kasakkain joukossa, ryösti Skrzetuskin
rakastetun, ruhtinatar Kurcewiczin. Se oli rakkautta! Koko sotajoukko
itki nähdessään Skrzetuskin parran epätoivon surusta harmaantuvan,
vaikka mies oli vasta vähän päälle kaksikymmentä vuotta. Mutta
arvatkaapas, mitä hän teki?

— Mistäpä minä sen arvaisin?

— Koska isänmaa oli vaarassa ja tarvitsi häntä, koska julma

Chmielnicki lähestyi voitokkaasti, niin hän ei lähtenyt etsimään
ruhtinatartaan. Hän uhrasi surunsa Jumalalle ja otti osaa kaikkiin
ruhtinas Jeremin johtamiin taisteluihin, kunnes Zbarazin luona
saavutti sellaisen maineen, että kaikki nyt ihaillen mainitsevat hänen
nimensä. Verratkaa nyt kuinka menetteli hän ja kuinka te, ja mikä
ero!

Kmicic vaikeni ja puri viiksiään. Wolodyjowski jatkoi:

— Mutta siksipä Jumala palkitsikin Skrzetuskia ja antoi hänelle

neidon. Heti Zbarazin taistelun jälkeen he menivät naimisiin, ja nyt
on heillä jo kolme lasta. Skrzetuski palvelee yhä armeijassa. Mutta te
olette harjoittanut väkivaltaa auttaen siten vihollista ja olitte vähällä
menettää henkenne, puhumattakaan siitä, että muutamia päiviä
sitten olitte vähällä päästä neidistännekin.

— Millä tavalla? — kysyi Kmicic nousten istumaan vuoteessaan. —

Mitä hänelle on tapahtunut?

— Ei mitään. Eräs mies vain on pyytänyt hänen kättään toivoen

häntä vaimokseen.
 Kmicic kalpeni, ja hänen sisäänpainuneet silmänsä alkoivat
säihkyä. Hän aikoi nousta, karkasi jo puoleksi pystyyn ja huusi:

— Kuka se lurjus oli? Sanokaa!

— Minä, — vastasi herra Wolodyjowski.

— Te? Te? — toisti Kmicic ihmetellen. — Tekö?…

— Minä juuri.

— Petturi! Sen saatte vielä maksaa!… Entä hän?… Kautta Jumalan,

kertokaa kaikki!… Suostuiko hän?

— Kielsi vähääkään arvelematta.

Seurasi taas hetken hiljaisuus. Kmicic hengitti raskaasti ja tuijotti

Wolodyjowskiin. Tämä jatkoi:

— Miksi sanotte minua petturiksi? Olenko minä muka teidän

veljenne tai puhemiehenne? Olenko syönyt sanani? Voitin teidät
kaksintaistelussa ja olisin voinut tehdä teille mitä hyvänsä.

— Vanhan tavan mukaan olisi toinen meistä maksanut tuon

verellään. Jollen miekalla olisi kyennyt, niin olisin ampunut, ja
vieköönpä minut sitten vaikka itse paholainen!

— Jos hän olisikin suostunut, en minä olisi kuitenkaan ryhtynyt

kanssanne toistamiseen kaksintaisteluun. Mitä syytä minulla olisi
ollut sellaiseen? Mutta tiedättekö miksi hän antoi kieltävän
vastauksen?

— Miksi? toisti Kmicic kuin kaiku.

 — Siksi, että rakastaa teitä.

Se meni yli sairaan voimien. Kmicic vaipui patjoille, hiki nousi

otsalle, ja hän makasi siinä hetken aikaa sanatonna.

- Kovasti olen vielä heikko, — virkkoi hän sitten. — Mistä te

tiedätte, että hän rakastaa minua?

— Onhan minulla silmät, joilla näen, ja järki, jolla ymmärrän.

Etenkin nyt, kun olen saanut rukkaset, käsitän kaikki niin hyvin.
Ensiksikin: kun kaksintaistelun jälkeen tulin ilmoittamaan hänelle,
että hän on vapaa, koska te olitte joutunut häviölle, oli hän pyörtyä,
ja sen sijaan että olisi osoittanut minulle kiitollisuuttaan, hän tuskin
huomasi minua. Toiseksi: kun Domaszewiczit kantoivat teidät tänne,
niin hän kohotti hellävaroen kuin äiti päätänne. Kolmanneksi: kun
minä häntä kosin, hän osoitti minut luotaan kuin koiran. Jolleivät
nämä seikat todista mielestänne tarpeeksi, on miekanisku lyhentänyt
järkeänne.

— Kunpa tuo olisi totta! — huudahti Kmicic heikosti. — Täällä

minun haavaani voidellaan jos jollakin… mutta minulle ei voisi löytää
sen parempaa balsamia kuin teidän sananne.

— Onko se petturi, joka tuo balsamia?

— Suokaa anteeksi. Päässäni ei ole tilaa sellaiselle onnelle, kuin

että hän vielä ottaisi minut…

— Minä sanoin, että hän rakastaa teitä, mutta ottamisesta en ole

mitään puhunut… Se on toinen asia!

— Jollei hän huoli minua miehekseen, niin ruhjon pääni seinään.

En voi elää ilman häntä!
 — Ehkä hän ottaisikin, jos te rehellisesti katuisitte rikoksianne.
Nythän on sota, voitte lähteä taistelutanterille, voitte tehdä rakkaalle
isänmaalle suuria palveluksia, voitte niittää urhoollisen soturin
kunniaa ja hankkia hyvän maineen. Kukapa olisi synnitön? Kenelläpä
ei olisi syytöksiä omallatunnollaan?… Jokaisella… Mutta jokaisella on
myös tie avoinna katumukseen ja parannukseen. Te olette tehnyt
itsenne syypääksi mielivaltaan, älkää enää tehkö sitä; te olette
rikkonut isänmaata vastaan aiheuttamalla sisäistä sekasortoa sodan
aikana, pelastakaa se nyt vaarasta; te olette tehnyt vääryyttä
kansalaisille, palkitkaa… Se on teille paljon parempi tie kuin ruhjoa
päänsä seinään.

Kmicic katsoi kunnioittavasti Wolodyjowskiin ja sanoi sitten:

— Te puhutte kuin tosi ystävä.

— En ole ystävänne, joskaan en, totta puhuen,

vihamiehennekään, mutta neitoa minun käy sääliksi, vaikka hän
antoikin minulle rukkaset, sillä tulin sanoneeksi hänelle ankaria
sanoja… Minä en rukkasten tähden mene hirteen, eiväthän ne olleet
ensimmäiset, enkä vihan kaunaa myöskään kanna. Jos voin auttaa
teidät hyvälle tielle, teen palveluksen isänmaalle, koska olette
oivallinen ja kokenut sotilas.

— Mutta onko minulla aikaa päästä tuolle hyvälle tielle? Monet

oikeudenkäynnit odottavat minua. Minun täytyy vuoteeltani suoraan
oikeuteen… Enkä tahdo paeta. Niin paljon oikeudenkäyntejä! Ja
oikeus tuomitsee varmasti syylliseksi.

— Tämä pelastaa niistä! — sanoi herra Wolodyjowski ottaen esille

käskykirjeitään.
 — Käskykirjelmä! — huudahti Kmicic. — Kenelle?

— Teille! Eikä teidän tarvitse mennä oikeuteen täällä, koska olette

tämän mukaan hetmanin oikeuden alainen. Kuunnelkaa, mitä
ruhtinas kirjoittaa minulle.

Ja herra Wolodyjowski luki Kmicicille Radziwillin yksityisen kirjeen,

huokasi sitten syvään, kiersi viiksiään ja sanoi:

— Kuten kuulitte, riippuu minusta, annanko käskykirjelmän teille

vai enkö.

Pelko ja toivo kuvastuivat Kmicicin kasvoilla.

— Entä mitä teette? — kysyi hän vaisulla äänellä.

— Minä annan, — vastasi Wolodyjowski.

Kmicic ei ensin virkkanut mitään. Hänen päänsä painui patjalle, ja

hän jäi hetkeksi tuijottamaan kattoon. Yht'äkkiä hänen silmänsä
alkoivat kostua, ja niissä tuiki tuntemattomat vieraat, kyynelet,
värisivät jo ripsissä.

— Lyötäköön minut säpäleiksi, — sanoi hän vihdoin, — jos

koskaan olen tavannut jalompaa miestä kuin te!… Jos te minun
tähteni olette saanut rukkaset ja jos, kuten arvelette, Oleńka
rakastaa minua, niin olisi joku muu tämän varmasti kostanut minulle
ja syössyt minut entistä syvemmälle… Mutta te ojennatte minulle
kätenne ja kiskotte minut kuin haudasta ylös!

— Minä tahdon uhrata omat etuni rakkaan isänmaan tähden, jolle

te vielä voitte tehdä suuria palveluksia. Mutta sen sanon teille, että
jos olisitte lainannut nuo kasakat Trubeckilta tai Chowanskilta, niin
siinä tapauksessa en olisi antanut teille käskykirjelmää. Olipa onni,
ettette ollut sitä tehnyt!

— Etevä esimerkki muille! — huudahti Kmicic. — Antakaa minun

puristaa kättänne! Suokoon Jumala, että saisin joskus tehdä teille
vastapalveluksen. Olen teille kiitollinen koko ikäni!

— No niin… Kuulkaahan sitten: Teidän ei siis tarvitse ilmaantua

minkään oikeusistuimen eteen, vaan ryhtyä työhön. Jos te olette
hyödyksi isänmaalle, niin kyllä täkäläinen aateli suo teille anteeksi,
sillä isänmaa on sen sydämellä… Te voitte vielä hyvittää rikoksenne,
voittaa hyvän maineen ja säteillä kunniaa kuin aurinko… Ja sitten
tiedän vielä erään neitosen, joka varmasti valmistaa teille lahjan
palkinnoksi.

— Hei! — huusi Kmicic innoissaan. — Miksi loikoilla täällä

sängyssä, kun vihollinen polkee isänmaata! Hei! Onko siellä ketään?
Poika, tuo saappaani!… Ei kuulu ketään!… Yksin olen!… Mutta
lyököön salama minut, jos näillä patjoilla kauan piehtaroin!

Herra Wolodyjowski hymyili tyytyväisesti ja virkkoi:

— Teidän henkenne on vielä ruumistanne paljon voimakkaampi.

Sitten hän nousi lähteäkseen, mutta Kmicic ei päästänyt häntä,

ennenkuin saisi tarjota hänelle viiniä.

Oli jo ilta, kun pienikasvuinen ritari läksi Lubiczista suunnaten

matkansa Wodoktyyn.

»Minä hyvitän parhaiten kovat sanani», ajatteli hän, »kun kerron

neidille, että Kmicic ei nouse ainoastaan vuoteeltaan, vaan myöskin
häpeästään… Hän ei ole mikään pohjaltaan pilaantunut mies, vaan
hän on julman tulinen. Minä tuon neidille varmasti lohdutuksen ja
luulen, että hän ottaa minut ystävällisemmin vastaan nyt kuin silloin,
kun itseäni tarjosin…»

Tässä tuo rehellinen ritari huokasi ja virkahti:

— Onkohan maailmassa neitoa, joka olisi minulle määrätty?

Tällaisissa mietteissä hän saapui Wodoktyyn. Tuttavamme,

pörröpäinen samogitialainen, riensi veräjälle, jonka hän avasi
verkalleen sanoen:

— Emäntä ei ole kotona.

— Onko hän lähtenyt jonnekin?

— On.

— Minne?

— En tiedä.

— Milloin hän palaa?

— En tiedä.

— Puhu ihmisiksi! Eikö hän sanonut milloin tulee?

— Jumala ties, ehk'ei koskaan, sillä vaunut olivat kirstuja täynnä.

Siitä pääteltiin, että pitkä on matka.

— Vai niin! — murahti herra Michal. — Tämän olen nyt saanut

aikaan!…
KYMMENES LUKU.

Tavallisesti, kun auringon lämpimät säteet alkavat tunkeutua

synkkien talvipilvien läpi, kun ensimmäiset silmikot puhkeavat puihin
ja vihertävä kevätlaiho alkaa nousta kosteille vainioille, silloin
paremman toivo herää myöskin ihmisten sydämissä. Mutta kevät
vuonna 1655 ei tuonut mukanaan Puolan valtakunnan
kovaakokeneille asukkaille tuota tavallista lohdutusta. Maan koko
itäinen raja pohjoisesta aina etelän villeille kentille saakka oli
ikäänkuin tulisen nauhan peitossa, eivätkä kevätsateet voineet paloa
sammuttaa, päinvastoin, se yhä leveni ja laajeni yli maan. Sitäpaitsi
nähtiin taivaalla pahaaennustavia merkkejä, jotka viittasivat vielä
suurempiin onnettomuuksiin. Taivaalla kiitävät pilvet muodostivat
korkeita torneja ja linnoituksia, jotka ukkosen jyristessä raukesivat
raunioiksi. Salama iski maahan, jota lumi vielä peitti, havumetsät
kellastuivat ja puitten oksat vääntyivät kummallisiksi, sairaalloisiksi
koukeroiksi; karjaa kaatui, lintuja putoili kuolleina maahan.
Huomattiinpa vihdoin auringossakin omituisia täpliä, jotka ikäänkuin
muodostivat käden, jossa oli omena, lävistetyn sydämen ja ristin.
Mielten levottomuus kasvoi, ja munkit pohtivat, mitä moiset merkit
saattoivat ennustaa. Kummallinen pelko valtasi kaikkien sydämet.