Download Full Version ebook - Visit ebooknice.

com

(Ebook) Temporal Logics in Computer Science:

Finite-State Systems by Stéphane Demri, Valentin
Goranko, Martin Lange ISBN 9781107028364,
1107028361
https://ebooknice.com/product/temporal-logics-in-computer-
science-finite-state-systems-6747622

Click the button below to download

DOWLOAD EBOOK

Discover More Ebook - Explore Now at ebooknice.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Start reading on any device today!

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason;

Viles, James ISBN 9781459699816, 9781743365571,
9781925268492, 1459699815, 1743365578, 1925268497
https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

ebooknice.com

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena

Alfredsson, Hans Heikne, Sanna Bodemyr ISBN 9789127456600,
9127456609
https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

ebooknice.com

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT

II Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-
math-1c-and-2c-2002-peterson-s-sat-ii-success-1722018

ebooknice.com

(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master

the SAT Subject Test: Math Levels 1 & 2) by Arco ISBN
9780768923049, 0768923042
https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-
arco-master-the-sat-subject-test-math-levels-1-2-2326094

ebooknice.com
(Ebook) Cambridge IGCSE and O Level History Workbook 2C -
Depth Study: the United States, 1919-41 2nd Edition by
Benjamin Harrison ISBN 9781398375147, 9781398375048,
1398375144, 1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044

ebooknice.com

(Ebook) Bounded Variable Logics and Counting: A Study in

Finite Models by Martin Otto ISBN 9781107167940,
9781316716878, 9781316731550, 9781316733486, 1107167949,
1316716872, 1316731553, 1316733483
https://ebooknice.com/product/bounded-variable-logics-and-counting-a-
study-in-finite-models-47637062

ebooknice.com

(Ebook) Logic as a Tool: A Guide to Formal Logical

Reasoning by Valentin Goranko ISBN 9781118880005,
1118880005
https://ebooknice.com/product/logic-as-a-tool-a-guide-to-formal-
logical-reasoning-5677508

ebooknice.com

(Ebook) Logic as a Tool: A Guide to Formal Logical

Reasoning by Valentin Goranko ISBN 9781118880005,
9781118880043, 1118880005, 1118880048
https://ebooknice.com/product/logic-as-a-tool-a-guide-to-formal-
logical-reasoning-6746800

ebooknice.com

(Ebook) Logic and Discrete Mathematics: A Concise

Introduction, Solutions Manual by Willem Conradie,
Valentin Goranko, Claudette Robinson ISBN 9781118762677,
1118762673
https://ebooknice.com/product/logic-and-discrete-mathematics-a-
concise-introduction-solutions-manual-5100454

ebooknice.com
Temporal Logics in Computer Science
Finite-State Systems

STÉPHANE DEMRI
Centre National de la Recherche Scientifique (CNRS), France

VA L E N T I N G O R A N KO
Stockholms Universitet

MART IN LANGE
Universität Kassel, Germany
 University Printing House, Cambridge CB2 8BS, United Kingdom

Cambridge University Press is part of the University of Cambridge.

www.cambridge.org
Information on this title: www.cambridge.org/9781107028364
© Stéphane Demri, Valentin Goranko and Martin Lange 2016

First published 2016

Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library
ISBN 978-1-107-02836-4 Hardback
 Contents

1 Introduction page 1
1.1 Temporal Logics and Computer Science: A Brief Overview 1
1.2 Structure and Summary of the Book Content 6
1.3 Using the Book for Teaching or Self-Study 12
PART I MODELS
2 Preliminaries and Background I 17
2.1 Sets and Relations 18
2.2 Some Fundamental Preliminaries 25
3 Transition Systems 35
3.1 Basic Concepts 37
3.2 Reachability 49
3.3 Bisimulation Relations 55
3.4 Bisimilarity 62
3.5 Trace Equivalence 71
3.6 Exercises 75
3.7 Bibliographical Notes 79
PART II LOGICS
4 Preliminaries and Background II 85
4.1 Preliminaries on Modal Logic 85
4.2 Logical Decision Problems 91
4.3 Expressive Power 93
4.4 Deductive Systems 93
5 Basic Modal Logics 100
5.1 The Basic Modal Logic BML 102
5.2 Renaming and Normal Forms 107
5.3 Modal and Bisimulation Equivalence 110
5.4 Model Checking 116
5.5 Satisfiability and the Tree Model Property 123
 5.6 The Basic Tense Logic BTL 131
5.7 Axiomatic Systems 135
5.8 Exercises 141
5.9 Bibliographical Notes 146
6 Linear-Time Temporal Logics 150
6.1 Syntax and Semantics on Linear Models 152
6.2 Logical Decision Problems 159
6.3 The Small Model Property 164
6.4 Decision Procedures 169
6.5 Adding Past-Time Operators 176
6.6 Invariance Properties 182
6.7 Extensions of LTL 185
6.8 An Axiomatic System for LTL 191
6.9 Exercises 196
6.10 Bibliographical Notes 206
7 Branching-Time Temporal Logics 209
7.1 A Hierarchy of Branching-Time Logics 211
7.2 Bisimulation Invariance 228
7.3 Model Checking 233
7.4 Some Fragments and Extensions of CTL∗ 241
7.5 Axiomatic Systems 252
7.6 Exercises 259
7.7 Bibliographical Notes 265
8 The Modal Mu-Calculus 271
8.1 Fixpoint Quantifiers 272
8.2 Fixpoint Iteration 282
8.3 The Structural Complexity of Formulae 289
8.4 Model-Checking Games 303
8.5 Bisimulation Invariance 309
8.6 The Second-Order Nature of Temporal Logics 313
8.7 Variants 315
8.8 Exercises 320
8.9 Bibliographical Notes 324
9 Alternating-Time Temporal Logics 329
9.1 Concurrent Multiagent Transition Systems 330
9.2 Temporal Logics for Concurrent Game Models 337
9.3 Logical Decision Problems 346
9.4 Exercises 352
9.5 Bibliographical Notes 355
PART III PROPERTIES
10 Expressiveness 361
10.1 Embeddings among Linear-Time Logics 363
10.2 Embeddings among Branching-Time Logics 376
10.3 Separation Results 385
10.4 Exercises 409
10.5 Bibliographical Notes 414
11 Computational Complexity 419
11.1 Proving Lower Bounds 421
11.2 Linear-Time Temporal Logics 435
11.3 Branching-Time Temporal Logics 445
11.4 An Overview of Completeness Results 453
11.5 Exercises 457
11.6 Bibliographical Notes 460
PART IV METHODS
12 Frameworks for Decision Procedures 467
12.1 A Brief Introduction to Three Methodologies 468
12.2 The Frameworks Compared 472
13 Tableaux-Based Decision Methods 476
13.1 A Generic Incremental Tableau Construction 479
13.2 Tableaux for LTL 498
13.3 Tableaux for TLR and CTL 514
13.4 Exercises 536
13.5 Bibliographical Notes 540
14 The Automata-Based Approach 543
14.1 Introduction to Nondeterministic Büchi Automata 546
14.2 From LTL Formulae to Automata 552
14.3 Introduction to Alternating Automata on Words 570
14.4 From LTL Formulae to Alternating Büchi Automata 581
14.5 Extensions of LTL 591
14.6 Tree Automata for Branching-Time Logics 598
14.7 Alternating Tree Automata and CTL 606
14.8 Exercises 615
14.9 Bibliographical Notes 621
15 The Game-Theoretic Framework 625
15.1 Parity Games 627
15.2 Constructions for Automata on Infinite Words 647
15.3 Model Checking 659
15.4 Satisfiability Checking 682
15.5 Exercises 705
15.6 Bibliographical Notes 711
References 716
Index 737
 1
Introduction

1.1 Temporal Logics and Computer Science: A Brief Overview 1

1.1.1 Historical Origins of Temporal Reasoning and Logics 1
1.1.2 The Role of Temporal Logics in Computer Science 3
1.1.3 The Influence of Computer Science on the Development of
Temporal Logics 5
1.2 Structure and Summary of the Book Content 6
1.3 Using the Book for Teaching or Self-Study 12

Temporal logics provide a generic logical framework for modelling and reasoning about
time and temporal aspects of the world. While stemming from philosophical considera-
tions and discussions, temporal logics have become over the past 50 years very useful and
important in computer science, and particularly for formal specification, verification and
synthesis of computerised systems of various nature: sequential, concurrent, reactive, dis-
crete, real time, stochastic, etc. This book provides a comprehensive exposition of the most
popular discrete-time temporal logics, used for reasoning about transition systems and com-
putations in them.

1.1 Temporal Logics and Computer Science: A Brief Overview

We begin with a brief historical overview of temporal reasoning and logics and their role
in computer science.

1.1.1 Historical Origins of Temporal Reasoning and Logics

The study of time and temporal reasoning goes back to the antiquity. For instance, some
of Zeno’s famous paradoxical arguments refer to the nature of time and the question of
infinite divisibility of time intervals. Perhaps the earliest scientific reference on temporal
reasoning, however, is Aristotle’s ‘Sea Battle Tomorrow’ argument in the Organon II – On
2 Introduction

Interpretation, Chapter 9, claiming that future contingents, i.e. statements about possible
future events which may or may not occur, such as ‘There will be a sea-battle tomorrow’,
should not be ascribed definite truth values at the present time. A few decades later the
philosopher Diodorus Cronus from the Megarian school illustrated the problem of future
contingents in his famous Master Argument where he, inter alia, defined ‘possible’ as ‘what
is or will ever be’ and ‘necessary’ as ‘what is and will always be’.
Philosophical discussions on temporality, truth, free will and determinism, and their
relationships continued during the Middle ages. In particular, William of Ockham held
that propositions about the contingent future cannot be known by humans as true or false
because only God knows their truth value now. However, he argued that humans can still
freely choose amongst different possible futures, thus suggesting the idea of a future-
branching model of time with many possible time-lines (histories), truth of propositions
being relativised to a possible actual history. This model of time is now often called Ock-
hamist or actualist. Later, several philosophers and logicians raised and analysed problems
relating temporality with nondeterminism, historical necessity, free will, God’s will and
knowledge, etc., proposing various different solutions. Still, little tangible progress in the
formal study of temporal reasoning occurred until some new ideas emerged in the late 19th
to early 20th century, including the developments of Minkowski’s four-dimensional space
time model and its application to Einstein’s relativity theory, of Reichenbach’s theory of
temporality and tenses in natural language, as well as some informal philosophical stud-
ies of temporal reasoning by C. Peirce, McTaggart, B. Russell, W. Quine, J. Findley, J.
Łukasiewicz and others. However, temporal logic as a formal logical system only emerged
in the early 1950s when the philosopher Arthur Prior set out to analyse and formalise such
arguments, leading him, inter alia, to the invention of (as Prior called it) Tense Logic, of
which he developed and discussed several formal systems. Prior’s seminal work initiated
the modern era of temporal logical reasoning, leading to numerous important applications
not only to philosophy, but also to computer science, artificial intelligence and linguistics.
Prior introduced the following basic temporal modalities:

P ‘It has at some time in the past been the case that …’
F ‘It will at some time in the future be the case that …’
H ‘It has always been the case that …’
G ‘It will always be the case that …’

Put together in a formal logical language they allow complex temporal patterns to be
expressed, for instance:

There is a solstice ∧ there is a lunar eclipse → GP ‘There is a solstice and a lunar eclipse’

means that if there is a solstice and there is a lunar eclipse now then it will always be the
case that there has been a solstice and a lunar eclipse at the same moment of time.
Subsequently, other temporal modalities were introduced, notably Nexttime and the
binary operators Since and Until in Kamp’s (1968) very influential doctoral thesis. The
 1.1 Temporal Logics and Computer Science: A Brief Overview 3

simplest and most common models of time are linear-ordered time flows and Prior’s tem-
poral operators, as well as Since and Until, have natural interpretation in them. The resulting
temporal logics for linear time are quite expressive. A classical expressiveness result, due
to Kamp (1968) states that the temporal logic with Since and Until is as expressive as first-
order logic. The temporal logic with operators for Next-time and Until interpreted on the
time flow of natural numbers, known as the linear-time temporal logic LTL, became since
the late 1970s the most popular temporal logic used in computer science.
Coming back to Prior’s philosophical studies of temporal reasoning: in his analysis of
Diodorus’s Master Argument Prior argued that its fallacy lay in Diodorus’s assumption that
whatever is, or is not, or will be, or will never be, has in the past been necessarily so – thus, in
effect assuming that the future is deterministic. Prior supported Aristotle’s view that ‘while
it is now beyond the power of men or gods to affect the past, there are alternative futures
between which choice is possible’. In order to resolve the problems pointed out by Aristotle
and Diodorus, Prior wanted, inter alia, to capture the logic of historical necessity. His philo-
sophical analysis and the quest for formalisation of the arguments for ‘the incompatibility
of foreknowledge (and fore-truth) and indeterminism’ lead him to consider two formalisa-
tions of temporal logic of branching time, reflecting the ‘Peircean’ and the ‘Ockhamist’ (or,
‘actualist’) views, underlying respectively the ideas behind the branching-time temporal
logics CTL and CTL∗ presented here. For further reading on the history of temporal rea-
soning and logics, see Øhrstrøm and Hasle (1995). In particular, for details on Prior’s views
and motivating analyses, see Prior (1967, Chapter VII) and Øhrstrøm and Hasle (1995,
Chapters 2.6 and 3.2). A broad but concise overview of temporal logics is Goranko and
Galton (2015).

1.1.2 The Role of Temporal Logics in Computer Science

Temporal aspects and phenomena are pervasive in computer and information systems.
These include: scheduling of the execution of programs by an operating system; concurrent
and reactive systems and, in particular, synchronisation of concurrent processes; real-time
processes and systems; hardware verification; temporal databases, etc. Many of these are
related to specification and verification of properties of transition systems and computations
in them. Formally, transition systems are directed graphs consisting of states and transitions
between them. They are used to model sequential and concurrent processes. There can be
different types of transitions (e.g. affected by different types of actions) which we indi-
cate by assigning different labels to them. Thus, more generally, we talk about labelled
transition systems. A state in a transition system may satisfy various properties: it can be
initial, terminal, deadlock, safe or unsafe, etc. One can describe state properties by formu-
lae of a suitable state-description language; on propositional level these are simply atomic
propositions. The set of such propositions that are true at a given state is encoded in the
label of that state, and a transition system where every state is assigned such label will be
called an interpreted transition system. In terms of the semantics of modal logics, transition
4 Introduction

systems are simply Kripke frames, and the labelling of states corresponds to a valuation of
the atomic propositions in such frames, so interpreted transition systems are just Kripke
models. A computation in a transition system is, intuitively, the observable trace of a run –
a sequence of states produced by following the transition relations in the system, viz. the
sequence of the labels of these states. It can be regarded as a record of all observable suc-
cessive intermediate results of the computing process.
Early work implicitly suggesting applications of temporal reasoning to modelling and
analysis of deterministic and stochastic transition systems is the theory of processes and
events in Rescher and Urquhart (1971, Chapter XIV). The use of temporal logic for speci-
fication and verification of important properties of reactive and concurrent systems, such as
safety, liveness and fairness, was first explicitly proposed in Amir Pnueli’s (1977) seminal
paper (see also Abrahamson 1979; Lamport 1980; Ben-Ari et al. 1981, 1983; Clarke and
Emerson 1981). In particular, Pnueli proposed and developed, together with Zohar Manna,
versions of the temporal linear-time logic LTL, in Manna and Pnueli (1979, 1981) as logi-
cal framework for deductive verification of such systems. Other influential early works on
temporal logics of programs and processes include Abrahamson (1979, 1980) and Kröger
(1987).
Since the late 1970s temporal logics have found numerous other applications in com-
puter science. The key for their success and popularity is that temporal logics are syn-
tactically simple and elegant, have natural semantics in interpreted transition systems, are
well expressive for properties of computations and – very importantly – have good com-
putational behaviour. Thus, they provide a very appropriate logical framework for formal
specification and verification of programs and properties of transition systems. Depending
on the type of systems and properties to specify and verify, two major families of temporal
logics have emerged: linear-time and branching-time logics. Manna and Pnueli (1992) is
a comprehensive reference on the early use of (linear time) temporal logics for specifica-
tion of concurrent and reactive systems, and Manna and Pnueli (1995) is its continuation
showing how that can be used to guarantee safety of such systems.
Two major developments, both starting in the 1980s, contributed strongly to the popular-
ity and success of temporal logics in computer science. The first one is the advancement of
model checking as a method for formal verification by Clarke and Emerson (1981), followed
by Clarke et al. (1983, 1986), and independently by Queille and Sifakis (1982a). Introduc-
tions to model checking can be found in Huth and Ryan (2000) and Clarke and Schlingloff
(2001), and for more comprehensive expositions, see Clarke et al. (2000) and Baier and
Katoen (2008). The second major development is the emergence of automata-based meth-
ods for verification, initially advocated in a series of papers by Streett (1982), Sistla, Vardi
and Wolper (1987), Vardi and Wolper (1986a,b) and Emerson and Sistla (1984), and fur-
ther developed in Sistla et al. (1987), Emerson and Jutla (1988), Muller et al. (1988), Streett
and Emerson (1989), Thomas (1990), Emerson and Jutla (1991), Vardi (1991), Vardi and
Wolper (1994), Vardi (1996, 1997), Wolper (2000), Kupferman et al. (2000), Löding and
Thomas (2000), etc. See also Vardi and Wilke (2007), Vardi (2007) and Grädel et al. (2002)
for broader overviews and further references.
 1.1 Temporal Logics and Computer Science: A Brief Overview 5

1.1.3 The Influence of Computer Science on the Development of Temporal Logics

The applications of temporal logics for specification and verification of computer systems
have in turn strongly stimulated the study of their expressiveness and computational com-
plexities and the development of efficient algorithmic methods for solving their basic logical
decision problems.
The first proof systems developed for temporal logics were Hilbert-style axiomatic
systems (see e.g. Rescher and Urquhart 1971; Goldblatt 1992; Reynolds 2001), as well
as tableaux-style calculi and some Gentzen-type systems (see e.g. Fitting 1983; Wolper
1985; Goré 1991; Schwendimann 1998b; Goré 1999). In particular, efficient and intuitively
appealing tableau-based methods for satisfiability testing of temporal formulae were devel-
oped in the early 1980s, by Wolper (1983, 1985) for the linear-time logic LTL, Ben-Ari,
Manna and Pnueli (1981) for the branching-time logic UB and Emerson and Halpern (1982,
1985) for the branching-time logic CTL. The increasing demand coming from the field of
formal verification for efficient algorithmic methods solving verification problems has led
to the subsequent development of new methods, based on automata (see earlier references)
and games (Stirling 1995, 1996; Stevens and Stirling 1998; Lange and Stirling 2000, 2001,
2002; Grädel 2002; Lange 2002a; see also Grädel et al. 2002 for a comprehensive cover-
age and further references). Other methods, such as temporal resolution, have also been
developed recently (see e.g. Fisher 2011).
Whereas traditional systems of logical deduction and tableaux-style calculi can be qual-
ified as direct methods, since they handle temporal formulae directly, the automata-based
and the game-theoretic approaches work by reducing the decision problems for tempo-
ral logics to decision problems about automata and games, respectively. For instance, the
automata-based approach is based on reducing logical to automata-based decision problems
in order to take advantage of known results from automata theory (D’Souza and Shankar
2012). The most standard target problems on automata used in this approach are the lan-
guage nonemptiness problem (checking whether an automaton admits at least one accepting
computation) and the language inclusion problem (checking whether the language accepted
by an automaton A is included in the language accepted by another automaton B). In a pio-
neering work, Richard Büchi (1962) introduced a class of automata on infinite words and
showed that these automata are equivalent in a precise sense to formulae in the monadic
second-order theory (MSO) of the structure (N, <), which eventually resulted in Büchi’s
proof of the decidability of that theory. Later this idea and the result were extended by Rabin
(1969) in his seminal paper to automata on trees and the decidability of the MSO theory of
the infinite binary tree (equivalently, the MSO theory of two successor functions). These
results provide the theoretical foundation of the automata-based decision methods in tem-
poral logics.
Likewise, in the game-theoretic approach, the question of whether a temporal formula
is satisfiable corresponds to the question of whether some designated player has a winning
strategy in an associated satisfiability game. More generally, the existence of winning strate-
gies of a given player in model-checking games and satisfiability-checking games provide
6 Introduction

uniform characterisations of the model checking and the satisfiability-testing problems for
many temporal logics. These game-theoretic characterisations are of particular interest in
the context of program verification.
As we demonstrate in this book, the methods for solving decision problems of temporal
logics, based on automata, tableaux and games are closely related, both conceptually and
technically, while each of them has pros and cons compared to the other. Because of the
conceptual elegance and technical power and convenience, the automata-based methods
have generally been favoured by the researchers in the area of verification and most of
the tools that have been implemented are based on automata. On the other hand, tableau-
based methods for model checking and satisfiability checking of temporal logics have so
far been less developed for practical purposes and less tested for industrial applications, but
are arguably more natural and intuitive from logical perspective, easier for execution by
humans and potentially more flexible and practically efficient, if suitably optimised.

1.2 Structure and Summary of the Book Content

With this book we aim to offer a comprehensive, uniform, technically precise and concep-
tually in-depth exposition of the field, focusing on the intrinsically logical aspects of it and
including both classical and recent results and methods of fundamental importance. The
book has been written with the specific intention to be suitable both as a comprehensive
graduate textbook and as a professional reference on the current state of the art in the field.
For that purpose, it presents an essentially self-contained and rigorous treatment of the
content, with precise definitions, statements and many detailed proofs, as well as numerous
examples and exercises. Many easy or routine proofs are omitted in the main text but are
explicitly left as exercises.
Except for the introductory chapter, we do not make references and credits to specific
results in the main text, but provide such references in the bibliographic notes at the end of
every chapter, where we also mention related topics not treated in the book and provide a
number of additional references.
The book consists of 15 chapters structured in four parts. Here we will give a concise
summary of their contents and will mention a few highlights in each chapter. More detailed
summaries can be found at the beginning of each chapter.

Part I, ‘Models’, presents the basic theory of the abstract models for the temporal logics
studied further, viz. transition systems and computations in them.

Chapter 2, ‘Preliminaries and Background I’, provides some preliminary material that
will be needed further, including basics of sets, binary relations and orders, fixpoint the-
ory, computational complexity and 2-player games on graphs. We do not intend to teach
this material here but rather to recall the most basic definitions, terminology and notation,
mainly for readers’ convenience, as a quick reference.
 1.2 Structure and Summary of the Book Content 7

Chapter 3, ‘Transition Systems’, introduces the basic concepts and facts related to transi-
tion systems, computations and important types of their properties. In particular, we provide
simple algorithms solving the basic reachability problems in transition systems. Besides, we
discuss transition systems as abstract models of the behaviour of real systems with respect
to transitions between states. A real transition system can be modelled by different abstract
transition systems and at different levels of detail. The chapter addresses the important ques-
tion of when two transition systems should be considered behaviourally equivalent, that is,
modelling essentially equivalent real transition systems. That question leads to the funda-
mental notion of bisimulation, the main versions of which are introduced and studied in the
chapter. In particular, we introduce bisimulation games and relate the existence of winning
strategies for one of the players in such games to the existence of bisimulation between the
transitions systems on which the game is played. This is the first place in the book where
games are considered and they will be one of its main themes.

Part II, ‘Logics’, is the core part of the book, presenting and studying the most impor-
tant temporal logics used for specification and verification of discrete transition systems
and many variations of them. In a relatively uniform manner we introduce the syntax and
semantics of each of these logical systems, discuss and illustrate their use for formal spec-
ification of properties of transition systems and computations, study their expressiveness
and the basic logical decision problems (model checking, satisfiability and validity test-
ing) for them, relate them to standard verification tasks and present algorithms for solving
some of these decision problems. The expositions are written from the primary perspec-
tive of computer science rather than from the perspective of pure modal and temporal log-
ics. Thus, we have emphasised the more relevant topics associated with transition systems,
such as expressiveness, bisimulation invariance, model checking, small model property and
deciding satisfiability, while other fundamental logical topics, such as deductive systems
and proof theory (except for short sections on axiomatic systems and derivations in them),
model theory, correspondence theory, algebraic semantics, duality theory, etc. are almost
left untouched here, but references are given in the bibliographic notes to other books where
they are treated in depth.
Following is a brief summary of the chapters in this part.

Chapter 4, ‘Preliminaries and Background II’, provides some common background,

terminology and notation related to logical decision problems and deductive systems. It
assumes some knowledge of propositional logic, upon which all temporal logics studied
here are built.

Chapter 5, ‘Basic Modal Logics’, is a concise introduction to the multimodal logic BML,
regarded here as the basic temporal logic for reasoning about interpreted transition systems.
Highlights on that chapter include: invariance of BML formulae under bisimulations and
characterising the existence of bounded bisimulations between interpreted transition sys-
tems with BML formulae; study of the logical decision problems of model checking and
8 Introduction

satisfiability testing for BML, model-checking games and an axiomatic system for BML.
Inter alia, we provide a simple optimal algorithm for testing satisfiability in BML that runs
in polynomial space. These are further adapted to the extension BTL of BML with past-time
operators. This chapter can also be viewed as a stepping stone towards the more expressive
and interesting branching-time temporal logics CTL, CTL∗ and the modal μ-calculus, pre-
sented further.

Chapter 6, ‘Linear-Time Temporal Logics’, is devoted to temporal logics for linear-time

models, representing single computations (i.e. infinite sequences of states generated by the
transition relations), rather than entire transition systems. Here we present and study in
detail the linear-time temporal logic LTL. We focus again on the logical and computational
properties of this logic, viz. its semantics, expressiveness, model checking and testing of
satisfiability and validity. We provide conceptually simple decision procedures for satis-
fiability testing and model-checking problems for LTL based on the so-called ultimately
periodic model property and discuss how these can be refined and transformed into optimal
decision procedures. (Alternative decision methods, essentially using the same property but
based on tableaux and automata, are developed further, respectively in Chapters 13 and 14.)
We also present and discuss here some of the most interesting extensions of LTL over linear
models: with past-time operators, automata-based operators, propositional quantification,
etc. At the end, we provide a complete axiomatic system for LTL and illustrate it with some
derivations.

Chapter 7, ‘Branching-Time Temporal Logics’, introduces and studies a variety of the

most important and expressive branching-time temporal logics extending BML with global
temporal operators capturing different reachability properties and with quantifiers over
paths/computations. These include, inter alia, the temporal logic of reachability TLR, the
more expressive computation tree logic CTL, and the full computation tree logic CTL∗
combining CTL with LTL, plus several fragments, variations and extensions of it. We estab-
lish the tree-model property for CTL∗ , extend the bisimulation invariance result to it and
discuss the expressiveness of the family of branching-time logics. In particular, we show
that formulae of TLR suffice to characterise every finite interpreted transition system up to
bisimulation equivalence. We then study the logical decision problems for CTL and CTL∗ .
We present the linear-time labelling algorithm for model checking CTL formulae and show
how model checking in CTL∗ can be reduced to repeated model checking of LTL formulae.
Lastly, we briefly present complete axiomatic systems for TLR and CTL and illustrate them
with some derivations.

Chapter 8, ‘The Modal Mu-Calculus’, presents the most expressive of all temporal logics
studied in this book. The modal μ-calculus Lμ extends BML with the fundamental syntac-
tic construct of a least fixpoint operator and its dual, the greatest fixpoint operator. Using
these, all previously studied temporal operators in the linear and branching-time logics can
be defined simply and elegantly. The chapter provides a detailed and technically involved
 1.2 Structure and Summary of the Book Content 9

exposition of the key syntactic and semantic concepts of the modal μ-calculus, including
the technical machinery needed to understand and evaluate its formulae, viz. approximants,
signatures and games. We present the embedding of all previously studied (single-action)
temporal logics, including CTL∗ , into the μ-calculus, as well as related topics such as modal
equation systems, model-checking games and results about structural complexity of formu-
lae. The chapter also offers a comparison with other logical formalisms, such as monadic
second-order logic and linear-time μ-calculus.

Chapter 9, ‘Alternating-Time Temporal Logics’, introduces concurrent game structures

as multiagent generalisation of the transition systems considered so far and the respective
multiagent extensions of the branching-time logics, known as alternating-time temporal
logics, which enable strategic reasoning in open and multiagent transition systems and have
recently been gaining increasing popularity. Formulae in the alternating-time temporal log-
ics can quantify over strategies of agents and over all computations consistent with a given
collective strategy of a coalition of agents, which is a refinement of the path quantifica-
tion in CTL∗ , thus enabling the expression of properties of the type ‘the group of agents
C has a collective strategy to achieve a given (LTL-definable) temporal objective on all
computations consistent with that strategy’. Following the general structure of the previ-
ous chapters, we present the syntax and semantics of the multiagent analogues ATL and
ATL∗ of the branching-time logics CTL and CTL∗ , discuss their expressiveness and study
their logical decision problems. In particular, we show that the linear-time labelling algo-
rithm for model checking CTL extends smoothly to ATL. We also generalise the notion of
bisimulation to alternating bisimulation and extend the bisimulation invariance property to
ATL. This chapter is relatively independent from the rest of the book, as we do not study
alternating-time temporal logics further, but it provides sufficient material for their further
study.

Part III, ‘Properties’, is dedicated to two fundamental generic questions about temporal
logics: their expressiveness and the computational complexity of their main logical decision
problems. We provide a uniform treatment by proposing a synthetic and unified approach
to both questions. These topics are also discussed in the rest of the book, but in a less
systematic way. A summary of Part III follows.

Chapter 10, ‘Expressiveness’, provides an in-depth study of the relationships and com-
parisons between the different temporal logics introduced in Part II with respect to the
temporal properties of interpreted transition systems that can, or cannot, be expressed by
formulae in a given particular logic. A way to address such questions is to look at the entire
spectrum of all temporal logics. We can naturally compare the expressiveness of two logics
by asking whether every property that is formalisable in one of them can also be formalised
in the other. This yields a preorder on all temporal logics of the same kind, i.e. with seman-
tics based on the same class of models. A temporal formula can be identified with the class
of (rooted) interpreted transition systems in which that formula is valid. Thus, a formula of
10 Introduction

one logic being expressible by a formula in another logic simply amounts to the two formu-
lae being equivalent in the usual logical sense, i.e. being valid in the same class of models.
The structure of this chapter reflects the two main kinds of results that are presented
there. The first two sections contain positive results of expressive inclusions, by means of
translations from one logic to another, where the two different families of logics – for linear
time and for branching time – are treated in separate sections. For instance, we show that
all the different major extensions of LTL introduced earlier – with automata-based oper-
ators (ETL), with propositional quantification (QLTL), the linear μ-calculus LTμ and the
industrial standard Property Specification Language (PSL) – have the same expressiveness.
With regards to branching-time logics, we present, for instance, the embedding of CTL∗
into Lμ . The last part of Chapter 10 presents negative expressiveness results of the kind that
some properties in one logic are not expressible in another. For instance, we show that LTL
is expressively weaker than its extensions mentioned earlier, pinpoint the expressive power
of CTL by separating it from the other branching-time logics and prove that higher degree
of fixpoint alternation in Lμ yields greater expressiveness.

Chapter 11, ‘Computational Complexity’, is devoted to analysing the computational

complexity of the main decision problems for temporal logics, which is of great impor-
tance for the assessment of their practical suitability as tools for formal verification. The
chapter proposes a unifying picture of the computational complexity of most of the tempo-
ral logics studied here, by characterising the complexities of their satisfiability and model-
checking problems. Complexity upper bounds for these problems are obtained by providing
respective decision procedures in the other chapters, by using either concretely described
algorithms presented there or by developing general decision methods such as semantic
tableaux, automata or games. These upper bounds are revisited in that chapter and matching
complexity lower bounds are established there, thus establishing the optimal complexities
of the respective decision problems. In particular, the chapter presents a hierarchy class
of problems, namely tiling problems and their variants involving games, thus providing a
uniform and powerful framework for obtaining lower bound complexity results.

In the last part, Part IV, ‘Methods’, we present three fundamental methods for solving
decision problems for temporal logics, namely: the tableaux-based, the games-based and
the automata-based methods, by devoting a chapter to each of these. These technical frame-
works are closely related and we discuss briefly their relationships and compare their pros
and cons in the brief opening Chapter 12, ‘Frameworks for Decision Procedures’, pro-
vided to help the understanding of the bridges between these methods. Here is a brief sum-
mary of the other chapters in this part.

Chapter 13, ‘Tableaux-Based Decision Methods’, provides a systematic and uniform

exposition of (a suitable adaptation) of the method of semantic tableaux, for constructive
satisfiability testing and model building for formulae of the logics BML, LTL, TLR and
CTL, by organising a systematic search for a satisfying model of the input set of formulae.
 1.2 Structure and Summary of the Book Content 11

We begin the chapter by sketching a generic tableaux construction for testing satisfiability,
illustrated with the basic modal logic BML. This construction builds on, and optimises
further, a blend of ideas going back to the incremental tableaux method introduced for
the propositional dynamic logic PDL in Pratt (1979, 1980) and further developed for LTL
by Wolper (1983, 1985) and for CTL by Emerson and Halpern (1985). That version of the
tableaux method differs essentially from the more traditional approach to tableaux-based
calculi presented in Fitting (1983) and Goré (1999), because of the need for special treat-
ment of the fixpoint definable temporal operators. Following the generic construction we
develop and illustrate with running examples the tableau constructions for the logics LTL
and CTL and present in detail proofs of their termination, soundness and completeness.
The most distinctive features of the tableaux building methodology presented in Chapter 13
are uniformity, conceptual simplicity and flexibility. That methodology, however, does not
extend readily to the more expressive logics CTL∗ and the Lμ , for which alternative deci-
sion methods, using automata and games, have been developed. Such alternative methods
are developed, also for LTL and CTL, respectively in the two subsequent chapters.

In Chapter 14, ‘The Automata-Based Approach’, we present in detail that approach for
a variety of temporal logics and classes of automata and compare the advantages and the
drawbacks of the different constructions. The automata-based approach consists in reduc-
ing logical decision problems to decision problems about automata and their languages and
takes advantage of known results and decision procedures from automata theory. Here tem-
poral logical formulae are represented by automata and their models are considered to be
of uniform nature, viz. infinite words (representing infinite computations, for linear-time
logics) or trees (representing tree-unfoldings of transition systems, for the branching-time
logics). Thus, the question of truth of a temporal formula in a model is equivalently replaced
by the question of acceptance of that model by the automaton representing the formula.
Respectively, the satisfiability problem for the formula is reduced to the language nonempti-
ness problem of its representing automaton. Based on this generic transformation, in this
chapter we develop the automata-based approach for solving logical decision problems for
several logics, mainly LTL and its variants (ETL, PSL) as well as the branching-time tem-
poral logic CTL (and its fragment BML) for which we present several reductions from
formulae to automata. To do so, we make use of notions from previous chapters but also we
provide some original reductions, e.g. using alternating automata, thus also illustrating the
diversity of automata-based methods. Moreover, we present several methods for LTL-like
logics, as well as analogous methods for CTL, as a generalisation of some of the methods for
LTL. The chapter does not deal with CTL∗ and with modal μ-calculus. Instead, we provide
decision methods based on verification games for those two logics in the next Chapter 15.

Chapter 15, ‘The Game-Theoretic Framework’, develops the approach to the main
decision problems for temporal logics based on special logical games. We present model-
checking games and satisfiability-checking games and use them to characterise the respec-
tive decision problems for temporal logics. These games are defined by means of an arena
12 Introduction

Figure 1.1 Dependencies between the different parts and chapters.

on which two players perform plays. Then we relate the model-checking problem – i.e. the
question of whether a given state of an interpreted transition system satisfies a temporal
formula – to the existence of a winning strategy for the proponent player in the respective
model-checking game. Likewise, the satisfiability problem corresponds to the question of
whether the proponent player has a winning strategy in the respective satisfiability game.
The chapter is divided into four sections. The first one builds the general game-theoretic
framework and concludes with the Main Reduction Theorem which shows how to obtain
reductions to parity games. In the next section we provide the necessary automata-theoretic
constructions for such reductions. The last two sections contain the actual definitions of
games for temporal logics.

1.3 Using the Book for Teaching or Self-Study

The book can be used for a variety of introductory and advanced courses. An introductory
course could cover most of Parts I and II (perhaps, excluding the chapter on the modal
μ-calculus and more advanced topics from the previous chapters). More advanced courses
could add quite independently parts from the chapters in Parts III and IV, see the depen-
dencies between the chapters in the book illustrated in Figure 1.1. The leftmost column
indicates the parts, and each chapter is identified by a keyword and by its number. Arrows
specify dependencies between the parts/chapters where a normal arrow indicates that the
chapter at the arrow’s source should be read before the chapter at the arrow’s target. A
dashed arrow means that it is helpful to read it first. The preliminary chapters 2 and 4 are
not represented in the figure; their content may be needed in the entire book.

Acknowledgements
First of all, we are indebted to many researchers who have contributed over the years to the
development of the area of temporal logics in computer science and have thus produced
 1.3 Using the Book for Teaching or Self-Study 13

much of the content of the book. We have provided detailed references at the end of each
chapter, but for such a huge and dynamic field of research omissions are inevitable, so we
also apologise here en bloc for all – certainly unintended, but nevertheless unfortunate –
omissions to give due credit.
We are grateful to many colleagues who have contributed in one way or another to our
work on the book, from proofreading preliminary parts to discussing its content or related
topics over the years. In particular, we thank Florian Bruse, Paul Gastin, Markus Latte,
Omer Mermelstein, Philippe Schnoebelen and Steen Vester. We also thank Ivan Danov for
the help with the design of the book’s website.
Valentin Goranko and Martin Lange have both been supported by visiting professorships
from the École Normale Supérieure de Cachan in France. Much of the work that has gone
into this book was done during these visits.
Special thanks are due to the representatives of Cambridge University Press: David
Tranah and Clare Dennison for their infinite patience and understanding during the mul-
tiple delays and extensions of the target dates for the completion of this book, as well as
their continual help and support.
Last but not least, we wish to thank our families for the understanding and moral support
during the many years of work on the book. In addition, Valentin is particularly grateful to
his partner Nina for the valuable technical support with the preparation of many figures.
Part I
Models
 2
Preliminaries and Background I

2.1 Sets and Relations 18

2.1.1 Operations on Sets 18
2.1.2 Binary Relations 18
2.1.3 Partial and Linear Orders 21
2.2 Some Fundamental Preliminaries 25
2.2.1 Fixpoints in Powerset Lattices 25
2.2.2 Complexity Classes 26
2.2.3 Two-Player Games 28

This chapter presents preliminaries on set-theoretical notions, binary relations, linear order-
ings, fixpoint theory and computational complexity classes. Mainly, we provide notations
for standard notions rather than giving a thorough introduction to these notions. More def-
initions are provided in the book, and we invite the reader to consult textbooks on these
subjects for further information. For instance, in Moschovakis (2006) any reader can find
material about set-theoretical notions, ordinals or fixpoints far beyond what is sketched in
this chapter. Still, we implicitly assume that the reader has basic set-theoretic background.
As stated already in Chapter 1, we do not intend to teach this material here but rather to
recall the most basic notions, terminology and notation. The current chapter is included for
the convenience of the reader as a quick reference.

Structure of the chapter. The chapter is divided into two sections. The first section con-
tains standard material on sets and relations. Section 2.1.1 presents standard set-theoretical
notions that are used throughout the book. Binary relations are ubiquitous structures in this
volume, and Section 2.1.2 is dedicated to standard definitions about them. In Section 2.1.3,
we provide basic definitions about partial and linear orders.
The second section contains material that is more specialised and needed for the devel-
opment of a theory of and algorithms for temporal logics. Section 2.2.1 presents the basics
of fixpoint theory; Chapter 8, which deals with the modal μ-calculus with fixpoint opera-
tors, uses some of the results stated herein. In Section 2.2.2, we recall standard complex-
ity classes defined via deterministic and nondeterministic time- and space-bounded Turing
machines. Other classes, in particular involving alternating Turing machines, are discussed
18 Preliminaries and Background I

in Chapter 11. Section 2.2.3 provides an introduction to 2-player zero-sum games of per-
fect information that are useful, for instance, in defining the game-theoretic approach to
temporal logics.

2.1 Sets and Relations

2.1.1 Operations on Sets
Throughout this book we use the standard notations for set-theoretical notions: membership
(∈), inclusion (⊆), strict (or proper) inclusion (⊂), union of sets (∪), intersection of sets (∩),
difference of sets (\) and product of sets (×). The empty set is denoted by ∅.
By X n we denote the product X × · · · × X with n factors. A disjoint union of the sets
X and Y is the set X Y = (X × {1}) ∪ (Y × {2}). The union and the intersection of an
 
indexed family (Xi )i∈I of sets are denoted by i∈I Xi and i∈I Xi , respectively.
N, Z, R and Q denote the sets of all natural numbers, integers, real numbers and rational
numbers, respectively. By P (X ) we denote the powerset of the set X, which is the family
of all subsets of X. By card(X ) we denote the cardinality of the set X.
A family (Xi )i∈I of subsets of Y is

r a cover of Y iff i∈I Xi = Y and
r a partition of Y iff it is a cover of Y such that for all i, j ∈ I, we have that i = j implies
Xi ∩ X j = ∅.

2.1.2 Binary Relations

A binary relation between sets X and Y is any subset of X × Y . In particular, a binary
relation in a set X is any subset of X 2 = X × X.
Let R ⊆ X × Y . Some notation: if (x, y) ∈ R, we will also write xRy or R(x, y) and say
that x is R-related to y.
Given sets X and Y and R ⊆ X × Y we define the domain of R,

dom(R) := {x ∈ X | there is y ∈ Y such that (xRy)},

and the range of R

ran(R) := {y ∈ Y | there is x ∈ X such that (xRy)}.

We recall some special relations on a set X:

r the empty relation ∅,
r the equality relation (also known as the identity relation or the diagonal relation):
EX := {(x, x) | x ∈ X} and
r the universal relation X 2 .

Since binary relations are sets themselves, all Boolean operations ∪, ∩ and \ apply to
them, too. Also, the complement of a relation R ⊆ X × Y is defined as R := (X × Y ) \ R.
 2.1 Sets and Relations 19

Note that we can always assume that two relations are defined between the same sets,
because if R1 ⊆ X1 × Y1 and R2 ⊆ X2 × Y2 , we can assume that R1 ⊆ (X1 ∪ X2 ) × (Y1 ∪ Y2 )
and R2 ⊆ (X1 ∪ X2 ) × (Y1 ∪ Y2 ).

Inverse and Composition

Let R ⊆ X × Y . The inverse of R is the relation R−1 ⊆ Y × X defined as
R−1 := {(y, x) | (y, x) ∈ R}.
Thus, we have xRy iff yR−1 x.
Let R ⊆ X × Y and S ⊆ Y × Z. The composition of R and S is the binary relation R ◦
S ⊆ X × Z defined by
R ◦ S := {(x, z) | there is y ∈ Y such that (xRy and ySz)}.
In particular, when R ⊆ X 2 , then R ◦ R is defined, and it is denoted by R2 .
The composition of binary relations is associative, i.e. (R ◦ S) ◦ T = R ◦ (S ◦ T ). How-
ever, the composition of binary relations is generally not commutative.
For any binary relations R, S, we have (R ◦ S)−1 = S−1 ◦ R−1 .

Some Special Binary Relations

A binary relation R ⊆ X is called
2

r reflexive if, for all x ∈ X, we have (x, x) ∈ R;

r irreflexive if, for all x ∈ X, we have (x, x) ∈ R;
r serial, or total, if, for all x ∈ X, there is y ∈ X such that (x, y) ∈ R;
r functional or deterministic if, for all x ∈ X, there is a unique y ∈ X such that (x, y) ∈ R;
r symmetric if, for all x, y ∈ X, we have (x, y) ∈ R iff (y, x) ∈ R;
r antisymmetric if, for all x, y ∈ X, we have that if (x, y) ∈ R and (y, x) ∈ R, then x = y;
r connected if, for all x, y ∈ X, either (x, y) ∈ R or (y, x) ∈ R;
r transitive if, for all x, y, z ∈ X, if (x, y) ∈ R and (y, z) ∈ R, then (x, z) ∈ R;
r equivalence relation if it is reflexive, symmetric and transitive;
r pre-order (or quasi-order) if it is reflexive and transitive;
r partial order if it is reflexive, transitive and antisymmetric, i.e. an antisymmetric
preorder;
r strict partial order, if it is irreflexive and transitive;
r linear order (or total order) if it is a connected partial order;
r strict linear (or total) order, if it is a connected strict partial order.

Example 2.1.1.
1. The divisibility relation in Z \ {0} and the relation ‘not longer than’ in the set of English
words are pre-orders, but not partial orders.
2. The divisibility relation in N \ {0} and the relation ⊆ in any power set P (X ) are partial
orders, but not linear orders, unless X has at most one element.
3. The relation ≤ in any of N, Z, Q, R is a linear order.
4. The relation < in any of N, Z, Q, R is a strict linear order.
20 Preliminaries and Background I

Many of the special binary relations listed can be characterised in a purely relational
language, without referring to the elements of their domains and ranges. For any set X and
binary relation R ⊆ X 2 , the following hold:

1. R is reflexive iff EX ⊆ R.
2. R is symmetric iff R−1 ⊆ R iff R−1 = R.
3. R is antisymmetric iff R−1 ∩ R ⊆ EX .
4. R is connected iff R ∪ R−1 = X 2 .
5. R is transitive iff R2 ⊆ R.

The graph of a function (mapping) f : X → Y is the binary relation Gf ⊆ X × Y defined

by Gf := {(x, f(x)) | x ∈ X}.
A relation R ⊆ X × Y is functional iff it is the graph of a function from X to Y .

Closure Operations
Recall that R denotes the binary relation obtained from the composition R ◦ R. More gen-
2

erally, let us define Ri for all i ∈ N, where R is a binary relation in X × X, as follows:

r R0 := EX , i.e. R0 is the diagonal relation on X and
r Ri+1 := (Ri ◦ R) for all i ∈ N.

We define the transitive closure of the binary relation R as follows:

∞

R+ := Ri .
i=1

We also define the reflexive and transitive closure R∗ of R as

∞

∗
R := Ri = R0 ∪ R+ .
i=0

It is easy to show that, indeed, R is transitive, while R∗ is reflexive and transitive.

+

Section 3.2 contains generalisations of this definition by considering families of binary

relations, as well as other parameters.

Equivalence Relations
Recall that an equivalence relation in a set X is a reflexive, symmetric and transitive relation
in X. For instance, the relations of equality (=) and congruence modulo n (≡n ) in Z are
equivalence relations. (Recall: m ≡n m for m, m ∈ Z iff n divides m − m .)
Let R ⊆ X 2 be an equivalence relation and x ∈ X. The subset of X: [x]R := {y ∈ X | xRy}
is called the equivalence class or the cluster of x generated by R.
For every equivalence relation R in a set X and x, y ∈ X the following hold:

1. x ∈ [x]R .
2. x ∈ [y]R implies [x]R = [y]R .
3. x ∈ [y]R implies [x]R ∩ [y]R = ∅.
 2.1 Sets and Relations 21

Let R ⊆ X 2 be an equivalence relation. The set X/R := {[x]R | x ∈ X} is called the

quotient-set of X generated by R.
The element [x]R , also denoted by x/R , is called the quotient-element of x generated
by R. The mapping ηR : X → XR defined by ηR (x) := x/R is called the canonical mapping
of X onto X/R .
Note that the canonical mapping ηR is surjective, i.e. for every y ∈ XR , there is some
x ∈ X such that ηR (x) = y.
Let f : X → Y . Then the following three statements hold:

1. The binary relation f̃ in X defined by

xf̃y iff f(x) = f(y)

is an equivalence relation in X, called the kernel equivalence of f.

2. The mapping g : X/f̃ −→ Y defined by

g(x/f̃ ) := f(x)

is a well-defined injection, such that f = gηf̃ .

To see this, let x/f̃ = y/f̃ . Then xf̃y, hence f(x) = f(y), i.e. g(x/f̃ ) = g(y/f̃ ), so g is
well defined. Conversely, if g(x/f̃ ) = g(y/f̃ ), then f(x) = f(y), i.e. xf̃y, hence x/f̃ = y/f̃ ,
so g is an injection. Finally, (gηf̃ )(x) = g(ηf̃ (x)) = g(x/f̃ ) = f(x).
3. Moreover, if f is surjective, then g is bijective. This follows immediately from the pre-
vious point.

These statements show, inter alia, that every mapping can be represented as a composi-
tion of a surjective mapping followed by an injective mapping.

Partitions
A partition of a set X is any family P of nonempty and pairwise disjoint subsets of X, the
union of which is X.
For every equivalence relation R ⊆ X 2 , the quotient-set {[x]R | x ∈ X} is a partition of
X.
If P is a partition of a set X, then the relation ∼P ⊆ X 2 defined by

x ∼P y iff x and y belong to the same member of P

is an equivalence relation in X. Thus, equivalence relations in X and partitions of X are two

faces of the same coin.

2.1.3 Partial and Linear Orders

Pre-Orders and Partial Orders
Recall that a pre-order is a reflexive and transitive relation, while a partial order is an anti-
symmetric pre-order. We consider a typical application of the quotient construction. Let R
be a pre-order in X. Then the following hold:
22 Preliminaries and Background I

1. The relation ∼ in X defined by

x ∼ y iff xRy and yRx

is an equivalence relation in X.
2. The relation R̃ in X/∼ , defined by

x/∼ R̃ y/∼ iff xRy,

is a well-defined partial order in X (called the partial order induced by R).

Let ≤ be a (fixed) partial order in the set X. Then X is called a partially ordered set,
abbreviated as poset. We will use standard notation, in particular x ≥ y for y ≤ x; x < y for
x ≤ y and x = y; and x > y for y < x.
Note that every subset Y of a poset X is a poset with respect to the restriction of the
partial order in X to Y .

Extremal Elements
Let X be a poset and Y ⊆ X. A lower (resp. upper) bound for Y in X is any x ∈ X such
that x ≤ y (resp. x ≥ y) for every y ∈ Y . The least of all upper bounds of Y , if it exists, is
also called the supremum of Y ; the greatest of all lower bounds of Y , if it exists, is called
the infimum of Y .
If a lower (resp. upper) bound of Y belongs to Y , it is called a least (resp. greatest)
element of Y .
Note the following:
r Neither infimum nor supremum of Y need to exist, but if one exists, it must be unique.
r If a least (greatest) element of Y exists, it is the infimum (supremum) of Y and therefore
is unique.
r However, the infimum (supremum) of Y need not belong to Y and therefore need not be
a least (resp. greatest) element of it.

An element x ∈ Y is called minimal in Y if there is no element of Y strictly less than x,

i.e. for every y ∈ Y , if y ≤ x, then x = y. It is called maximal in Y if there is no element of
Y strictly greater than x, i.e. for every y ∈ Y , if y ≥ x, then x = y.
Note that the least (resp. greatest) element of Y , if it exists, is minimal (resp. maximal).

Example 2.1.2. The poset N with respect to divisibility has a least element, viz. 1, but no
maximal (hence no greatest) elements.
However, the poset {2, 3, 4, . . .} has no least element with respect to divisibility, while
it has infinitely many minimal elements, viz. all primes.

Trees
A tree is defined as a pair Tr = (X, R), where X is a nonempty set of nodes and R is a
binary relation on X such that there is a designated element r of X (the root) such that
 2.1 Sets and Relations 23

Figure 2.1 A tree with 12 branches.

R−1 (r) = ∅, R∗ (r) = X (i.e. every element of X is reachable by an R-path from r) and for
every x ∈ X \ {r}, R−1 (x) is a singleton set.
If (x, x ) ∈ R, then we say that x is the parent of x and x is a child of x. If (x, x ) ∈ R+ ,
then we say that x is an ancestor of x . The elements of X are called the nodes of the tree
Tr = (X, R).
A branch Br is a subset of X such that the restriction of (X, R) to Br, i.e. (Br, R ∩ Br ×
Br), is a total order and that for any proper extension Br ⊂ Y , (Y, R ∩ Y × Y ) is not a total
order. Depending on the ‘order type’ of (Br, R ∩ Br × Br), a branch can be represented by a
maximal sequence x0 , x1 , . . . such that x0 = r, and for every i ≥ 1, (xi , xi+1 ) ∈ R. Figure 2.1
contains a tree with 12 branches.
The level of x ∈ X, denoted by |x|, is the distance from the root, i.e. the number of R-
transitions from the root to x. In particular, |r| = 0. Equivalently, |x| = n iff there is a branch
Br = x0 , x1 , . . . , xn , . . . such that x = xn . The nodes x of X such that R(x) = ∅ are called
the leaves.
A tree Tr = (X, R) over N is a tree such that X is a subset of N∗ (set of finite sequences
of natural numbers), and if x · i ∈ X, where x ∈ N∗ and i ∈ N, then (x, x · i) ∈ R and
x · (i − 1) ∈ X if i > 1.
Let  be a (finite) alphabet. A -labelled tree is a pair (Tr, f) such that Tr = (X, R)
is a tree and f : X →  is a map. We often refer to f as a labelled tree, leaving its domain
implicit.

Well-Ordered Sets
An ascending (resp. strictly ascending) chain in a poset (X, ≤) is any, finite or infinite,
sequence x1 ≤ x2 ≤ . . . (resp. x1 < x2 < . . .) of elements of X. A descending (resp. strictly
descending) chain is defined analogously.
A poset is called well founded if it contains no infinite strictly descending chains.
24 Preliminaries and Background I

Every finite poset is well founded. Similarly, the linearly ordered set (N, <) is well
founded. The poset (P (X ), ⊆), where X is any infinite set, is not well founded.
A well-founded linear order is called well-order (or complete order).

Example 2.1.3. (N, ≤) is a well-order, while (Z, ≤), (Q, ≤) and (R, ≤) are not. The lex-
icographic order in N2 defined by (x1 , y1 ) ≤ (x2 , y2 ) iff x1 < x2 or (x1 = x2 and y1 ≤ y2 )
is a well-order in N2 .

A poset (X, <) is well founded iff every nonempty subset of X has a minimal element.
In particular, (X, <) is a well-order iff every nonempty subset of X has a least element.

Theorem 2.1.4. Let (X, ≤) be a well-founded set and Y ⊆ X be such that for every x ∈ X,
if all elements of X less than x belong to Y , then x itself belongs to Y . Then Y = X. 
Proof. Assume the contrary, i.e. X \ Y = ∅. Then X \ Y has a minimal element x. Then
all elements of X less than x belong to Y , hence x must belong to Y which is a
contradiction. 

Order Types and Ordinal Numbers

Two ordered sets (X, ≤) and (Y, ≤ ) have the same order type when the sets are order
isomorphic. This means that there is a bijection f : X → Y such that
r for all x, y ∈ X, x ≤ y implies f(x) ≤ f(y) and
r for all x , y ∈ Y , x ≤ y implies f−1 (x ) ≤ f−1 (y ).

Ordinals are defined as equivalence classes of isomorphic well-ordered sets; so, when
(X, ≤) is a well-ordered set, its order type can be understood as the unique ordinal isomor-
phic to (X, ≤); see the following basics on ordinals (see also Rosenstein 1982; Moschovakis
2006). Ordinals can be defined alternatively by considering a well-ordered set for each ordi-
nal that canonically represents the equivalence class of well-ordered sets corresponding to
the ordinal. So, the ordinals can be more conveniently defined inductively by the following:
r The empty set (in that context, usually written 0) is an ordinal.
r If α is an ordinal, then α ∪ {α} (in that context, usually written α + 1) is an ordinal.

r If X is a set of ordinals, then α∈X α is an ordinal.

The ordering is defined by β < α iff β ∈ α. An ordinal α is a successor ordinal iff there
exists an ordinal β such that α = β + 1. An ordinal which is not 0 or a successor ordinal is
a limit ordinal. The first limit ordinal is written ω.
Addition, multiplication and exponentiation of ordinals can be defined inductively. Here
is the definition for addition:
r α + 0 = α,
r α + (β + 1) = (α + β ) + 1 and
r α + β = sup{α + γ : γ < β}, where β is a limit ordinal. Note that here, sup corresponds
to set union. Limit ordinals are sometimes denoted by κ.
 2.2 Some Fundamental Preliminaries 25

Multiplication and exponentiation are defined similarly. The principle behind the definition
of addition is transfinite induction, which is an extension of induction to well-ordered sets.
In a nutshell, let p(α) be a property defined for all ordinals α. Suppose that if p(β ) for all
the ordinals β < α, then p(α) holds true. Transfinite induction guarantees that the property
p(α) is true for all the ordinals α. Such a principle can also be used to define operations
on ordinals, as done for addition, for example. Indeed, the class of ordinals is known to be
well ordered, and therefore Theorem 2.1.4 applies to it. In particular, any nonempty set of
ordinals has a least element. Usually, proving p(α) for all α can be divided into three cases:
(1) prove that p(0) holds; (2) prove that for all ordinals α, if p(α) holds, then p(α + 1)
holds; (3) prove that for limit ordinal κ, if, for all α < κ, we have p(α) holds, then p(κ )
holds. Similar cases appear for defining objects/sets/operations indexed by ordinals.
An ordinal that cannot be put in one-to-one correspondence with any smaller ordinal is
called a cardinal. Furthermore, assuming the Axiom of Choice, every set X can be put in a
one-to-one correspondence with an ordinal. The least ordinal for which there is a one-to-one
correspondence with X is a cardinal, called the cardinality of X and denoted card(X ).

2.2 Some Fundamental Preliminaries

2.2.1 Fixpoints in Powerset Lattices
Given a set X, a mapping F : P (X ) → P (X ) is called an operator on X. An operator F on
X can be iterated upwards α times, denoted F α , for any ordinal α. The upwards iteration
F α is defined by transfinite induction on α as follows (Y ⊆ X):
r F 0 (Y ) := Y ,
r F α+1 (Y ) := F (F α (Y )) and

r F α (Y ) = β<α F β (Y ) for limit ordinals α.

Likewise, F can be iterated downwards α times, denoted Fα , for any ordinal α, and the
iteration Fα is defined by transfinite induction on α as follows:
r F0 (Y ) := Y ,
r Fα+1 (Y ) := F (Fα (Y )) and

r Fα (Y ) = β<α Fβ (Y ) for limit ordinals α.

Definition 2.2.1. An operator F : P (X ) → P (X ) is monotone if, for all Y1 , Y2 ⊆ X, if

Y1 ⊆ Y2 then F (Y1 ) ⊆ F (Y2 ). ∇

Definition 2.2.2. Given an operator F : P (X ) → P (X ), a set Y ⊆ X is

r a fixpoint of F if F (Y ) = Y ,
r a pre-fixpoint of F if F (Y ) ⊆ Y ,
r a post-fixpoint of F if F (Y ) ⊇ Y ,
r a least fixpoint if Y is a fixpoint and Y ⊆ Z for every fixpoint Z,
r a least pre-fixpoint if Y is a pre-fixpoint and Y ⊆ Z for every pre-fixpoint Z,
Exploring the Variety of Random
Documents with Different Content
Thus educational systems become the chief enemies of education,
and seats of learning the chief obstacles to the growth of knowledge,
while in an otherwise stagnant or decadent society these tendencies
sooner or later get the upper hand and utterly corrupt the social
memory. The power of the professor is revealed not so much by the
things he teaches, as by the things he fails or refuses to teach.
History is full of examples. How many religions have not perished
from ritual sclerosis, how many sciences have not been degraded
into pseudo-sciences or games! Logic has been just examinable
nonsense for over two thousand years. The present economic chaos
in the world has been indirectly brought about by the policy adopted
by the professors of economics forty or fifty years ago, to suit their
own convenience. For they then decided that they must escape from
the unwelcome attentions of the public by becoming more ‘scientific’;
i.e. they ceased to express themselves in plain language and took to
mathematical formulae and curves instead; with the result that the
world promptly relapsed into its primitive depths of economic
ignorance. So soon as the professors had retired from it, every
economic heresy and delusion, which had been exposed and
uprooted by Adam Smith, at once revived and flourished. In one
generation economics disappeared completely from the public ken
and the political world, and the makers of the Peace Treaties of 1919
were so incapable of understanding an economic argument that not
even the lucid intelligence of Mr Keynes could dissuade them from
enacting the preposterous conditions which rendered impossible the
realization of their aims.[A] Nor was it so very long ago that, in order
to save the Mathematical Tripos at Cambridge, it had to be recast,
because it had degenerated into an intellectual jig-saw puzzle,
wholly unrelated to the applications of mathematics to the other
sciences. To avoid jealousies, I hasten to add that the University of
Oxford, which has organized itself as an asylum for lost causes,
skilfully cultivates, by means of its classical and historical studies, a
backward-looking bias in its alumni. The true ‘Greats’ man is meant
to go down indelibly imbued with the conviction that in matters of
morals and politics nothing of importance has been discovered or
said since Plato and Aristotle, and that nothing else matters.
Clearly then we cannot take for granted that in any society
knowledge can progress without limits, nor can we count on our
academic institutions to save us from stagnation and decay, even in
matters of knowledge. All institutions are social mechanisms, and all
mechanisms need a modicum of intelligent supervision, in the
absence of which they become dangerous engines of destruction.
 IV
It appears then that we can extract no guarantee of progress either
from the nature of Man or from the nature of human institutions.
There is no law of progress, if by law be meant a superior power
able to coerce the creatures that are said to ‘obey’ it. Neither can we
extract from history any proof of the superiority of civilized man over
his uncivilized ancestors. Such progress as has been attained has
been achieved only by the active co-operation of the progressive
organisms: every step has been fought for, and progress has ceased
whenever effort ceased, or was switched off into different directions.
Consequently, modern man has no right to ‘boast himself far better
than his fathers’—in intrinsic quality. Intrinsically, i.e. apart from the
effects of culture and social training, it is probable that he is slightly
inferior in capacity to his own ancestors, while very markedly inferior
to the great races of antiquity (like the Greeks) in their hey-day. Nor
is there any reason to suppose that his moral nature has changed
materially. Modern man may be a little tamer and better-tempered,
because he has been herded together much more closely than
primitive man, and city life, even in slums, demands, and produces,
a certain ‘urbanity.’ For many generations those who would not pack
tight and could not stand the strain of constantly exhibiting ‘company
manners’ and accommodating their action to those of their fellows,
must have fled away into the wilds, where they could be
independent, or have eliminated themselves in other ways, e.g. by
committing murder. It is probable that the social history of Iceland,
settled as it was by unbridled individualists who would not brook any
form of organized government, might throw some light on this
process of taming the individual.
Nevertheless there is little doubt that, in the main, humanity is still
Yahoo-manity. Alike in mentality and in moral, modern man is still
substantially identical with his palæolithic ancestors. He is still the
irrational, impulsive, emotional, foolish, destructive, cruel, credulous,
creature he always was. Normally the Yahoo in him is kept under
control by the constant pressure of a variety of social institutions; but
let anything upset an established social order, and the Yahoo comes
to the front at once. The history of the past fifty years abundantly
proves that man is still capable of atrocities equal to any in his
record. Not only have we lived through the greatest political and the
deadliest natural convulsion, the Great War and the Tokio
earthquake, but the Russian Revolution has outdone the French and
Landru the legendary Bluebeard, while for mingled atrocity and
baseness the murders of Rasputin and of Alexander of Serbia are
unsurpassed in history. The painful truth is that civilization has not
improved Man’s moral nature. His moral habits are still mainly
matters of custom, and the effect of moral theories is nugatory
everywhere. Thus civilization is not even skin deep; it does not go
deeper than the clothes.
 V
Clearly it is risky to expose the inelastic nature of so stubbornly
conservative a creature to new conditions at a rapid rate. He may not
be able to adapt himself quickly enough, and his old reactions, which
did little or no harm before, may become extremely dangerous. Yet
this is just what has happened. Science has exposed the palæolithic
savage masquerading in modern garb to a series of physical and
mental shocks which have endangered his equilibrium. It has also
enormously extended his power and armed him with a variety of
delicate and penetrating instruments which have often proved edge
tools in his hands and which the utmost wisdom could hardly be
trusted to use aright. Under these conditions the fighting instinct
ceases to be an antiquated foible, like the hunting instinct, and
becomes a deadly danger. No wonder the more prescient are
dismayed at the prospect of the old savage passions running amok
in the full panoply of civilization!
 VI
Nor is this the final item in our tale of woe. A third and most sinister
fact which has to be faced is that Civilization, as at present
constituted, is very definitely a deteriorating agency, conducing to the
degeneration of mankind. This effect of Civilization is nothing new,
but has been operating, it would seem, from the beginning, though
not probably as intensively as now: its discovery, however, is very
recent. It is quite indirect, unintended, and fortuitous, but cumulative,
and in the long run has probably been a chief cause in the decay of
States and civilizations, as well as an important factor in the arrest of
biological development which we have had to recognize.
A simple and easily observable sociological fact is at the bottom of
the mischief. The different classes in a society have different birth-
rates and death-rates, and the differences between these yield their
several net rates of increase or decrease. Now, whereas under the
conditions of savage life class differences can hardly exist, or, at
least cannot be accentuated, so that the whole tribe flourishes or
perishes together, and among barbarians the upper classes have a
very great advantage and the tribe recruits itself chiefly from the
children of the chiefs, because the conditions of life are so severe
that the lower classes are not able to rear many children; in civilized
societies these conditions are reversed. It is found that though both
birth-rates and death-rates grow as we descend the social scale, so
does the net rate of increase. Indeed, the highest or ruling class
nowhere appears to keep up its numbers without considerable
recruitment from below. So society, as at present organized, is
always dying off at the top, and proliferating at the bottom, of the
social pyramid.
The disastrous consequences of this sort of social organization may
easily be apprehended, with a little reflection. (1) All societies, even
those whose social structure is most rigid, have need of ability,
discover it, and reward it by social promotion. But (2) as this
promotion means passing into a class with a relatively inadequate
rate of reproduction, the biological penalty attaching to social
promotion is racial extinction. Thus (3) the ultimate reward of merit is
sterilization, and society appears to be an organization devoted to
the suicidal task of extirpating any ability it may chance to contain, by
draining it away from any stratum in which it may occur, promoting it
into the highest, and there destroying it. It is exactly as though a
dairyman should set in motion apparatus for separating the cream
from the milk, and then, as it rose, skim it off, and throw it away!
At present it is calculated that the highest classes in the chief
civilized societies only reproduce themselves to the extent of fifty per
cent. of their number in each generation, so that the hereditary ability
of half of them is lost in each generation. But even then the
remainder is largely wasted. It is churned into froth and scum by
social forces. For neither now nor at any time has social intelligence
shown itself equal to devising a training for the youth of the highest
classes that would provide them with adequate stimuli to develop
their faculties, and to lead a strenuous life of social service. The
children of the rich are tempted to live for ‘society’ in the narrower
sense, which means frittering away one’s life on a round of vacuous
amusement; and they rarely resist the temptation.
Naturally it is difficult to trace the accumulation of ability in the upper
social strata which is theoretically to be expected. On the other hand,
in some subjects at any rate, the symptoms of a world-wide dearth of
ability are becoming unmistakable. The Great War, though it made
abundantly manifest the prevalence of incompetents in high places,
did not reveal the existence either of a great general or of a great
statesman anywhere.
It is superfluous to insist either on the fatuity of a social organization
such as this, or on the certainty of racial degeneration which it
entails: but it may be well to draw attention to the rapidity with which
these degenerative processes are at present sapping the vitality and
value of our civilized races. The failure to reproduce does not, as in
former times, affect merely the aristocracy in the highest social
strata; it has spread to the whole of the professional and middle
classes, and to most classes of skilled labour. It is not too much to
say that, with the exception of the miners, none of the desirable
elements in the nation are doing their bit to keep up the population,
and that its continued growth is mainly due to the unrestrained
breeding of the casual labourers and the feeble-minded.
In the rest of the population its increase is checked by birth-control
and the postponement of marriage, neither of which affects the
undesirables. They are too stupid, reckless, and ignorant to practise
the former, and have nothing to gain by the latter. Also, to make it
quite certain that they shall form a true ‘proletariate,’ the wisdom of
our rulers ordains that a knowledge of birth-control shall be a (fatal)
privilege reserved for the intelligent and well-to-do. They instruct the
police to prevent it from penetrating to the poor and stupid—
apparently from the mistaken idea that the State needs plenty of
cheap labour and cheap cannon-fodder. So child-bearing remains
compulsory for the wretched women of the poor, whereas elsewhere
only those women produce children who desire them, and natural
selection is thus allowed gradually to eliminate the temperament of
the unwilling (and, therefore, probably less competent) mother.
The dysgenic effects of this class-discrimination are further
intensified by other tendencies: (1) The advance of medicine and
hygiene has enormously diminished selective mortality in all classes,
and improved the chances of weaklings to survive and leave
descendants. (2) The advance of philanthropy preserves them,
especially in the lower classes, where formerly the mortality was
largely selective and a high death-rate both counteracted an
excessive birth-rate and increased the value of the survivors. The
emotional appeal of ‘baby-saving’ goes so directly to the heart of
civilized man that his head never reflects whether the particular baby
is worth saving, and whether a baby from a different breed and with
a better pedigree would not be better worth having. (3) Modern
obstetrics save the lives of thousands of women, whose physique is
such that in former times they would inevitably have died in child-
birth. The result is that child-birth is becoming more difficult. Also
babies brought up on the bottle, which has an irresistible attraction
for microbes of all sorts, are apt to be less healthy than those
nourished in the more primitive manner.
(4) Lastly, the bastardizing, which used formerly to provide for a
considerable infusion of the blood of the upper classes into the
lower, has now practically ceased. Since the merry days of King
Charles II, very few noble families of royal descent have been added
to the peerage.
 VII
Our civilization, therefore, carries within it the seeds of its own decay
and destruction, and it does not require high prophetic gifts to predict
the future of a race which goes the way marked out for it by such
perversely suicidal institutions. It cannot improve, but must
degenerate, and the only question would seem to be whether the
decadence of Man will leave him viable as a biological species. At
present it looks very much as though his blind leaders would lead
their blinder followers from catastrophe to catastrophe, through
imperialist world-wars to class-wars and to race-wars: but even if, by
some miraculous rally of human intelligence, these convulsions
should be averted, the prospect will not really be improved. The
violent destruction of the human race by war will only be more
dramatic: it will not be more fatal than its gradual decay as its arts
and sciences slowly fossilize, or peter out, in an overwhelming flood
of feeble-mindedness.
 VIII
This is the one alternative. We shall get to it, if we go on as we are
going: but it is not our doom. The alternative is to exercise the
danger by an adequate reform of human nature and of human
institutions. This again seems attainable in at least two ways.
The first, and more paradoxical, of these would make a direct frontal
attack on the palæolithic Yahoo, and try to bring about his moral
reformation. The means for this purpose are ready to hand. Christian
ethics have been in being, as a moral theory, for nearly two thousand
years. If the Yahoo could be really christianized, he would at any rate
cease to cut his own throat in cutting his neighbour’s. And it is
astonishing how much scientific support is forthcoming for the
paradoxes of Christian ethics. It is an historical fact that the meek
have a knack of inheriting the earth after their lords and masters
have killed each other off, and that passive resistance wears out the
greatest violence, and conscientious objection defeats the craftiest
opportunism, if only you can get enough of them. It is a biological
fact that the rabbit survives better than the tiger; and the same would
appear to be true of the human ‘rabbit’ and the Nietzschean ‘wild
beast.’ Intrinsically, therefore, Christian ethics might be well worth
trying.
I wish I could believe it likely that this policy will be tried. But the
palæolithic Yahoo has been dosed with Christian ethics for two
thousand years, and they have never either impressed or improved
him. Their paradoxes give him a moral shock, and he has not brains
enough to grasp their rationality. He will exclaim rather with the
gallant admiral in the House of Commons, when justly indignant at
the unheard-of notion that a ‘moral gesture’ of a Labour Government
might be the best policy, “Good God, sir, if we are to rely for our air
security on the Sermon on the Mount, all I can say is, ‘God help us!’”
Besides, the proposal to put Christian principles into practice would
be bitterly opposed by all the Churches in Christendom.[B]
It may be more prudent, therefore, to try a safer though slower way,
that of the eugenical reform and reconstruction of our social
organization. As to the possibilities in this direction, I incline to be
much more hopeful than either Mr Haldane or Mr Russell. Mr
Haldane despises eugenics, because he is looking for the more
spectacular advent of the ‘ectogenetic baby,’ to be the Saviour of
mankind. But he might not arrive, or be seriously delayed in
transmission, or fail to come up to Mr Haldane’s expectations; and,
meanwhile, we cannot afford to wait.
Mr Russell distrusts eugenics, because he fears that any eugenical
scheme put into practice will be ‘nobbled’ by our present ruling rings,
and perverted into an instrument to consolidate their power. He
thinks that dissent from dominant beliefs and institutions will be
taken as proof of imbecility, and sterilized accordingly,[C] and that the
result would merely be to spread over all the world the hopeless
uniformity and commonplaceness of the ideals and practice of the
American business man, as depicted by Mr Sinclair Lewis.
This prognostication would be very plausible, if we supposed
eugenics to be introduced into the social structure from above,
privily, and in small doses, and by way of administrative order, as
under the existing Acts to check the spread of feeble-mindedness.
But this method would be impracticable. It would not generate
anything like the social momentum necessary to carry through any
radical reform. To make it effective, it would have to be backed by a
powerful, enthusiastic, and intelligent public sentiment. This
presupposes that the public has been biologically educated to
appreciate the actual situation, and has been thoroughly wrought up
about the fatuity of our social order, and understands what is wrong
with it. If it understands that much, it can also be made to see that it
is fantastic to expect to leap to the Ideal State by a social revolution.
No one now knows what the institutions of an Ideal State would be
like, nor how they would work. We only know that they will have to
be evolved out of our present institutions, even as the Superman has
to be evolved out of the primitive Yahoo. In either case, the process
will be gradual, and its success will depend upon details, on taking
one step after another at the right rate in the right direction, making a
new adjustment here, overcoming an old difficulty there, removing
obstacles, smoothing over the shell-holes and scars dating from
Man’s lurid past, and, in general, feeling one’s way systematically
and scientifically to better things. Such a mode of progression may
seem unheroic, but it has the great advantage that it is unlikely to go
irretrievably wrong. If we know from the outset that we are tentatively
feeling our way, we shall always be on the look out for traps and
possibilities of going astray, trying out the value of our policies by
their results, and willing to retrace our steps when we have made a
false one.
The social temper, therefore, will become far more intelligent and
reasonable than it has been hitherto. It will be slow to dogmatize,
and will regard the toleration of differences of opinion as among the
cardinal principles of a sanely progressive social order. For as we
can no longer assume, with Plato and the other Utopians, that
perfection may be postulated, provision has always to be made for
the improvement of the social order. It can never be accepted as
absolutely good, but must always be regarded as capable, in
principle, of being bettered. Even the best of established institutions
are only good relatively to the alternatives to which they showed
themselves superior: under changed conditions they may become
inferior, and may fail us, or ruin us, if we do not make haste to
transform them into something better fitted to the new conditions.
Hence the social order must be plastic, and must never be allowed
to grow rigid. There must always be room in it for experiments that
have a reasonable prospect of turning out to be improvements. For
progress will depend on the timely adoption of such novelties.
But society has no means of commanding them at will. It has to wait
till they occur to some one. As biological variations have to arise
spontaneously before they can be selected, so valuable new ideas
have to occur in a human mind before they can be tried and
approved. Society cannot originate discoveries, it can only refrain
from so organizing itself as to stamp them out when they occur. It is
vitally necessary, therefore, that we should beware of suppressing
variations, whether of thought or of bodily endowment, that may
prove to be valuable.
Also, of course, we shall have to realize that our whole procedure is
essentially experimental, and all that this implies. We do not know, at
the outset, what would be the best obtainable type, either of man or
of society; true, but we mean to find out. Nor is it unreasonable to
expect to do so as we go along. We start with a pretty shrewd
suspicion that certain types, say the feeble-minded, the sickly, the
insane, are undesirable, and that no good can come of coddling and
cultivating them: we similarly are pretty sure that certain other types,
say the intelligent, healthy, and energetic, are inherently superior to
the former. We try, therefore, to improve and increase the better
types. How precisely, and how most effectively we do not quite know,
though we can make pretty good preliminary guesses. So we try.
That will entail experimentation in a variety of directions, with ‘control
experiments,’ and a modicum of mistakes. But our mistakes will not
be fatal, because if we advance tentatively and with intelligent
apprehension, we shall realize them in time, and shall not feel bound
to persist in any course that yields unsatisfactory results.
It is really one of the great advantages of eugenics that it cannot
proceed upon any cut-and-dried scheme, but will have to be guided
by the results of experiment and the fruits of experience, each of
which will be followed and discussed by an intensely interested
public. For the difficulties of eugenics are all difficulties of detail, and
intelligent attention to detail may overcome them all. Thus the
dysgenical working of civilized society, which has come about
unintentionally through the unfortunate convergence of a number of
tendencies, may be altered similarly, by changing the incidence of
social forces.
 IX
If scientific eugenics can put a stop to the contra-selection incidental
to civilization, Man will recover the plasticity and the progressiveness
he once possessed, and will be able to evolve further—in whatever
direction seems to him best. We need not take alarm at this
possibility, for with his superior knowledge he may surely be trusted
to make a better job of his evolution than the Lemur and the
Pithecanthropus, who were our progenitors and managed to evolve
into modern man.
But the process will necessarily be a slow one, even though a
comprehensive scheme of eugenics will be providing simultaneously
two sources of improvement, by the elimination of defectives at the
bottom of the social scale, and by the increase of ability at the top.
As, moreover, time presses, and sheer destruction may overtake us
before eugenics have made much difference, it would be highly
desirable if some means could be found to accelerate the change of
heart required. For this purpose, I am much less inclined to put my
trust in the advance of pharmacology than Mr Haldane and Mr
Russell.[D] Hitherto new drugs have only meant new vices,
sometimes (like cocaine) of so fascinating a character as to distract
the whole police force from their proper function of repressing crime.
So it seems legitimate to be very sceptical about moral
transformation scenes to be wrought by pills and injections.
On the other hand there does seem to be a science from the
possible progress of which something of a sensational kind might not
unreasonably be expected. It is, moreover, the science most directly
concerned with affairs of this sort. Psychology, the science of human
mentality, is, by common consent, in a deplorably backward state. It
has remained a ground for metaphysical excursions and a
playground for the arbitrary pedantries of classificatory systematists.
Its efforts to become scientific have only led it to ape assumptions
and to borrow notions found to be appropriate in sciences with
widely different problems and objects. The results, as the
psychologists themselves confess, are meagre and disappointing;
which, of course, only proves that the borrowed notions are
inappropriate and incapable of making Psychology into an effective
science. But if psychologists should take it into their heads to settle
down to business, to recognize the primary obligation of every
science to develop methods and conceptions capable of working
upon its subject-matter, and so tried to authenticate their ‘truth’ after
the ordinary fashion of the other sciences, namely by the pragmatic
test of successful working, some surprising effects might be elicited
even from the actual human mind.
For there is reason to suppose that its present organization is very
far from being the best of which it is capable. It has come about in a
very haphazard manner, and we are not at present making anything
like an adequate use of all our powers. Hence by changing the
gearing and re-arranging the traditional coupling, so to speak, of our
faculties, improvements might conceivably be wrought which would
seem to us to border on the miraculous. Thus a pragmatically
efficient Psychology might actually invert the miracle of Circe, and
really transform the Yahoo into a man.
 X
I have endeavoured in this very summary sketch to show that the
doom of Tantalus is by no means unconditional, and that he can
save himself if he chooses, and that by no superhuman effort, but
merely by recognizing facts that are right before his nose and well
within his comprehension, and by a little clear thinking upon their
import. But I would not presume to predict that he will save himself:
history affords no unambiguous guide. It seems to show that
something worse and something better than what actually happens
is always conceivable, and that neither our hopes nor our fears are
ever fully realized. If so, poor Tantalus, hoping against hope, fearing
against reason, may muddle along for a good while yet, without
repeating either his ancient error of imagining that he could sup with
the gods, or his modern folly of using his reason, as Goethe’s
Mephistopheles declared, only to become more bestial than any
beast!
 FOOTNOTES:
[A] The most absurd perhaps was the clause, appearing
in all the Peace Treaties, which made ‘reparations’ a first
charge on all the assets of the defeated countries. This,
of course, completely destroyed their credit, and
incapacitated them from raising a loan, forcing them to
have recourse to progressive inflation, and so into
bankruptcy.
[B] This does not mean, of course, that there are no
Christians in the Churches, but only that they are not in
control of these institutions.
[C] Icarus, p. 49.
[D] cf. Daedalus, p. 34; Icarus, p. 54.
 TRANSCRIBER’S NOTE:
Obvious typographical errors have been corrected.
*** END OF THE PROJECT GUTENBERG EBOOK TANTALUS ***

Updated editions will replace the previous one—the old editions

will be renamed.

Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright
in these works, so the Foundation (and you!) can copy and
distribute it in the United States without permission and without
paying copyright royalties. Special rules, set forth in the General
Terms of Use part of this license, apply to copying and
distributing Project Gutenberg™ electronic works to protect the
PROJECT GUTENBERG™ concept and trademark. Project
Gutenberg is a registered trademark, and may not be used if
you charge for an eBook, except by following the terms of the
trademark license, including paying royalties for use of the
Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such
as creation of derivative works, reports, performances and
research. Project Gutenberg eBooks may be modified and
printed and given away—you may do practically ANYTHING in
the United States with eBooks not protected by U.S. copyright
law. Redistribution is subject to the trademark license, especially
commercial redistribution.

START: FULL LICENSE

Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com