Philosophies Puzzles And Paradoxes A

Statisticians Search For Truth Yudi Pawitan

Youngjo Lee download

https://ebookbell.com/product/philosophies-puzzles-and-paradoxes-
a-statisticians-search-for-truth-yudi-pawitan-youngjo-
lee-56829298

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Philosophy On Tap Pintsized Puzzles For The Pub Philosopher

Isbn9781444397017 Google7nluoj9b8wic Matt Lawrence

https://ebookbell.com/product/philosophy-on-tap-pintsized-puzzles-for-
the-pub-philosopher-isbn9781444397017-google7nluoj9b8wic-matt-
lawrence-47122090

Philosophies Of Structural Safety And Reliability Vladimir Raizer

https://ebookbell.com/product/philosophies-of-structural-safety-and-
reliability-vladimir-raizer-47242730

Philosophies Of Religion A Global And Critical Introduction Timothy

Knepper

https://ebookbell.com/product/philosophies-of-religion-a-global-and-
critical-introduction-timothy-knepper-48775950

Philosophies Of Work In The Platonic Tradition A History Of Labor And

Human Flourishing Jeffrey Hanson

https://ebookbell.com/product/philosophies-of-work-in-the-platonic-
tradition-a-history-of-labor-and-human-flourishing-jeffrey-
hanson-50225796
Philosophies Of Nature After Schelling Iain Hamilton Grant

https://ebookbell.com/product/philosophies-of-nature-after-schelling-
iain-hamilton-grant-50674308

Philosophies Of Place An Intercultural Conversation

https://ebookbell.com/product/philosophies-of-place-an-intercultural-
conversation-51898672

Philosophies Of Happiness A Comparative Introduction To The

Flourishing Life Diana Lobel

https://ebookbell.com/product/philosophies-of-happiness-a-comparative-
introduction-to-the-flourishing-life-diana-lobel-51904458

Philosophies Of Hospitality And Tourism Giving And Receiving Prokopis

A Christou

https://ebookbell.com/product/philosophies-of-hospitality-and-tourism-
giving-and-receiving-prokopis-a-christou-51927172

Philosophies Of Art And Beauty Introducing Aesthetics Hugh Bredin

Liberato Santorobrienza

https://ebookbell.com/product/philosophies-of-art-and-beauty-
introducing-aesthetics-hugh-bredin-liberato-santorobrienza-51975320
 Philosophies, Puzzles
and Paradoxes
Unlike mathematics, statistics deals with real-world data and involves a higher degree of subjectivity
due to the role of interpretation. Interpretation is shaped by context as well as the knowledge, prefer-
ences, assumptions and preconceptions of the interpreter, leading to a variety of interpretations of
concepts as well as results. Philosophies, Puzzles and Paradoxes: A Statistician’s Search for Truth
thoroughly examines the distinct philosophical approaches to statistics – Bayesian, frequentist and
likelihood – arising from different interpretations of probability and uncertainty. These differences
are highlighted through numerous puzzles and paradoxes and illuminated by extensive discussions of
the background philosophy of science.
Features
• Exploration of the philosophy of knowledge and truth and how they relate to deductive and induc-
tive reasoning, and ultimately scientific and statistical thinking.
•  Discussion of the philosophical theories of probability that are wider than the standard Bayesian
and frequentist views.
•  Exposition and examination of Savage’s axioms as the basis of subjective probability and Bayesian
statistics.
•  Explanation of likelihood and likelihood-based inference, including the controversy surrounding
the likelihood principle.
• Discussion of fiducial probability and its evolution to confidence procedure.
• Introduction of extended and hierarchical likelihood for random parameters, with the recognition
of confidence as extended likelihood, leading to epistemic confidence as an objective measure of
uncertainty for single events.
• Detailed analyses and new variations of classic paradoxes, such as the Monty Hall puzzle, the para-
dox of the ravens, the exchange paradox, and more.
• Substantive yet non-technical, catering to readers with only introductory exposure to the theory of
probability and statistics.
This book primarily targets statisticians in general, including both undergraduate and graduate stu-
dents, as well as researchers interested in the philosophical basis of probability and statistics. It is also
suitable for philosophers of science and general readers intrigued by puzzles and paradoxes.
Yudi Pawitan graduated with a PhD in statistics in 1987 from the University of California at Davis and has been
a professor of biostatistics since 2001 at the Department of Medical Epidemiology and Biostatistics, Karolinska
Institutet, Stockholm, Sweden. He has worked in many areas of statistical applications, including time series
analyses and medical imaging, and for the last 20 years in the modelling and analysis of high-throughput genetic
and molecular data with applications in cancer. He has published more than 200 peer-reviewed research papers,
split about equally between methodology and applied publications. He is the author of the monograph In All
Likelihood (2001) and co-author of Generalized Linear Models with Random Effects (2017) together with Youngjo
Lee and John Nelder, both covering likelihood-based statistical modelling and inference. Philosophy of science,
statistical puzzles and paradoxes have been lifelong interests.

Youngjo Lee graduated with a PhD in statistics in 1983 from Iowa State University. He is currently a professor
emeritus of statistics at Seoul National University, an endowed-chair professor of data and knowledge service
engineering at Dankook University, and a vice president of the Korean Academy of Science and Technology.
Alongside the late John Asworth Nelder, he is an originator of hierarchical generalized models and h-likelihood,
having co-authored over 200 peer-reviewed research papers on the application of h-likelihood in various statisti-
cal areas. He is also a co-author of monographs on h-likelihood theory and applications. Furthermore, he has
developed related software and is currently extending h-likelihood procedures to deep neural networks.
Philosophies, Puzzles
and Paradoxes
A Statistician’s Search for Truth

Yudi Pawitan and Youngjo Lee

First edition published 2024
by CRC Press
2385 Executive Center Drive, Suite 320, Boca Raton, FL 33431, U.S.A.

and by CRC Press

4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

CRC Press is an imprint of Taylor & Francis Group, LLC

© 2024 Yudi Pawitan and Youngjo Lee

Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot as-
sume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have
attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright
holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowl-
edged please write and let us know so we may rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted,
or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, includ-
ing photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the
Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are
not available on CCC please contact [email protected]

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for
identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Pawitan, Yudi, author. | Lee, Youngjo, author.

Title: Philosophies, puzzles and paradoxes : a statistician’s search for truth / Yudi Pawitan and Youngjo Lee.
Description: First edition. | Boca Raton : CRC Press, 2024. | Includes bibliographical references and index. |
Summary: “Mathematics is focused on formal manipulation of abstract concepts, while statistics deals with
real-world data and involves a higher degree of subjectivity due to the role of interpretation. Interpretation is
shaped by context as well as the knowledge, biases, assumptions or preconceptions of the interpreter, leading
to a variety of potential interpretations of concepts as well as results. This book thoroughly examines the
distinct philosophical approaches to statistics--Bayesian, frequentist, and likelihood--arising from different
interpretations of probability and uncertainty. These differences are highlighted through a variety of puzzles
and paradoxes”-- Provided by publisher.
Identifiers: LCCN 2023039429 (print) | LCCN 2023039430 (ebook) | ISBN 9781032377391 (paperback) |
ISBN
9781032377407 (hardback) | ISBN 9781003341659 (ebook)
Subjects: LCSH: Probabilities--Philosophy. | Statistics--Philosophy. | Statistical decision. | Subjectivity.
Classification: LCC QA273.A35 P395 2024 (print) | LCC QA273.A35 (ebook) |
DDC 519.501--dc23/eng/20231221
LC record available at https://lccn.loc.gov/2023039429
LC ebook record available at https://lccn.loc.gov/2023039430

ISBN: 978-1-032-37740-7 (hbk)

ISBN: 978-1-032-37739-1 (pbk)
ISBN: 978-1-003-34165-9 (ebk)

DOI: 10.1201/9781003341659

Typeset in CMR10
by KnowledgeWorks Global Ltd.

Publisher’s note: This book has been prepared from camera-ready copy provided by the authors.
Contents

Foreword xi

Preface xv

List of Abbreviations xvii

List of Puzzles and Paradoxes xix

Introduction and Summary xxi

I How Do We Know What We Believe Is True? 1

1 Philosophical Theories of Knowledge and Truth 3

1.1 The Rationalists . . . . . . . . . . . . . . . . . . . . . . 5
1.2 The Empiricists . . . . . . . . . . . . . . . . . . . . . . 10
1.3 The Positivists and the Verification Principle . . . . . . 15

2 Deduction and Induction 19

2.1 Reasoning Solely Based on Induction . . . . . . . . . . 23
2.2 Reasoning Solely Based on Deduction . . . . . . . . . . 26
2.3 Complementary Induction and Deduction . . . . . . . . 28
2.4 Abduction . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Category-Based Induction . . . . . . . . . . . . . . . . 33

v
vi Contents

3 Hilbert’s Broken Dream: Limitations of Deductive

Reasoning 39
3.1 Is Euclidean Geometry True? . . . . . . . . . . . . . . . 40
3.2 Gödel’s Incompleteness Theorems . . . . . . . . . . . . 44

4 ‘Real’ Scientific Process 49

4.1 Theory vs Observations . . . . . . . . . . . . . . . . . . 49
4.2 Paradigm Shift . . . . . . . . . . . . . . . . . . . . . . . 50

II Probability and Inverse-Probability Inference 57

5 The Rise of Probability 59

5.1 Mathematical Probability . . . . . . . . . . . . . . . . . 60
5.2 Countable vs Finite Additivity . . . . . . . . . . . . . . 64
5.3 What Is Probability? . . . . . . . . . . . . . . . . . . . 67
5.4 Pascal’s Wager . . . . . . . . . . . . . . . . . . . . . . . 70

6 Philosophical Theories of Probability 75

6.1 Classical Theory . . . . . . . . . . . . . . . . . . . . . . 76
6.2 Logical Theory . . . . . . . . . . . . . . . . . . . . . . . 80
6.3 Frequency Theory . . . . . . . . . . . . . . . . . . . . . 83
6.4 Propensity Theory . . . . . . . . . . . . . . . . . . . . . 86
6.5 Subjective Theory . . . . . . . . . . . . . . . . . . . . . 90
6.6 Consensus theory . . . . . . . . . . . . . . . . . . . . . 96

7 Rereading Savage 101

7.1 The Axioms . . . . . . . . . . . . . . . . . . . . . . . . 102
7.2 Qualitative Probability and Axiom 6 . . . . . . . . . . 109
7.3 Conditional Probability . . . . . . . . . . . . . . . . . . 111
7.4 Utility, Axiom 7 and The Theorem . . . . . . . . . . . 112
7.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 114
Contents vii

8 The Inverse Probability Method 119

8.1 Bayes’s Essay . . . . . . . . . . . . . . . . . . . . . . . 120
8.2 Bayesian Learning . . . . . . . . . . . . . . . . . . . . . 123
8.3 Laplace’s Memoir . . . . . . . . . . . . . . . . . . . . . 123
8.4 Testimony Puzzle . . . . . . . . . . . . . . . . . . . . . 125
8.5 The Sunrise Problem . . . . . . . . . . . . . . . . . . . 127
8.6 Qualitative Plausible Reasoning . . . . . . . . . . . . . 130

9 What Prior? 135

9.1 The Principle of Insufficient Reason: Uniform Prior . . 136
9.2 Axiomatic Basis of the Principle of Insufficient Reason 140
9.3 Invariant Prior∗ . . . . . . . . . . . . . . . . . . . . . . 141
9.4 Multiparameter Case∗ . . . . . . . . . . . . . . . . . . . 143

III Likelihood and Likelihood-Based Inference 147

10 Likelihood 149
10.1 Classical Likelihood . . . . . . . . . . . . . . . . . . . . 149
10.2 Exchange Paradox . . . . . . . . . . . . . . . . . . . . . 154
10.3 Prosecutor’s Fallacy . . . . . . . . . . . . . . . . . . . . 155
10.4 Likelihood Principle . . . . . . . . . . . . . . . . . . . . 157
10.5 Controversy . . . . . . . . . . . . . . . . . . . . . . . . 160

11 P-values and Confidence 167

11.1 Significance Test via Randomization . . . . . . . . . . . 167
11.2 P-values and Fiducial Probability . . . . . . . . . . . . 169
11.3 Confidence Procedures and Coverage Probability . . . . 172
11.4 Can We Say with Complete Confidence that the Sun Will
Rise Forever? . . . . . . . . . . . . . . . . . . . . . . . . 175
viii Contents

12 Extended Likelihood 179

12.1 Random Parameters . . . . . . . . . . . . . . . . . . . . 179
12.2 Hierarchical Likelihood∗ . . . . . . . . . . . . . . . . . . 182
12.3 Confidence Is an Extended Likelihood . . . . . . . . . . 185
12.4 Wallet Game Paradox . . . . . . . . . . . . . . . . . . . 189

13 Epistemic Confidence 195

13.1 Non-epistemic Confidence and the Dutch Book . . . . . 197
13.2 Epistemic Confidence Theorem . . . . . . . . . . . . . . 199
13.3 Epistemic Confidence and Objective Bayesianism . . . 204
13.4 Probability, Likelihood, Extended Likelihood, H-likelihood
and Confidence . . . . . . . . . . . . . . . . . . . . . . . 206

IV Puzzles and Paradoxes 209

14 Paradoxes of Savage’s Axioms 211

14.1 Allais’s Paradox . . . . . . . . . . . . . . . . . . . . . . 211
14.2 Ellsberg’s Paradox . . . . . . . . . . . . . . . . . . . . . 217

15 Fallacious Fallacies 225

15.1 Conjunction Fallacy . . . . . . . . . . . . . . . . . . . . 226
15.2 Inclusion Fallacy . . . . . . . . . . . . . . . . . . . . . . 227
15.3 Statistical and Scientific Reasoning . . . . . . . . . . . 228
15.4 Probability and Likelihood-Based Reasoning . . . . . . 230
15.5 Degree of Confirmation . . . . . . . . . . . . . . . . . . 234
15.6 Traces of Likelihood-Based Reasoning in Infants . . . . 235
15.7 Discussion: The Rewards of Being Right . . . . . . . . 237

16 Monty Hall Puzzle and the Three Prisoners Paradox 241

16.1 The Puzzle and the Paradox . . . . . . . . . . . . . . . 241
16.2 Likelihood-based Solution . . . . . . . . . . . . . . . . . 243
Contents ix

16.3 Probability or Likelihood? . . . . . . . . . . . . . . . . 244

16.4 Three Prisoners Paradox: A Guard Involved . . . . . . 245
16.5 Three Prisoners Paradox: Breaking News . . . . . . . . 247
16.6 The Case of a Talkative Guard . . . . . . . . . . . . . . 249
16.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 251

17 The Lottery Paradox and the Cold Suspect Puzzle 253

17.1 The Lottery Paradox . . . . . . . . . . . . . . . . . . . 253
17.2 Probability in Court . . . . . . . . . . . . . . . . . . . . 255
17.3 DNA Fingerprinting . . . . . . . . . . . . . . . . . . . . 257
17.4 Database Search: The Cold Suspect Puzzle . . . . . . . 258
17.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 265

18 Paradox of the Ravens 269

18.1 The Paradox . . . . . . . . . . . . . . . . . . . . . . . . 269
18.2 Data, Models and Their Likelihoods . . . . . . . . . . . 271
18.3 Multinomial Models for Various Sampling Plans . . . . 273
18.4 Large-Sample Evidence . . . . . . . . . . . . . . . . . . 278
18.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 279

19 The Exchange Paradox 283

19.1 The Paradox and Its Variants . . . . . . . . . . . . . . 283
19.2 Before vs After Opening the Envelope . . . . . . . . . . 285
19.3 Likelihood, Frequentist and Bayesian Explanations . . . 286
19.4 Actual vs Counterfactual Envelopes . . . . . . . . . . . 290
19.5 Risk Aversion and Confidence-Weighted Utility . . . . . 291
19.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 295

Bibliography 299

Index 311
Foreword

This is a remarkable book: wide-ranging, ambitious, challenging and

profound but also intriguing, fascinating and original. Despite the fact
that I have been thinking about the foundations of statistics for a long
time, in particular as to how those foundations relate to what we want
to do and achieve as practising statisticians, I learned a lot from reading
it and also unlearned some things I thought I knew.
Yudi Pawitan and Youngjo Lee are well-respected statisticians who,
sometimes working together but also apart, have both made major con-
tributions to our understanding of the role of likelihood in statistical
inference and to its continuing development as a statistical tool. The
book they have now written together is a philosophical treatise based
on their deep understanding of likelihood and related statistical concepts
such as confidence, and they have extended the scope of these tools to
address evidential paradoxes often treated in philosophy, such as, for
example, Hempel’s paradox of the ravens.
It was John Nelder who first drew my attention to the importance of
Yudi’s by now classic book on likelihood (1) and it was through my
friendship with John that I got to know of the ground-breaking work
Youngjo was doing with John in extending likelihood to analysing hier-
archical generalised linear models (2). Eventually the three collaborated
to produce a book that did for hierarchical models (3) what John’s fa-
mous book (4) with Peter McCullagh did for generalised linear models.
John died in 2010 but his two coauthors have continued their collabo-
ration and this has now taken this surprising turn away from modelling
itself to consider more foundational issues.
The book has four parts. Part I covers a view of epistemology from the
perspective of the statistician. Statisticians unfamiliar with the philoso-
phy of science will find it a valuable introduction to that field. Philoso-
phers of science, whether or not they are in agreement with this survey
of their field, should find it a useful introduction to how statisticians
might think about it.

xi
xii Foreword

Part II addresses probability and inverse probability. The former is the

province of mathematics. Even if there are disagreements as to what
probability means (and the book presents the various theories), its de-
velopment can proceed largely axiomatically. However, any student of
probability must occasionally wonder how anyone would ever know what
a probability could be. The inverse passage from evidence to probability
that such knowledge would involve is in the domain of statistics and
raises many difficulties, which are dealt with here.
Part III is a thorough survey of likelihood and likelihood-based inference,
a field in which the authors, as already noted, are experts. In what was
a surprise to me, confidence emerges as a fundamental related concept.
(One can only wonder what R.A. Fisher would think about a term of
Neyman’s he despised being related to a concept such as likelihood,
which he regarded as being very much his own!)
Part IV uses the understanding gained in the earlier part to probe and
resolve various notorious paradoxes. Quite apart from anything else, this
section is very great fun, providing new insights on old favourites such
as the Monty Hall Problem.
I very much enjoyed reading this highly original work and recommend
it to all who are interested in statistical inference, evidence and the
philosophy of science.

Stephen Senn
Edinburgh

References

1. Pawitan, Y. (2001). In All Likelihood: Statistical Modelling and In-

ference Using Likelihood. Oxford: Clarendon Press.

2. Lee, Y. and Nelder, J.A. (1996). Hierarchical generalized linear mod-

els (with discussion). Journal of the Royal Statistical Society: Series
B (Methodology), 58, 619–678.
Foreword xiii

3. Lee, Y., Nelder, J.A. and Pawitan, Y. (2006). Generalized Linear

Models with Random Effects: Unified Analysis via H-likelihood. 1st
Edition. Boca Raton: Chapman & Hall/CRC.

4. McCullagh, P. and Nelder, J.A. (1983). Generalized Linear Models.

1st Edition. London: Chapman & Hall.
Preface

In the face of numerous blogs, Wikipedia entries, podcasts, YouTube

videos, chatbots, etc., why a book on philosophies, puzzles and para-
doxes? During most of our career, which started in the dark ages before
the internet, we have actually been thinking about these things and the
statistical issues underneath them, but never seriously enough to write
anything about them. But now that we are at the end of our career,
we realize that it would have been good for us to know and understand
these issues earlier. Unfortunately, they are not the kind of topic found in
standard textbooks, and it is difficult to put high confidence in generic
blogs or Wikipedia entries. The many decades of experience we have
gained as professional statisticians have helped us formulate and convey
these topics with unique perspectives to new generations of statisticians.
In our effort to illuminate some poorly lit terrain of statistical inference,
we realize that we’re not offering easy-to-digest five-minute blog pieces.
To paraphrase Mark Twain, we don’t have time to write short pieces, so
we write long ones instead. Our only defense is that most of the chapters
here are based on many years of thinking, discussing and contemplating,
so we hope the readers would indulge us by giving more than five minutes
of their precious time. So, after all these meanderings, why a book? We
believe that a book – perhaps even in its old-fashioned physical form –
is still the best medium for reflective reading that these topics deserve.
Our primary interest is in the philosophical foundation of probability and
statistics. It is a well-trodden area and some topics, such as the philo-
sophical disagreements between the frequentist and Bayesian schools,
have been debated for nearly a century. Serious students of statistics
are likely to be perplexed by these disagreements, as they do not occur
in other areas of mathematics or scientific disciplines such as physics,
chemistry or biology. Unfortunately, although the passionate arguments
have largely died down, it’s actually not due to any consensus. Statisti-
cians still argue about how to interpret the confidence interval and the
role of P-value. Whatever truth that forms the foundation of our subject,

xv
xvi Preface

it must be of the elusive kind that Oscar Wilde said to be rarely pure
and never simple. Our recent realization of confidence as an extended
likelihood, which leads to the concept of epistemic confidence, might
contribute to a consensus based on the common ground between the
epistemic non-Bayesianism and objective Bayesianism.
Many topics are presented via puzzles and paradoxes that highlight the
nature of probability and statistical inference. The existence of a paradox
is a warning that there is something incomplete, if not wrong, in our
reasoning. Paradoxes have always been important in the development of
mathematics; for instance, Russell’s paradox about the set of all sets that
are not members of themselves led to the development of set theory, logic
and the foundation of mathematics. Similarly, in statistics, discussion
and resolution of paradoxes may lead to a clarification of concepts and
an enlargement of vocabularies that will strengthen the foundation of
our subject.
Our book is an eclectic, some might even say idiosyncratic, collection
of topics that reflect our interests. We believe that most chapters are
readable by nonspecialists who have some knowledge of the elementary
theory of probability and statistics, perhaps after one semester each.
Several topics are indeed advanced, but the corresponding discussions
rarely require deep knowledge or special technical competence. We have
never used the book for any course, though we imagine that it could be
useful for a discussion course.
Last but not least, we acknowledge the contributions of many colleagues
and former students who, over the years, we have both challenged and,
at times, tortured with these puzzles and paradoxes. Like jokes, puzzles
and paradoxes are enjoyable for sharing. But unlike jokes, our under-
standing of puzzles and paradoxes improves with sharing. We thank
Lars Rönnegård for careful reading of the draft and helpful sugges-
tions, Zheng Ning and Jiangwei Sun for discussion and critical read-
ing of several chapters on the paradoxes, and Jaehyuk Kwon and Dr.
Hangbin Lee for editorial assistance. We have tried to present all facts,
ideas and points of view accurately, but nobody is perfect, so we wel-
come corrections, comments or clarifications; these will be collated in
https://www.meb.ki.se/sites/yudpaw/book/.

Yudi Pawitan and Youngjo Lee

Stockholm and Seoul
List of Abbreviations

CA countable additivity
CP conditionality principle
DF99 Donnelly and Friedman (1999)
FA finite additivity
HGLM hierarchical generalized linear model
GLM generalized linear model
iid independent and independently distributed
LLN law of large numbers
LP likelihood principle
MLE maximum likelihood estimate
RDF Ramsey and de Finetti
SLLN strong law of large numbers
SLP strong likelihood principle
SP sufficiency principle
VNM von Neumann and Morgenstern (1947)
WLP weak likelihood principle
WLLN weak law of large numbers

xvii
List of Puzzles and Paradoxes

all-or-nothing exchange paradox, 284

Allais’s paradox, 211
Bertrand’s paradox, 78
book paradox, 79
boy-or-girl paradox, 62
coin paradox, 79, 137
cold suspect puzzle, 258
conjunction fallacy, 226
Ellsberg’s paradox, 217
exchange paradox, 154, 283
factual reverse exchange paradox, 290
inclusion fallacy, 227
induction problem, 127, 175
lottery paradox, 253
magnified exchange paradox, 286
marginalization paradox, 144
Monty Hall puzzle, 241
probability dilution paradox, 145,175, 206
prosecutor’s fallacy, 155
ravens paradox, 269
reverse exchange paradox, 285
satellite conjunction paradox, 145, 206
Stein’s paradox, 145
stopping rule paradox, 161
sunrise problem, 127, 175
testimony puzzle, 125
ticket paradox, 152, 187
three prisoners paradox, 242
wallet game paradox, 189
wine-water paradox, 78

xix
Introduction and Summary

Even though we’re writing a book for nonspecialists, we still realize that
it would be impossible to satisfy the varying levels of all potential au-
diences. Advanced readers may find some materials too trivial, while
others may find them too advanced. Some sections are starred as warn-
ing signs that they require advanced mathematical knowledge or use
concepts that appear in later chapters. Nevertheless, at the cost of some
repetitiveness, we have tried to make the chapters as self-contained as
possible so that the reader can jump around following their interest. All
the chapters in Part IV on the paradoxes, except perhaps Chapter 14
on the Allais and Ellsberg paradoxes, can be read independently from
the rest of the book.
When we perform data analysis, there is a feeling that we’re searching
for the truth – the true model, the true parameter, etc. Mathematicians
try to find and prove the true propositions; scientists in various fields
work to discover the true laws of nature, the true causes of diseases, etc.
Yet it has taken philosophers many centuries to understand the nature
of truth and the methods to establish it. Chapters 1–4 are summaries
of our attempt to absorb what (some) philosophers have thought and
written about knowledge and truth.
Our primary interest is statistics and statistical reasoning. This interest
leads immediately to probability, not just mathematically, but also philo-
sophically. Although there is a consensus on mathematical probability
around Kolmogorov’s system of axioms, interestingly, even philosophers
disagree on the meaning of probability. To some extent, but not fully,
the Bayesian-frequentist divide reflects the philosophical disagreements.
Philosophers spend their time thinking about these things more than
statisticians do, so we find it worthwhile to read their literature closely.
This is the basis of Chapters 5 and 6. We were rather surprised to dis-
cover, for example, that what we understand as the ‘frequentist philoso-
phy’ actually does not follow from the ‘frequency theory’ of probability.

xxi
xxii Introduction and Summary

We have focused on orthodox concepts of uncertainty: probability, likeli-

hood and confidence. Other concepts have been proposed, such as fuzzy
logic, modal logic, belief functions, possibility and plausibility measures,
etc. There is a large literature on these topics that is almost independent
of probability, suggesting the enormous depth of the notion of uncer-
tainty. Within probability-based inference, we cover in depth the basis
of Bayesian inference by going into Savage’s axioms in Chapter 7, his-
torical records from Bayes and Laplace in Chapter 8 and the problem of
prior selection in Chapter 9. Our own professional work and experience
revolve around likelihood (Chapter 10), the central concept of statistical
model and inference first explicitly recognized by Fisher in 1921.
During the 1930s, Fisher proposed fiducial probability, which was meant
to give probability-based inference without having to specify prior distri-
bution. Since it is not a proper probability, the fiducial idea led to many
controversies, but it motivated the development of the confidence pro-
cedure (Chapter 11). In recent years, likelihood has been fundamentally
extended to deal with random parameters (Chapter 12). The recognition
that confidence is an extended likelihood leads to the epistemic confi-
dence concept in Chapter 13. The rest of the book (Chapters 14–19)
is dedicated to various puzzles and paradoxes. Short summaries of the
chapters are as follows.
Chapter 1 covers philosophical theories of knowledge and truth, an area
in which philosophers have traditionally been interested. They have the
advantage of being professional truth seekers, so it seems natural that
we should try to learn from what they have come up with. Two enduring
schools of thought emerged since the Age of Enlightenment: the Conti-
nental rationalists, who emphasized reason, and the British empiricists,
who emphasized the role of experience and observation. Hume’s struggle
with induction is particularly relevant for statisticians.
In Chapter 1 we see the recurring themes of necessary and contingent
truths along with deductive and inductive reasoning to uncover them.
In Chapter 2 we go into more detail. Deduction, favoured by ratio-
nalists, uses pure reason to prove propositions based on known truths.
It relies on assumed propositions for its foundation, but the validity of
these premises can’t always be observed. Induction, rooted in empiri-
cism, involves evidence and support through experimentation. Inductive
reasoning, though commonly used in daily thinking, remains tricky and
doesn’t necessarily improve with age. It’s characterized by uncertainty
due to incomplete information, leading to potential pitfalls. Clarifying
Introduction and Summary xxiii

the limits of both deductive and inductive logic is crucial for establishing
reliable knowledge.
Throughout the centuries, mathematics has been seen as a source of
objective truth achieved through deductive reasoning from self-evident
axioms. Chapter 3 discusses Hilbert’s dream of automating all math-
ematical truths; it’s a dream shattered by Gödel’s incompleteness the-
orems. Together, the theorems highlight that, within a formal system,
there are undecidable statements and true statements that cannot be
proven, thus requiring reliance on other sources like intuition or empir-
ical evidence. Uncertainty in deduction calls for additional methods to
determine the truth. In geometry, different sets of axioms can lead to
or represent different spatial models. Similarly, models in science and
statistics, though not necessarily true, are valuable for understanding
reality and making predictions.
The complex interplay between theory and observation in the scientific
process is covered in Chapter 4. Some philosophers argue that theories
or thoughts precede evidence, while others emphasize starting with ob-
servations to establish theories. Kuhn’s model outlines stages of normal
science, extraordinary research, paradigm adoption, and the establish-
ment of new axioms. Paradigm shifts occur when dominant theories
clash with new evidence or anomalies, leading to the adoption of new
paradigms.
Chapter 5 charts the rise of probability as a model and a tool for dealing
with uncertainty. Probability encompasses both deductive and inductive
reasoning, offering a language to handle uncertainty. Its historical de-
velopment involved diverse interpretations, from representing chances
in games to epistemic degrees of belief. Kolmogorov’s axioms laid the
foundation for mathematical probability, distinguishing it from measure
theory by introducing conditional probability and independence.
Although most statisticians follow Kolmogorov’s mathematical proba-
bility, there is a wide range of philosophical theories and interpreta-
tions. Chapter 6 covers: (i) classical, (ii) logical, (iii) subjective, (iv)
frequency, (v) propensity and (vi) consensus theories of probability. The
Bayesians adhere to the subjective theory, while the frequentists follow
a blend of frequency and propensity theories. The consensus theory is
a novel theory that gives the most compelling objective probability of
single events.
xxiv Introduction and Summary

Chapter 7 evaluates Savage’s theory of subjective probability, which

is based on axioms of rational preferences over acts and consequences.
We describe in detail the seven axioms: weak ordering, sure-thing princi-
ple, state independence, consequence independence, non-triviality, small-
event continuity and strong dominance. The meaning and implications
of each axiom are discussed, as well as how they relate to other the-
ories of decision-making. The theory highlights the power and limit of
axiomatic systems, and how and whether they can be used to guide
rational decisions under uncertainty.
Early writers like Pascal and Leibniz used probability to quantify un-
certainty, often relying on simple reasoning similar to the principle of
insufficient reason. In Chapter 8 we discuss Bayes’s 1763 Essay To-
wards Solving a Problem in the Doctrine of Chances and Laplace’s 1774
Memoir on the Probability of Causes by Events, which laid the foun-
dation for the so-called inverse probability method that dominated the
19th-century applied probability and statistics. Their idea solved Hume’s
struggle with induction, providing a rational approach to incorporating
new information in probability estimation.
The Bayesian school is founded on subjective probability and the use
of priors. The subjective probability can be justified by the coherency
axiom, an attractive axiom even for non-Bayesians. However, the choice
of a prior distribution remains problematic for most non-Bayesians. In
Chapter 9 we discuss how different choices of the prior distribution lead
to different variants of Bayesianism, with ‘objective Bayesians’ using
non-subjective and non-informative priors.
In Chapter 10 we describe the concept of likelihood and explain how
it differs from probability. Likelihood is central in statistical modelling
and inference. Maximum likelihood estimates (MLEs) are widely used
in practice, but in orthodox statistical inference, likelihood requires
probability-based calibration. The key role of likelihood in statistical
inference is based on its property of containing all the evidence about
a parameter in the data, a result implied by the likelihood principle.
However, a normative interpretation of the strong likelihood principle
that demands equivalent likelihoods to yield identical inferences leads
to controversy, especially among frequentists.
Significance testing and P-value concepts are covered in Chapter
11. Fiducial probability, introduced by Fisher via the P-value as a
function of the parameter, resembles a standard probability. However,
Introduction and Summary xxv

crucially, it lacks full additivity, which causes problems when dealing

with many-to-one transformations. Although the fiducial idea has been
abandoned in statistical practice, it motivated the widely accepted con-
fidence interval procedure and the more recent confidence distribution
concept.
In Chapter 12 we describe extended likelihood and hierarchical like-
lihood as the unifying tools for dealing with complex models that in-
volve fixed and random parameters. We explain how extended likelihood
is different from classical likelihood. We show the difficulties that can
arise when using extended likelihood, primarily due to the lack of in-
variance with respect to the transformation of the random parameters.
Hierarchical likelihood is introduced as a special extended likelihood
that avoids these problems and gives optimal estimators and predic-
tors for fixed and random parameters. Confidence is shown to be an
extended likelihood; this connection expands confidence’s utility beyond
the confidence-interval context.
Chapter 13 introduces the concept of epistemic confidence, an objec-
tive sense of uncertainty in unique events. For confidence intervals, it is
the sense of confidence in a computed or observed confidence interval.
Orthodox frequentist inference denies epistemic confidence, treating the
confidence level as applying to the procedure rather than the interval.
Epistemic confidence is essential for decision making and is naturally
embraced by Bayesians. To make non-Bayesian confidence epistemic, we
follow Fisher’s proposal that a probability is epistemic if there is no rel-
evant subset, but we use the confidence concept instead of probability.
The epistemic confidence density and implied prior mirror the Bayesian
posterior and prior, offering a non-Bayesian alternative to Bayesian infer-
ence. However, unlike Bayesian approaches, epistemic confidence doesn’t
require an explicit prior and isn’t a probability.
In Chapter 14 we discuss Allais’s and Ellsberg’s paradoxes, two well-
known paradoxes associated with Savage’s axioms (Chapter 7), par-
ticularly the sure-thing principle. The paradoxes highlight the descrip-
tive/predictive vs normative roles of the axiomatic system. In the former
role, the axioms should be able to predict human rational decisions, so
a common violation of any of the axioms raises the question of whether
the system is rich enough to capture real human behaviour. However, in
its normative role, the axioms act as a guide to human decision making,
where a violation would imply that it is the human decision that needs
to be corrected in order to follow the axioms.
xxvi Introduction and Summary

We use probability to quantify uncertainties in our reasoning for mak-

ing judgements and decisions. Intriguingly, when used mathematically,
probability-based thinking can sometimes produce starkly different con-
clusions compared to when used statistically. Our aim in Chapter 15
is to discuss a well-known fallacy called the conjunction fallacy and a
related inclusion fallacy. In the spirit of this book, they might be con-
sidered paradoxes, where seemingly valid probability-based reasoning
leads to a contradiction viz-á-viz statistical reasoning. We will contrast
two different modes of reasoning, one captured by probability and the
other by likelihood, and suggest that our seemingly irrational behaviour
is often due to decision making based on the likelihood. Thus, from a
likelihood perspective, we are still acting rationally.
The Monty Hall puzzle has been solved and dissected in many ways, but
always using probabilistic arguments, so it is considered a probability
puzzle. In Chapter 16 the puzzle is set up as an orthodox statisti-
cal problem involving an unknown parameter, a probability model and
an observation. This means we can compute a likelihood function, and
the decision to switch corresponds to choosing the maximum likelihood
solution. We also describe an earlier version of the puzzle in terms of
three prisoners: two to be executed and one released. Unlike goats and
cars, these prisoners are sentient beings that can think about exchanging
punishments. When two of them do that, however, we have a paradox
where it is advantageous for both to exchange their punishments with
each other.
Dealing with uncertainty is never easy. Although probability as the main
tool for dealing with uncertainty is well developed, we still have to deal
with many probability-related puzzles and paradoxes. In Chapter 17
we describe the lottery paradox, which shows the logical problem of
accepting uncertain statements based on high probability. In the cold
suspect puzzle, the differences between the schools of inference – the
frequentist, Bayesian and likelihood schools – lead to conflicting quan-
tification of forensic evidence in the assessment of cold suspects from
database searches.
In Hempel’s paradox of the ravens, using seemingly logical reasoning,
seeing a non-black non-raven – such as a red pencil – is considered sup-
porting evidence that all ravens are black. This sounds non-sensical, but
previous Bayesian analyses and Hempel himself accepted the paradoxical
conclusions. Chapter 18 describes likelihood-based analyses of various
statistical models that show that the paradox feels paradoxical because
Introduction and Summary xxvii

there are natural models in which observing a red pencil tells us nothing
about the colour of ravens.
We have explored the various interpretations of probability in Chapter
6. An intriguing paradox that arises in the context of probability for
single events is the exchange paradox in Chapter 19. You are asked
to choose and open one of two envelopes that contain some amount
of money. Seemingly correct probability-based reasoning suggests that
exchanging is always better, which is absurd, since you have just chosen
one at random. The paradox highlights the role of likelihood, extended
likelihood, confidence and subjective utility to deal with the uncertainty
for single events.
 Part I

How Do We Know What

We Believe Is True?
1
Philosophical Theories of Knowledge and
Truth

Whoever undertakes to set himself up as a judge of Truth and

Knowledge is shipwrecked by the laughter of the gods. – Edmund
Burke (1729–1797), Anglo-Irish politician-philosopher.

‘I’d believe it when I see it with my own eyes!’ That sounds like a
gold standard for verifying beliefs, but it’s only for a hyperskeptic or
an exasperated person. Certainly not the ultimate standard, as such a
standard would limit our knowledge base severely. Let’s consider a few
questions, some of which are intrinsically factual and directly verifiable,
and some not: Did O.J. Simpson kill his ex-wife? Is Socrates mortal?
Are protons immortal, or do they decay? Is there – or was there ever –
life on Mars? Is there life somewhere else in the universe? Do the angles
inside an arbitrary triangle always add up to two right angles? Does
mask-wearing help reduce the spread of covid-19 infections? Does the
mRNA vaccine work against the virus? Does ivermectin work? Even
verifiable ones, such as the last three questions, require very careful and
rigorous steps to answer, not verifiable just by our eyes. You may feel
that you already ‘know’ the answers to some of the questions, but how
do we know what we believe is true?
To answer the question, let’s turn to the philosophers; they’re the pro-
fessional thinkers who spend their days contemplating the meaning of
beauty, ethics, the purpose of life, or even the art of travel and the
sources of value in economic commodities. We wish to find revelations
from the ones that have thought deeply over several centuries about
knowledge and truth. The wish turns out to be in the easier-said-than-
done category. Perhaps it is not just the vigorous verbal gymnastics
that makes us woozy. Unlike reductionist scientists, who are happy
with incremental progress, philosophers have a tendency to build an
all-encompassing theory of knowledge, but often with too few concrete
examples.

3
4 Philosophical Theories of Knowledge and Truth

But, of course, we couldn’t give up and let you down, dear readers. After
some perseverance, we identified recurring ideas that appeared over the
centuries and made sure that we absorb the essential philosophical ideas
critically. By way of apology, due to natural time constraints in this
sojourn outside our comfort zone, we have focused unevenly on a few
celebrated (and dead) philosophers whose names are synonymous with
the study of knowledge and truth.
During the late 17th and early 18th centuries, Europe went through a
special time known as the Age of Enlightenment. It was a period char-
acterized by a focus on reason and the pursuit of knowledge and marked
by significant advances in science, philosophy and politics. Many writers
and philosophers from that era are still read today, similar to those who
appeared during the Golden Age of Philosophy in Greece during the fifth
and fourth centuries BC, such as Socrates, Plato and Aristotle. Perhaps
it’s not surprising that probability theory also emerged during this pe-
riod, and many major figures of the Enlightenment era contributed to
probability theory.
The Enlightenment philosophers fall roughly into two camps: continen-
tal rationalists and British empiricists. Rationalists – such as Descartes,
Leibniz, Kant or Hegel – were either mathematicians themselves or
highly influenced or even enthralled by mathematics and science. They
emphasized the role of reason and logic to arrive at knowledge and truth.
On the other hand, empiricists, such as Locke, Berkeley or Hume, were
less influenced by mathematics and placed more emphasis on experience
as the source of knowledge.
Those definitions are, of course, rough and simplistic. Rationalists do
not claim that knowledge is independent of experience, but rather that
reasons provide a framework that structures the experience, or provide
concepts that give more information than experience alone. And perhaps
most importantly, reason is the final arbiter of truth. On the other hand,
empiricists would consider two kinds of experience, sensory and reflec-
tive, where the latter is not far from rationalist reason. So in reality, in
any single person, there is no clean division between the two traditions:
Descartes had empiricist ideas, whereas, for issues such as ethics, Locke
weighed reason and experience equally. Hume also recognized the power
of deductive logic in establishing ‘Relations of Ideas’ such as those found
in mathematics.
The Rationalists 5

1.1 The Rationalists

René Descartes (1596–1650), considered the father of modern western

philosophy, is best known for his Meditations on First Philosophy (1641),
where he sought to provide a foundation for knowledge and sure belief.
Gottfried Leibniz (1646–1716) perceptively identified two kinds of truth:
necessary and contingent.1 A necessary truth is one whose negation leads
to a contradiction. Mathematics is the obvious place for necessary truths,
which can be demonstrated by reason alone. On the other hand, a con-
tingent truth is not necessarily true, but is nonetheless true. Negation
of a contingent truth does not lead to a contradiction. True empirical
facts are always contingent. For instance, at present the sky is blue, but
of course it doesn’t have to be: it could be gray and rainy, especially as
we’re writing this very sentence in Ireland.
Truth is established based on the great principle of rationalism: the prin-
ciple of sufficient reason, which states that there is a reason or cause for
everything, including the existence of objects, the occurrence of events,
and the truth or untruth of propositions. It’s actually an ancient prin-
ciple already known to the Greeks. Necessary truths must be true in
order to avoid contradiction. Contingent truths must also have some
underlying cause or reason. If your car breaks down – an annoying con-
tingent truth – you’d be sure there must be a reason, which a mechanic
should be able to find out and fix. We’d have a similar reaction to bigger
problems, such as the ozone hole over Antarctica or the mass death of
coral reefs in the world’s oceans. The rationalist goal is to uncover the
cause or explanation of any phenomenon through reason and inquiry.
We inherit this optimistic view of science to this day.
If everything has a reason, then there must be a chain of reasons going all
the way to the ultimate reason. Being a devout Christian, for Descartes
the ultimate reason is God, whose existence he proved deductively (Sec-
tion 2.2). For Leibniz, it is the ‘monad,’ the fundamental building block
1
During the 20th century, we saw a formalization in modal logic in terms of neces-
sary and possible truths with its colourful possible world semantics. Thus, a necessary
truth is true in all possible worlds; a possible truth is true in at least one possible
world; a contingent truth is true in some but not all possible worlds. Gödel’s theorem
on God’s existence in Section 2.2 is expressed in modal logic.
6 Philosophical Theories of Knowledge and Truth

of the universe, all reality and knowledge. It is not clear what a monad
is; philosophers did not even seem to agree with each other about it.
In deductive logic, all propositions have to rely on previously established
true propositions, so to break the infinite regress, we have to start with
propositions that do not require a proof. These are the axioms, whose
nature of truth remained somewhat mysterious and contentious until
the 20th century. The early rationalists simply assumed that, by the
power of reason, we could recognize the axioms as self-evidently true.
In modern mathematics, this position is no longer common. Axioms are
now considered true only by definition or by construction.
Immanuel Kant (1724–1804) is widely regarded as one of the most in-
fluential figures in western philosophy. We shall give him a lot of space
befitting his importance. Writing about 150 years after Descartes’s Med-
itations, in his Critique of Pure Reason (1781; second edition 1787) he
lamented the state of philosophy (‘metaphysics’) which, he said, was
ruled by ‘the antiquated and rotten constitution of dogmatism.’ To bring
things into order, he introduced the concepts of (i) analytic vs synthetic
knowledge, and (ii) a priori vs a posteriori knowledge. (Kant used the
word ‘judgement’ or ‘cognition,’ not exactly the same as knowledge, but
for uniformity we will use ‘knowledge.’)
A priori knowledge is independent of any experience whatsoever, as op-
posed to a posteriori knowledge, which must be established by sensory
experience. Analytic knowledge is tautological; it’s true by construction
or by definition, not by checking against reality. For instance, the sen-
tence ‘All bachelors are unmarried’ is analytic. It is certain, but the
certainty comes with a price: It does not contain any factual informa-
tion. On the other hand, synthetic knowledge is non-tautological: It’s
built from a combination of two or more unrelated concepts in a new
way, so it’s not necessarily true. For instance, ‘All bachelors are happy’
is synthetic, since happiness is not built in the concept of bachelorhood,
and its truth needs to be checked against reality. Kant argued strongly
that the statement ‘7+5=12’ is synthetic, supposedly because the idea
of ‘12’ is not contained in ‘7’ or ‘5.’ It’s a highly questionable view, since
surely no one can find any factual evidence that can negate the truth
of 7+5=12. Indeed, the 20th-century logical positivists considered it an-
alytic, as it is a consequence of Peano’s axioms of arithmetic together
with the definition of the ‘+’ operation (Hempel, 1945a).
The Rationalists 7

It may help us to understand Kant’s theory more easily if we put his

analytic-synthetic concepts in parallel to Leibniz’s necessary-contingent
truths. We might concede a slight difference between analytical and nec-
essary truths: Axioms are analytically true, as they are true by defini-
tion. Whereas, given axioms, theorems are necessarily true, since their
negation implies a contradiction. But we could also argue that theorems
are true by construction, hence analytic. In any case, the synthetic and
the contingent truths look and smell the same.
Kant’s great philosophical question was whether synthetic a priori
knowledge is possible. To be a priori, knowledge has to be necessarily and
universally true, and innate in the human mind, that is, not abstracted
from experience. But, to be synthetic, it has to combine unrelated con-
cepts, so it’s not necessarily true. So ‘synthetic a priori knowledge’ ac-
tually sounds contradictory, but he argued that we do have synthetic a
priori knowledge. In mathematics, such knowledge is contained in the
principal rules of logic and, at the time, Euclid’s axioms in number
theory and geometry. To some extent, he continued, such knowledge is
also found in the basic laws of physics, e.g., Newton’s laws or the law
of conservation of mass. In effect, Kant elevated the lame ‘self-evident’
qualification of axioms to a philosophically sophisticated ‘synthetic a
priori’ label.
His ultimate goal was to reconstruct metaphysics using only synthetic
a priori knowledge, with the expectation that it would be as pure and
universal as mathematics. He identified such axiomatic knowledge in our
understanding of space and time, and our grasp of cause and effect, which
are crucial for organizing and making sense of our sensory experiences.
Kant’s ideas about space and time have been understood as innate forms
of perception, known as outer and inner intuition.
As these innate abilities are not learned from experience, they are pre-
sumably present at birth. Due to the lack of clear examples, it is uncer-
tain whether his metaphysical concept of synthetic a priori knowledge
falls under the category of what is commonly called instinct. For spiders,
the ability to weave a web must be synthetic a priori, since it’s com-
pletely innate and independent of experience. It’s coded in their genes.
Similarly, human linguistic skill, especially grammar, is an instinct, an
innate ability encoded in the genes that shape our brain. Exposure alone
does not explain children’s depth of understanding and mastery of the
grammar of their native language. Our intuitive grasp of cause and effect
8 Philosophical Theories of Knowledge and Truth

also appears to be encoded in the brain: Damage to the frontal cortex

can hamper or even erase our causal reasoning.
But all these instincts are products of evolution, a long-term adaptation
based on experience living in a specific natural environment. During crit-
ical periods in childhood, our plastic brain needs environmental input in
order to develop properly. For example, a deaf child will grow up mute
due to lack of sound input. In ground-breaking studies that eventually
led to the Nobel Prize, David Hubel and Torsten Wiesel in the 1960s
showed that, during its third to eighth week, a kitten must receive vi-
sual stimulation to avoid blindness. Hence, our mental faculty is not a
priori in the Kantian sense, which requires absolute independence from
experience. The main problem with Kant’s theory is that it depends
so much on mental cognition, but the knowledge of neuroscience in the
18th century was virtually nonexistent.
Yet, Kant’s theory of knowledge survives to the present day, with asso-
ciated books and papers still being published, imposing itself on those
genuinely interested in the philosophy of knowledge. He was an ener-
getic obscurer of language: Reading him, if you have to, should only be
done when your brain is in its freshest state. He loved befuddling words
like ‘apodeictic’ or ‘noumena.’ The word ‘transcendental’ appears more
than 500 times in the Critique, often in unexpected combinations with
no obvious meaning. Here is the reason:

I apply the term transcendental to all knowledge which is not so much

occupied with objects as with the mode of our cognition of these objects,
so far as this mode of cognition is possible a priori.

Thus, ‘transcendental aesthetic’ refers to our knowledge of space and

time, while ‘transcendental analytic’ refers to our knowledge of cause
and effect. Subsequent philosophers disagree with what he meant in his
basic pronouncements, for instance, on the nature of space. His funda-
mental views of the axioms in geometry and our sense of space and time
proved to be incorrect, as clearly seen after the establishment of non-
Euclidean geometry and Einstein’s theories of relativity. Logical pos-
itivists in the 20th century do not consider mathematical axioms as
synthetic, but analytic; see more below. Any physicist today will have a
completely different intuition about space and time compared to 18th-
century physicists, so our mind does not have an a priori notion for space
and time.
The Rationalists 9

With the ‘transcendental idealism’ (or ‘ı̀deality’ in some translations),

Kant claimed that our mind actively constructs our experience and im-
poses its own concepts and categories on sensory input from the external
world. In other words, all empirical objects are mind-dependent: What
we experience is not the ‘world in itself’; we cannot know ‘things in them-
selves,’ but only our mind’s representation of it. To some extent a similar
‘idealism’ also appears in N.R. Hanson’s and T. Kuhn’s contention in
the 20th century that our observations are ‘theory-laden.’ That is, our
effort to understand the world is shaped by previous theories or models
created in our mind to make sense of it.
We can view Kant’s idealism from two opposing perspectives. First,
from the biological perspective, it is actually very close to how modern
neuroscience views our understanding of the world as a product of our
brain’s processing of external input. It is well known that our brain has
a ‘mind’ of its own, which, disturbingly, can be independent of reality.
One of the most dramatic examples is the phenomenon of the phantom
limb (Doidge, 2008), where an amputee still feels the presence of the am-
putated limb and even feels pain on it. But, from the second perspective,
at least in physics, to say that we cannot know ‘things in themselves’
sounds anti-scientific, since the purpose of science is to explain nature
objectively, i.e., independently of our mind. The scientific view is more
like this: There is no limit – or, we do not know the limit – in our
understanding of nature. The relativity and quantum theories have re-
vealed strange properties of nature that are far beyond a priori human
cognition. Physicists would say, that’s the way nature is.
In summary, the primary lesson we get from the rationalists is a spe-
cial respect for the ‘mind,’ with its powerful reasoning, truth-preserving
deductive logic and independent existence from the body. Descartes’s
model of mind-body duality survived well into the 20th century. In this
model, the brain is an inert physical machine run by the lively mind or
the soul – the ghost in the machine. Alas, the ghost was never found. In
recent times, neuroscience has broken this Cartesian duality and shown
that the mind is just a result of brain anatomy and processes. This
implies that the mind has no special access to the truth a priori.
10 Philosophical Theories of Knowledge and Truth

1.2 The Empiricists

With his emphasis on experience and observation as a source of knowl-

edge, Aristotle is considered the father of empiricism. His 10-volume
History of Animals alone contained detailed observations and descrip-
tions of over 500 species of animals, covering their anatomy, reproduc-
tion, behaviour, habitats, etc. In addition, he wrote four other volumes
on animals and at least eight volumes on physics.
However, Aristotle’s credibility as an empiricist is undermined by glaring
unforced errors that could be checked easily, such as men have more teeth
than women,2 and by fanciful theories such as this from The Generation
of Animals:

Again, more males are born if copulation takes place when north than
when south winds are blowing. For in the latter case the animals pro-
duce more secretion, and too much secretion is harder to concoct; hence
the semen of the males is more liquid, and so is the discharge of the
catamenia [menstruation].

Bertrand Russell (2016, p. 7) remarked sarcastically that ‘although [Aris-

totle] was twice married, it never occurred to him to verify this statement
by examining his wives’ mouths.’ Was Russell just cherry picking? A
classicist, Peter Gainsford, checked 21 statements Aristotle made about
teeth: he was 83% correct.3 That’s an excellent mark for a student in
an exam, but perhaps not good enough for a reference text by a great
philosopher. For instance, which of the following statements from The
History of Animals would you trust: (i) Elephants have two tusks: large
and bent upward in the case of males, small and bent downward in the
case of females. (ii) Elephants have teeth at birth?
Did Aristotle not care about the truth? In the passage on male vs female
births in The Generation of Animals, he actually wrote: ‘Observed facts
confirm what we have said.’ So, he was aware of the ‘facts’ and he was
not making things up. The problem was obviously in the poor standard
of proof of empirical statements. It’s an interesting historical question
2
Full quote from The History of Animals: ‘Males have more teeth than females, in
the cases of humans, sheep, goats, and pigs. In other species an observation has not
yet been made.’
3
http://kiwihellenist.blogspot.com/2017/09/aristotles-errors.html.
The Empiricists 11

in itself, but in any case it highlights that the meticulous fact checking
in the culture of modern science is actually an achievement we cannot
take for granted.
In Metaphysics, Aristotle discussed an early version of the principle of
sufficient reason, also called the principle of causality: Things exist or
happen for a reason, which can be discovered by observation. This was
a move away from Plato’s idealism, which considers the ultimate reality
to be an eternal non-material realm – the so-called Platonic world – of
perfect Ideas and Forms. For an extreme empiricist, all knowledge comes
from experience. But real empiricists, such as Aristotle or Hume do
consider mathematics, particularly geometry, as a product of thinking.
Aristotle also believed in ultimate reasons that are not verifiable by
experience or observation.
Aristotle had an interesting view on knowledge that is based on convinc-
ing others rather than oneself. In his Rhetoric he wrote that a speaker
has three ways to persuade his audience: by the power of his personal
character, by appealing to their emotion, or by providing arguments.
Of the arguments, he provided a demonstrative/deductive logical argu-
ment or an inductive argument based on examples or analogy. As for
the demonstrative arguments, they must be based on some organizing
principles – the axioms. He spent a lot of time explaining how to ar-
rive at the principles, where he emphasized the role of experience and
induction.
In the Enquiry Concerning Human Understanding, a remarkable 121-
page book first published in 1748, Hume (1711–1776) wrote (p. 16)
what’s clearly a reaction to the continental rationalists and emphasized
the importance of sensory experience:

All ideas, especially abstract ones, are naturally faint and obscure: the
mind has but a slender hold of them: they are apt to be confounded
with other resembling ideas.

On the contrary, all impressions, that is, all sensations, either outward
or inward, are strong and vivid: the limits between them are more ex-
actly determined. When we entertain, therefore, any suspicion that a
philosophical term is employed without any meaning or idea (as is but
too frequent), we need but enquire, from what impression is that sup-
posed idea derived? And if it be impossible to assign any, this will serve
to confirm our suspicion.
12 Philosophical Theories of Knowledge and Truth

For scientists, it is easy to side with Hume. It’s ironic that Kant credited
Hume for waking him up from his ‘dogmatic slumber,’ as he became a
prolific inventor of obscure philosophical terms.
Hume (p. 18) categorized all subjects of human reasoning into two dif-
ferent types: (i) Relations of Ideas, which include geometry, logic and
arithmetic, where every affirmation is either intuitively or demonstra-
tively certain. The truth here is ‘discoverable by the mere operation of
thought, without dependence on what is anywhere existent in the uni-
verse.’ (ii) Matters of Fact, a completely different subject from the first,
since their negation never implies a contradiction, so they are not nec-
essary truth. This division has close parallels to Leibniz’s necessary and
contingent truths.
Hume clearly recognized the need for special reasoning to deal with facts,
which we now see as inductive reasoning. Though he himself failed to
solve it, it is instructive to follow closely Hume’s thoughts and open
struggle. The principle of sufficient reason would only state that there
must be a reason underlying the matters of fact, but it does not provide
any method of reasoning. Hume realized that a matter of fact must be
established by another fact; all reasoning concerning them relies on the
causal relationship; and the causal knowledge comes not from thinking
a priori, but from experience.
To appreciate Hume’s difficulties, it’s instructive to recognize the differ-
ent layers of ‘Facts’ in terms of verifiability. There are simple facts, the
kind reported in newspapers, that are easily verifiable and those that are
not. The correlation between smoking and cancer is a fact that is easy to
verify, but whether ‘smoking causes cancer’ is not so easy to verify. Also
easy to verify is a claim that when John was very sick with covid-19, he
improved after taking ivermectin, but it’d be impossible to verify that
it was ivermectin that cured him. It could just be due to the passage
of time. It’s more feasible, though still a nontrivial effort via a rigorous
clinical trial, to verify whether ivermectin has any curative benefit in
a group of people. Seeing a flock of black ravens is a simple fact, but
the claim that ‘all ravens are black’ is no longer a verifiable fact. In-
deed, these claims should actually be considered ideas, more specifically
‘contingent ideas,’ whose truths are only contingent.
In general, facts that are unverified, difficult or virtually impossible
to verify directly, such as the composition of the Earth’s core or the
properties of black holes or the existence of dark matter, resemble
The Empiricists 13

theories – hence ideas – rather than mere ‘Facts.’ All statistical hypothe-
ses and scientific theories, and their unverified predictions, are contingent
ideas; calling them ‘Facts’ just sounds wrong, even though we use facts
to establish them or in the future some of them might be considered
facts. So Hume’s hard separation of Ideas and Facts is too simplistic:
Conceptually, in fact, the trickiest reasoning problem is found in the
Relations between Facts and Ideas.
Mathematics and deductive logic are used to establish necessary truths
based on the Relations of Ideas. But what kind of reasoning can be
used to establish the truth of a contingent idea? Once we leave deduc-
tion, we’re on an uncertain ground. Unlike deduction, induction is not
truth-preserving: Correct premises/facts do not necessarily lead to cor-
rect conclusions. As these conclusions are often in the form of theories
that aim to explain the facts, we may say that theories are underde-
termined by facts. Even the most bizarre conspiracy theory is based on
facts, albeit selective. Epidemiologists have 74 names for different biases
lurking when we try to draw conclusions from data (Delgado-Rodriguez
and Llorca, 2004). Not all of these biases are distinct, but they nonethe-
less highlight a high number of potential pitfalls that we can encounter
when analysing observational data. To make matters worse, in some
cases, induction is discontinuous: adding a new fact could lead to a
dramatically different conclusion; just think of the crime stories. As a
consequence, overall, there is no inductive method that can guarantee
the truth of its conclusions.
Hume spent a large portion of his book on ideas or propositions in which
we gain confidence from repeated occurrences of simple verifiable facts.
It’s a pleasant monotone property of a certain type of induction problems
that actually makes them amenable to a proper deductive analysis.
Let’s consider a modern version of his example. You sometimes feel
nauseous, bloated and constipated. You guess it must be food related,
but since you’re an omnivore and have never heard of coeliac disease, you
can’t figure out the cause. After some trials and errors, you’re suspicious
of bread and now start paying a close attention. The next time you eat
it, you notice the symptoms. Aha, that’s the culprit! Whenever you eat
bread again, you are going to expect the same experience, though that’s
not necessarily true because the symptoms maybe due to some types of
bread only. Your confidence in the cause of your problem will grow the
more times it occurs. Here comes his key question: If causal knowledge
14 Philosophical Theories of Knowledge and Truth

comes from experience, what kind of reasoning is used to establish the

knowledge? Hume (p. 25) seemed genuinely puzzled:

But if you insist that the inference is made by a chain of reasoning,

I desire you to produce that reasoning. The connexion between these
[cause-effect] propositions is not intuitive. There is required a medium,
which may enable the mind to draw such an inference, if indeed it be
drawn by reasoning and argument. What that medium is, I must confess,
passes my comprehension; and it is incumbent on those to produce it,
who assert that it really exists, and is the origin of all our conclusions
concerning matter of fact.

In other words, he was asking an inductive inference question, which was

not yet available until Bayes and Laplace came to the scene later in the
century (Chapter 8). Hume (pp. 26–28) could only argue the following:

In reality, all arguments from experience are founded on the similarity

which we discover among natural objects, and ...

From causes which appear similar we expect similar effects. This is

the sum of all our experimental conclusions.

Now it seems evident that, if this conclusion were formed by reason,

it would be as perfect at first, and upon one instance, as after ever so
long a course of experience. But the case is far otherwise.

It is only after a long course of uniform experiments in any kind [our

emphasis], that we attain a firm reliance and security with regard to a
particular event.

Hume was often said to be skeptical of induction, but he actually stated

explicitly that he was not a skeptic:

Now where is that process of reasoning which, from one instance, draws
a conclusion, so different from that which it infers from a hundred in-
stances that are nowise [sic] different from that single one?

This question I propose as much for the sake of information, as with an

intention of raising difficulties. I cannot find, I cannot imagine any such
reasoning. But I keep my mind still open to instruction, if any one will
vouchsafe to bestow it on me.

As an agent, I am quite satisfied in the point; but as a philosopher,

who has some share of curiosity, I will not say scepticism, I want to
learn the foundation of this inference.
The Positivists and the Verification Principle 15

Let’s call this Hume’s challenge. However, due to a lack of inductive

inference methodology in his time, Hume (p. 31) could only come up
with a simplistic and wholly unsatisfactory explanation to the induction
problem by appealing to the ‘principle of Custom or Habit’:

All inferences from experience, therefore, are effects of custom, not of

reasoning.

The growth of statistics and scientific methods can be seen as a system-

atic effort to provide rigorous methodologies to answer Hume’s challenge.

1.3 The Positivists and the Verification Principle

During the 19th century, the positivists – such as Auguste Comte and
John Stuart Mill – continued to focus on empirical evidence, but with
more emphasis on scientific methods, particularly in the natural and so-
cial sciences, as the basis for understanding and explanation. In the 20th
century, positivism was further developed by the philosophers of the Vi-
enna Circle into logical positivism or logical empiricism. The main point
of view is that philosophy must use the same empirical methodology as
other sciences. In other words, there is no methodology that is unique to
philosophy. Verification is taken as a central principle, to the point that
what cannot be verified empirically is meaningless. Only mathematical
and linguistic questions can be answered without reference to empiri-
cal support. Traditional metaphysical questions are out as far as true
knowledge is concerned. Again, here we see the clear division between
necessary and contingent truths.
Although positivists are all empiricists, compared to early empiricists
such as Hume, the logical positivists pay more attention to mathemat-
ics and logic and place more emphasis on linguistics. The latter is due
to Wittgenstein’s influence. Non-Euclidian geometry, the axiomatization
of mathematics, Gödel’s incompleteness theorems (1931), etc., have con-
tributed to the new perspectives. Euclid’s axioms are no longer seen as
self-evidently true a priori, but simply as assumptions for an abstract
spatial model. Pure geometry, arithmetic, etc. are seen as a collection
of analytic statements, which are only true by definition. So, strictly
16 Philosophical Theories of Knowledge and Truth

speaking, we can’t say that Euclid was right or wrong, but only whether
his geometry is consistent or not; for further discussion, see, e.g., Hempel
(1945a). When geometry is applied to the real world, it becomes ap-
plied mathematics, part of the empirical sciences, where assumptions
and models have to be checked against reality. Hence, assumptions are
no longer analytic statements, but contingent or synthetic.
The views of logical positivists are recognizably scientific, but some
philosophers such as Karl Popper or Bertrand Russell criticized them as
being too narrow. Who will deal with questions of ethics, aesthetics or
the meaning of life? The verification principle is also often interpreted
loosely and hence unspecific. What exactly does it mean? In the cer-
tain world of mathematics, an analytic statement is ‘verified’ when it is
demonstrated in a proof, which must consist of a finite number of steps
only. Completeness and finiteness of the proof are essential, not optional,
requirements. So, ideally, we should also have the same requirements for
contingent truth. But, in reality, as we have described before, there are
many levels of verification.
Commonly solvable things in life often satisfy the requirements, though
not always. If you’re asked to verify your identity by the authorities,
you can do so completely in a finite number of steps, e.g., you can show
your passport or utility bill. If there is something wrong with your car,
a good mechanic can usually find the cause in a finite time and verify
it by fixing it. Political opponents during an election campaign would
verify each other’s claims, and it’s in their interest to do it completely
and immediately.
However, if something is wrong with your body – a much more com-
plex system – a doctor may take a long time to get a proper diagnosis;
in some cases, there may not be a definitive diagnosis. And, the puta-
tive diagnosis may not be verified if the supposed treatment does not
work. Or you may get a diagnosis with the adjective ‘idiopathic,’ such
as idiopathic scoliosis, which actually means ‘unknown cause.’
In court, a place where truth is supposed to be paramount, verification
is also not a simple matter. Did O.J. Simpson kill his ex-wife? Whose
evidence or arguments do you trust? The defence’s or the prosecutor’s?
And what verdict is correct? In the criminal court, Simpson was declared
not guilty. But the subsequent civil court ruled against him. ‘Verification’
is judged with the qualitative threshold of ‘beyond reasonable doubt’ in
the criminal court, but only with the ‘balance of probability’ in the
The Positivists and the Verification Principle 17

civil court. In both courts, various technicalities apply regarding what

evidence can be included, etc.
Is science better at verification? Let’s take an example: How do we verify
that the core of the Earth consists of iron? We can never verify it directly.
Is it enough to verify indirectly, for instance, by inferring it from the
Earth’s magnetic field? Much more is now known about the Earth’s
core: its composition, temperature, pressure, etc., all necessarily derived
from indirect evidence. Indirect verification in science is indeed common,
so while scientists often consider their theories as ‘verified,’ the theories
are only considered provisionally true.
How would you verify the continental drift theory, as proposed in 1912
by Alfred Wagener, that the continents started as a single land mass and
later moved apart? If you treat the continents as jigsaw puzzle pieces,
some pieces, such as South America and West Africa, or East Africa and
India, would fit pretty well. But, although you can’t imagine how such a
fit could occur by accident, it was only considered as circumstantial evi-
dence, not sufficient to verify such a remarkable claim. Wagener himself
proposed various indirect evidence, but for 50 years geologists rejected
the theory because there was no good causal/mechanistic explanation.
Keith Runcorn’s research on paleomagnetism and polar wandering in
the 1950s revived the continental drift theory. The convincing ‘verifica-
tion’ came only when Harry Hess showed evidence of seafloor spreading
around 1960.
Even if a theory has only a finite number of consequences, not all can
or will be checked for the theory to be considered verified. A rich the-
ory, such as the general theory of relativity, generates many testable
predictions over a long period of time. To validate his general theory
of relativity, Einstein made three predictions: (i) the anomaly of the
perihelion movement of mercury; (ii) the deflection of light during solar
eclipse; (iii) the gravitational redshift. These were verified in 1915, 1919
and 1954, respectively. At what point is the theory considered ‘verified’ ?
Physicists paid attention to Einstein’s theory after the verification of the
first prediction, and Einstein became a global celebrity-scientist after the
second verification. The gravitational redshift has continued to be ver-
ified until recently, involving better precision or data from more and
more galaxies. Other predictions of the theory– such as the expanding
universe, the black holes and the big bang – came later, each time requir-
ing more and more time to verify. So, in general, ‘complete verification’
is not required in science.
18 Philosophical Theories of Knowledge and Truth

Even a simple statement such as ‘All humans are mortal’ is not strictly
verifiable, since it will require the verifier to wait until the end of the
human race. This violates the completeness and finiteness requirements.
We cannot say ‘All humans so far are mortal.’ Who knows, among those
alive today, some may be immortal when scientists discover the secret of
immortality. Do we have to qualify the statement to something like ‘All
humans born before the year 1800 are mortal’ ? This is of course silly,
since we might as well say ‘All humans that have died are mortal,’ so its
truth is no longer contingent, but necessary and contains no information.
How do we verify that mRNA vaccines work against covid-19? First of
all, it is not required that the statement is true for all patients, for all
variants of the virus and at all times. As long as it is better than placebo
by a certain margin, for a relevant variant of the virus at the present
time, the statement is considered verified by the authorities and, we
hope, by the public. Even with these lower standards, the clinical trials
required to verify the statement convincingly were a multi-billion dollar
effort.
In summary, like Hume with his ‘Matters of Fact,’ logical positivists
overestimate our ability to verify, so tying meaning and truth to veri-
fication becomes problematic. Instead of ‘verified,’ Carnap introduced
the notion ‘confirmed,’ where confirmation does not require a full veri-
fication (Section 8.6). At the other extreme, Popper suggested that, to
be meaningful, a scientific hypothesis does not have to be ‘verified’ or
even ‘verifiable’; it only needs to be ‘falsifiable.’ This view is closer to
the scientists’ real modus operandi with their provisional theories.
2
Deduction and Induction

To sum up our philosophical explorations in the previous chapter, we can

point to the recurring theme of necessary and contingent truths, along
with the deductive and inductive reasoning that go with them. We shall
now go a little deeper into these modes of reasoning. Deduction is the
primary tool of the rationalist. In this mode of reasoning, one applies
pure reason to prove correct propositions and to disprove incorrect ones,
starting with central propositions presumed to be true. Euclid’s geom-
etry is seen as the epitome of a deductive axiomatic system. Induction
is also known as analogy, examples or generalization. It is based on the
empiricist tradition developed by Locke and Hume, and later by statis-
ticians as well as natural and social scientists. With induction, one does
not ‘prove’ an objective truth, but only provides evidence or support
through experimentation and observations. Philosophers seem to recog-
nize another form of reasoning called ‘abduction,’ but as we explain in
Section 8.6, it is just a form of induction.
Deductive reasoning is highly attractive as it leads to certain conclu-
sions. This might explain why we’re drawn to generalizations, which can
then function as basic premises in the deductive process. It might be
argued that in a deductive process, no new information is produced,
only seemingly tautological consequences of old information. Whereas
in induction, the explanatory hypothesis can be a new creative inference
suggested by the data, so the process generates new knowledge. But let’s
think more carefully: Theorems in geometry are consequences of the ax-
ioms, but they constitute vast new knowledge. We’re all interested in
our bank statements, especially in the current balance, even though it is
just computed from other numbers following the deductive rules of arith-
metic. So, deduction does produce interesting new information. What’s
true is that deduction does not produce new factual knowledge; this is
fine, as not all knowledge is factual. Results from thinking alone can
produce astonishingly rich knowledge beyond what’s suggested by ob-
servations. In physics, starting with fundamental laws as axioms, one

19
20 Deduction and Induction

can derive an amazing array of novel theories and predictions, such as

the movement of planets, E = mc2 , gravitational waves, black holes, etc.
Among the questions at the start of Chapter 1, there is one simple one:
Is Socrates mortal? If this is Socrates the philosopher, then we know
that he is dead, so we can answer in the affirmative. But you may have
a neighbour named Socrates who is still very much alive. Yet, without
waiting for the news of his death, we also know he is mortal. We have
intuitively used a basic deductive reasoning known as Aristotle’s syl-
logism, a logical argument where a conclusion is drawn based on two
assumed propositions: (i) a major premise, taken from universal laws,
scientific theories, axioms in mathematics, basic propositions in philos-
ophy, creeds in religions, etc.; (ii) a minor premise, stating a specific
instance, evidence, data or observation. Thus, in this case we intuitively
use a general proposition as the major premise

G: All humans are mortal,

and the minor premise is that Socrates is a human. To obtain a specific

proposition as the conclusion,

E: Socrates is mortal.

As long as we accept the two premises, we must accept the mortality of

Socrates; there is no need for direct evidence.
In deductive reasoning, specific propositions follow from the general
proposition. So, if we want to establish a specific proposition E, we
look for a general proposition G whose truth can be easily accepted. On
the other hand, the specificity of E means that it can often function as
a testable prediction of the general proposition. Logically, if E is false,
then G is falsified. If there existed an immortal human, then G cannot
be true. (Here we see the formality of a deductive system, as it depends
on or assumes unambiguous definitions. In reality, our idea of mortality
may vary depending on how we view our existential self: If we could
preserve our cell/genome indefinitely so that we can be cloned at any
time in the future, would we be considered immortal? This has already
been achieved with some animals, such as dogs or even camels. What if,
additionally, we could also find a way to preserve our core personality,
memories and consciousness?)
Deduction and Induction 21

Although deductive reasoning is natural in mathematics, it also forms

a large part of physics, with known or accepted laws taking the role of
general propositions. Highly accurate knowledge of the planetary mass,
positions, motion, etc. can be derived from Newton’s laws. Let’s consider
the question: Is the proton immortal? Matter consists of protons, neu-
trons and electrons. Theoretically, the laws of conservation would imply
that electrons are immortal, but there are no laws against the decay of
protons and neutrons. Experimental evidence has indeed shown that free
neutrons outside the nucleus decay within minutes. So far, there is no
evidence to show the decay of protons. Which fact would violate the laws
of physics: the immortality or the mortality of protons? Mathematically,
the so-called Grand Unified Theories actually predict that protons must
decay. It is, in fact, the strongest testable prediction of the theories.
In deductive reasoning, as long as G is assumed true, E must be true.
But how do we know that G is true? Observing that E is true does
not establish the truth of G. General propositions such as G cannot be
entirely observed because we can’t wait to see the death of all human
beings. What we can observe is evidence (specific event) of G. For exam-
ple, we see that a particular person, named Socrates, died at some point
and we can confirm his death officially. This is the problem we face in
deduction: the construction of the basic propositions and the fact that
the very foundation of our deduction – the truthfulness of basic propo-
sitions – can be in doubt. Early rationalist philosophers simply claimed
that there are self-evident true general propositions.
For ‘Matters of Fact,’ Hume argued against the existence of innate, a
priori true propositions, positing that all human knowledge is founded
solely on experience that serves as evidence, not from theoretical de-
ductions. Furthermore, inductive reasoning and the belief in causality
cannot be rationally justified via deduction. Instead, our trust in causal-
ity and induction results from our customs and mental habits and is only
attributable to our experience of the ‘constant conjunction’ of events.
This is because we can never actually perceive the causality of events. We
can only perceive that two events are conjoined. Hume’s skepticism had
a great impact on the development of justifiable epistemic knowledge.
Unlike Hume’s skepticism, which challenged the possibility of attaining
certain knowledge, fallibilism does not reject the existence of knowl-
edge. Rather, fallibilists acknowledge that our current understanding
of the world may be incomplete or mistaken, and that even our most
firmly held beliefs are subject to revision in light of new evidence or
Random documents with unrelated
content Scribd suggests to you:
"I am," remarked Toby, as he lifted his glass, "a prophet in a small
way. Old boy, your hand. To the health of our double marriage--and
no heeltaps."

Archie finished his glass.

CHAPTER XII.

ARS AMORIS.

'Tis very easy to make love;

A smile--a pressure of the hand.
A reference to the stars above,
A "fly with me to some far land,"
A sigh as soft as coo of dove,
A kiss--the rest she'll understand.

Mr. Gelthrip, thinking no one but himself knew anything, had

contradicted his clerical superior on some point connected with the
introduction of printing into England, and the vicar in great wrath
had carried off his dogmatic curate to the library in order to prove
his case. The two elder ladies were talking about Sir Rupert as Mrs.
Valpy had met him a few months previously, and Mrs. Belswin was
trying to find out all about her quondam husband, in order to
strengthen her position as much as possible. At present she knew
that she was entirely at the mercy of Sir Rupert, so if she could
discover something detrimental to his character it might serve as a
weapon against him. The scheme which she hoped to carry through
with the assistance of Ferrari, was a dangerous one; and moreover,
she was doubtful if the Italian would consent to aid her; therefore
she was anxious to try all other methods of coercing her husband
before resorting to the last and most terrible expedient. She was a
clever woman, was Mrs. Belswin, and the instinct for discovery,
which she inherited from her savage grandparents, made her
wonderfully acute in cross-examining simple Mrs. Valpy, who not
comprehending the subtlety of her companion, told all she knew
about the baronet in the most open manner. The result was not
gratifying to Mrs. Belswin; for with all her dexterity in twisting, and
turning and questioning, and hinting, she discovered nothing likely
to compromise Sir Rupert in any way.

"It's no use," she thought, with a feeling of despair in her heart,

"Rupert has it all his own way, and I can do nothing--nothing except-
---"

She smiled significantly, and simple Mrs. Valpy, seeing that the
companion was looking at Toby and her daughter, who were
amusing themselves at the piano, misinterpreted the smile, and
therefore spoke according to her misinterpretation.

"They'll make a very happy couple, won't they, Mrs. Belswin?"

Mrs. Belswin, thus being appealed to, started, smiled politely, and
assented with much outward show of interest to the remark of the
old lady.

"It's so nice for Toby to have his home here," pursued Mrs. Valpy,
with much satisfaction; "because, you know, our place is not far
from the vicarage, so I shall not be parted from my daughter."

The other woman started, and laid her hand on her breast, as if to
still the beating of her heart.

"Yes; it would be a terrible thing to part with your only child," she
said in a low voice. "I know what the pain of such a separation is."
"You have parted from your child, then?" said Mrs. Valpy,
sympathetically.

Mrs. Belswin clutched her throat, and gave an hysterical laugh.

"Well, no; not exactly;" she said, still in the same low voice; "but--
but my little daughter--my little daughter died many years ago."

It was very hard for her to lie like this when her daughter was only a
few yards away, chatting to Maxwell at the window; but Mrs. Belswin
looked upon such necessary denial as punishment for her sins, and
accepted it accordingly.

"I'm very sorry," observed Mrs. Valpy, with well-bred condolence.

"Still, time brings consolation."

"Not to all people."

"Oh, yes, I think so. Besides, now you have that dear girl, Kaituna,
and she seems very fond of you."

"Yes."

She could say no more. The strangeness of the situation excited her
to laughter, to that laughter which is very near tears, and she was
afraid to speak lest she should break down.

"And then Sir Rupert will be so glad to find his daughter has such a
good friend."

The mention of the hated name restored Mrs. Belswin to her usual
self, and with a supercilious glance at the blundering woman who
had so unconsciously wounded her, she answered in her ordinary
manner--

"I hope so! But I'm afraid I shall not have an opportunity of seeing
Sir Rupert at once, as I go to town shortly, on business."
"But you will return?"

"Oh, yes! of course I shall return, unless some unforeseen

circumstances should arise. We are never certain of anything in the
future, you know, Mrs. Valpy."

"No, perhaps not! At all events I think you will like Sir Rupert."

Mrs. Belswin sneered.

"Oh, do you think so?"

"I'm certain. Such a gentlemanly man. Quite young for his age. I
wonder he does not marry again."

"Perhaps he had enough of matrimony with his first wife," said Mrs.
Belswin, coolly.

"Oh, he was devotedly attached to her."

"Was he, indeed?"

"Yes! Simply worshipped her. She died in New Zealand when Kaituna
was a baby, I believe, and Sir Rupert told me how this loss had
overshadowed his life."

"Hypocrite!" murmured Mrs. Belswin, between her clenched teeth.

The conversation was becoming a little difficult for her to carry on,
as she dare not disclose herself yet, and did not care about
exchanging complimentary remarks on the subject of a man she
detested so heartily.

At this moment Toby struck a chord on the piano, and Tommy burst
out laughing, so, with ready wit, Mrs. Belswin made this interruption
serve as an excuse to break off the conversation.
"The young people seem to be merry," she said to Mrs. Valpy, and
rising to her feet, "I must go over and see what the joke is about."

Mrs. Valpy nodded sleepily, feeling somewhat drowsy after her

dinner, so Mrs. Belswin, seeing she did not mind being left to her
own devices, walked across to the piano and interrupted the two
lovers, for which interruption, however, they did not feel profoundly
grateful.

"Won't you sing something?" asked the companion, addressing Toby,

"or you, Miss Valpy?"

"Oh, my songs are too much of the orthodox drawing-room' type,"

replied Miss Valpy, disparagingly. "Now Toby is original in his ditties.
Come, let's have a little chin-music, Toby!"

"Wherever do you learn such slang?" said Mrs. Belswin, with a smile.

"Toby."

"I! Oh, how can you? I speak the Queen's English."

"Do you really?" said Tommy, laughing. "Well, I at present speak the
President's American, so go right along, stranger, and look slippy
with the barrel organ."

"If your mother hears you," remonstrated Mrs. Belswin, "she will----"

"Yes, I know she will," retorted Tommy, imperturbably; "but she's

asleep and I'm awake, very much so. I say, Mrs. Belswin, where's
Kaituna?"

"I think she's walking on the lawn with Mr. Maxwell."

"As a chaperon you should hunt them out," said Miss Valpy,
mischievously.
"Suppose I give the same advice to your mother," replied Mrs.
Belswin, dryly.

"Don't," said Toby, in mock horror; "as you are strong be merciful."

"Certainly, if you sing something."

"What shall I sing?"

"Anything," said Tommy, sitting down, "except that new style of

song, all chords and no tune."

Toby laughed mischievously and began to sing--

"If I mashed her would she kiss me?

No! no! no!
If I bolted would she miss me?
No! no! no!
She knows I haven't got a rap;
Besides, there is the other chap--
At him, not me, she sets her cap;
No! no! no!"

"Mr. Clendon," said Tommy, in a tone of dignified rebuke, "we don't

want any music-hall songs. If you can't sing something refined, don't
sing at all."

"I must collect my ideas first," replied Toby, running his fingers over
the piano. "Wait till the spirit moves me."

Mrs. Belswin had resumed her seat near the sleeping form of Mrs.
Valpy, and was thinking deeply, though her thoughts, judging from
the savage expression in her fierce eyes, did not seem to be very
agreeable ones, while Tommy leaned over the piano watching Toby's
face as he tried to seek inspiration from her smiles.
Outside on the short dry grass of the lawn, Kaituna was strolling,
accompanied by Archie Maxwell. The grass extended for some
distance in a gentle slope, and was encircled by tall trees, their
heavy foliage drooping over the beds of flowers below. Beyond, the
warm blue of the sky, sparkling with stars, and just over the
trembling tree-tops the golden round of the moon. A gentle wind
was blowing through the rustling leaves, bearing on its faint wings
the rich odours of the flowers, and the lawn was strewn with aerial
shadows that trembled with the trembling of the trees. Then the
white walls of the vicarage, the sloping roof neutral tinted in the
moonlight, the glimmer of the cold shine on the glass of the upstair
windows, and below, the yellow warm light streaming out of the
drawing-room casements on the gravelled walk, the lawn beyond,
and the figures of the two lovers moving like black shadows through
the magical light. A nightingale began to sing deliciously, hidden in
the warm dusk of the leaves, then another bird in the distance
answered the first. The hoot of an owl sounded faintly through the
air, the sharp whirr of a cricket replied, and all the night seemed full
of sweet sounds.

Kaituna sat down on a bench placed under the drawing-room

windows, and Archie, standing beside her, lighted a cigarette after
asking and obtaining the requisite permission. The voices of the
vicar and his curate sounded in high dispute from the adjacent
library; there was a murmur of conversation from within, where Mrs.
Belswin was talking to the other lovers, and at intervals the sharp
notes of the piano struck abruptly through the voices, the songs of
the nightingale, and the charm of the night.

"What I miss very much in the sky here," said Kaituna, looking up at
the stars, "is the Southern Cross."

"Yes; I have seen it myself," replied Archie, removing his cigarette.

"You know I have travelled a great deal."

"And intend to travel still more!"

"Perhaps."

"You don't seem very sure, Mr. Maxwell. What about South
America?"

"I thought I had told you that I had changed my mind about South
America."

Kaituna flushed a little at the significance of his words, and cast

down her eyes.

"I believe you said something about putting off your journey till the
end of the year."

"I'll put it off altogether, if a certain event takes place."

"And that certain event?"

"Cannot you guess?"

Duplicity on the part of the woman, who knew perfectly well the
event to which the young man referred.

"No, I am afraid I can't."

"Miss Pethram--Kaituna, I----"

"Hush! Mr. Clendon is singing."

It was only to gain time for reflection, as she knew that a declaration
of love trembled on his lips, but with feminine coquetry could not
help blowing hot to his cold.

And Toby was singing a bold martial song, with a curious

accompaniment like the trotting of a horse--a song which thrilled
through the listeners, with its fierce exultation and savage passion.
 On God and his prophet I seven times called me;
I opened the Koran--the omen appalled me;
I read it--thou wast to be bride to another;
I knew my betrayer, 'twas him I called brother,
Zulema! Zulema!

I sprang on my steed as he waited beside me,

Then rode through the desert with Allah to guide me;
Fierce blew the sirocco, its terrors were idle;
I galloped till dawn to be first at your bridal
Zulema! Zulema!

I rode to the tent-door, your father's tribe knew me;

They dreamed of the glory they'd gain if they slew me;
I dashed through the cowards--I met my betrayer,
He fell from his saddle, and I was his slayer,
Zulema! Zulema!

You ran from your dwelling--your father's spears missed me;

You sprang to my saddle with fervour to kiss me;
We broke through the press of your kinsfolk, my foemen;
I won thee, Zulema, so false was the omen;
Zulema! Zulema!

"Ah!" said Archie, with a long breath, when the fierce cry had rung
out for the last time, "that is the way to win a bride."

Kaituna thought so too, although she did not make any remark, but
the shrill savagery of the song had stirred her hereditary instincts
profoundly, and even in the dim moonlight Archie could see the
distension of her nostrils, and the flash of excitement that sparkled
in her eyes. It gave him an idea, and throwing himself on his knees,
he began to woo her as fiercely and as freely as ever her dusky
ancestors had been wooed in the virgin recesses of New Zealand
woods.

"Kaituna, I love you! I love you. You must have seen it; you must
know it. This is no time for timid protestations, for doubtful sighing.
Give me your hands." He seized them in his strong grasp. "I am a
man, and I must woo like a man. I love you! I love you! I wish you
to be my wife. I am poor, but I am young, and with you beside me, I
can do great things. Say that you will marry me."

"But my father!"

He sprang to his feet, still holding her hands, and drew her forcibly
towards him.

"Your father may consent--he may refuse. I do not care for his
consent or his refusal. Say you will be my wife, and no human being
shall come between us. I have no money. I will gain a fortune for
you. I have no home--I will make one for you. Youth, love, and God
are on our side, and we are made the one for the other. You must
not say no! You shall not say no. You are the woman needed to
complete my life; and God has given you to me. Lay aside your
coquetry, your hesitations, your fears. Speak boldly to me as I do to
you. Let no false modesty--no false pride--no maidenly dread come
between us. I love you, Kaituna. Will you be my wife?"

There was something in this akin to the fierce wooing of primeval

man. All the artificial restraints of civilisation were laid aside. The
doubts, the fears, the looks, the shrinkings, all these safeguards and
shields of nervous natures had vanished before this whirlwind of
passion, which bore down such feeble barriers set between man and
woman. As his eyes ardent with love, passionate with longing,
flashed into her own she felt her bosom thrill, her blood rush rapidly
through her veins, and, with an inarticulate cry, wherein all the
instincts she had inherited from her Maori ancestors broke forth, she
flung herself on his heaving breast.

"Kaituna!"

"Yes! yes! take me I take me! I am yours, and yours only."

 CHAPTER XIII.

EXIT MRS. BELSWIN.

She smiles she laughs! she talks of this and that--

To all appearances a very woman.
Ah! but that phrase bears deep interpretation--
"A very woman" is a treacherous thing;
Her smile's a lie--a lie to hide the truth,
For when the time is ripe for all her schemes
"A very woman" slips her smiling mask,
And lo! behold, a look which means, "You die."

One who has been in strange lands, and ventured his life in far
countries, is by no means anxious to court again the dangers he has
so happily escaped. The traveller, telling his tales by his lately gained
fireside, shudders as he remembers the perils he has dared, the
risks he has encountered, and is thankful for his present safety, so
thankful indeed that he is unwilling to place his life for the second
time at the disposal of chance.

It was somewhat after this fashion that Mrs. Belswin viewed her
present security in contrast to her past jeopardy. She had been a
free-lance, and adventuress, an unprotected woman at the mercy of
the world, so hard and pitiless to such unfortunates; but now she
had found a home, a refuge, a daughter's love, a bright oasis in the
desert of affliction, and she dreaded to be driven out of this peaceful
paradise, which held all that made her life worth having, into a
stormy world once more. Through perils more deadly than those of
savage lands, through storms more terrible than those of the ocean,
she had passed into a haven of tranquillity; but now that she was
tasting of the pleasures of hope and repose, it seemed as though
she would once more be driven forth to battle with her fellow-
creatures.

Her quondam husband held her fate in his hand. He had right and
might on his side, and she knew that she could expect no mercy
from one whom she had so deeply wronged. Had the positions been
reversed she felt that she would not have scrupled to enforce the
powers she possessed, and, therefore, never for a moment dreamed
that her husband would act otherwise. All she knew was that she
was now in Paradise, that she enjoyed her daughter's affection,
ignorant as that daughter was of the mother's identity, and that the
husband of her youth, and the father of her dearly-loved child would
expel her from this hardly won Paradise as soon as he discovered
her therein.

This being the case, she did not waste time in asking for a mercy not
likely to be granted, but set herself to work to find out some means
of retaining her position in defiance of her husband's enmity and
hatred. After her conversation with Mrs. Valpy, she saw that Rupert
Pethram had glossed over the affair of the divorce in order to avoid
all suspicion of scandal against himself and the mother of his child,
for he was unwilling that the child should suffer for the sin of her
parent. This was certainly a point in her favour, as by threatening to
denounce the whole affair if she was not allowed to retain her
position she could force him to acquiesce in her demand, in order to
avoid scandal.

But then if he, though keeping the terrible affair secret from the
outside world, told Kaituna all about her mother's disgrace, thus
destroying the love which the girl had for the memory of one whom
she thought was dead--it would be too terrible, as she could urge
nothing in extenuation of her sin, and would be forced to blush
before her own child. No, nothing could be done in that way. Should
she throw herself on the mercy of the man she had wronged? Alas!
she knew his stern nature well enough to be aware of the hopeless
folly of such an attempt. Looking at the whole affair in whatever way
that suggested itself to her fertile brain, she saw no means of
retaining her position, her child or her newly-found respectability,
except by enlisting the sympathy of Ferrari and----

But it was too terrible. It was a crime. Guilty as she was, to do this
would render her still more guilty. Even if she succeeded in getting
her husband out of the way, and it was not discovered by the law,
there was still Ferrari to be reckoned with. It would give him a
strong hold over her, which he would use to force her into marriage,
and then she would be still separated from her child, so that the
crime she contemplated would be useless.

To see this woman raging up and down her bedroom was a pitiful
sight. Flinging herself on her knees she would pray to God to soften
the heart of her husband, then, realising how futile was the hope,
she would start to her feet and think again of the crime she
contemplated committing with the assistance of her Italian lover. She
raged, she wept, she sighed, she implored. Her mood changed with
every tick of the clock; from hope she fell into despair; from despair
she changed once more to hope--tears imprecations, prayers,
threats, she tried them all in their turn, and the result was always
the same--absolute failure. She was dashing herself in vain against
an adamantine wall, for in her calmer moments she saw how
helpless she was against the position held by her husband--a
position approved of by law, approved of by the world. She could do
nothing, and she knew it.

Still, Ferrari!

Yes, she would go up and see him, for perhaps he could solve the
riddle which thus perplexed her so terribly. He would demand his
price, she knew him well enough for that. Well, she would pay it in
order to still retain possession of her child. Let her accomplish her
present desire and the future would take care of itself. So, Mrs.
Belswin, summoning all her philosophy to her aid, composed her
features, and told Kaituna that she was going up to London on
business.

"But papa will be here next week," said the girl in dismay.

"Yes; I'm sorry to go at such a time, dear," replied Mrs. Belswin, with
an immovable countenance, "but it is a very important matter that
takes me away."

"You will be back again soon?"

"In a fortnight at the least."

"Oh, I'm glad of that," said Kaituna, with a flush; "you know I want
you to help me gain papa's consent to my marriage with Archie."

Mrs. Belswin smiled bitterly as she kissed her daughter, knowing how
weak was the reed upon which the girl leaned. She ask Rupert
Pethram to consent to the marriage--she dare to demand a favour of
the man she had wronged for the child she had forsaken! She
almost laughed as she thought of the terrible irony of the situation,
but, restraining herself with her usual self-command, bade the girl
hope for the best.

"Your father must like Mr. Maxwell, he is such a charming young

fellow," she said encouragingly, "and as you love him so dearly, Sir
Rupert, for the sake of your happiness, may perhaps overlook his
want of money."

"But you will speak to papa, Mrs. Belswin?"

"Yes; if I see your father on my return I will certainly speak to him."

"How strangely you talk," said Kaituna, rather puzzled; "if you come
back in a fortnight you will be sure to see papa."
"Of course, dear! of course. I was only thinking that some
unforeseen accident----"

"Oh, no, no!"

"Kaituna, you love your father very dearly."

"Very, very dearly. He is all I have in the world."

It required all Mrs. Belswin's self-restraint to prevent her then and

there throwing herself into the girl's arms and telling her all. Such a
course, however, would have been worse than madness, so she was
forced to crush down her maternal feelings.

After this interview with Kaituna, she departed for London--departed

for the possible commission of a crime, and as the carriage left
Thornstream she looked back with a sigh to the girl standing on the
terrace.

"Perhaps I shall never see her again," she said, with a groan,
throwing herself back in her seat. "But no; that will never happen;
even if Rupert does turn me out of the house he will not tell Kaituna
anything to destroy her belief in her mother, so I shall some day
meet her with her husband."

Her lips curled as she said this, knowing well that Sir Rupert would
never give his consent to the marriage, and then she clenched her
hands with a frown.

"He must consent to the marriage--Kaituna's heart is set on it. He

can destroy my happiness, but I'll kill him before he destroys that of
my child."

And with this firm determination she left her husband's house--the
house in which she should have reigned a happy mistress and
mother, and the house into which she had crept like a disguised
thief, the house which she, in the mad instinct of her savage nature,
intended to deprive of its master.

While waiting on the railway platform for the London train, she saw
Samson Belk.

The relations between these two were peculiar. Ever since he had
seen her at his mother's cottage, Belk had followed her everywhere
like her shadow, much to Mrs. Belswin's astonishment, for, candid in
all things to herself, she could not conceive how a handsome young
man could leave younger women for one verging on middle age. Yet
such was the case. This bucolic man had fallen passionately in love,
and adored her with all the sullen ardour of his obstinate nature. He
was slow-witted, dull-headed, and it took a long time for an idea to
penetrate into his brain, but once the idea was there, nothing could
get it out again. This woman, so different from all he had known,
who spoke in a commanding way, who flashed her eyes fiercely on
all, as if they were her slaves, had, without a word, without a sign,
brought to his knees this uncultured man, who knew nothing of the
deference due to the sex, and whose only attributes were great
physical strength and a handsome exterior. Formerly, owing to these
advantages, he had gained admiration from all women, and in return
had treated them with brutal indifference, or scarcely veiled
contempt; but now the positions were reversed, and he was the
abject slave of this imperious queen, who looked down at him with
disdain. It was a case of Samson like wax in the hands of Delilah--of
Hercules subjugated by Omphale; and Samson Belk, with all his
virile strength, his handsome face, his stalwart figure, was crouching
like a dog at the feet of Mrs. Belswin.

He looked somewhat haggard as he came towards her and took off

his hat, Mrs. Belswin nodding coldly to him in return.

"Well, Mr. Belk," she said, indifferently, "what are you doing here?"

"I heard you were going to town, madam."

"Yes? How can that possibly concern you?" Belk stood twisting his
hat round and round in a sheepish manner.

"I thought I might be of service to you," he stammered, looking at

her portmanteau.

"Thank you, but there is no need. The porters will attend to all that,"
replied the lady, graciously. "But you don't look very well, Mr. Belk. I
suppose you've been drinking."

Candour was Mrs. Belswin's strong point, and looking at Belk as an

inferior animal, she treated him accordingly, but he seemed in
nowise displeased at her bluntness.

"No; I haven't been drinking, madam."

"That's just as well. You know Sir Rupert returns next week, and if
he found you to be dissipated, he'd dismiss you on the spot."

"Would he?" said Belk, sullenly. "Let him if he likes. You seem to
know Sir Rupert, madam."

"I? No; but I have heard about him."

"He's a hard man, what I've seen of him."

Mrs. Belswin was not going to discuss this subject with a servant like
Belk, so she turned indifferently away as the train came into the
station, and left him standing there, looking in sullen admiration at
her graceful form in the dark garments she now affected.

When she was safely installed in a first-class carriage, her rustic

admirer, who had seen personally after her luggage, appeared at the
window with some newspapers.

"You'll want them to read, madam," he said awkwardly, as she

thanked him. "I hope you'll have a pleasant journey."
"Thank you, Mr. Belk, I hope I shall."

"You'll be coming back soon I hope?"

He blurted out this question with a deep flush, and Mrs. Belswin
stared at him with undisguised astonishment She could not
understand the reason of this man's deference, for she judged it
impossible that he could be so deeply in love with her as all his
actions seemed to denote. Good-natured, however, when not
crossed in any way, she replied politely, as the train moved off--

"I shall return in a fortnight."

"If you don't," muttered Belk, as the long line of carriages

disappeared, "I'll follow you up to London."

"Good heavens!" said Mrs. Belswin, throwing herself back in her

seat, "what on earth can the man see in me to admire? I'm not a
vain woman. I never was a vain woman, and why that handsome
young fellow should leave youth to run after age is more than I can
understand. It's flattering; very much so; but," continued the lady,
struck by a sudden thought, "if Ferrari met my new admirer, I'm
afraid there would be trouble."

She laughed at the idea, and taking up the Telegraph began to read,
but suddenly laid it down with a nervous start.

"Ferrari loves me! Belk loves me! I love neither, but only my child.
Rupert stands between me and my happiness. Which of these men
will remove him out of my path? Ferrari--a subtle Italian, Belk--a
brutal Saxon. Humph! The fox and the lion over again--craft and
strength! I can depend on them both, and Rupert----"

She struck her hands together with a triumphant laugh.

"Rupert Pethram, you are marching blindfolded into a trap."

 CHAPTER XIV.

SIGNOR FERRARI DECLINES.

"Number One is the greater number; if I assisted Number Two it would become the
lesser."

Signor Ferrari was a gentleman who knew how to make himself

thoroughly comfortable; and, in order to do so, squandered his
earnings in a most spendthrift fashion. At present he was receiving a
very handsome salary for his singing in Sultana Fatima, therefore he
denied himself nothing in the way of luxury. He was a true Bohemian
in every action of his life, and accepted his fluctuating fortunes with
the utmost equanimity. If he fared badly on dry bread and water one
day, he was hopeful of oysters and champagne the next; and when
the feast of Dives was before him, made the most of it in eating and
drinking, so as to recompense himself for all future deprivations,
which would be the lot of poverty-stricken Lazarus.

While his voice lasted he was well aware that he could command an
excellent income which satisfied him completely; for when he grew
old and songless he was quite prepared to return to Italy, and live
there the happy-go-lucky life of his youth on polenta and sour wine.
In his impulsive southern fashion he loved Mrs. Belswin madly; but,
strangely enough, it never for a moment occurred to him to save
money against his possible marriage with her. If he starved, she
would starve; if he made money, she would share it; and if she
objected to such a chequered existence, Signor Ferrari was quite
confident enough in his own powers of will and persuasion to be
satisfied that he could force her to accept his view of the matter.
This was the Ferrari philosophy, and no bad one either as times go,
seeing that a singer's livelihood depends entirely upon the caprice of
the public. As long as he could get enough to eat, be the food rich
or plain, a smoke, and plenty of sleep, the world could go hang for
all he cared. He lived in the present, never thought about the past,
and let the future take care of itself; so altogether managed to
scramble through life in a leisurely, selfish manner eminently
egotistical in fashion.

At present, being in the heyday of life, he was dining with Dives,

which was happiness enough in itself; but, in order that nothing
should be wanting to complete his felicity, he had received a letter
from Mrs. Belswin, telling him of her contemplated arrival. Under
these circumstances he had nothing left to wish for, and lounging on
the sofa in his sitting-room in a state of blissful contentment awaited
the coming of his fair friend.

"Buõno," said the signor, with smiling satisfaction, folding up the

letter and putting it in his pocket, "the singing-bird returns to its
nest. This time I will clip its wings, so that it flies not again. Per
Bacco, the kind heart of Stephano surprises himself, for who would
let his bird fly as he has done? But I fear not the jealousy, offspring
of suspicion. Ecco! she loves but me, and comes again to the nest.
And what a nest! Cospetto! My Lucrezia will be hard to please if she
likes not this palazzo del amor."

It was a very pretty nest indeed, from a lodging-house point of view,

although its incongruity of colouring and furnishing would have
driven an artist out of his mind; but then the signor was not exacting
in the way of harmonious effect, and, provided his dwelling was
fairly comfortable, felt completely satisfied. Lying on the sofa, he
looked complacently at the furniture, covered with painfully bright
blue satin, at the scarlet curtains, the green wall-paper, and at all
the wax flowers, Berlin wool mats, and gimcrack ornaments with
which the room was adorned. Ferrari had added to this splendid
furnishing an excellent piano for professional purposes, and
numerous photographs, principally feminine, of his artistic friends;
so that he conceived himself to be housed in a princely fashion.

It was three o'clock by the incorrect French timepiece on the tawdry

mantelpiece, and Ferrari was getting somewhat impatient, as Mrs.
Belswin had mentioned two o'clock as the time of her arrival; but
with his accustomed philosophy he manifested no anger at the
delay.

"La Donna é mobile," he hummed, shrugging his shoulders, as he

strolled towards the piano. "Women are always late; it is one of their
charming follies. Ah! EH! EE! Diavolo! my voice is bad this day.
These English fogs are down my throat Ah! Eh! EE! Dio! What a
note! Voce del oca.

"Ask not the stars the fate they deal.

Read in my eyes the love I feel."

"That's a good song, that serenade to Fatima. It shows off my voice.

I'll sing it to exercise my high notes."

He did so, and was just in the middle of the first verse when Mrs.
Belswin made her appearance, upon which he stopped abruptly, and
came forward to greet her with theatrical effusion.

"Stella dora! once more you shine," he cried, seizing her hands, with
a passionate look in his dark eyes. "Oh, my life! how dear it is to see
thee again."

"You missed me then, Stephano?" said Mrs. Belswin, sinking wearily

into a chair.

"Missed thee, carissima!" exclaimed the Italian, throwing himself on

his knees before her and kissing her hand; "by this, and this, and
this again, I swear that all has been dark to me without the light of
thine eyes. But you will not leave me again, angela mia. Thou hast
come back for ever to be my wife."

Mrs. Belswin drew her hand away sharply and frowned, for in her
present irritable state of mind the exaggerated manner of Ferrari
jarred on her nerves.

"Do be sensible, Stephano," she said in a vexed tone. "You are

always acting."

"How can that be acting, cruel one, which is the truth?" replied
Ferrari, reproachfully, rising from his knees. "Thou knowst my love,
and yet when I speak you are cold. Eh, Donna Lucrezia, is your
heart changed?"

"My heart remains as It always was, my friend; but I've come up to

see you on business----"

"Oh, business!" interrupted Stephano, suspiciously. "Cospetto! You

want once more to leave me."

"For a time; yes."

"Oh, for a time; yes!" echoed Ferrari, mockingly. "Amica mia, you
have a strange way of speaking to him who adores you. Dio, you
play with me like a child. I love you, and wish you for my wife. You
say 'yes,' and depart for a time. Now return you to me and again
say, 'Stephano, I leave you for a time.'"

"I made no promise to be your wife," said Mrs. Belswin, angrily, "nor
will I do so unless you help me now."

"Help you! and in what way? Has the little daughter been cruel? You
wish me to speak as father to her."
"I wish you to do nothing of the sort. My daughter is quite well, and
I was perfectly happy with her."

"And without me," cried Ferrari, jealously; upon which Mrs. Belswin
made a gesture of irritation.

"We can settle that afterwards," she said, drawing off her gloves:
"meanwhile let us talk sense. I shall be up in town for a fortnight."

"And you stay, cara?"

"At an hotel in the Strand. I'll give you the address before I leave."

"Bene! I will then have you to myself for two weeks."

"It all depends on whether you will help me in what I wish to do."

"Ebbene! Is it il marito?"

Mrs. Belswin nodded, and the Italian burst out laughing.

"Povero diavolo. He has then come again."

"No! but he arrives next week."

"How pleased you are," said Ferrari, mockingly. "Oh, yes, he will be
so sweet to behold you."

"That's the very question! I don't want him to see me."

"Then return not to the little daughter."

"I must! I must!" cried Mrs. Belswin in despair. "I can't give up my
child after meeting her again. Twenty years, Stephano, and I have
not seen her; now I am beside her every day. She loves me--not as
her mother, but as her friend. I can't give up all this because my
husband is returning."
Signor Ferrari shrugged his shoulders and lighted a cigarette.

"But there is nothing more you can do," he said, spreading out his
hands with a dramatic gesture, "eh, carrissima? Think of what is this
affair. Il marito has said to you, 'Good-bye.' The little daughter thinks
you to be dead. If then you come to reveal yourself, il marito--eh,
amica mia! it is a trouble for all."

"What can I do?"

"Nothing! oh no, certainly! You have beheld the little daughter for a
time. Now you are to me again. I say, Stella 'dora, with me remain
and forget all."

"No, I will not! I will not!" cried Mrs. Belswin, savagely, rising to her
feet. "Cannot you see how I suffer? If you love me as you say, you
must see how I suffer. Give up my child, my life, my happiness! I
cannot do it."

"Dio! you cannot make the miracles."

"I can! I must! Do you think I will stay with you while my child calls
me?"

"With me you must stay, my Norma. I love thee. I will not leave you
no more."

"You can't stop me."

"Ebbene," said Ferrari, conscious that he held the advantage. "Go,

then, and see how il marito will behold you."

Mrs. Belswin felt her helplessness, and clenched her hands with a
savage cry of despair, that seemed to be torn out of her throbbing
heart. Up and down the gaudy room she paced, with her face
convulsed with rage, and her fierce eyes flashing with an unholy fire,
while Ferrari, secure in his position, sat quietly near the window,
smoking leisurely. His self-possession seemed to provoke her, ready
as she was to vent her impotent anger on anything, and, stopping
abruptly she poured forth all her anger.

"Why do you sit there smiling, and smiling, like a fool?" she shrieked,
stamping her foot. "Can you not suggest something? Can you not do
something?"

"Eh, carissima, I would say, 'Be quiet' The people below will hear you
cry out."

"Let them! What do I care? I am a desperate woman, Ferrari, and I

am determined to keep my position beside my child. I will stop at
nothing--nothing--not even murder!"

"Murder!"

Signor Ferrari let the cigarette drop from his fingers, and jumped up
with a cry of dismay looking pale and unnerved. She saw this, and
lashing him with her tongue, taunted him bitterly.

"Yes, murder, you miserable! I thought you were a brave man; but I
see I made a mistake. You love me! You want to be my husband!
No, no, no! I marry a brave man--yes, a brave man; not a coward!"

Ferrari winced, with an angry glitter in his eyes.

"Eh, Lucrezia. You think I am a brave man if I go to assassin il

marito. Cospetto! I am an Italian; but the Italians are not fools. If
another man loved you, and would take you away, I would kill him--
yes! But il marito--eh, that is not quite the same. I kill him and you
return to the little daughter for always. What gain to me, carissima?
I kill him, and your law gives me the rope. What gain to me? No,
Donna Lucrezia. Do what you love. Stab him with a stiletto, or give
the poison, I say nothing; but as for me to obey--Dio, the life is not
trouble to me yet."
"You are afraid."

He bounded across the room, and seized her roughly by the wrist.

"Devil-woman, I have no fear! You lie to speak so I You lie, figlia

inferna."

"Then why do you refuse to help me?"

"Per Bacco, I am no assassin. Il marito is not an enemy to me. To

you he is hateful. Revenge yourself as it pleases; but I--cospetto.
You ask too much."

He flung her away from him with a gesture of anger, and began to
walk about the room. Mrs. Belswin remained silent, savagely
disappointed at the failure of her plan, and presently Ferrari began
to talk again in his rapid, impulsive fashion.

"If there was any gain. Yes. But I see not anything. I would work
against myself. You know that, Signora Machiavelli. Ah, yes; I am not
blind, cara mia. While il marito lives, you are mine. He will keep you
from the little daughter. But he dies--eh, and you depart."

"No, no! I swear----"

"I refuse your swearing. They are false. Forget, il marito--forget the
little daughter! You are mine, mia moglie, and you depart not again."

Mrs. Belswin laughed scornfully, and put on her gloves again with
the utmost deliberation. Then, taking up her umbrella, she moved
quickly towards the door; but not so quickly as to prevent Ferrari
placing himself before her.

"Where go you?" demanded the Italian, between his clenched teeth.

"To find a braver man than Stephano Ferrari."

"No; you will find no one."

Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com