Visit https://ebookfinal.

com to download the full version and

explore more ebooks

Stochastic Processes An Introduction 2nd Edition

Peter Watts Jones

_____ Click the link below to download _____

https://ebookfinal.com/download/stochastic-processes-
an-introduction-2nd-edition-peter-watts-jones/

Explore and download more ebooks at ebookfinal.com

Here are some suggested products you might be interested in.
Click the link to download

Introduction to Hospitality Operations An Indispensible

Guide to the Industry 2nd Edition Peter Jones

https://ebookfinal.com/download/introduction-to-hospitality-
operations-an-indispensible-guide-to-the-industry-2nd-edition-peter-
jones/

Applied Probability and Stochastic Processes 2nd Edition

Frank Beichelt

https://ebookfinal.com/download/applied-probability-and-stochastic-
processes-2nd-edition-frank-beichelt/

An Introduction to Stochastic Modeling Fourth Edition

Pinsky

https://ebookfinal.com/download/an-introduction-to-stochastic-
modeling-fourth-edition-pinsky/

Probability and Stochastic Processes A Friendly

Introduction for Electrical and Computer Engineers 2nd
Edition Roy D. Yates
https://ebookfinal.com/download/probability-and-stochastic-processes-
a-friendly-introduction-for-electrical-and-computer-engineers-2nd-
edition-roy-d-yates/
An Elementary Introduction To Stochastic Interest Rate
Modeling 2nd Edition Edition Nicolas Privault

https://ebookfinal.com/download/an-elementary-introduction-to-
stochastic-interest-rate-modeling-2nd-edition-edition-nicolas-
privault/

Stochastic processes with applications to finance 2nd ed.

Edition Kijima

https://ebookfinal.com/download/stochastic-processes-with-
applications-to-finance-2nd-ed-edition-kijima/

Stochastic Processes and Models David Stirzaker

https://ebookfinal.com/download/stochastic-processes-and-models-david-
stirzaker/

Applied Stochastic Processes 1st Edition Ming Liao

(Author)

https://ebookfinal.com/download/applied-stochastic-processes-1st-
edition-ming-liao-author/

Stochastic Finance An Introduction in Discrete Time 3rd

Edition Hans Föllmer

https://ebookfinal.com/download/stochastic-finance-an-introduction-in-
discrete-time-3rd-edition-hans-follmer/
Stochastic Processes An Introduction 2nd Edition Peter
Watts Jones Digital Instant Download
Author(s): Peter Watts Jones, Peter Smith
ISBN(s): 9781420099607, 1420099604
Edition: 2nd
File Details: PDF, 2.36 MB
Year: 2009
Language: english
 Texts in Statistical Science

Stochastic
Processes
An Introduction
Second Edition

Peter W. Jones
Peter Smith

K10004_FM.indd 1 9/3/09 12:51:59 PM

 CHAPMAN & HALL/CRC
Texts in Statistical Science Series
Series Editors
Bradley P. Carlin, University of Minnesota, USA
Julian J. Faraway, University of Bath, UK
Martin Tanner, Northwestern University, USA
Jim Zidek, University of British Columbia, Canada

Analysis of Failure and Survival Data Epidemiology — Study Design and

P. J. Smith Data Analysis, Second Edition
The Analysis of Time Series — M. Woodward
An Introduction, Sixth Edition Essential Statistics, Fourth Edition
C. Chatfield D.A.G. Rees
Applied Bayesian Forecasting and Time Series Extending the Linear Model with R — Generalized
Analysis Linear, Mixed Effects and Nonparametric Regression
A. Pole, M. West and J. Harrison Models
Applied Nonparametric Statistical Methods, J.J. Faraway
Fourth Edition A First Course in Linear Model Theory
P. Sprent and N.C. Smeeton N. Ravishanker and D.K. Dey
Applied Statistics — Handbook of GENSTAT Generalized Additive Models:
Analysis An Introduction with R
E.J. Snell and H. Simpson S. Wood
Applied Statistics — Principles and Examples Interpreting Data — A First Course
D.R. Cox and E.J. Snell in Statistics
A.J.B. Anderson
Applied Stochastic Modelling, Second Edition
An Introduction to Generalized
B.J.T. Morgan Linear Models, Third Edition
Bayesian Data Analysis, Second Edition A.J. Dobson and A.G. Barnett
A. Gelman, J.B. Carlin, H.S. Stern Introduction to Multivariate Analysis
and D.B. Rubin C. Chatfield and A.J. Collins
Bayesian Methods for Data Analysis, Introduction to Optimization Methods and Their
Third Edition Applications in Statistics
B.P. Carlin and T.A. Louis B.S. Everitt
Beyond ANOVA — Basics of Applied Statistics Introduction to Probability with R
R.G. Miller, Jr. K. Baclawski
Computer-Aided Multivariate Analysis, Introduction to Randomized Controlled Clinical
Fourth Edition Trials, Second Edition
A.A. Afifi and V.A. Clark J.N.S. Matthews
A Course in Categorical Data Analysis Introduction to Statistical Inference and Its
T. Leonard Applications with R
M.W. Trosset
A Course in Large Sample Theory
T.S. Ferguson Introduction to Statistical Methods for
Clinical Trials
Data Driven Statistical Methods
T.D. Cook and D.L. DeMets
P. Sprent
Large Sample Methods in Statistics
Decision Analysis — A Bayesian Approach
P.K. Sen and J. da Motta Singer
J.Q. Smith
Linear Models with R
Elementary Applications of Probability Theory,
J.J. Faraway
Second Edition
H.C. Tuckwell Logistic Regression Models
J.M. Hilbe
Elements of Simulation
B.J.T. Morgan

K10004_FM.indd 2 9/3/09 12:52:00 PM

 Markov Chain Monte Carlo — Sampling Methodologies with Applications
Stochastic Simulation for Bayesian Inference, P.S.R.S. Rao
Second Edition Statistical Analysis of Reliability Data
D. Gamerman and H.F. Lopes M.J. Crowder, A.C. Kimber,
Mathematical Statistics T.J. Sweeting, and R.L. Smith
K. Knight Statistical Methods for Spatial Data Analysis
Modeling and Analysis of Stochastic Systems O. Schabenberger and C.A. Gotway
V. Kulkarni Statistical Methods for SPC and TQM
Modelling Binary Data, Second Edition D. Bissell
D. Collett Statistical Methods in Agriculture and Experimental
Modelling Survival Data in Medical Research, Biology, Second Edition
Second Edition R. Mead, R.N. Curnow, and A.M. Hasted
D. Collett Statistical Process Control — Theory and Practice,
Multivariate Analysis of Variance and Repeated Third Edition
Measures — A Practical Approach for Behavioural G.B. Wetherill and D.W. Brown
Scientists Statistical Theory, Fourth Edition
D.J. Hand and C.C. Taylor B.W. Lindgren
Multivariate Statistics — A Practical Approach Statistics for Accountants
B. Flury and H. Riedwyl S. Letchford
Pólya Urn Models Statistics for Epidemiology
H. Mahmoud N.P. Jewell
Practical Data Analysis for Designed Experiments Statistics for Technology — A Course in Applied
B.S. Yandell Statistics, Third Edition
Practical Longitudinal Data Analysis C. Chatfield
D.J. Hand and M. Crowder Statistics in Engineering — A Practical Approach
Practical Statistics for Medical Research A.V. Metcalfe
D.G. Altman Statistics in Research and Development,
A Primer on Linear Models Second Edition
J.F. Monahan R. Caulcutt
Probability — Methods and Measurement Stochastic Processes: An Introduction,
A. O’Hagan Second Edition
Problem Solving — A Statistician’s Guide, P.W. Jones and P. Smith
Second Edition Survival Analysis Using S — Analysis of
C. Chatfield Time-to-Event Data
Randomization, Bootstrap and Monte Carlo M. Tableman and J.S. Kim
Methods in Biology, Third Edition The Theory of Linear Models
B.F.J. Manly B. Jørgensen
Readings in Decision Analysis Time Series Analysis
S. French H. Madsen

K10004_FM.indd 3 9/3/09 12:52:00 PM

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2010 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Version Date: 20130920

International Standard Book Number-13: 978-1-4398-7122-5 (eBook - PDF)

This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but
the author and publisher cannot assume 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 acknowledged 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, including 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, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright
Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a
variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to
infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
 Contents

Preface ix

1 Some Background on Probability 1

1.1 Introduction 1
1.2 Probability 1
1.3 Conditional probability and independence 5
1.4 Discrete random variables 7
1.5 Continuous random variables 9
1.6 Mean and variance 10
1.7 Some standard discrete probability distributions 12
1.8 Some standard continuous probability distributions 14
1.9 Generating functions 17
1.10 Conditional expectation 21
1.11 Problems 24

2 Some Gambling Problems 29

2.1 Gambler’s ruin 29
2.2 Probability of ruin 29
2.3 Some numerical simulations 33
2.4 Duration of the game 34
2.5 Some variations of gambler’s ruin 38
2.5.1 The infinitely rich opponent 38
2.5.2 The generous opponent 38
2.5.3 Changing the stakes 39
2.6 Problems 39

3 Random Walks 45
3.1 Introduction 45
3.2 Unrestricted random walks 46
3.3 The general probability distribution of a walk 48
3.4 First returns of the symmetric random walk 50
3.5 Problems 52

4 Markov Chains 59
4.1 States and transitions 59
4.2 Transition probabilities 60

v
vi CONTENTS
4.3 General two-state Markov chains 64
4.4 Powers of the general transition matrix 66
4.5 Gambler’s ruin as a Markov chain 73
4.6 Classification of states 76
4.7 Classification of chains 83
4.8 Problems 86

5 Poisson Processes 93
5.1 Introduction 93
5.2 The Poisson process 93
5.3 Partition theorem approach 96
5.4 Iterative method 97
5.5 The generating function 98
5.6 Variance in terms of the probability generating function 100
5.7 Arrival times 101
5.8 Summary of the Poisson process 103
5.9 Problems 104

6 Birth and Death Processes 107

6.1 Introduction 107
6.2 The birth process 107
6.3 Birth process: Generating function equation 110
6.4 The death process 112
6.5 The combined birth and death process 115
6.6 General population processes 119
6.7 Problems 122

7 Queues 131
7.1 Introduction 131
7.2 The single-server queue 132
7.3 The stationary process 134
7.4 Queues with multiple servers 140
7.5 Queues with fixed service times 144
7.6 Classification of queues 147
7.7 A general approach to the 𝑀 (𝜆)/𝐺/1 queue 147
7.8 Problems 151

8 Reliability and Renewal 157

8.1 Introduction 157
8.2 The reliability function 157
8.3 Exponential distribution and reliability 159
8.4 Mean time to failure 160
8.5 Reliability of series and parallel systems 161
8.6 Renewal processes 163
8.7 Expected number of renewals 165
CONTENTS vii
8.8 Problems 167

9 Branching and Other Random Processes 171

9.1 Introduction 171
9.2 Generational growth 171
9.3 Mean and variance 174
9.4 Probability of extinction 176
9.5 Branching processes and martingales 179
9.6 Stopping rules 182
9.7 The simple epidemic 184
9.8 An iterative solution scheme for the simple epidemic 186
9.9 Problems 188

10 Computer Simulations and Projects 195

Answers and Comments on End-of-Chapter Problems 203

Appendix 211

References and Further Reading 215

Index 217
 Preface

This textbook was developed from a course in stochastic processes given by the au-
thors over many years to second-year students studying Mathematics or Statistics at
Keele University. At Keele the majority of students take degrees in Mathematics or
Statistics jointly with another subject, which may be from the sciences, social sci-
ences or humanities. For this reason the course has been constructed to appeal to
students with varied academic interests, and this is reflected in the book by including
applications and examples that students can quickly understand and relate to. In par-
ticular, in the earlier chapters, the classical gambler’s ruin problem and its variants
are modeled in a number of ways to illustrate simple random processes. Specialized
applications have been avoided to accord with our view that students have enough to
contend with in the mathematics required in stochastic processes.
Topics can be selected from Chapters 2 to 9 for a one-semester course or mod-
ule in random processes. It is assumed that readers have already encountered the
usual first-year courses in calculus and matrix algebra and have taken a first course
in probability; nevertheless, a revision of relevant basic probability is included for
reference in Chapter 1. Some of the easier material on discrete random processes is
included in Chapters 2, 3, and 4, which cover some simple gambling problems, ran-
dom walks, and Markov chains. Random processes continuous in time are developed
in Chapters 5 and 6. These include Poisson, birth and death processes, and general
population models. Continuous time models include queues in Chapter 7, which has
an extended discussion on the analysis of associated stationary processes. The book
ends with two chapters on reliability and other random processes, the latter including
branching processes, martingales, and a simple epidemic. An appendix contains key
mathematical results for reference.
There are over 50 worked examples in the text and 205 end-of-chapter problems
with hints and answers listed at the end of the book.
Mathematica𝑇 𝑀 is a mathematical software package able to carry out complex
symbolic mathematical as well as numerical computations. It has become an integral
part of many degree courses in Mathematics or Statistics. The software has been used
throughout the book to solve both theoretical and numerical examples and to produce
many of the graphs.
R is a statistical computing and graphics package which is available free of charge,
and can be downloaded from:

http://www.r-project.org

ix
x PREFACE
Like R, S-PLUS (not freeware) is derived from the S language, and hence users
of these packages will be able to apply them to the solution of numerical projects,
including those involving matrix algebra presented in the text. Mathematica code
has been applied to all the projects listed by chapters in Chapter 10, and R code to
some as appropriate. All the Mathematica and R programs can be found on the Keele
University Web site:
http://www.scm.keele.ac.uk/books/stochastic processes/
Not every topic in the book is included, but the programs, which generally use
standard commands, are intended to be flexible in that inputs, parameters, data, etc.,
can be varied by the user. Graphs and computations can often add insight into what
might otherwise be viewed as rather mechanical analysis. In addition, more compli-
cated examples, which might be beyond hand calculations, can be attempted.
We are grateful to staff of the School of Computing and Mathematics, Keele Uni-
versity, for help in designing the associated Web site.
Finally, we would like to thank the many students at Keele over many years who
have helped to develop this book, and to the interest shown by users of the first edition
in helping us to refine and update this second edition.

Peter W. Jones
Peter Smith
Keele University
 CHAPTER 1

Some Background on Probability

1.1 Introduction

We shall be concerned with the modeling and analysis of random experiments us-
ing the theory of probability. The outcome of such an experiment is the result of a
stochastic or random process. In particular we shall be interested in the way in which
the results or outcomes vary or evolve over time. An experiment or trial is any sit-
uation where an outcome is observed. In many of the applications considered, these
outcomes will be numerical, sometimes in the form of counts or enumerations. The
experiment is random if the outcome is not predictable or is uncertain.
At first we are going to be concerned with simple mechanisms for creating random
outcomes, namely games of chance. One recurring theme initially will be the study of
the classical problem known as gambler’s ruin. We will then move on to applications
of probability to modeling in, for example, engineering, medicine, and biology. We
make the assumption that the reader is familiar with the basic theory of probability.
This background will however be reinforced by the brief review of these concepts
which will form the main part of this chapter.

1.2 Probability

In random experiments, the list of all possible outcomes is termed the sample space,
denoted by 𝑆. This list consists of individual outcomes or elements. These elements
have the properties that they are mutually exclusive and that they are exhaustive.
Mutually exclusive means that two or more outcomes cannot occur simultaneously:
exhaustive means that all possible outcomes are in the list. Thus each time the exper-
iment is carried out one of the outcomes in 𝑆 must occur. A collection of elements of
𝑆 is called an event: these are usually denoted by capital letters, 𝐴, 𝐵, etc. We denote
by P(𝐴) the probability that the event 𝐴 will occur at each repetition of the random
experiment. Remember that 𝐴 is said to have occurred if one element making up 𝐴
has occurred. In order to calculate or estimate the probability of an event 𝐴 there are
two possibilities. In one approach an experiment can be performed a large number of
times, and P(𝐴) can be approximated by the relative frequency with which 𝐴 occurs.
In order to analyze random experiments we make the assumption that the conditions
surrounding the trials remain the same, and are independent of one another. We hope

1
2 SOME BACKGROUND ON PROBABILITY
that some regularity or settling down of the outcome is apparent. The ratio
the number of times a particular event 𝐴 occurs
total number of trials
is known as the relative frequency of the event, and the number to which it ap-
pears to converge as the number of trials increases is known as the probability of
an outcome within 𝐴. Where we have a finite sample space it might be reasonable
to assume that the outcomes of an experiment are equally likely to occur as in the
case, for example, in rolling a fair die or spinning an unbiased coin. In this case the
probability of 𝐴 is given by
number of elements of 𝑆 where 𝐴 occurs
P(𝐴) = .
number of elements in 𝑆
There are, of course, many ‘experiments’ which are not repeatable. Horse races are
only run once, and the probability of a particular horse winning a particular race may
not be calculated by relative frequency. However, a punter may form a view about
the horse based on other factors which may be repeated over a series of races. The
past form of the horse, the form of other horses in the race, the state of the course, the
record of the jockey, etc., may all be taken into account in determining the probability
of a win. This leads to a view of probability as a ‘degree of belief’ about uncertain
outcomes. The odds placed by bookmakers on the horses in a race reflect how punters
place their bets on the race. The odds are also set so that the bookmakers expect to
make a profit.
It is convenient to use set notation when deriving probabilities of events. This leads
to 𝑆 being termed the universal set, the set of all outcomes: an event 𝐴 is a subset
of 𝑆. This also helps with the construction of more complex events in terms of the
unions and intersections of several events. The Venn diagrams shown in Figure 1.1
represent the main set operations of union (∪), intersection (∩), and complement
(𝐴𝑐 ) which are required in probability.
∙ Union. The union of two sets 𝐴 and 𝐵 is the set of all elements which belong to
𝐴, or to 𝐵, or to both. It can be written formally as
𝐴 ∪ 𝐵 = {𝑥∣𝑥 ∈ 𝐴 or 𝑥 ∈ 𝐵 or both}.
∙ Intersection. The intersection of two sets 𝐴 and 𝐵 is the set 𝐴∩𝐵 which contains
all elements common to both 𝐴 and 𝐵. It can be written as
𝐴 ∩ 𝐵 = {𝑥∣𝑥 ∈ 𝐴 and 𝑥 ∈ 𝐵}.
∙ Complement. The complement 𝐴𝑐 of a set 𝐴 is the set of all elements which
belong to the universal set 𝑆 but do not belong to 𝐴. It can be written as
𝐴𝑐 = {𝑥 ∕∈ 𝐴}.
So, for example, in an experiment in which we are interested in two events 𝐴
and 𝐵, then 𝐴𝑐 ∩ 𝐵 may be interpreted as ‘only 𝐵’, being the intersection of the
complement of 𝐴 and 𝐵 (see Figure 1.1d): this is alternatively expressed in the
difference notation 𝐵∖𝐴 meaning 𝐵 but not 𝐴. We denote by 𝜙 the empty set, that
is the set which contains no elements. Note that 𝑆 𝑐 = 𝜙. Two events 𝐴 and 𝐵 are
PROBABILITY 3
U U U U

A A A

B B B

(a) (b) (c) (d)

Figure 1.1 (a) the union 𝐴 ∪ 𝐵 of 𝐴 and 𝐵; (b) the intersection 𝐴 ∩ 𝐵 of 𝐴 and 𝐵; (c) the
complement 𝐴𝑐 of 𝐴: 𝑆 is the universal set; (d) 𝐴𝑐 ∩ 𝐵 or 𝐵∖𝐴

said to be mutually exclusive if 𝐴 and 𝐵 have no events in common 𝐴 and 𝐵 so that

𝐴 ∩ 𝐵 = 𝜙, the empty set: in set terminology 𝐴 and 𝐵 are said to be disjoint sets.
The probability of any event satisfies however calculated the three axioms
∙ Axiom 1: 0 ≤ P(𝐴) ≤ 1 for every event 𝐴
∙ Axiom 2: P(𝑆) = 1
∙ Axiom 3: P(𝐴∪𝐵) = P(𝐴)+P(𝐵) if 𝐴 and 𝐵 are mutually exclusive (𝐴∩𝐵 =
𝜙)
Axiom 3 may be extended to more than two mutually exclusive events, say 𝑘 of them
represented by
𝐴1 , 𝐴2 , . . . , 𝐴𝑘
where 𝐴𝑖 ∩ 𝐴𝑗 = 𝜙 for all 𝑖 ∕= 𝑗. This is called a partition of 𝑆 if
∙ (a) 𝐴𝑖 ∩ 𝐴𝑗 = 𝜙 for all 𝑖 ∕= 𝑗,
𝑘
∪
∙ (b) 𝐴𝑖 = 𝐴1 ∪ 𝐴2 ∪ . . . ∪ 𝐴𝑘 = 𝑆,
𝑖=1
∙ (c) P(𝐴𝑖 ) > 0.
In this definition, (a) states that the events are mutually exclusive, (b) that every event
in 𝑆 occurs in one of the events 𝐴𝑖 , and (c) implies that there is a nonzero probability
that any 𝐴𝑖 occurs. It follows that
𝑘
∑
1 = P(𝑆) = P(𝐴1 ∪ 𝐴2 ∪ ⋅ ⋅ ⋅ ∪ 𝐴𝑘 ) = P(𝐴𝑖 ).
𝑖=1

Theorem
∙ (a) P(𝐴𝑐 ) = 1 − P(𝐴);
∙ (b) P(𝐴 ∪ 𝐵) = P(𝐴) + P(𝐵) − P(𝐴 ∩ 𝐵).
(a) Axiom 3 may be combined with Axiom 2 to give P(𝐴𝑐 ), the probability that the
4 SOME BACKGROUND ON PROBABILITY
complement 𝐴 occurs, by noting that 𝑆 = 𝐴 ∪ 𝐴𝑐 . This is a partition of 𝑆 into the
𝑐

mutually exclusive exhaustive events 𝐴 and 𝐴𝑐 . Thus

1 = P(𝑆) = P(𝐴 ∪ 𝐴𝑐 ) = P(𝐴) + P(𝐴𝑐 ),
giving
P(𝐴𝑐 ) = 1 − P(𝐴).
(b) For any sets 𝐴 and 𝐵
𝐴 ∪ 𝐵 = 𝐴 ∪ (𝐵 ∩ 𝐴𝑐 ),
and
𝐵 = (∩𝐵) ∪ (𝐵 ∩ 𝐴𝑐 ),
in which 𝐴 and 𝐵 ∩ 𝐴𝑐 are disjoint sets, and 𝐴 ∩ 𝐵 and 𝐵 ∩ 𝐴𝑐 are disjoint sets.
Therefore, by Axiom 3,
P(𝑎 ∪ 𝐵) = P(𝐴) + P(𝐵 ∩ 𝐴𝑐 ),
and
P(𝐵) = P(𝐴 ∩ 𝐵) + P(𝐵 ∩ 𝐴𝑐 ).
Elimination of P(𝐵 ∩ 𝐴𝑐 ) between these equations leads to
P(𝐴 ∪ 𝐵) = P(𝐴) + P(𝐵) − P(𝐴 ∩ 𝐵) (1.1)
as required.

Example 1.1. Two distinguishable fair dice 𝑎 and 𝑏 are rolled and the values on the uppermost
faces noted. What are the elements of the sample space? What is the probability that the sum
of the face values of the two dice is 7? What is the probability that at least one 5 appears?

We distinguish first the outcome of each die so that there are 6 × 6 = 36 possible outcomes
for the pair. The sample space has 36 elements of the form (𝑖, 𝑗) where 𝑖 and 𝑗 take all integer
values 1, 2, 3, 4, 5, 6, and 𝑖 is the outcome of die 𝑎 and 𝑗 is the outcome of 𝑏. The full list is
𝑆={ (1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6),
(2, 1), (2, 2), (2, 3), (2, 4), (2, 5), (2, 6),
(3, 1), (3, 2), (3, 3), (3, 4), (3, 5), (3, 6),
(4, 1), (4, 2), (4, 3), (4, 4), (4, 5), (4, 6),
(5, 1), (5, 2), (5, 3), (5, 4), (5, 5), (5, 6),
(6, 1), (6, 2), (6, 3), (6, 4), (6, 5), (6, 6) },
and they are all assumed to be equally likely since the dice are fair. If 𝐴1 is the event that the
sum of the dice is 7, then from the list,
𝐴1 = {(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1)}
which occurs for 6 elements out of 36. Hence
6
𝑃 (𝐴1 ) = 36
= 16 .
The event that at least one 5 appears is the list
𝐴2 = {(1, 5), (2, 5), (3, 5), (4, 5), (5, 1), (5, 2), (5, 3), (5, 4), (5, 5), (5, 6), (6, 5)},
CONDITIONAL PROBABILITY AND INDEPENDENCE 5
which has 11 elements. Hence
11
𝑃 (𝐴2 ) = 36
.

Example 1.2 From a well-shuffled pack of 52 playing cards a single card is randomly drawn.
Find the probability that it is a heart or an ace.

Let 𝐴 be the event that the card is an ace, and 𝐵 the event that it is a heart. The event 𝐴 ∩ 𝐵
is the ace of hearts. We require the probability that it is an ace or a heart, which is P(𝐴 ∪ 𝐵).
However, since one of the aces is a heart the events are not mutually exclusive. Hence, we
must use eqn (1.1). It follows that
the probability that an ace is drawn is P(𝐴) = 4/52,
the probability that a heart is drawn is P(𝐵) = 13/52 = 1/4,
the probability that the ace of hearts is drawn is P(𝐴 ∩ 𝐵) = 1/52.
From (1.1)
4 1 1 16 4
P(𝐴 ∪ 𝐵) = P(𝐴) + P(𝐵) − P(𝐴 ∩ 𝐵) = + − = = .
52 4 52 52 13
This example illustrates events which are not mutually exclusive. The result could also be
obtained directly by noting that 16 of the 52 cards are either hearts or aces.
In passing note that 𝐴 ∩ 𝐵 𝑐 is the set of aces excluding the ace of hearts, whilst 𝐴𝑐 ∩ 𝐵 is
the heart suit excluding the ace of hearts. Hence
3 12 3
P(𝐴 ∩ 𝐵 𝑐 ) = , P(𝐴𝑐 ∩ 𝐵) = = .
52 52 13

1.3 Conditional probability and independence

If the occurrence of an event 𝐵 is affected by the occurrence of another event 𝐴 then
we say that 𝐴 and 𝐵 are dependent events. We might be interested in a random ex-
periment with which 𝐴 and 𝐵 are associated. When the experiment is performed, it
is known that event 𝐴 has occurred. Does this affect the probability of 𝐵? This prob-
ability of 𝐵 now becomes the conditional probability of 𝐵 given 𝐴, which is now
written as P(𝐵∣𝐴). Usually this will be distinct from the probability P(𝐵). Strictly
speaking, this probability is conditional since we must assume that 𝐵 is conditional
on the sample space occurring, but it is implicit in P(𝐵). On the other hand the con-
ditional probability of 𝐵 is restricted to that part of the sample space where 𝐴 has
occurred. This conditional probability is defined as
P(𝐴 ∩ 𝐵)
P(𝐵∣𝐴) = , P(𝐴) > 0. (1.2)
P(𝐴)
In terms of counting, suppose that an experiment is repeated N times, of which 𝐴
occurs N(𝐴) times, and 𝐴 given by 𝐵 occurs N(𝐵 ∩ 𝐴) times. The proportion of
times that 𝐵 occurs is
N(𝐵 ∩ 𝐴) N(𝐴 ∩ 𝐵) N
= ,
N(𝐴) N N(𝐵)
6 SOME BACKGROUND ON PROBABILITY
which justifies (1.2).
If the probability of B is unaffected by the prior occurrence of 𝐴, then we say that
𝐴 and 𝐵 are independent or that
P(𝐵∣𝐴) = P(𝐵),
which from above implies that
P(𝐴 ∩ 𝐵) = P(𝐴)P(𝐵).
Conversely, if P(𝐵∣𝐴) = P(𝐵), then 𝐴 and 𝐵 are independent events. Again this
result can be extended to 3 or more independent events.

Example 1.3 Let 𝐴 and 𝐵 be independent events with P(𝐴) = 14 and P(𝐵) = 32 . Calculate
the following probabilities: (a) P(𝐴 ∩ 𝐵); (b) P(𝐴 ∩ 𝐵 𝑐 ); (c) P(𝐴𝑐 ∩ 𝐵 𝑐 ); (d) P(𝐴𝑐 ∩ 𝐵);
(e) P((𝐴 ∪ 𝐵)𝑐 ).

Since the events are independent, then P(𝐴 ∩ 𝐵) = P(𝐴)P(𝐵). Hence

(a) P(𝐴 ∩ 𝐵) = 14 ⋅ 32 = 16 .
(b) The independence 𝐴 and 𝐵 𝑐 follows by eliminating P(𝐴 ∩ 𝐵) between the equations
P(𝐴 ∩ 𝐵) = P(𝐴)P(𝐵) = P(𝐴)[1 − P(𝐵 𝑐 )]
and
P(𝐴) = P[(𝐴 ∩ 𝐵 𝑐 ) ∪ (𝐴 ∩ 𝐵)] = P(𝐴 ∩ 𝐵 𝑐 ) + P(𝐴 ∩ 𝐵).
Hence
P(𝐴 ∩ 𝐵 𝑐 ) = P(𝐴)P(𝐵 𝑐 ) = P(𝐴)[1 − P(𝐵)] = 41 (1 − 32 ) = 1
12
.
𝑐 𝑐
(c) Since 𝐴 and 𝐵 are independent events,
3 1
P(𝐴𝑐 ∩ 𝐵 𝑐 ) = P(𝐴𝑐 )P(𝐵 𝑐 ) = [1 − P(𝐴)][1 − P(𝐵)] = 4
⋅ 3
= 14 .
(d) Since 𝐴𝑐 and 𝐵 are independent events, P(𝐴𝑐 ∩ 𝐵) = P(𝐴𝑐 )P(𝐵) = [1 − 14 ] 23 = 12 .
(e) P((𝐴 ∪ 𝐵)𝑐 ) = 1 − P(𝐴 ∪ 𝐵) = 1 − P(𝐴) − P(𝐵) + P(𝐴 ∩ 𝐵) by (1.1). Hence
1 2 1
P((𝐴 ∪ 𝐵)𝑐 ) = 1 − P(𝐴) − P(𝐵) + P(𝐴)P(𝐵) = 1 − 4
− 3
+ 6
= 14 .

Example 1.4. For three events 𝐴, 𝐵, and 𝐶, show that

P(𝐴 ∩ 𝐵∣𝐶) = P(𝐴∣𝐵 ∩ 𝐶)P(𝐵∣𝐶),
where P(𝐶) > 0.

By using (1.2) and viewing 𝐴 ∩ 𝐵 ∩ 𝐶 as (𝐴 ∩ 𝐵) ∩ 𝐶 or 𝐴 ∩ (𝐵 ∩ 𝐶),

P(𝐴 ∩ 𝐵 ∩ 𝐶) = P(𝐴 ∩ 𝐵∣𝐶)P(𝐶) = P(𝐴∣𝐵 ∩ 𝐶)P(𝐵 ∩ 𝐶).
Hence
P(𝐵 ∩ 𝐶)
P(𝐴 ∩ 𝐵∣𝐶) = P(𝐴∣𝐵 ∩ 𝐶) = P(𝐴∣𝐵 ∩ 𝐶)P(𝐵∣𝐶)
P(𝐶)
by (1.2) again.

A result known as the law of total probability or the partition theorem will
DISCRETE RANDOM VARIABLES 7
S

A1 A2

A4 A5

A3

Figure 1.2 Schematic set view of a partition of 𝑆 into 5 events 𝐴1 , . . . , 𝐴5 .

be used extensively later, for example, in the discrete gambler’s ruin problem (Sec-
tion 2.1) and the Poisson process (Section 5.2). Suppose that 𝐴1 , 𝐴2 , . . . , 𝐴𝑘 repre-
sents a partition of 𝑆 into 𝑘 mutually exclusive events in which, interpreted as sets,
the sets fill the space 𝑆 but with none of the sets overlapping. Figure 1.2 shows such
a scheme. When a random experiment takes place one and only one of the events can
take place.
Suppose that 𝐵 is another event associated with the same random experiment (Fig-
ure 1.2). Then 𝐵 must be made up of the sum of the intersections of 𝐵 with each of
the events in the partition. Some of these will be empty but this does not matter. We
can say that 𝐵 is the union of the intersections of 𝐵 with each 𝐴𝑖 . Thus
𝑘
∪
𝐵= 𝐵 ∩ 𝐴𝑖 ,
𝑖=1

but the significant point is that any pair of these events is mutually exclusive. It
follows that
∑𝑘
P(𝐵) = P(𝐵 ∩ 𝐴𝑖 ). (1.3)
𝑖=1
Since, from equation (1.2),
P(𝐵 ∩ 𝐴𝑖 ) = P(𝐵∣𝐴𝑖 )P(𝐴𝑖 ),
equation (1.3) can be expressed as
𝑘
∑
P(𝐵) = P(𝐵∣𝐴𝑖 )P(𝐴𝑖 ),
𝑖=1

which is the law of total probability or the partition theorem.

1.4 Discrete random variables

In most of the applications considered in this text, the outcome of the experiment
will be numerical. A random variable usually denoted by the capital letters 𝑋, 𝑌 ,
or 𝑍, say, is a numerical value associated with the outcome of a random experiment.
8 SOME BACKGROUND ON PROBABILITY
If 𝑠 is an element of the original sample space 𝑆, which may be numerical or sym-
bolic, then 𝑋(𝑠) is a real number associated with 𝑠. The same experiment, of course,
may generate several random variables. Each of these random variables will, in turn,
have sample spaces whose elements are usually denoted by lower case letters such as
𝑥1 , 𝑥2 , 𝑥3 , . . . for the random variable 𝑋. We are now interested in assigning proba-
bilities to events such as P(𝑋 = 𝑥1 ), the probability that the random variable 𝑋 is
𝑥1 and P(𝑋 ≤ 𝑥2 ), the probability that the random variable is less than or equal to
𝑥2 .
If the sample space is finite or countably infinite on the integers (that is, the ele-
ments 𝑥0 , 𝑥1 , 𝑥2 , . . . can be counted against integers, say 0, 1, 2, . . .) then we say that
the random variable is discrete. Technically, the set {𝑥𝑖 } will be a countable subset
𝒱, say, of the real numbers ℛ. We can represent the {𝑥𝑖 } generically by the variable
𝑥 with 𝑥 ∈ 𝒱. For example, 𝒱 could be the set
{0, 12 , 1, 23 , 2, 25 , 3, . . .}.
In many cases 𝒱 consists simply of the integers or a subset of the integers, such as
𝒱 = {0, 1} or 𝒱 = {0, 1, 2, 3, . . .}.
In the random walks of Chapter 3, however, 𝒱 may contain all the positive and neg-
ative integers
. . . − 3, −2, −1, 0, 1, 2, 3, . . . .
In these integer cases we can put 𝑥𝑖 = 𝑖.
The function
𝑝(𝑥𝑖 ) = P(𝑋 = 𝑥𝑖 )
is known as the probability mass function. The pairs {𝑥𝑖 , 𝑝(𝑥𝑖 )} for all 𝑖 in the
sample space define the probability distribution of the random variable 𝑋. If 𝑥𝑖 =
𝑖, which occurs frequently in applications, then 𝑝(𝑥𝑖 ) = 𝑝(𝑖) is replaced by 𝑝𝑖 . Since
the 𝑥 values are mutually exclusive and exhaustive then it follows that
∙ (a) 0 ≤ 𝑝(𝑥𝑖 ) ≤ 1 for all 𝑖,
∑∞ ∑
∙ (b) 𝑝(𝑥𝑖 ) = 1, or in generic form 𝑝(𝑥) = 1,
𝑖=0 𝑥∈𝒱
𝑘
∑
∙ (c) P(𝑋 ≤ 𝑥𝑘 ) = 𝑝(𝑥𝑖 ), which is known as the distribution function.
𝑖=0

Example 1.5. A fair die is rolled until the first 6 appears face up. Find the probability that the
first 6 appears at the 𝑛-th throw.

Let the random variable 𝑁 be the number of throws until the first 6 appears face up. This is
an example of a discrete random variable 𝑁 with an infinite number of possible outcomes
{1, 2, 3, . . .} .
1
The probability of a 6 appearing for any throw is 6
and of any other number appearing is 56 .
CONTINUOUS RANDOM VARIABLES 9
Hence the probability of 𝑛 − 1 numbers other than 6 appearing followed by a 6 is
( 5 )𝑛−1 ( 1 ) 5𝑛−1
P(𝑁 = 𝑛) = = ,
6 6 6𝑛
which is the probability mass function for this random variable.

1.5 Continuous random variables

In many applications the discrete random variable, which for example might take
the integer values 1, 2, . . ., is inappropriate for problems where the random variable
can take any real value in an interval. For example, the random variable 𝑇 could be
the time measured from time 𝑡 = 0 until a light bulb fails. This could be any value
𝑡 ≥ 0. In this case 𝑇 is called a continuous random variable. Generally, if 𝑋 is a
continuous random variable there are mathematical difficulties in defining the event
𝑋 = 𝑥: the probability is usually defined to be zero. Probabilities for continuous
random variables may only be defined over intervals of values as, for example, in
P(𝑥1 < 𝑋 < 𝑥2 ).
We define a probability density function (pdf) 𝑓 (𝑥) over −∞ < 𝑥 < ∞ which
has the properties:
∙ (a) 𝑓 (𝑥) ≥ 0, (−∞ < 𝑥 < ∞);
∫ 𝑥2
∙ (b) P(𝑥1 ≤ 𝑋 ≤ 𝑥2 ) = 𝑓 (𝑥)𝑑𝑥 for any 𝑥1 , 𝑥2 such that −∞ < 𝑥1 < 𝑥2 <
𝑥1
∞;
∫ ∞
∙ (c) 𝑓 (𝑥)𝑑𝑥 = 1.
−∞

A possible graph of a density function 𝑓 (𝑥) is shown in Figure 1.3. By (a) above
the curve must remain nonnegative, by (b) the probability that 𝑋 lies between 𝑥1

f(x)

x1 x2 x

Figure 1.3 A probability density function.

and 𝑥2 is the shaded area, and by (c) the total area under the curve must be 1 since
P(−∞ < 𝑋 < ∞) = 1.
We define the (cumulative) distribution function (cdf) 𝐹 (𝑥) as the probability
10 SOME BACKGROUND ON PROBABILITY
that 𝑋 is less than or equal to 𝑥. Thus
∫ 𝑥
𝐹 (𝑥) = P(𝑋 ≤ 𝑥) = 𝑓 (𝑢)𝑑𝑢.
−∞
It follows from (c) above that
𝐹 (𝑥) → 1 as 𝑥 → ∞,
and that ∫ 𝑥2
P(𝑥1 ≤ 𝑥 ≤ 𝑥2 ) = 𝑓 (𝑢)𝑑𝑢 = 𝐹 (𝑥2 ) − 𝐹( 𝑥1 ).
𝑥1
A possible cdf is shown in Figure 1.4.
F(x)

Figure 1.4 A (cumulative) distribution function.

Example 1.6. Show that

{
1/(𝑏 − 𝑎) 𝑎≤𝑥≤𝑏
𝑓 (𝑥) =
0 for all other values of 𝑥
is a possible probability density function. Find its cumulative distribution function.

The function 𝑓 (𝑥) must satisfy conditions (a) and (c) above. This is the case since 𝑓 (𝑥) ≥ 0
and ∫ ∞ ∫ 𝑏
1
𝑓 (𝑥)𝑑𝑥 = 𝑑𝑥 = 1.
−∞ 𝑎
𝑏 − 𝑎
Also its cumulative distribution function 𝐹 (𝑥) is given by
∫ 𝑥
1 𝑥−𝑎
𝐹 (𝑥) = 𝑑𝑥 = for 𝑎 ≤ 𝑥 ≤ 𝑏.
𝑎
𝑏−𝑎 𝑏−𝑎
For 𝑥 < 𝑎, 𝐹 (𝑥) = 0 and for 𝑥 > 𝑏, 𝐹 (𝑥) = 1.
The pdf 𝑓 (𝑥) is the density function of the uniform distribution.

1.6 Mean and variance

The mean (or expectation or expected value), E(𝑋), of a discrete random variable
𝑋 is defined as
∑∞
𝝁 = E(𝑋) = 𝑥𝑖 𝑝(𝑥𝑖 ),
𝑖=0
MEAN AND VARIANCE 11
where 𝑝(𝑥𝑖 ) = P(𝑋 = 𝑥𝑖 ), and, for a continuous random variable 𝑋 by
∫ ∞
𝝁 = E(𝑋) = 𝑥𝑓 (𝑥)𝑑𝑥,
−∞

where 𝑓 (𝑥) is the probability density function. It can be interpreted as the weighted
average of the values of 𝑋 in its sample space, where the weights are either the
probability function or the density function. It is a measure which may be used to
summarize the probability distribution of 𝑋 in the sense that it is a central value. In
the discrete case the summation over ‘all 𝑥’ includes both finite and infinite sample
spaces.
A measure which is used in addition to the mean as a summary measure is the
variance of 𝑋 denoted by V(𝑋) or 𝝈 2 . This gives a measure of variation or spread
(dispersion) of the probability distribution of 𝑋, and is defined by
𝝈2 = V(𝑋)
= E[(𝑋 − E(𝑋))2 ] = E[(𝑋 − 𝝁)2 ]
⎧ ∞
 ∑ ∑


⎨ (𝑥𝑖 − 𝝁)2 𝑝(𝑥𝑖 ) or (𝑥 − 𝝁)2 𝑝(𝑥), if 𝑋 is discrete,
= ∫𝑖=0∞ 𝑥∈𝒱

 2

⎩ (𝑥 − 𝝁) 𝑓 (𝑥)𝑑𝑥, if 𝑋 is continuous.
−∞

The variance is the mean of the squared deviations of each value of 𝑋 from the
central value 𝝁. In order to give a measure of variation which is in the same units as
the mean, the square root 𝝈 of V(𝑋) is used, namely
√
𝝈 = V(𝑋).
This is known as the standard deviation (sd) of 𝑋.
A function of a random variable is itself a random variable. If ℎ(𝑋) is a function
of the random variable 𝑋, then it can be shown that the expectation of ℎ(𝑋) is given
by
⎧ ∞

 ∑ ∑

⎨ ℎ(𝑥𝑖 )𝑝(𝑥𝑖 ) or ℎ(𝑥)𝑝(𝑥), if 𝑋 discrete
E[ℎ(𝑋)] = 𝑖=0
∫ ∞ 𝑥∈𝒱



⎩ ℎ(𝑥)𝑝(𝑥)𝑑𝑥, if 𝑋 continuous.
−∞

It is relatively straightforward to derive the following results for the expectation

and variance of a linear function of 𝑋:
E(𝑎𝑋 + 𝑏) = 𝑎E(𝑋) + 𝑏 = 𝑎𝝁 + 𝑏,

V(𝑎𝑋 + 𝑏) = E[(𝑎𝑋 + 𝑏 − 𝑎𝝁 − 𝑏)2 ]

= E[(𝑎𝑋 − 𝑎𝝁)2 ] = 𝑎2 E[(𝑋 − 𝝁)2 ] = 𝑎2 V(𝑋),
where 𝑎 and 𝑏 are constants. Note that the translation of the random variable does
not affect the variance. Also
V(𝑋) = E(𝑋 2 ) − 𝝁2 ,
12 SOME BACKGROUND ON PROBABILITY
which is sometimes known as the computational formula for the variance. These re-
sults hold whether 𝑋 is discrete or continuous and enable us to interpret expectation
and variance as operators on 𝑋. Of course, in all cases mean and variance only exist
where the summations and infinite integrals are finite.
For expectations, it can be shown more generally that
[ 𝑘 ] 𝑘
∑ ∑
E 𝑎𝑖 ℎ𝑖 (𝑋) = 𝑎𝑖 E[ℎ𝑖 (𝑋)],
𝑖=1 𝑖=1

where 𝑎𝑖 , 𝑖 = 1, 2, . . . , 𝑘 are constants and ℎ𝑖 (𝑋), 𝑖 = 1, 2, . . . , 𝑘 are functions of

the random variable 𝑋.

1.7 Some standard discrete probability distributions

In this section we shall look at discrete random variables 𝑋 with probability mass
function 𝑝(𝑥𝑖 ) or 𝑝𝑖 , where 𝑥𝑖 takes integer values. Each of the random variables
considered are numerical outcomes of independent repetitions (or trials) of a simple
experiment knowns as a Bernoulli experiment. This is an experiment where there
are only two outcomes: a ‘success’ (𝑋 = 1) or a ‘failure’ (𝑋 = 0). The value of
the random variable 𝑋 is used as an indicator of the outcome which may also be
interpreted as the presence or absence of a particular characteristic. For example, in
a single coin toss 𝑋 = 1 is associated with the occurrence, or the presence of the
characteristic, of a head and 𝑋 = 0 with a tail, or the absence of a head.
The probability function of this random variable may be expressed in algebraic
terms as
𝑝𝑖 = 𝑝𝑖 𝑞 1−𝑖 , over the two values 𝑖 = 0, 1,
where 𝑝 = P(𝑋 = 1) is known as the parameter of the probability distribution, and
𝑞 = 1 − 𝑝. We say that the random variable has a Bernoulli distribution.
The expected value of 𝑋 is easily seen to be
𝝁 = E(𝑋) = 0 × 𝑞 + 1 × 𝑝 = 𝑝,
and the variance of 𝑋 is
𝝈 2 = V(𝑋) = E(𝑋 2 ) − 𝝁2 = 02 × 𝑞 + 12 × 𝑝 − 𝑝2 = 𝑝𝑞.
Suppose now that we are interested in random variables associated with indepen-
dent repetitions of Bernoulli experiments each with a probability of success, 𝑝. Con-
sider first the probability distribution of a random variable 𝑋 (a different 𝑋 from
that defined above) which is the number of successes in a fixed number of indepen-
dent trials, 𝑛. Independence of the trials gives the probability of any sequence of 1’s
and 0’s without order, which are mutually exclusive, leading to 𝑋 = 𝑘 as 𝑝𝑘 𝑞 𝑛−𝑘 .
The number of ways in which 𝑥 successes can be arranged in 𝑛 trials is the binomial
expression ( )
𝑛! 𝑛
or .
𝑘!(𝑛 − 𝑘)! 𝑘
Since each of these mutually exclusive sequences occurs with probability 𝑝𝑘 𝑞 𝑛−𝑘 ,
SOME STANDARD DISCRETE PROBABILITY DISTRIBUTIONS 13
the probability function of this random variable is given by
( )
𝑛 𝑘 𝑛−𝑘
𝑝(𝑘) or 𝑝𝑘 = 𝑝 𝑞 , 𝑘 = 0, 1, 2, . . . , 𝑛. (1.4)
𝑘
This is the 𝑘-th term in the binomial expansion of (𝑝 + 𝑞)𝑛 , and for this reason is
known as the binomial distribution with parameters 𝑛 and 𝑝. A consequence of this
observation is that
𝑛 𝑛 ( )
∑ ∑ 𝑛 𝑘 𝑛−𝑘
𝑝𝑘 = 𝑝 𝑞 = (𝑝 + 𝑞)𝑛 = 1.
𝑘
𝑘=0 𝑘=1

The mean and variance may be easily shown to be 𝑛𝑝 and 𝑛𝑝𝑞 respectively, which is
𝑛 times the mean and variance of the Bernoulli distribution.
Instead of fixing the number of trials, suppose now that the number of successes,
𝑟, is fixed, and that the sample size required in order to reach this fixed number is
the random variable 𝑋: this is sometimes called inverse sampling. Then in the case
𝑟 = 1, the probability that 𝑘 − 1 failures is followed by 1 success is

𝑝𝑘 = 𝑞 𝑘−1 𝑝, 𝑘 = 1, 2, . . . .,

which is the geometric probability function with parameter 𝑝. Note that successive
probabilities form a geometric series with common ratio 𝑞 = 1 − 𝑝. Note that the
sample space is now countably infinite. After some algebra it can be shown that the
mean is given by 𝝁 = 1/𝑝,and the variance by 𝝈 2 = 𝑞/𝑝2 .
The geometric probability distribution possesses an interesting property known as
the ‘no memory’, which can be expressed by

P(𝑋 > 𝑎 + 𝑏∣𝑋 > 𝑎) = P(𝑋 > 𝑏),

where 𝑎 and 𝑏 are positive integers. What this means is that if a particular event has
not occurred in the first 𝑎 repetitions of the experiment, then the probability that it
will occur in the next 𝑏 repetitions is the same as in the first 𝑏 repetitions of the
experiment. The result can be proved as follows, using the definition of conditional
probabilty in Section 1.3:
P(𝑋 > 𝑎 + 𝑏 ∩ 𝑋 > 𝑎) P(𝑋 > 𝑎 + 𝑏)
P(𝑋 > 𝑎 + 𝑏∣𝑋 > 𝑎) = = .
P(𝑋 > 𝑎)) P(𝑋 > 𝑎)
Since P(𝑋 > 𝑥) = 𝑞 𝑥 ,

𝑞 𝑎+𝑏
P(𝑋 > 𝑎 + 𝑏∣𝑋 > 𝑎) = = 𝑞 𝑏 = P(𝑋 > 𝑏).
𝑞𝑎
The converse is also true, but the proof is not given here.
In the case where 𝑟 > 1, the probability function of the number of trials may be
derived by noting that 𝑋 = 𝑘 requires that the 𝑘-th trial results in the 𝑟-th success
and that the remaining 𝑟 − 1 successes may occur in any order in the previous 𝑘 − 1
trials. Arguments based on counting the number of possible sequences of 1’s and 0’s,
14 SOME BACKGROUND ON PROBABILITY
and independence of trials, lead to
( )
𝑘 − 1 𝑟 𝑘−𝑟
𝑝𝑘 = 𝑝 𝑞 , 𝑘 = 𝑟, 𝑟 + 1, . . . .
𝑟−1
This is known as a Pascal or negative binomial distribution. Its mean is 𝑟/𝑝 and its
variance 𝑟𝑞/𝑝2 , which are respectively 𝑟 times the mean and 𝑟 times the variance of
the geometric distribution. Hence a similar relationship exists between the geometric
and the Pascal distributions as between the Bernoulli and the binomial distributions.
The binomial random variable arises as the result of observing 𝑛 independent iden-
tically distributed Bernoulli random variables, and the Pascal by observing 𝑟 sets of
geometric random variables.
Certain problems involve the counting of the number of events which have oc-
curred in a fixed time period. For example, the number of emissions of alpha parti-
cles by an X-ray source or the number of arrivals of customers joining a queue. It
has been found that the Poisson distribution is appropriate in modeling these counts
when the underlying process generating them is considered to be completely random.
We shall spend some time in Chapter 5 defining such a process which is known as a
Poisson process.
As well as being a probability distribution in its own right, the Poisson distribu-
tion also provides a convenient approximation to the binomial distribution to which
it converges when 𝑛 is large and 𝑝 is small and 𝑛𝑝 = 𝛼, a constant. This is a situa-
tion where rounding errors would be likely to cause computational problems, if the
numerical probabilities were to be calculated.
The Poisson probability function with parameter 𝛼 is
𝑒−𝛼 𝛼𝑘
𝑝𝑘 = , 𝑘 = 0, 1, 2, . . .
𝑘!
with mean and variance both equal to 𝛼.
The discrete uniform distribution with integer parameter 𝑛 has a random variable
𝑋 which can take the values 𝑟, 𝑟 + 1, 𝑟 + 2, . . . 𝑟 + 𝑛 − 1 with the same probability
1/𝑛 (the continuous uniform distribution was introduced in Example 1.6). It is easy
to show that the mean and variance of 𝑋 are given by
𝝁 = 21 (𝑛 + 1), 𝝈 2 = V(𝑋) = 1
12 (𝑛
2
− 1).
A simple example of the uniform discrete distribution is the fair die in which the
faces 1, 2, 3, 4, 5, 6 are equally likely to appear each with probability 61 .

1.8 Some standard continuous probability distributions

The exponential distribution is a continuous distribution which will be used in sub-
sequent chapters to model the random variable, say 𝑋, which is the time to a partic-
ular event. In a Poisson process, we shall see that this is the time between successive
occurrences of the event of interest. For example, the inter-arrival time of customers
in a queue, or the lifetime or the time to failure of a component, where the failure
rate 𝛼 is constant, in reliability theory, can be modelled by exponential distributions.
SOME STANDARD CONTINUOUS PROBABILITY DISTRIBUTIONS 15
The density function is
{
𝛼𝑒−𝛼𝑥 , 𝑥 ≥ 0
𝑓 (𝑥) =
0 𝑥<0
where 𝛼 is a positive parameter. It is a density function since
∫ ∞ ∫ ∞
𝑓 (𝑥)d𝑥 = 𝛼𝑒−𝛼𝑥 d𝑥 = 1.
−∞ 0

The mean and variance are given by

∫ ∞
1
𝝁 = E(𝑋) = 𝛼𝑥𝑒−𝛼𝑥 𝑑𝑥 = ,
0 𝛼
and ( )
1 2 −𝛼𝑥
∫ ∞
1
𝝈 2 = V(𝑋) = 𝑥− 𝛼𝑒 𝑑𝑥 = 2 .
0 𝛼 𝛼
The distribution function 𝐹 (𝑥) of this random variable is given by
∫ 𝑥 ∫ 𝑥
𝐹 (𝑥) = 𝑓 (𝑢)d𝑢 = 𝛼𝑒−𝛼𝑦 𝑑𝑦 = 1 − 𝑒−𝛼𝑥 .
−∞ 0

A closed form for 𝐹 (𝑥) is not always possible for continuous random variables, and
tables are sometimes necessary. Like the geometric distribution considered in the last
section, the exponential distribution possesses the ‘no memory’ property
P(𝑋 > 𝑎 + 𝑏∣𝑋 > 𝑎) = P(𝑋 > 𝑏), 𝑎, 𝑏 > 0. (1.5)
This may be proved using the definition of conditional probability in Section 1.3, and
noting that P(𝑋 > 𝑥) = 𝑒−𝛼𝑥 . It can be shown that if (1.5) is true, then 𝑋 has an
exponential distribution.
A random variable 𝑋 has a normal distribution with mean 𝝁 and variance V(𝑋)
or 𝝈 2 , if its probability density function is
[ ]
1 (𝑥 − 𝝁)2
𝑓 (𝑥) = √ exp − (−∞ < 𝑥 < ∞). (1.6)
𝝈 2𝜋 2𝝈 2
That ∫ ∞
𝑓 (𝑥)𝑑𝑥 = 1
−∞
follows from the standard integral
∫ ∞
2 √
𝑒−𝑢 𝑑𝑢 = 𝜋.
−∞

For the normal distribution it can be verified that the mean and variance are
∫ ∞ ∫ ∞
𝑥𝑓 (𝑥)𝑑𝑥 = E(𝑋) = 𝝁, (𝑥 − 𝝁)2 𝑓 (𝑥)𝑑𝑥 = 𝝈 2 .
−∞ −∞

The normal distribution is often denoted by 𝑁 (𝝁, 𝝈 2 ).

The gamma distribution depends on the properties of the gamma function Γ(𝑛)
16 SOME BACKGROUND ON PROBABILITY
defined by ∫ ∞
Γ(𝑛) = 𝑥𝑛−1 𝑒−𝑥 𝑑𝑥.
0
Integration by parts produces the recurrence formula
Γ(𝑛) = (𝑛 − 1)Γ(𝑛 − 1),
since
∫ ∞
𝑑𝑒−𝑥
Γ(𝑛) = − 𝑥𝑛−1 𝑑𝑥
0 𝑑𝑥
∫ ∞ 𝑛−1
𝑑𝑥
= [−𝑥𝑛−1 𝑒−𝑥 ]∞
0 + 𝑒−𝑥 𝑑𝑥
0 𝑑𝑥
∫ ∞
𝑛−2 −𝑥
= (𝑛 − 1) 𝑥 𝑒 𝑑𝑥 = (𝑛 − 1)Γ(𝑛 − 1).
0
When 𝑛 is an integer it follows that
Γ(𝑛) = (𝑛 − 1)!.
The gamma distribution has two parameters 𝑛, 𝛼, and has the density function
𝛼𝑛 𝑛−1 −𝛼𝑥
𝑓 (𝑥) = 𝑥 𝑒 , 𝑥 > 0.
Γ(𝑛)
Note that by setting 𝑛 = 1, we obtain the exponential distribution. The mean is given
by ∫ ∞ ∫ ∞
𝛼𝑛 𝛼𝑛
E(𝑋) = 𝑥𝑛 𝑒−𝛼𝑥 𝑑𝑥 = 𝑦 𝑛 𝑒−𝑦 𝑑𝑦,
Γ(𝑛) 0 Γ(𝑛)𝛼𝑛+1 0
after the change of variable 𝑦 = 𝛼𝑥. Hence
Γ(𝑛 + 1) 𝑛
E(𝑋) = = .
𝛼Γ(𝑛) 𝛼
It can be shown that the variance of the gamma distribution is 𝑛/𝛼2 . Note that the
mean and variance are respectively 𝑛 times the mean and variance of the exponential
distribution with parameter 𝛼.
It will be proved later that if 𝑋1 , 𝑋2 , . . . , 𝑋𝑛 are independent identically dis-
tributed exponential random variables, then
𝑛
∑
𝑌 = 𝑋𝑖
𝑖=1

has a gamma distribution with parameters 𝛼, 𝑛.

Another distribution arising in reliability (Chapter 8) is the Weibull distribution
which depends on two positive parameters 𝛼, 𝛽, and has density
𝛽
𝑓 (𝑥) = 𝛼𝛽𝑥𝛽−1 𝑒−𝛼𝑥 , 𝑡 > 0.
This enables more complex lifetime data to be modeled, especially where a constant
failure rate is not a reasonable assumption. Note that setting 𝛽 = 1 gives the expo-
nential distribution. After some algebra, it may be shown that the mean and variance
GENERATING FUNCTIONS 17
of the Weibull distribution are
) (
1 1
𝝁= 1Γ +1 ,
𝛼𝛽 𝛽
[ ( ) { ( )}2 ]
2 1 2 1
𝝈 = 2 Γ +1 − Γ +1 .
𝛼𝛽 𝛽 𝛽

1.9 Generating functions

In this section we are going to consider two generating functions. The main purpose
is to use their uniqueness properties to identify the probability distributions of func-
tions of random variables. The first is the moment generating function (mgf). This
function depends on a dummy variable 𝑠 and uses the series expansion of 𝑒𝑠𝑥 to gen-
erate the moments E(𝑋 𝑟 ), (𝑟 ≥ 1) of the random variable 𝑋. The expected value
E(𝑋 𝑟 ) is called the 𝑟-th moment of the random variable 𝑋 about zero. The power
series expansion of 𝑒𝑠𝑋 is given by
(𝑠𝑋)2 (𝑠𝑋)𝑟
𝑒𝑠𝑋 = 1 + 𝑠𝑋 + + ⋅⋅⋅ + + ⋅⋅⋅,
2! 𝑟!
a series which converges for all 𝑠𝑋. The mgf is obtained by taking expected values
of both sides of this equation,
[ ]
𝑠𝑋 (𝑠𝑋)2 (𝑠𝑋)𝑟
𝑀𝑋 (𝑠) = E(𝑒 ) = E 1 + 𝑠𝑋 + + ⋅⋅⋅+ + ⋅⋅⋅ . (1.7)
2! 𝑟!
We mentioned in Section 1.6 that the expected value of a finite sum of functions was
equal to the sum of the expected values, and that E(𝑎𝑋) = 𝑎E(𝑋). We now wish
to apply this result to an infinite sum, making the assumption that this result holds (it
does so under fairly general conditions). Thus we assume that
∞
𝑠2 𝑠𝑟 ∑ 𝑠𝑟
𝑀𝑋 (𝑠) = 1 + 𝑠E(𝑋) + E(𝑋 2 ) + ⋅ ⋅ ⋅ + E(𝑋 𝑟 ) + ⋅ ⋅ ⋅ = E(𝑋 𝑟 ).
2! 𝑟! 𝑟=0
𝑟!

The coefficient of 𝑠𝑟 /𝑟! is therefore the 𝑟-th moment of 𝑋. Taking successive deriva-
tives of the mgf with respect to 𝑠, and then setting 𝑠 = 0 we can obtain these mo-
ments. For example,
′ ′′
𝑀𝑋 (0) = E(𝑋) = 𝝁, 𝑀𝑋 (0) = E(𝑋 2 ),
and the variance is therefore given by (see Section 1.6)
𝝈 2 = 𝑀 ′′ (0) − [𝑀 ′ (0)]2 .
Let 𝑋 have a gamma distribution with parameters 𝑛, 𝛼. Then
∫ ∞ 𝑛
𝛼
𝑠𝑋
𝑀𝑋 (𝑠) = E(𝑒 ) = 𝑒𝑠𝑥 𝑥𝑛−1 𝑒−𝛼𝑥 𝑑𝑥,
0 Γ(𝑛)
∫ ∞ 𝑛
𝛼
= 𝑥𝑛−1 𝑒−𝑥(𝛼−𝑠) 𝑑𝑥.
0 Γ(𝑛)
18 SOME BACKGROUND ON PROBABILITY
The change of variable 𝑥 = 𝑢/(𝛼 − 𝑠) leads to the result
( )𝑛
𝛼
𝑀𝑋 (𝑠) = provided that 𝑠 < 𝛼.
𝛼−𝑠
Now consider the result quoted but not proved in the previous section on the distribu-
tion of independent and identically distributed (iid) exponential random variables
𝑋1 , 𝑋2 , . . . , 𝑋𝑛 .
We may now use the two results above to prove this. We first need to note that
since the random variables are independent, it follows that
( ) ( ) ( ) ( )
E 𝑔1 (𝑋1 )𝑔2 (𝑋2 ) ⋅ ⋅ ⋅ 𝑔𝑛 (𝑋𝑛 ) = E 𝑔1 (𝑋1 ) E 𝑔2 (𝑋2 ) ⋅ ⋅ ⋅ E 𝑔𝑛 (𝑋𝑛 ) .
∑
Let 𝑌 = 𝑛𝑖=1 𝑋𝑖 . Then
∑𝑛
𝑀𝑌 (𝑠) = E(𝑒𝑠𝑌 ) = E(𝑒𝑠 𝑖=1 𝑋𝑖 )
∏𝑛 𝑛
∏
= E( 𝑒𝑠𝑋𝑖 ) = E(𝑒𝑠𝑋𝑖 ) (using the result above)
𝑖=1 𝑖=1
𝑛
∏
= 𝑀𝑋𝑖 (𝑠),
𝑖=1

but since the 𝑋𝑖 s are identically distributed, they have the same mgf 𝑀𝑋 (𝑠). Hence
[ ]𝑛
𝑀𝑌 (𝑠) = 𝑀𝑋 (𝑠) .

Of course this result holds for any iid random variables. If 𝑋 is exponential, then
𝛼
𝑀𝑋 (𝑠) = ,
𝛼−𝑠
and ()𝑛
𝛼
𝑀𝑌 (𝑠) = ,
𝛼−𝑠
which is the mgf of a gamma-distributed random variable.

Example 1.7 𝑛 rails are cut to a nominal length of 𝑎 meters, with an error exponentially
distributed with parameter 𝛼 independently of each other. The rails are welded end-to-end to
form a continuous rail of nominal length 𝑛𝑎. Find the expected length and variance of the
composite rail.

Let 𝑋𝑖 be the random variable representing the length of rail 𝑖. Let 𝑌 be the random variable
of the length of the whole rail so that
𝑌 = 𝑋1 + 𝑋2 + ⋅ ⋅ ⋅ + 𝑋𝑛 .
Since the random variables are independent and identically distributed, and the errors expo-
nentially distributed, the moment generating function is (see above)
𝑀𝑌 (𝑠) = [𝑀𝑋 (𝑠)]𝑛 ,
GENERATING FUNCTIONS 19
where 𝑋 is exponentially distributed. Therefore
( 𝛼 )𝑛 ( 𝑠 −𝑛
)
𝑀𝑌 (𝑠) = = 1−
𝛼−𝑠 𝛼
𝑛 1 𝑛(𝑛 − 1) 2
= 1+ 𝑠+ 𝑠 + ⋅⋅⋅.
𝛼 2 𝛼2
Hence the expected value of the length of the composite rail and its variance are given by
𝑛 𝑛(𝑛 − 1) 𝑛
E(𝑌 ) = , 𝝈 2 = E(𝑌 2 ) − [E(𝑌 )]2 = − 2.
𝛼 𝛼2 𝛼
The latter result is an example of the result that
V(𝑋1 + 𝑋2 + ⋅ ⋅ ⋅ + 𝑋𝑛 ) = V(𝑋1 + V(𝑋2 ) + ⋅ ⋅ ⋅ + V(𝑋𝑛 ),

provided 𝑋1 , 𝑋2 , . . . , 𝑋𝑛 are independent.

Moment generating functions may be defined for both discrete and continuous
random variables but the probability generating function (pgf) is only defined for
integer-valued random variables. To be specific, we consider the probability distri-
bution 𝑝𝑛 = P(𝑁 = 𝑛), where the random variable 𝑁 can only take the values
0, 1, 2, . . .. (We use the notation 𝑝𝑛 rather than 𝑝(𝑛) in this context since it conforms
to the usual notation for coefficients in power series.) Again it is expressed in terms
of a power series in a dummy variable 𝑠, and is defined as the expected value of 𝑠𝑁 :
∞
∑
𝐺𝑁 (𝑠) = E(𝑠𝑁 ) = 𝑝 𝑛 𝑠𝑛 ,
𝑛=0

provided the right-hand side exists. The question of uniqueness arises with generating
functions, since we deduce distributions from them. Later we shall represent the
probability generating function by 𝐺(𝑠) without the random variable subscript. It
can be shown (see Grimmett and Welsh (1986)) that two random variables 𝑋 and
𝑌 have the same probability generating function if and only if they have the same
probability distributions.
In many cases the pgf will take a closed form: this will occur if the outcomes
can only be a finite number. In others, such as in birth and death processes (Chapter
6), in which there is, theoretically, no upper bound to the population size, the series
for the pgf will be infinite. However,
∑∞ in all cases the coefficient of 𝑠𝑛 will give the
probability 𝑝𝑛 . If 𝐺𝑁 (𝑠) = 𝑛=0 𝑝𝑛 𝑠𝑛 , then
∞
d𝐺𝑁 ∑
= 𝐺′𝑁 (𝑠) = 𝑛𝑝𝑛 𝑠𝑛 , (0 ≤ 𝑠 ≤ 1),
d𝑠 𝑛=1
∞
d2 𝐺𝑁 ′′
∑
= 𝐺𝑁 (𝑠) = 𝑛(𝑛 − 1)𝑝𝑛 𝑠𝑛−2 , (0 ≤ 𝑠 ≤ 1).
d𝑠2 𝑛=2
The pgf has the following properties:
∞
∑
∙ (a) 𝐺𝑁 (1) = 𝑝𝑛 = 1.
𝑛=0
20 SOME BACKGROUND ON PROBABILITY
∞
∑
∙ (b) 𝐺′𝑁 (1) = 𝑛𝑝𝑛 = E(𝑁 ) = 𝝁, the mean.
𝑛=0
∑∞
∙ (c) 𝐺′′𝑁 (1) = 𝑛(𝑛 − 1)𝑝𝑛 = E(𝑁 2 ) − E(𝑁 ), so that the variance is given by
𝑛=0

V(𝑁 ) = 𝝈 2 = E(𝑁 2 ) − 𝝁2 = 𝐺′′ (1) + 𝐺′ (1) − [𝐺′ (1)]2 .

(𝑚)
∙ (d) 𝐺𝑁 (1) = 𝑑𝑚 𝐺(1)/𝑑𝑠𝑚 = E[𝑁 (𝑁 − 1) . . . (𝑁 − 𝑚 + 1)], which is called
the factorial moment of 𝑁 .
∙ (e) If 𝑁1 , 𝑁2 , . . . , 𝑁∑
𝑟 are independent and identically distributed discrete random
variables, and 𝑌 = 𝑟𝑖=1 𝑁𝑖 , then
[ 𝑟 ]
[ ∑𝑟 ] ∏
𝐺𝑌 (𝑠) = E[𝑠𝑌 ] = E 𝑠 𝑖=1 𝑁𝑖 = E (𝑠𝑁𝑖 )
𝑖=1
𝑟
∏
= 𝐺𝑁𝑖 (𝑠),
𝑖=1

and since the 𝑁𝑖 ’s are identically distributed

𝐺𝑌 (𝑠) = [𝐺𝑋 (𝑠)]𝑟 ,
which is similar to the result for moment generating functions.

Example 1.8 The random variable 𝑁 has a binomial distribution with parameters 𝑚, 𝑝. Its
probability function is given by
( )
𝑚 𝑛 𝑚−𝑛
𝑝(𝑛) = 𝑝𝑛 = P(𝑁 = 𝑛) = 𝑝 𝑞 , 𝑛 = 0, 1, 2, . . . , 𝑚
𝑛
(see Eqn(1.4)). Find its pgf.

The pgf of 𝑁 is
𝑚 ( ) 𝑚 ( )
∑
𝑛 𝑚 𝑛 𝑚−𝑛 ∑ 𝑚
𝐺𝑁 (𝑠) = 𝐺(𝑠) = 𝑠 𝑝 𝑞 = (𝑝𝑠)𝑛 𝑞 𝑚−𝑛
𝑛 𝑛
𝑛=0 𝑛=0
= (𝑞 + 𝑝𝑠)𝑚
using the binomial theorem. It follows that
𝐺′ (𝑠) = 𝑚𝑝(𝑞 + 𝑝𝑠)𝑚−1 ,
𝐺′′ (𝑠) = 𝑚(𝑚 − 1)𝑝2 (𝑞 + 𝑝𝑠)𝑚−2 .
Using the results above, the mean and variance are given by
𝝁 = 𝐺′ (1) = 𝑚𝑝,
and
𝝈 2 = 𝐺′′ (1) + 𝐺′ (1) − [𝐺′ (1)]2 = 𝑚(𝑚 − 1)𝑝2 + 𝑚𝑝 − 𝑚2 𝑝2 = 𝑚𝑝𝑞.
CONDITIONAL EXPECTATION 21
The Bernoulli distribution is the binomial distribution with 𝑛 = 1. Hence its pgf is
𝐺𝑋 (𝑠) = 𝑞 + 𝑝𝑠.
Consider 𝑛 independent and ∑identically distributed Bernoulli random variables
𝑋1 , 𝑋2 , . . . , 𝑋𝑛 , and let 𝑌 = 𝑛𝑖=1 𝑋𝑖 . Then from the results above
𝐺𝑌 (𝑠) = [𝐺𝑋 (𝑠)]𝑛 = (𝑞 + 𝑝𝑠)𝑛 ,
which is again the pgf of a binomial random variable.
It is possible to associate a generating function with any sequence {𝑎𝑛 }, (𝑛 =
0, 1, 2, . . .) in the form
∑∞
𝐻(𝑠) = 𝑎𝑛 𝑠𝑛
𝑛=0
provided that the series converges in some interval containing the origin 𝑠 = 0. Un-
like the pgf this series need not satisfy the conditions 𝐻(1) = 1 for a probability dis-
tribution nor 0 ≤ 𝑎𝑛 ≤ 1. An application using such a series is given Problem 1.24,
and in Section 3.3 on random walks.

1.10 Conditional expectation

In many applications of probability we are interested in the possible values of two or
more characteristics in a problem. For this we require two or more random variables
which may or may not be independent. We shall only consider the case of two dis-
crete random variables 𝑋 and 𝑌 which form a two-dimensional random variable de-
noted by (𝑋, 𝑌 ), which can take pairs of values (𝑥𝑖 , 𝑦𝑗 ) (𝑖 = 1, 2, . . . ; 𝑗 = 1, 2, . . .)
(either sequence may be finite or infinite) with joint probability (mass) function
𝑝(𝑥𝑖 , 𝑦𝑗 ), which is now a function of two variables. As in Section 1.4, the proba-
bilities must satisfy
∙ (a) 0 ≤ 𝑝(𝑥𝑖 , 𝑦𝑗 ) ≤ 1 for all (𝑖, 𝑗),
∑∞ ∑∞
∙ (b) 𝑝(𝑥𝑖 , 𝑦𝑗 ) = 1.
𝑖=1 𝑗=1

The random variables 𝑋 and 𝑌 are said to be independent, if and only if,
𝑝(𝑥𝑖 , 𝑦𝑗 ) = 𝑞(𝑥𝑖 )𝑟(𝑦𝑗 ) for all 𝑖 and 𝑗, (1.8)
where, of course,
∞
∑ ∞
∑
𝑞(𝑥𝑖 ) = 1, 𝑟(𝑦𝑗 ) = 1.
𝑖=1 𝑗=1

If the random variable 𝑍 = 𝐻(𝑋, 𝑌 ) is a function of the random variables 𝑋 and

𝑌 , then we will state without proof that the expected value of 𝑍 is given by
∞ ∑
∑ ∞
E(𝑍) = 𝐻(𝑥𝑖 , 𝑦𝑗 )𝑝(𝑥𝑖 , 𝑦𝑗 ).
𝑖=1 𝑗=1
22 SOME BACKGROUND ON PROBABILITY
We can also consider the conditional expectation of 𝑋 for a given value 𝑦𝑗 of 𝑌 .
This is defined as
∑∞ ∑∞
E(𝑋∣𝑌 = 𝑦𝑗 ) = 𝑥𝑖 P(𝑋 = 𝑥𝑖 ∣𝑌 = 𝑦𝑗 ) = 𝑥𝑖 𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 )
𝑖=1 𝑖=1

where 𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 ) is the probability that 𝑋 = 𝑥𝑖 occurs given that 𝑌 = 𝑦𝑗 has
occurred. We can view this expectation as a function of 𝑦𝑖 . We can also interpret
it as the same function of the random variable 𝑌 : we denote this function of 𝑌 as
E(𝑋∣𝑌 ). Similarly
∑∞
E(𝑌 ∣𝑋 = 𝑥𝑖 ) = 𝑦𝑗 𝑝𝑌 (𝑦𝑗 ∣𝑥𝑖 )
𝑗=1

and E(𝑌 ∣𝑋) can be defined. The conditional probability 𝑝(𝑥𝑖 ∣𝑦𝑗 ) is given by (see
Section 1.4)
𝑝(𝑥𝑖 , 𝑦𝑗 )
𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 ) = (𝑝𝑌 (𝑦𝑗 ) ∕= 0),
𝑝𝑌 (𝑦𝑗 )
where
∑∞
𝑝𝑌 (𝑦𝑗 ) = 𝑝𝑋 (𝑥𝑖 , 𝑦𝑗 )
𝑖=1
is the marginal probability distribution of 𝑌 .
As we have observed the expectations E(𝑋∣𝑌 ) and E(𝑌 ∣𝑋) are random vari-
ables which are functions of 𝑌 and 𝑋 respectively. However, if 𝑋 and 𝑌 are inde-
pendent random variables, then 𝑝(𝑥𝑖 , 𝑦𝑗 ) = 𝑞(𝑥𝑖 )𝑟(𝑦𝑗 ) by Eqn(1.9) so that
𝑝(𝑥𝑖 , 𝑦𝑗 ) 𝑞(𝑥𝑖 )𝑟(𝑦𝑗 )
𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 ) = = ∑∞ = 𝑞(𝑥𝑖 ),
𝑝𝑌 (𝑦𝑗 ) 𝑖=1 𝑞(𝑥𝑖 )𝑟(𝑦𝑗 )
and, similarly,
𝑝𝑌 (𝑦𝑖 ∣𝑥𝑗 ) = 𝑟(𝑦𝑗 ).
Note that, for independent random variables, the marginal probability distributions
with respect to 𝑌 and 𝑋 are 𝑟(𝑦𝑖 ) and 𝑞(𝑥𝑖 ) respectively. Hence
∞
∑ ∞
∑
E(𝑋∣𝑌 ) = 𝑥𝑖 𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 ) = 𝑥𝑖 𝑞(𝑥𝑖 ) = E(𝑋).
𝑖=1 𝑖=1

By a similar argument, E(𝑌 ∣𝑋) = E(𝑌 ).

On the other hand if 𝑋 and 𝑌 are dependent random variables, then E(𝑋∣𝑌 ) and
E(𝑌 ∣𝑋) can themselves have expectations. Thus
∞
(∞ )
∑ ∑ 𝑝(𝑥𝑖 , 𝑦𝑗 )
E(E(𝑋∣𝑌 )) = 𝑥𝑖 𝑝𝑌 (𝑦𝑗 ),
𝑗=1 𝑖=1
𝑝𝑌 (𝑦𝑗 )

since 𝑝(𝑦𝑗 ) is the probability that 𝑌 = 𝑦𝑗 occurs. Hence

∞ ∑
∑ ∞
E(E(𝑋∣𝑌 )) = 𝑥𝑖 𝑝(𝑥𝑖 , 𝑦𝑗 ) = E(𝑋).
𝑗=1 𝑖=1
CONDITIONAL EXPECTATION 23
Similarly
E(E(𝑌 ∣𝑋)) = E(𝑌 ).

Example 1.9. A bird lays 𝑌 eggs in a nest, where 𝑌 has a binomial distribution with parame-
ters 𝑚 and 𝑝. Each egg hatches with probability 𝑟 independently of the other eggs. If 𝑋 is the
number of young in the nest, find the conditional expectations E(𝑋∣𝑌 ), the expected value
that an egg hatches given 𝑦 eggs have been laid, and E(𝑌 ∣𝑋), the expected value that 𝑦 eggs
were laid given 𝑥 eggs hatching.

Let 𝑦𝑗 = 𝑗, (0 ≤ 𝑗 ≤ 𝑚) be possible values of 𝑌 . Since 𝑌 is binomial, using the notation

above, ( )
𝑚 𝑗
𝑝𝑌 (𝑦𝑗 ) = 𝑝 (1 − 𝑝)𝑚−𝑗 , (0 ≤ 𝑗 ≤ 𝑚).
𝑗
Let 𝑥𝑖 = 𝑖, (0 ≤ 𝑖 ≤ 𝑗) be possible numbers of the young in the nest. Then the conditional
mass function
( )
𝑗 𝑖
𝑝𝑋 (𝑥𝑖 ∣𝑦𝑗 ) = P(𝑋 = 𝑥𝑖 ∣𝑌 = 𝑦𝑗 ) = 𝑟 (1 − 𝑟)𝑗−𝑖 , (0 ≤ 𝑗 ≤ 𝑚, 0 ≤ 𝑖 ≤ 𝑗),
𝑖
which is the number of ways in which 𝑖 items can be chosen from 𝑗 items. This distribution is
also binomial but with parameters 𝑗 and 𝑖. Hence
𝑗 𝑗 ( )
∑ ∑ 𝑗
E(𝑋∣𝑌 = 𝑦𝑗 ) = 𝑖𝑝(𝑥𝑖 ∣𝑦𝑗 ) = 𝑖 𝑟 𝑖 (1 − 𝑟)𝑗−𝑖
𝑖
𝑖=1 𝑖=1
𝑗 ( )( )𝑖
𝑗
∑ 𝑗 𝑟
= (1 − 𝑟) 𝑘
𝑖 1−𝑟
𝑖=1
𝑟 𝑟
( )𝑗−1
= (1 − 𝑟)𝑗 . .𝑗 1 +
1−𝑟 1−𝑟
= 𝑗𝑟 = 𝑦𝑗 𝑟.
Hence E(𝑋∣𝑌 ) is the random variable 𝑟𝑌 , (0 ≤ 𝑌 ≤ 𝑚).
To obtain the conditional expectation E(𝑌 ∣𝑋), start from Eqn(1.2) which can be written as
P(𝑌 ∩ 𝑋) = P(𝑋∣𝑌 )P(𝑌 ) = P(𝑋 ∩ 𝑌 ) = P(𝑌 ∣𝑋)P(𝑋).
Hence the conditional mass function
P(𝑋 = 𝑥𝑖 ∣𝑌 = 𝑦𝑗 )P(𝑌 = 𝑦𝑗 )
𝑝𝑋 (𝑦𝑗 , 𝑥𝑖 ) = P(𝑌 = 𝑦𝑗 ∣𝑋 = 𝑥𝑖 ) =
P(𝑋 = 𝑥𝑖 )
𝑝(𝑥𝑖 ∣𝑦𝑗 )𝑝(𝑦𝑗 )
= .
𝑝𝑋 (𝑥𝑖 )
where
𝑚 ( ) ( )
∑ 𝑠 𝑖 𝑠−𝑖 𝑚 𝑠
𝑝𝑋 (𝑥𝑖 ) = P(𝑋 = 𝑥𝑖 ) = 𝑟 (1 − 𝑟) . 𝑝 (1 − 𝑝)𝑚−𝑠
𝑖 𝑠
𝑠=𝑖
( 𝑟 )𝑖 𝑚 ( )( )
𝑚
∑ 𝑠 𝑚
= (1 − 𝑝) 𝜌𝑠
1−𝑟 𝑖 𝑠
𝑠=𝑖
24 SOME BACKGROUND ON PROBABILITY
where
𝑝(1 − 𝑟)
𝜌= .
1−𝑝
It follows that (𝑗 )(𝑚)
𝑖 𝑗
𝜌𝑗
𝑝(𝑦𝑗 ∣𝑥𝑖 ) = ∑𝑚 (𝑠)(𝑚) .
𝑠=𝑖 𝑖 𝑠
𝜌𝑠
Finally
∑𝑚 (𝑗 )(𝑚)
𝑗=𝑖
𝑗 𝜌𝑗
E(𝑌 ∣𝑋 = 𝑥𝑖 ) = ∑ 𝑚 (𝑠𝑖)(𝑚𝑗) = 𝑢(𝑖, 𝑟, 𝑝, 𝑚) = 𝑢(𝑥𝑖 , 𝑟, 𝑝, 𝑚),
𝑠=𝑖 𝑖 𝑠
𝜌𝑠
say. Hence the conditional expectation E(𝑌 ∣𝑋) is the random variable 𝑢(𝑋, 𝑟, 𝑝, 𝑚).

1.11 Problems

1.1. The Venn diagram of three events is shown in Figure 1.5. Indicate on the diagram the

A B

Figure 1.5

following events:
(a) 𝐴 ∪ 𝐵; (b) 𝐴 ∪ (𝐵 ∪ 𝐶); (c) 𝐴 ∩ (𝐵 ∪ 𝐶); (d) (𝐴 ∩ 𝐶)𝑐 ; (e) (𝐴 ∩ 𝐵) ∪ 𝐶 𝑐 .

1.2. In a random experiment, 𝐴, 𝐵, 𝐶 are three events. In set notation write down expressions
for the events:
(a) only 𝐴 occurs;
(b) all three events 𝐴, 𝐵, 𝐶 occur;
(c) 𝐴 and 𝐵 occur but 𝐶 does not;
(d) at least one of the events 𝐴, 𝐵, 𝐶 occurs;
(e) exactly one of the events 𝐴, 𝐵, 𝐶 occurs;
(f) not more than two of the events occur.

1.3. For two events 𝐴 and 𝐵, P(𝐴) = 0.4, P(𝐵) = 0.5, and P(𝐴 ∩ 𝐵) = 0.3. Calculate
(a) P(𝐴 ∪ 𝐵); (b) P(𝐴 ∩ 𝐵 𝑐 ); (c) P(𝐴𝑐 ∪ 𝐵 𝑐 ).

1.4. Two distinguishable fair dice 𝑎 and 𝑏 are rolled. What are the elements of the sample
PROBLEMS 25
space? What is the probability that the sum of the face values of the two dice is 9? What is the
probability that at least one 5 or at least one 3 appears?

1.5. Two distinguishable fair dice are rolled. What is the probability that the sum of the faces
is not more than 6?

1.6. A probability function {𝑝𝑛 }, (𝑛 = 0, 1, 2, . . .) has a probability generating function

∞
∑ 1
𝐺(𝑠) = 𝑝𝑛 𝑠𝑛 = 41 (1 + 𝑠)(3 + 𝑠) 2 .
𝑛=0

Find the probability function {𝑝𝑛 } and its mean.

1.7. Find the probability generating function 𝐺(𝑠) of the Poisson distribution (see Section 1.7)
with parameter 𝛼 given by
𝑒−𝛼 𝛼𝑛
𝑝𝑛 = , 𝑛 = 0, 1, 2, . . . .
𝑛!
Determine the mean and variance of {𝑝𝑛 } from the generating function.

1.8. A panel contains 𝑛 warning lights. The times to failure of the lights are the indepen-
dent random variables 𝑇1 , 𝑇2 , . . . , 𝑇𝑛 which have exponential distributions with parameters
𝛼1 , 𝛼2 , . . . , 𝛼𝑛 respectively. Let 𝑇 be the random variable of the time to first failure, that is
𝑇 = min{𝑇1 , 𝑇2 , . . . , 𝑇𝑛 }.
∑𝑛
Show that 𝑇 has an exponential distribution with parameter 𝑗=1
𝛼𝑗 . Show also that the
∑𝑛
probability that the 𝑖-th panel light fails first is 𝛼𝑖 /( 𝑗=1 𝛼𝑗 ).

1.9. The geometric probability function with parameter 𝑝 is given by

𝑝(𝑥) = 𝑞 𝑥−1 𝑝, 𝑥 = 1, 2, . . .
where 𝑞 = 1−𝑝 (see Section 1.7). Find its probability generating function. Calculate the mean
and variance of the geometric distribution from its pgf.

1.10. Two distinguishable fair dice 𝑎 and 𝑏 are rolled. What are the probabilities that:
(a) at least one 4 appears;
(b) only one 4 appears;
(c) the sum of the face values is 6;
(d) the sum of the face values is 5 and one 3 is shown;
(e) the sum of the face values is 5 or only one 3 is shown?

1.11. Two distinguishable fair dice 𝑎 and 𝑏 are rolled. What is the expected sum of the face
values? What is the variance of the sum of the face values?

1.12. Three distinguishable fair dice 𝑎, 𝑏, and 𝑐 are rolled. How many possible outcomes are
there for the faces shown? When the dice are rolled, what is the probability that just two dice
show the same face values and the third one is different?

1.13. In a sample space 𝑆, the events 𝐵 and 𝐶 are mutually exclusive, but 𝐴 and 𝐵, and 𝐴
and 𝐶 are not. Show that
P(𝐴 ∪ (𝐵 ∪ 𝐶)) = P(𝐴) + P(𝐵) + P(𝐶) − P(𝐴 ∩ (𝐵 ∪ 𝐶)).
26 SOME BACKGROUND ON PROBABILITY
From a well-shuffled pack of 52 playing cards a single card is randomly drawn. Find the
probability that it is a club or an ace or the king of hearts.

1.14. Show that ⎧

⎨ 0
 𝑥<0
𝑓 (𝑥) = 1/(2𝑎) 0≤𝑥≤𝑎

⎩ −(𝑥−𝑎)/𝑎
𝑒 /(2𝑎) 𝑥>𝑎
is a possible probability density function. Find the corresponding probability function.

1.15. A biased coin is tossed. The probability of a head is 𝑝. The coin is tossed until the first
head appears. Let the random variable 𝑁 be the total number of tosses including the first head.
Find P(𝑁 = 𝑛), and its pgf 𝐺(𝑠). Find the expected value of the number of tosses.

1.16. The 𝑛 random variables 𝑋1 , 𝑋2 , . . . , 𝑋𝑚 are independent and identically distributed

each with a gamma distribution with parameters 𝑛 and 𝛼. The random variable 𝑌 is defined
by
𝑌 = 𝑋1 + 𝑋2 + ⋅ ⋅ ⋅ + 𝑋𝑚 .
Using the moment generating function, find the mean and variance of 𝑌 .

1.17. A probability generating function with parameter 0 < 𝛼 < 1 is given by

1 − 𝛼(1 − 𝑠)
𝐺(𝑠) = .
1 + 𝛼(1 − 𝑠)
Find 𝑝𝑛 = P(𝑁 = 𝑛) by expanding the series in powers of 𝑠. What is the mean of the
probability function {𝑝𝑛 }?

1.18. Find the moment generating function of the random variables 𝑋 which has the uniform
distribution {
1/(𝑏 − 𝑎), 𝑎 ≤ 𝑥 ≤ 𝑏,
𝑓 (𝑥) =
0, for all other values of 𝑥.
Deduce E(𝑋 𝑛 ).

1.19. A random variable has a normal distribution 𝑁 (𝝁, 𝝈 2 ). Find its moment generating
function.

1.20. Find the probability generating functions of the following distributions, in which 0 <
𝑝 < 1:
(a) Bernoulli distribution: 𝑝𝑛 = 𝑝𝑛 (1 − 𝑝)𝑛 , (𝑛 = 0, 1);
(b) geometric distribution: 𝑝𝑛 = 𝑝(1 − 𝑝)𝑛−1 , (𝑛 = 1, 2, . . .);
(c) negative binomial distribution with parameter 𝑟 expressed in the form:
( )
𝑟+𝑛−1 𝑟
𝑝𝑛 = 𝑝 (1 − 𝑝)𝑛 , (𝑛 = 0, 1, 2, . . .)
𝑟−1
where 𝑟 is a positive integer. In each case find also the mean and variance of the distribution
using the probability generating function.

1.21. A word of five letters is transmitted by code to a receiver. The transmission signal is
PROBLEMS 27
weak, and there is a 5% probability that any letter is in error independently of the others. What
is the probability that the word is received correctly? The same word is transmitted a second
time with the same errors in the signal. If the same word is received, what is the probability
now that the word is correct?

1.22. A random variable 𝑁 over the positive integers has the probability distribution with
𝛼𝑛
𝑝𝑛 = P(𝑁 = 𝑛) = − , (0 < 𝛼 < 1; 𝑛 = 1, 2, 3, . . .).
𝑛 ln(1 − 𝛼)
What is its probability generating function? Find the mean of the random variable.

1.23. The source of a beam of light is a perpendicular distance 𝑑 from a wall of length 2𝑎,
with the perpendicular from the source meeting the wall at its midpoint. The source emits
a pulse of light randomly in a direction 𝜃, the angle between the direction of the pulse and
the perpendicular, chosen uniformly in the range − tan−1 (𝑎/𝑑) ≤ 𝜃 ≤ tan−1 (𝑎/𝑑). Find
the probability distribution of 𝑥 (−𝑎 ≤ 𝑥 ≤ 𝑎), where the pulses hit the wall. Show that its
density function is given by
𝑑
𝑓 (𝑥) =
2(𝑥2 + 𝑑2 ) tan−1 (𝑎/𝑑)
(this the density function of a Cauchy distribution). If 𝑎 → ∞, what can you say about the
mean of this distribution?

1.24. Suppose that the random variable 𝑋 can take the integer values 0, 1, 2, . . .. Let 𝑝𝑗 and
𝑞𝑗 be the probabilities
𝑝𝑗 = P(𝑋 = 𝑗), 𝑞𝑗 = P(𝑋 > 𝑗), (𝑗 = 0, 1, 2, . . .).
Show that, if
∞ ∞
∑ ∑
𝐺(𝑠) = 𝑝 𝑗 𝑠𝑗 , 𝐻(𝑠) = 𝑞𝑗 𝑠𝑗 ,
𝑗=0 𝑗=0

then (1 − 𝑠)𝐻(𝑠) = 1 − 𝐺(𝑠).

Show also that E(𝑋) = 𝐻(1).
 CHAPTER 2

Some Gambling Problems

2.1 Gambler’s ruin

Consider a game of chance between two players 𝐴, the gambler and 𝐵, the opponent.
It is assumed that at each play, 𝐴 either wins one unit from 𝐵 with probability 𝑝
or loses one unit to 𝐵 with probability 𝑞 = 1 − 𝑝. Similarly, 𝐵 either wins from
𝐴 or loses to 𝐴 with probabilities 𝑞 or 𝑝. The result of every play of the game is
independent of the results of previous plays. The gambler 𝐴 and the opponent 𝐵
each start with a given number of units and the game ends when either player has
lost his or her initial stake. What is the probability that the gambler loses all his or
her money or wins all the opponent’s money assuming that an unlimited number of
plays are possible? This is the classic gambler’s ruin problem. In a simple example
of gambler’s ruin, each play could depend on the spin of a fair coin in which case
𝑝 = 𝑞 = 21 .
The problem will be solved by using results from conditional probability which
then leads to a difference equation and we shall have more to say about methods of
solution later. There are other questions associated with this problem, such as how
many plays are expected before the game finishes. In some games the player might
be playing against a casino which has a very large (effectively infinite) initial stake.

2.2 Probability of ruin

The result of each play of the game is a (modified) Bernoulli random variable (Sec-
tion 1.7) which can only take the values −1 and +1. After a series of plays, we are
interested in the current capital or stake of 𝐴, the gambler. This is simply the initial
capital of 𝐴 plus the sum of the values of the Bernoulli random variables generated
by these plays.We are also interested in how the random variable which represents
the current capital changes or evolves with the number of plays. This is measured at
discrete points when the result of each play is known.
Suppose that 𝐴 has an initial capital of 𝑘 units and 𝐵 starts with 𝑎 − 𝑘, where 𝑎
and 𝑘 are positive integers and 𝑎 > 𝑘. If 𝑋𝑛 is a random variable representing 𝐴’s
stake after 𝑛 plays (or at time point 𝑛), then 𝑋0 = 𝑘. If 𝑋𝑛 = 0, then the gambler 𝐴
is ruined (note that we must have 𝑛 ≥ 𝑘), whilst if 𝑋𝑛 = 𝑎 (𝑛 ≥ 𝑎 − 𝑘) then 𝐵 is
ruined, and in both cases the game terminates. Our initial objective is the derivation
of P(𝑋𝑛 = 0) for all 𝑛 ≥ 𝑘.
The sequence of random variables 𝑋0 , 𝑋1 , 𝑋2 . . . represents what is known as a

29
Exploring the Variety of Random
Documents with Different Content
eyes gone absent-minded while he rapidly conversed with the person at the
other end of the telephone.
‘Come on,’ said Mr. Thorpe, laying hold of Jocelyn’s arm.
He took him away to the hotel. The hotel was the Carlton. ‘Know me at
the Carlton,’ said Mr. Thorpe, who in the first year of his widowerhood,
before he felt justified in beginning to court Mrs. Luke, had sometimes
consoled himself with the cooking of the Carlton. And thus it was that Mrs.
Luke presently found herself too at the Carlton, for Jocelyn, who no more
than Mr. Thorpe would leave the neighbourhood of Scotland Yard, was
concerned for his mother, left alone at Almond Tree Cottage. So Mr. Thorpe
sent the car back for her, and also for the necessary luggage. He couldn’t
quite see himself appearing next morning at the Carlton in the dinner-jacket
he put on every night at Abergeldie because of the butler.

§
She arrived at one in the morning. Mr. Thorpe by that time had taken
three bedrooms, and a sitting-room.
‘I can’t pay,’ said the unhappy Jocelyn on seeing these arrangements.
‘But I can,’ said Mr. Thorpe.
‘I don’t know why——’ began Jocelyn, shrinking under the
accumulating weight of obligations.
‘But I do,’ said Mr. Thorpe, cutting him short.
Mrs. Luke never forgot that pink sitting-room at the Carlton, for it was
there that Jocelyn, walking up and down it practically demented, cast
himself adrift from her for ever. And yet what had she done but try to help
him? What had she ever done all his life but love him, and try to help him?
‘There’s been too much of that—there’s been too much of that,’ Jocelyn
raved, when she attempted, faintly, for she was exhausted, to defend herself.
She soon gave up. She soon said nothing more at all, but sat crying
softly, the tears dropping unnoticed on her folded hands.
Before this, however, while the car was fetching her from South Winch,
Mr. Thorpe, bracing himself to his plain and unshirkable duty, invited
Jocelyn into the sitting-room he had engaged, and ordered whiskies and
sodas. These he drank by himself, while Jocelyn, his head sunk on his chest,
sat stretched full length in a low chair staring at nothing; and having drunk
the whiskies, Mr. Thorpe felt able to perform his duty.
Which he did; and in a series of brief sentences described the girl’s state
of mind when he accidentally found her down by his fence, and how it was
the idea of being left alone with Jocelyn’s mother till the summer that she
couldn’t stand, because she simply couldn’t stand his mother. Frightened of
her. Scared stiff. Just simply couldn’t stand her.
At this Jocelyn, roused from his stupor, looked round at Mr. Thorpe with
heavy-eyed amazement.
‘Couldn’t stand my mother?’ he said in tones of wonder, his mouth
remaining open, so much was he surprised.
‘That’s the ticket,’ said Mr. Thorpe; and drank more whiskey.
He then, after explaining that he wasn’t an orator, told Jocelyn in a
further series of brief sentences that it was unnatural for wives to live with
their mothers-in-law instead of with their husbands, that his wife knew and
felt this, and that she was, besides, having been brought up on the Bible and
being otherwise ignorant of life, genuinely and deeply shocked at what she
regarded as his disobedience to God’s laws.
‘But my mother,’ said Jocelyn, ‘has been nothing but——’
‘Sees red about your mother, that girl does,’ interrupted Mr. Thorpe.
‘But why?’ said Jocelyn, sitting up straight now, his brows knitted in the
most painful bewilderment.
‘Don’t ask me,’ said Mr. Thorpe; and drank more whiskey.
He then told Jocelyn, in a third and last series of brief sentences, for after
that not only had he said his say but the young man didn’t seem able to
stand any more, that if—no, when—his wife was restored to him, he had
better see to it that his mother was as far off and as permanently off as
possible; and then, Jocelyn by this time looking the very image of
wretchedness, he gave him, poor young devil, the bit of comfort of telling
him that his wife had only meant to leave him till she knew he was in
Cambridge, and that then she had been going to join him there, and live in
some rooms somewhere near him. It wasn’t him she was running from, it
was his mother.
‘All that girl asked,’ said Mr. Thorpe, bringing his fist, weighty now with
whiskey, down shatteringly on the table, ‘was a couple of rooms, and you
sometimes in them. A girl in a thousand. If she’d been as ugly as sin she’d
still have been a treasure to any man. But look at her—look at her, I say.’
‘Oh, damn you!’ shouted Jocelyn, springing to his feet, unable to bear
any more, ‘Damn you—damn you! How dare you, how dare you, when it’s
you—you——’
And he came towards Mr. Thorpe, his arms lifted as if to strike him; but
he suddenly dropped them to his sides, and turning away gripped hold of
the chimneypiece, and, laying his head on his hands, sobbed.

§
Charles Moulsford, then, was right, and the Lukes suffered. So did Mr.
Thorpe, for it was all his fault really. He was amazed at the ease and
swiftness with which he had slipped away from being evidently and
positively a decent man into being equally evidently and positively an evil-
doer. That he had done evil, and perhaps irreparable evil, was plain. Yet its
beginning was after all quite small. He had only helped the girl to go to her
father. Such an act hadn’t deserved this tremendous punishment. Mr.
Thorpe couldn’t help feeling that fate was behaving unfairly by him. If all
his impulses and indiscretions throughout his life had been punished like
this, where would he have been by now?
But that was neither here nor there. This terrible thing had happened, and
it was his fault. Without him she couldn’t have budged; and, weighed down
by his direct responsibility, when Jocelyn advanced on him with his fists
uplifted ready to strike him he rather hoped he would actually do it, and
when instead the poor devil broke down and began to cry, Mr. Thorpe was
very unhappy indeed. Perhaps he hadn’t been quite tactful in the things he
had said to him. Perhaps he had been clumsy. Whiskey was tricky stuff. He
had only meant——
Then Margery arrived, with her white face and great, scared eyes, and
found her son standing there holding on to the chimneypiece and crying,
and—well, Mr. Thorpe felt he had overdone the getting even business
altogether, and discovered with a shock that he could no longer regard
himself as a decent man.
He went away to his bedroom, leaving them alone. He didn’t know what
they were saying to each other, but he could hear that Jocelyn seemed to be
talking a good deal. Couldn’t stop, the poor devil couldn’t; went on and on.
 Mr. Thorpe sat down to think out plans, the ceaseless sound of that voice
in his ears. It was he who had lost the girl, and it was he who was going to
find her. If Scotland Yard found her first so much the better, but he wasn’t
going to sit still till they did, he was going off on his own account next
morning. He’d begin by sending Margery home, who was doing no good
here, he could tell by the sounds coming through the door, pack Jocelyn,
who was doing no good here either raving like that, off to Cambridge
because of the remote chance that the girl was going to be able after all to
do what she said and join him there, and he himself would meanwhile make
a bee-line for her father.
Pinner was the man. Pinner was the point to start from. Pinner and
Woodles. She had said his name was Pinner, and that he lived at Woodles.
Woodles? Funny sort of name that, thought Mr. Thorpe, trying to cheer
himself up by being amused at it. The sounds coming through the door
weren’t very cheering. Raving, the poor young devil was,—raving at his
mother. Mr. Thorpe feared he had perhaps been quite beastly tactless, telling
him of Sally’s not being able to stand his mother. He felt very
uncomfortable about it, sitting there with those sounds in his ears. And
meanwhile the night was slipping along, and where was that girl?
There were so many possible answers to this question, and all of them so
very unpleasant, that Mr. Thorpe couldn’t, he found, sit quiet in his chair.
Three o’clock. Fourteen hours now since last she was seen....
He got up and walked about. In the next room he could hear Jocelyn
doing the same thing. No—dash it all, thought Mr. Thorpe after listening for
some time to the ceaseless voice, he couldn’t be allowed to go on at his
mother like that. He’d had close on a couple of hours of it. All very well
being heartbroken, all very well being out of one’s senses, but he couldn’t
be allowed——
Mr. Thorpe opened the door and went in. There was Jocelyn, striding
about the room, up and down, round and round, enough to make one giddy
just to see him, his words pouring out, his face convulsed, and there sitting
looking at him, not saying a word, with tears rolling down her face, was his
mother.
No—damn it all—there were limits——
‘Better shut up now, eh?’ said Mr. Thorpe firmly to the demented young
man. ‘Said all there’s to say long ago, I bet. Won’t help, you know—this
sort of thing.’
‘I’m telling my mother—I’m making it clear to her once and for all,’
raved Jocelyn, who indeed no longer had the least control of himself, ‘that
if I ever find Sally never again as long as I live shall she come between us,
never shall she set foot——’
‘Oh, shut up. We know all that, don’t we, Margery. Who’s going to come
between you, you silly young ass? Look here—no good crying, you know,’
said Mr. Thorpe, going to Mrs. Luke and putting his arm round her. It
seemed natural. For two pins he would have kissed her. Habit. Can’t get
away from habits.
But Mrs. Luke didn’t appear to know he was there. Her eyes, from which
the tears dropped slowly and unnoticed, were fixed only on Jocelyn.
‘He’s so tired—so tired,’ she kept on whispering to herself. ‘Oh, my
darling—you’re so tired.’

§
It was Mr. Pinner’s turn next day to have a bad time, and he had it. He
had a most miserable day, from noon on, when the same car that had
brought Sally drew up in front of his shop, and a stout elderly gentleman
with a red face and a bristly moustache got out, and came and spent half an
hour with him.
What a half hour that was; but all of a piece with the life he seemed now
to be living. The day before there had been first Sally, and then Mr. Luke,
and now there was this gentleman. Mr. Luke had soon been pacified, and
only wanted to be getting home again, but the stout gentleman came in and
sat down square to it, and at the end of half an hour Mr. Pinner felt as if he
had been turned inside out, and wouldn’t ever be able to look himself in the
face again.
For Sally hadn’t gone home, and it was his fault that she hadn’t. These
were the facts; the gentleman said so. Terrible, terrible, thought Mr. Pinner,
shrinking further than ever into his trousers. The first fact was terrible
enough, but the second seemed even worse to Mr. Pinner. Responsibility,
again—and he who had supposed when he got Sally safely married that he
had done with it for good and all!
 At first he had tried to make a stand and hold up his head, and had said
politely—nothing lost by manners,—‘Excuse me, sir, but are you by any
chance the gentleman my daughter mentioned to me as ’er father-in-law?’
And when the gentleman, after a minute, said he was, Mr. Pinner told him
that in that case it was he who was responsible for her loss, for it was he
who had lent her the car in which she had left her husband.
Wasn’t this true? Anybody would have thought so; but before Mr. Pinner
could say knife the boot had been put on the other leg, and he found that it
was his fault and his only that she was lost, because he hadn’t, as the
gentleman said was his plain duty, taken her back himself to the very door.
Mr. Pinner, constitutionally unable not to feel guilty if anybody told him
loud enough that he was, at once saw the truth of this. Terrible. Awful.
Fancy. Yes, indeed—a daughter like that. Yes, indeed—any daughter, but a
daughter like that, a daughter in a million. No, indeed—he didn’t know how
he came not to do such a thing——
And the more Mr. Thorpe cross-examined him about the details of that
seeing-off at the station, the more did Mr. Pinner’s conduct appear criminal;
for, under Mr. Thorpe’s searching questions, Mr. Pinner somehow began to
be sure the lady in the carriage hadn’t been a lady at all, but something quite
different, something terrible and wicked, who had carried Sally off into the
sort of place one doesn’t mention. He remembered her black eyes, and how
they rolled——
‘Rolled, eh?’ said Mr. Thorpe, who was snatching at Mr. Pinner’s words
almost before they appeared, trembling, on the edge of his mouth.
Yes—rolled. And bold-looking, she was too,—bold-looking, and pat as
you please at answering. Not Mr. Pinner’s idea at all of a modest woman.
Yes, and the compartment smelt of scent, now he came to think of it—yes,
he dared say it was cheap scent. And powdered, her face was—he had
remarked on it to himself, after the train had gone.
Thus did Mr. Thorpe’s own fears get by cross-examination into Mr.
Pinner’s mind, and by the end of the half hour Mr. Pinner was as much
convinced as Mr. Thorpe that Sally had fallen into the hands of somebody
of whom Mr. Thorpe used an expression that Mr. Pinner wouldn’t have
soiled his lips with for any sum one cared to mention. And then, after
swearing at him, and asking him what sort of a father he thought he was,
and Mr. Pinner, who by this time was wishing with all his heart that he
wasn’t a father at all, tremblingly begging him not to blaspheme, Mr.
Thorpe went away.
‘What ’ad I better do now, sir?’ Mr. Pinner asked, following him out on
to the steps in much distress, clinging to him in spite of his horrifying
language.
‘You? What can you do? You’ve done your damnedest——’
‘Sir, sir——’
And he got into his car, and Mr. Pinner heard him tell the chauffeur to
drive like the devil to London and go to Liverpool Street Station; and it
seemed as if in a flash the street were empty, and he alone.

§
That afternoon Mr. Pinner himself arrived at Liverpool Street Station—
an anxious little man in his Sunday clothes, his blue eyes staring with
anxiety. He couldn’t just stay in his shop, and as likely as not never hear
anything more, either one way or the other. He must do something. He must
ask questions. Nobody would tell him if Sally were found or not, if he
didn’t. She herself might some day perhaps drop him a line, but she wasn’t
much of a one for writing, and besides he had been harsh to her. ‘Don’t
believe you loves me,’ she had said, crying bitterly when he scolded her so
and wouldn’t let her stay with him. Love her? He loved her dearly. She was
all he had in the world. If anything had happened to that girl——
He timidly stopped a porter, and began to inquire. The porter, who was
busy, stared at him and hurried on. He then tried a guard, who said, ‘Eh?’
very loud, looked past him along the platform, waved a green flag, jumped
on to a train, and departed.
He then tried another porter; several porters; and at last, more timid than
ever by this time, approached a ticket-collector.
Nobody seemed to have time for Mr. Pinner. His trousers were against
him. So was his hat; so was everything he said and did. The ticket-collector,
who didn’t like shabbiness and meekness, ignored him. He knew perfectly
well who Mr. Pinner was talking about, for the whole station was invariably
aware of any of the Duke’s family passing through it, and everybody the
day before had seen Lady Laura and the young lady. Mr. Pinner hadn’t got
beyond his first words of description before the ticket-collector knew what
he was driving at, but he only looked down his long nose at the flushed
little man in the corkscrew trousers, and said nothing. Give a thing like that
information about her ladyship’s movements? Not much.
Yet this same ticket-collector, only an hour or two before, had been wax
in the gloved hands of Mr. Thorpe, and with these words had parted from
him:
‘Thank you, sir. Don’t mention it, sir. No trouble at all. Yes—a very
striking young lady indeed, sir. Her ladyship was going to Goring House for
a couple of days, so the chauffeur told me. Much obliged, sir. Yes, sir—
Lady Laura Moulsford. That’s right, sir—the Duke of Goring’s daughter.’
This same ticket-collector had said all that; and to Mr. Pinner he said not
a word. He merely down his long nose looked at him, and when the little
man explained that he was the fair young lady’s father he looked at him
more glassily than ever. So that presently for very shame Mr. Pinner
couldn’t go on standing there asking questions that got no answers, and
after lingering awhile uncertainly in the ticket-collector’s neighbourhood,
for something told him that this man could throw light on Sally’s
disappearance if he would, he went sorrowfully, but unresentfully, away.
Presently he found himself in South Winch. He seemed to have drifted
there, not knowing what to do or where to go next, and unable to bear the
thought of his lonely shop and of nobody’s letting him know about
anything. He had thought it fine and peaceful at first to be independent and
at last alone, but it didn’t seem so now. He missed his wife. Nobody now to
mind what he did, good or bad. Nobody.
In South Winch he sought out the grocer, so as to get Jocelyn’s address,
preferring him to the Post Office because the smell of currants and bacon
made him feel less lonely, and, having followed the directions the grocer
gave him, found the road and the house, and opened the white gate with
deferential trepidation. Timidly at the door he asked if he might say a word
to Mr. Luke, and the little maid, at once at ease with his sort of clothes,
inquired pleasantly if Mrs. Luke wouldn’t do just as well; better, suggested
the little maid, because she was there, and Mr. Jocelyn wasn’t. In fact she
offered Mrs. Luke to Mr. Pinner, she pressed her upon him,—a lady he
wouldn’t have dreamt of disturbing if left to himself.
So that Mr. Pinner, without apparently in the least wanting to, found
himself in a beautiful drawing-room, and there by the fire sat a lady, leaning
back on some cushions as though she were tired.
At first he thought she was asleep, and he was beginning to feel
extremely awkward when she turned her head and looked at him.
A pale lady. A very pale lady; with a face that seemed all eyes.
‘Beg pardon, mum,’ said Mr. Pinner, wishing he hadn’t come.
The lady went on looking at him. She didn’t move. Her hands were
hanging down over the arms of the chair as though she were tired. She just
turned her head, but didn’t move else.
‘It’s about Sally,’ said Mr. Pinner. ‘ ’Appened to be passin’, and thought
I’d——’
He stopped, for now he came to think of it he didn’t rightly know what
he had thought.
The lady leant forward in her chair. ‘Do you know where she is?’ she
asked quickly.
‘No, mum. Do you?’ asked Mr. Pinner.
‘No,’ said the lady in a queer sort of voice, her head drooping.
Mr. Pinner stood there very awkward indeed.
‘Are you her father?’ she asked, after a minute.
‘That’s right, mum,’ said Mr. Pinner.
Then she got up and came across to him.
‘I’m afraid you are very unhappy,’ she said, looking at him.
‘That’s right, mum,’ said Mr. Pinner.
She held out her hand, her eyes on his face.
He shook it respectfully, but without enthusiasm.
‘Why, you’re cold,’ she said.
‘That’s right, mum,’ said Mr. Pinner.
‘Won’t you come to the fire and get warm?’ she said; and before he had
time to consider what he ought to do next, Mr. Pinner found himself sitting
on the edge of the low chair the lady pushed up for him, warming his knees
and not saying anything.
The lady talked a little. She had some nice hot tea made for him, and
while he drank it talked a little, and said she was sure they would hear good
news soon, and he mustn’t worry, because she was sure....
 Then she fell silent too, and they sat there together looking into the fire;
and it was funny, thought Mr. Pinner, how just to sit there quietly, and know
she was sorry too about everything, seemed to make him feel better. A kind
lady; a good lady. What did Sally mean, saying he wouldn’t be able to stand
her either, if he knew her? The only thing wrong with her that Mr. Pinner
could see, was that she looked so ill. Half dead, thought Mr. Pinner.
And after being with her he had more courage to go back to the lonely
shop, and she promised faithfully to let him know the minute there was any
news, and again told him not to worry and everything would come all right,
and he went away comforted.
And she, watching him as he trotted off down to the gate, felt somehow
comforted too; not quite so lonely; not quite so lost.

§
Meanwhile Mr. Thorpe, having lunched and tidied and generally
freshened himself up, was on the steps of Goring House, asking for Lady
Laura Moulsford.
‘Her ladyship is hout,’ said the footman haughtily, for he knew at once
when Mr. Thorpe added the word Moulsford that he was what the footman
called not one of Our Lot. No good his having a car waiting there, and a fur
coat, and suède gloves; he simply wasn’t one of Our Lot. And the footman,
his head thrown back, looked at Mr. Thorpe very much as the ticket-
collector was at that moment looking at Mr. Pinner.
‘Out, eh?’ said Mr. Thorpe. ‘When will she be in?’
‘Her ladyship didn’t say,’ said the footman, his head well back.
‘You’ve got a young lady here of the name of Luke. She in?’
‘Mrs. Luke is hout,’ said the footman, beginning to shut the door.
‘Is anybody in?’ asked Mr. Thorpe, getting angry.
‘The family is hout,’ said the footman; and was going to shut the door
quite when Mr. Thorpe went close up to him and damned him. And because
Mr. Thorpe’s temper was quick and hot he damned him thoroughly, and the
footman, as he heard the familiar words, strongly reminiscent not only of
Lord Streatley but also of the different sergeants he had had during the war,
who, however unlike each other to look at, were identical to listen to,
thought he must be one of Lady Laura’s friends after all, and began to open
the door again; and Mr. Thorpe advancing, damning as he went and saying
things about flunkeys that were new to the footman, entered that marble hall
which had struck such a chill into Sally’s unaspiring soul.
The butler appeared. The butler was suave where the footman had been
haughty. He had heard some of the things Mr. Thorpe was saying as he
hurried from his private sitting-room into the echoing hall, and had no
doubt that he was a friend of the family’s.
Lady Laura had been in to lunch, but had gone out again; Mrs. Luke was
motoring with Lord Charles—who the devil was he, Mr. Thorpe wondered
—down to Crippenham, where she was going to stay the night. Her
ladyship had had a telegram from his lordship to that effect, and she herself
was going down the following morning.
‘Where’s Crippenham?’ asked Mr. Thorpe.
The butler was surprised. Up to that moment he had taken Mr. Thorpe
for a friend, if an infrequent one, of Lady Laura’s.
‘His Grace’s Cambridgeshire seat,’ he said, in his turn with hauteur. ‘His
Grace is at present in residence.’
‘Crikey!’ thought Mr. Thorpe. ‘Got right in with the Duke himself, has
she?’ And he felt fonder of Sally than ever.

§
At this point Mr. Thorpe, who had been behaving so well, began to
behave less well. The minute the pressure of anxiety was relaxed, the
minute, that is, that he no longer suffered, he became callous to the
sufferings of the Lukes; and instead of at once letting them know what he
had discovered he kept it to himself, he hugged his secret, and deferred
sending till some hours later a telegram to each of them saying, ‘Hot on her
tracks.’
Quite enough, thought Mr. Thorpe, as jolly again as a sand-boy, and
immediately unable to imagine the world other than populated by sand-boys
equally jolly,—quite enough that would be to go on with, quite enough to
make them both feel better. If he told them more, they’d get rushing off to
Crippenham and disturbing the Duke’s house-party. The whole thing should
now be allowed to simmer, said Mr. Thorpe to himself. Sally should be
given a fair field with her duke, and not have relations coming barging in
and interrupting.
But what a girl, thought Mr. Thorpe, slapping his knee—he was in his
car, on the way to his club—what a girl. She only had to meet dukes for
them to go down like ninepins at her feet. Apart from her beauty, what
spirit, what daring, what initiative, what resource! It had been worth all the
anxiety, this magnificent dénouement. Safe, and sounder than ever. A
glorious girl; and he too had at once seen how glorious she was, and at
once, like the Duke, fallen at her feet. That girl, thought Mr. Thorpe, who
began to believe she would rise triumphant even over a handicap like
Jocelyn, might do anything, might do any mortal thing,—no end at all, there
wasn’t, to what that girl couldn’t do. And, glowing, he telephoned to
Scotland Yard, and later on, after having had his tea and played a rubber of
bridge, sent his telegrams.
Then he went quietly home. Things should simmer. Things must now be
left to themselves a little. He went quietly home to Abergeldie, and didn’t
let Mrs. Luke know he was there. Her feelings, he considered, were
sufficiently relieved for the present by his telegram; things must now be
allowed to simmer. And he took a little walk in his shrubbery, and then had
a hot bath, and dressed, and dined, ordering up a pint of the 1911 Cordon
Rouge, and sat down afterwards with a great sigh of satisfaction by his
library fire.
He smoked, and he thought; and the only thing he regretted in the whole
business was the rude name he had called Lady Laura Moulsford to that
fool Pinner. But, long as he smoked and thought, it never occurred to him to
resent, or even to criticise, the conduct of the Moulsford family. Strange as
it may seem, considering that family’s black behaviour, Mr. Thorpe dwelt
on it in his mind with nothing but complacency.
 XV

§
At Crippenham next morning it was very fine. London and South Winch
were in a mist, but the sun shone brightly in Cambridgeshire, and the Duke
woke up with a curiously youthful feeling of eagerness to get up quickly
and go downstairs. He knew he couldn’t do anything quickly, but the odd
thing was that for years and years he hadn’t wanted to, and that now
suddenly he did want to; and just to want to was both pleasant and
remarkable.
He had been thinking in the night,—or, rather, Charles’s thoughts, placed
so insistently before him, had sunk in and become indistinguishable from
his own; and he had thought so much that he hadn’t gone to sleep till nearly
five. But then he slept soundly, and woke up to find his room flooded with
sunshine, and to feel this curiously agreeable eagerness to be up and doing.
The evening before, when Charles came in from the garden and packed
his bewitching guest off to bed, he had been very cross, and had listened
peevishly to all his son was explaining and pointing out; not because he
wasn’t interested, or because he resented the suggestions being made, but
simply because the moment that girl left the room it was as if the light had
gone out,—the light, and the fire. She needn’t have obeyed Charles. Why
should she obey Charles? She might have stayed with him a little longer,
warming him by the sight of her beauty and her youth. The instant she went
he felt old and cold; back again in the condition he was in before she
arrived, dropped back again into age and listlessness, and, however stoutly
he pretended it wasn’t so, into a deathly chill.
Now that, thought the Duke, himself surprised at the difference his
guest’s not being in the room made, was what had happened to David too
towards the end. They didn’t read it in the Lessons in church on Sundays,
but he nevertheless quite well remembered, from his private inquisitive
study of the Bible in his boyhood, how they covered David when he was
old with clothes but he got no heat, and only a young person called the
Shunammite was able, by her near presence, to warm him. The Duke didn’t
ask such nearness as had been the Shunammite’s to David, for he, perhaps
because he was less old, found all he needed of renewed life by merely
looking at Sally; but he did, remembering David while Charles talked, feel
aggrieved that so little as this, so little as merely wishing to look at her,
should be taken from him, and she sent to bed at ten o’clock.
So he was cross, and pretended not to understand, and anyhow not to be
interested. But he had understood very well, and in the watches of the night
had come to his decision. At his age it wouldn’t do to be too long coming to
decisions; if he wished to secure the beautiful young creature—Charles said
help, but does not helping, by means of the resultant obligations, also
secure?—he must be quick.
He rang for his servant half an hour before the usual time. He wanted to
get up, to go to her again, to look at her, to sit near her and have her
fragrant, lovely youth flowing round him. The mere thought of Sally made
him feel happier and more awake than he had felt for years. Better than the
fortnight’s cure of silence and diet at Crippenham was one look at Sally,
one minute spent with Sally. And she was so kind and intelligent, as well as
so beautiful—listening to every word he said with the most obvious
interest, and not once fidgeting or getting sleepy, as people nowadays
seemed to have got into the habit of doing. It was like sitting in the sun to
be with her; like sitting in the sun on a warm spring morning, and freshness
everywhere, and flowers, and hope.
Naturally, having found this draught of new life the Duke wasn’t going
to let it go. On the contrary, it was his firm intention, with all the strength
and obstinacy still in him, to stick to Sally. How fortunate that she was poor,
and he could be the one to help her. For she, owing all her happiness to him,
couldn’t but let him often be with her. Charles had said it would be both
new and desirable to do something in one’s life for nothing; but the Duke
doubted if it were ever possible, however much one wished to, to do
anything for nothing. In the case of Sally it was manifestly impossible.
Whatever he did, whatever he gave, he would be getting far more back; for
she by her friendship, and perhaps affection, and anyhow by her presence,
would be giving him life.
‘Come out into the garden, my dear,’ he said, when he had been safely
helped downstairs—the stairs were each time an adventure—putting his
shaking hand through her arm. ‘I want to see your hair in the sun, while I
talk to you.’
 And leading him carefully out, Sally thought, ‘Poor old gentleman,’ and
minded nothing at all that he said. Her hair, her eyes, all that Oh my ain’t
you beautiful business, of which she was otherwise both sick and afraid,
didn’t matter in him she called the Jewk. He was just a poor old gentleman,
an ancient and practically helpless baby, towards whom she felt like a
compassionate mother; and when he said, sitting in the sunny sheltered seat
she had lowered him on to and taking her hand and looking at her with his
watery old eyes, that he was going to give her Crippenham, and that the
only condition he made was that he might come and do a rest-cure there
rather often, she smiled and nodded as sweetly and kindly as she smiled and
nodded at everything else he said.
Like the croonings of a baby were the utterances of the Duke in Sally’s
ears; no more meaning in them, no more weight to be attached to them, than
that. Give her Crippenham? Poor old gentleman. Didn’t know what he was
talking about any more, poor old dear. She humoured him; she patted his
arm; and she wished to goodness Laura would be quick and come and take
her to her husband.
Sally now longed to get to Jocelyn as much as if she had passionately
loved him. He was her husband. He was the father of the little baby. Her
place was with him. She had had enough of this fleshpot business. She was
homesick for the things she knew,—plain things, simple things, duties she
understood. Kind, yes; kind as kind, the picks were, and they meant well;
but she had had enough. It wasn’t right it wasn’t, at least it wasn’t right for
her, to live so fat. What would her father have said if he had seen her in the
night in Laura’s bedroom, among all that lot of silver bottles and brushes
and laces and silks, and herself in a thin silk nightgown the colour of skin,
making her look stark naked? What would he have said if he had seen her
having her breakfast up there as though she were ill,—and such a breakfast,
too! Fleshpots, he’d have said; fleshpots. And he would have said, Sally,
strong if inaccurate in her Bible, was sure, that she had sold her husband for
a mess of fleshpots.
This was no life for her, this was no place for her, she thought, her head
bowed and the sun playing at games of miracles with her hair while the
Duke talked. She drew impatient patterns with the tip of her shoe on the
gravel. She hardly listened. Her ear was cocked for the first sounds of
Laura. She ached to have done with all this wasting of time, she ached to be
in her own home, getting on with her job of looking after her man and
preparing for her child. ‘Saturday today,’ she mused, such a lovely look
coming into her eyes that the Duke, watching her, was sure it was his
proposed gift making her divinely happy. ‘We’d be ’avin’ shepherd’s pie for
dinner—or p’raps a nice little bit of fish....’
And, coming out of that pleasant dream with a sigh, she thought,
‘Oughtn’t never to ’ave met none of these ’ere. All comes of runnin’ away
from dooty.’
Apologetically she turned her head and looked at the Duke, for she had
forgotten him for a moment, besides having been thinking on lines that
were hardly grateful. Poor old gentleman—still keeping on about giving her
Crippenham. Crippenham? She’d as soon have the cleaning of Buckingham
Palace while she was about it as of that great, frightening house—or, come
to that, of a prison.
But how like a bad dream it was, being kept there with the morning
slipping past, and she unable to reach him across the gulf of his deafness.
By eleven o’clock she was quite pale with unhappiness, she could hardly
bear it any longer. Would she have to give manners the go-by and take to
her heels once more? This time, though, there would be no kind father-in-
law to lend her a car; this time she would have to walk,—walk all the way,
and then when she got there find Jocelyn unaided. And the old gentleman
kept on and on about Crippenham being hers, and everything in it....
’E’s nothin’ but a nimage,’ she said to herself in despair. ‘Sits ’ere like a
old idol. Wot do ’e know about a married woman’s dooties?’
‘Where’s Charles?’ asked the Duke.
Sally shook her head. She hadn’t seen a sign of him that morning.
‘I want him to get my solicitor down—no time to lose,’ said the Duke.
‘You’re to have the place lock, stock and barrel, my dear, such as it is—
servants and all.’
Servants and all? Poor old gentleman. Why, she wouldn’t know which
end of a servant to start with. She with servants? And these ones here who,
however hard she tried up there in the bedroom, wouldn’t make friends.
They called her Madam. She Madam? Oh, my gracious, thought Sally,
shrinking in horror from such a dreadful picture.
‘It’s a hole of a place,’ went on the Duke, ‘and quite unworthy of you,
but we can have more bathrooms put in, and it’ll do till we find something
you like better. And Charles tells me you married rather suddenly, and
haven’t got anywhere to go to at present. He also says you have to live close
to Cambridge, because of your husband’s studies. And he also says, and I
entirely agree with him, my dear, that you oughtn’t to be in Cambridge
itself, but somewhere more secluded—somewhere where you won’t be seen
quite so much, somewhere hidden, in fact. Now I think, I really do think,
that Crippenham, in spite of all its disadvantages, does exactly fulfil these
requirements. And I want you to have it, my dear—to take it as my wedding
present to you, and to live in it very happily, and bless it and make it
beautiful by your presence.’
Thus the Duke.
‘ ’E don’t ’alf talk,’ thought Sally, quivering to be gone.

§
Charles, on being sent for by the Duke, was nowhere to be found. That
was because he was in South Winch. He had gone off at daybreak in his car,
and at the very moment his father woke up to the fact of his absence and
asked where he was, he was standing in the drawing-room at Almond Tree
Cottage, his eyes fixed eagerly on the door, waiting for Mrs. Luke.
He hadn’t been able to sleep for thinking of her. Somehow he had got it
into his head that she, more than her son, would suffer through Sally’s
disappearance, and be afraid. Because, thought Charles, she would feel that
it was from her the girl had run, and that any misfortune that might happen
to her would be, terribly, laid at her door. For two whole days and two
whole nights that unfortunate woman must have gone through torture. What
Charles couldn’t understand was why he hadn’t thought of this before.
Indeed his and Laura’s conduct had been utterly unpardonable. The least he
could now do, he thought, as he lay wide awake throughout the night, was
to get to South Winch without losing a minute, and put Mrs. Luke out of her
misery, and beg her forgiveness.
She was in the garden when he arrived. The little maid, staring at the
card he asked her to take to her mistress, said she would fetch her, and
ushered him into the drawing-room, where he waited with the books, the
bright cushions, the Tiepolo, and two withered tulips in a glass from which
nearly all the water had dried away; and while he waited he fought with a
feeling he considered most contemptible, in face of the facts, that he was
somehow on an errand of mercy, and arriving with healing in his wings,—
that he was somehow a benefactor.
Sternly he told himself he ought to feel nothing but shame; sternly he
tried to suppress his glow of misplaced self-satisfaction. There was nothing
good about him and Laura in this business. They had, the pair of them, been
criminally impulsive and selfish. He knew it; he acknowledged it. Yet here
he was, secretly glowing, his eyes watching the door, as much excited as if
he were going to bestow a most magnificently generous, unexpected
present.
Then it opened, and Mrs. Luke came in. He was sure it was Mrs. Luke,
for no one else could look so unhappy; and the glow utterly vanished, and
the feeling of shame and contrition became overwhelming.
‘She’s safe,’ said Charles quickly, eager to put a stop at once to the
expression in her eyes. ‘She’s at my father’s. She’s going to Cambridge
today to your son. She’s been with us the whole time——’
And he went to her, and took her hand and kissed it.
‘If it weren’t so ridiculous,’ he said, his face flushed with painful
contrition, still holding her hand and looking into her heavy, dark-ringed
eyes, ‘I’d very much like to go down on my knees to you, and beg your
pardon.’

§
And while Charles was in South Winch, Laura was in Cambridge,
dealing with Jocelyn. She, like Charles, had become conscious of the
sufferings of the Lukes, and, like him, was obsessed by them and lost in
astonishment that she hadn’t thought of them sooner; but for some obscure
reason, or instinct, her compunctions and her sympathies were for Jocelyn
rather than for his mother, and after a second sleepless night, during which
she was haunted by the image of the unfortunate young husband and greatly
tormented, she went down, much chastened, to Cambridge by the first
possible train, with only one desire now, to put him out of his misery and
beg his forgiveness.
So that Jocelyn, sitting doing nothing, his untouched breakfast still
littering the table, sitting bent forward in the basket-chair common to the
rooms of young men at Cambridge, his thin hands gripped so hard round his
knees that the knuckles showed white, his ears strained for the slightest
sound on the staircase, his eyes hollow from want of sleep, sitting as he had
sat all the previous afternoon after getting Mr. Thorpe’s telegram and most
of the night, sitting waiting, listening, and perhaps for the first time in his
life, for his mother had not included religious exercises in his early
education, doing something not unlike praying, did at last hear a woman’s
step crossing Austen’s Court, hesitating at what he felt sure was his corner,
then slowly coming up his staircase, and hesitating again at the first floor.
All the blood in his body seemed to rush to his head and throb there. His
heart thumped so loud that he could hardly hear the steps any more. He
struggled out of his low chair and stood listening, holding on to it to steady
himself. Would they come up higher? Yes—they were coming up. Yes—it
must be Sally. Sally—oh, oh, Sally!
He flew to the door, pulled it open, and saw—Laura.
‘It’s all right,’ she panted, for the stairs were steep and she was fat, ‘it is
—about Sally—don’t look so——’ she stopped to get her breath—‘so
dreadfully disappointed. She’s safe. If you’ll—oh, what stairs——’ she
pressed her hand to her heaving bosom—‘come with me, I’ll—take you—to
her——’
And having got to the top, she staggered past him into his room, and
dropped into the basket-chair, and for a minute or two did nothing but gasp.
But how difficult she found him. Jocelyn, whose reactions were always
violent, behaved very differently from the way his mother at that moment
was behaving, placed in the same situation of being asked forgiveness by a
Moulsford. Instead of forgiving, of being, as Laura had pictured, so much
delighted at the prospect of soon having Sally restored to him that he didn’t
mind anything, he appeared to mind very much, and quarrelled with her.
She, accustomed to have everything she did that was perhaps a little wrong
condoned and overlooked by all classes except her own, was astonished.
Here she was, doing a thing she had never done before, begging a young
man to forgive her, and he wouldn’t. On the contrary, he rated her. Rated
her! Her, Laura Moulsford. She knew that much is forgiven those above by
those below, and had frequently deplored the practice as one that has
sometimes held up progress, but now that the opposite was being done to
herself she didn’t like it at all.
 ‘Oh, what a nasty disposition you’ve got!’ she cried at last, when Jocelyn
had been telling her for ten impassioned minutes, leaning against the
chimney-piece and glowering down at her with eyes flashing with
indignation, what he thought of her. ‘I’m glad now, instead of sorry, for
what I did. At least Sally has had two days less of you.’
‘If you’re going to rag me as well——’ began Jocelyn, taking a quick
step forward as if to seize and shake this fat little incredibly officious
stranger,—so like him, his mother would have said, to waste time being
furious instead of at once making her take him to Sally.
But Laura, unacquainted with his ways, was astonished.
Then he pulled himself up. ‘It’s not you I’m cursing really at all,’ he
said. ‘It’s myself.’
‘Well, I don’t mind that,’ said Laura, smiling.
‘I’ve got the beastliest temper,’ said Jocelyn.
‘So I see,’ said Laura.
‘Do you think,’ he asked, for in spite of his anger he was all soft and
bruised underneath after his two days of fear, and when the fat stranger
smiled there was something very motherly about her, ‘I shall ever get over
it?’
‘Perhaps if you try—try hard.’
‘But—look here, I don’t care what you say—what business had you to
make away with my wife?’
‘Now you’re beginning all over again.’
‘Make away with my wife, smash up everything between me and my
mother——’
‘Oh, oh——’ interrupted Laura, stopping up her ears, and bowing her
head before the storm.

§
It was ten more minutes before she got him out of his rooms and into a
taxi.
‘We’ve lost twenty minutes,’ she said, looking at her watch. ‘You’ve lost
twenty kisses you might have had——’
 ‘For God’s sake don’t rag me!’ cried Jocelyn, gripping her by the arm
and bundling her into the taxi.
‘But what,’ asked Laura, who had tumbled in a heap on the seat, yet who
didn’t mind being thrown in because she knew she deserved worse than
that, ‘what else can one do with a creature like you?’
And she told him very seriously, as they heaved along towards
Crippenham, that the real mistake had been Sally’s marrying beneath her.
‘Beneath her?’ repeated Jocelyn, staring.
‘Isn’t it apparent?’ said Laura. ‘Angels should only marry other angels,
and not descend to entanglements with perfectly ordinary——’
‘No, I’m damned if I’m ordinary,’ thought Jocelyn. ‘And who the devil
is she, anyhow?’
‘Bad-tempered,’ continued Laura.
‘Yes, I’m beastly bad-tempered,’ he admitted.
‘Conceited——’
‘I swear I’m not conceited,’ he said.
‘Aren’t you?’ said Laura, turning her head and scrutinising him with
bright, mocking eyes.
And then, coming swift and silent as an arrow along the road towards
their taxi, she saw her father’s car.
‘Oh, stop!’ she cried, leaping to her feet and thrusting as much of herself
as would go through the window. ‘Here’s my father—yes, and Sally. Stop—
oh, stop!’ she cried, frantically waving her arms.

§
It had been decreed by Fate that Jocelyn should be reunited to Sally in
the middle of the road just beyond Waterbeach, at the point where the lane
to Lyddiatt’s Farm turns off; for such was the Duke’s desire to help his
lovely friend and such his infatuation, that he had actually broken his rule
of never emerging from Crippenham, once he got there, till the day
appointed for his departure, and was himself taking her to Ananias to hand
her over in person to her husband, afterwards lunching with the Master,—a
thing unheard of, this lunching, for the Duke disliked the Master’s politics
and the Master disliked the Duke’s, but what wouldn’t one do to further the
interests, by saying a good word for them, of the young couple?
This he had arranged that morning before coming downstairs, his
amazed servant telephoning the message and receiving the Master’s
hypocritical expressions of pleasure in return, for apart from the Duke’s
politics the Master was no fonder of a deaf guest than anybody else; and
just as Sally, on that garden seat, was coming to the end of her patience and
submissiveness and was seriously thinking of jumping up and taking to her
heels, the parlourmaid appeared on the path; and when she was quite close
she stood still, and opened her mouth very wide, and roared out that the car
was at the door; and the Duke, with a final pat of benediction, bade Sally
fetch her hat, and come with him to her husband.
So there it was that they met,—the taxi and the Rolls Royce, Laura and
Jocelyn, Sally and the Duke. And on the Swaffham Prior side of
Waterbeach, where the crooked signpost points to Lyddiatt’s Farm, the dull,
empty road was made radiant for a moment that day by happiness.
‘Stop! Stop!’ cried Laura, frantically waving.
‘Sally! Oh—oh, Sally!’ shouted Jocelyn, standing up too, and trying too,
behind Laura, to wave.
The chauffeur recognised Laura, and pulled up as soon as he could; the
taxi pulled up with a great grinding of its brakes; Jocelyn jumped out of one
door, and Laura of the other; and both ran.
‘Why,’ said Sally, who didn’t know what had happened, turning her head
and looking in astonishment at the two running figures coming along
behind, ‘why,’ she said, forgetting the Duke was deaf, ‘ ’ere is Mr. Luke
——’
And in another instant Jocelyn was there, up on the step of the car,
leaning over the side, dragging her to him with both arms, hugging her to
his heart, and kissing her as if there were no one in the world except
themselves.
‘Sally—oh, my darling! Oh, Sally—oh, oh, Sally!’ cried Jocelyn, raining
kisses on her between each word. ‘How could you—why did you—oh, yes
—I know, I know—I’ve been a beast to you—but I’m not going to be any
more—I swear, I swear——’
‘Now don’t, Mr Luke,’ Sally managed to say, stifled though she was,
‘don’t get swearin’ about it——’
 And pulling her head away from him she was able to attend to the
proprieties, and introduce him.
‘My ’usband,’ introduced Sally, looking over his arm, which was round
her neck, at the old man beside her. ‘The Jewk,’ she said, turning her face
back to Jocelyn, who took no notice of the introduction, who didn’t indeed
hear, because the moment she turned her face—oh, her divine, divine little
face!—back to him, he fell to kissing it again.
And Laura, coming panting up just then, got up on the step on the other
side of the car, and shouted in her father’s ear, who could always hear
everything she said, ‘This is Jocelyn Luke, Father—Sally’s husband.’
And the Duke said, ‘I thought it must be.’
 XVI

§
Now the end of this story, which is only the very beginning of Sally, the
merest introduction to her, for it isn’t to be supposed that nothing more
happened in her life,—the end of it is that she did as she was told about
Crippenham, and if the Duke had been less than ninety-three there would
have been a scandal.
But after ninety there is little scandal. The worst that was said of the
Lukes was that they had got hold of the old man, and nobody who saw
Sally believed that. Indeed, the instant anyone set eyes on her the Duke’s
behaviour was accounted for, and after five minutes in her company it
became crystal clear that she was incapable of getting hold of anybody. So
young, so shy, so acquiescent,—absurd to suppose she ever had such a thing
as an ulterior motive. And the husband, too; impossible to imagine that
silent scholar, also so young, and rather shy too, or else very sulky,—
impossible to imagine him plotting. On the contrary, he didn’t seem to like
what had happened to him much, and showed no signs whatever either of
pleasure or gratitude. But of Jocelyn no one thought long. He was without
interest for the great world. He was merely an obscure young man at
Cambridge, somebody the Duke’s amazing beauty had married.
Sally did, then, as she was told about Crippenham. It was given her, and
she took it; or rather, for her attitude was one of complete passivity, it
became hers. But she had an unsuspected simple tenacity of purpose, which
was later to develop disconcertingly, and she refused to live anywhere
except in the four-roomed cottage in the corner of the garden, built years
before as a playhouse for Laura and Charles.
On this one point she was like a rock; a polite rock, against which
persuasions, though received sweetly and amiably, should beat in vain. So
the Duke had the little house fitted up with every known labour-saving
appliance, none of which Sally would use because of having been brought
up to believe only in elbow-grease, and two bathrooms, one for her and one
for Jocelyn; and he attached such importance to these bathrooms, and he
insisted so obstinately on their being built, that Sally could only conclude
the picks must need a terrible lot of washing. Whited sepulchres they must
be, she secretly thought; looking as clean as clean outside, fit to eat one’s
dinner off if it came to that, but evidently nothing but show and take-in.
The Duke, much concerned at first, settled down to this determination of
Sally’s, and explained it to himself by remembering Marie-Antoinette. She
had her Trianon. She too had played, as Sally wished to play, at being
simple. He consoled himself by speaking of the cottage as Little Trianon; a
name Sally accepted with patience, though she told Jocelyn—who was so
much stunned at the strange turn his life had taken that she found she could
be quite chatty with him, and he never corrected, and never even said
anything back—she wouldn’t have thought of herself. Some day, the Duke
was sure, the marvellous child would grow up and get tired of her Trianon,
and then, when she wanted to move into the house, she should find
Versailles all ready for her, and very different from what it used to be.
So, on the excuse of seeing to the alterations, he was hardly ever away
from Crippenham, and if he had been less than ninety-three there would
certainly have been a scandal.
But Jocelyn, who woke up after the wild joy and relief of being reunited
to Sally to find himself the permanent guest of a duke, didn’t know whether
to be pleased or annoyed. The problems of his and Sally’s existence were
solved, it was true, but he wasn’t sure that he didn’t prefer the problems. He
rubbed his eyes. This was fantastic. It had no relation to real life, which was
the life of hard work and constant progress in his cloister at Ananias. Also,
its topsy-turviness bewildered him. Here was the Duke, convinced that
Sally had married beneath her, and so unshakably convinced that Jocelyn
had enormous difficulty in not beginning to believe it too. He couldn’t help
being impressed by the Duke. He had never met a duke before, never come
within miles of meeting one, and was impressed. That first afternoon, when
he had been carried off in the Rolls Royce to Crippenham, he had spent the
time between luncheon and tea shut up in the old man’s study being
upbraided for having taken advantage, as he was severely told, of Sally’s
youth and inexperience and motherlessness to persuade her into a marriage
which was obviously socially disastrous for her; and he couldn’t even if he
had wished to, which he certainly didn’t, tell him about Mr. Pinner, because
he couldn’t get through the barrier of his deafness. There the old man had
sat, with beetling brows and great stern voice, booming away at him hour
after hour, and there Jocelyn had sat, young, helpless, silent, his forehead
beaded with perspiration, listening to a description, among other things, of
the glories which would have been Sally’s if he hadn’t inveigled her into
marrying him. And so sure was the Duke of his facts, and so indignant, that
gradually Jocelyn began to think there was something in it, and every
moment felt more of a blackguard. In the old man’s eyes, he asked himself,
would there be much difference between him and Pinner? And was there, in
anybody’s eyes, much difference? More education; that was all. But of
family, in the Duke’s sense, he had as little as Pinner, and if Pinner had been
to a decent school, as Jocelyn had, and then gone to Cambridge—no,
Oxford for Pinner—he would probably have cut quite as good a figure, if
not in science then in something else; perhaps as a distinguished cleric.
He sat dumb and perspiring, feeling increasingly guilty; and if he could
have answered back he wouldn’t have, because the Duke made him feel
meek.
This meekness, however, didn’t last. It presently, after a period of
bewilderment, gave way to something very like resentment, which in its
turn developed into a growing conviction that he had become just a cat’s
paw,—he who, if left to himself, could have done almost anything.
Naturally he didn’t like this. But how, for the moment, could he help it?
Sally was going to have a baby. They had to live somewhere. It was really
heaven-sent, the whole thing. Yet—Sally, whom he had been going to
mould, was moulding him. Unconsciously; nothing to do with any intention
or desire of her own. And what she was moulding him into, thought
Jocelyn, as he drove himself backwards and forwards every day between
Crippenham and Cambridge, between his domestic life and his work,
between the strange mixture of emotions at the one end and the clear peace
and self-respect at the other, turning over in his mind with knitted brows, as
he drove, all that had happened to him in the brief weeks since he had
added Sally to his life—what she was moulding him into was a cat’s paw.
Yes. Just that.
Were all husbands cat’s paws?
Probably, thought Jocelyn.

§
 Mrs. Luke also reacted to the Moulsfords in terms of meekness. Hers,
however, lasted. She found them permanently dazzling. Besides, there was
nothing to be done. Jocelyn had gone; she had lost him for ever; he would
never come back, she very well knew, to the old life of dependence on her.
And if he must go, if she must lose him, there really was no one in the
world she would more willingly lose him to than the Duke of Goring. For
certainly it was a splendid, an exalted losing.
When she had had time to think after that visit from Lord Charles—he
had, she considered, a curious attractiveness—and was more herself again,
when she had recovered a little from the extreme misery she had gone
through and began not to feel quite so ill, she found it easy to forgive her
mauvais quart d’heure. The Moulsfords were heaping benefits on her boy.
They were settling all his difficulties. That morning when she was so
unhappy, Lord Charles had been most delightfully kind and sympathetic,
and had told her that the Duke, his father, intended to help the young
couple,—‘You know my son won this year’s Rutherford Prize,’ she had
said. ‘Indeed I do,’ he had answered in his charming, eager way, adding
how much interested his father was in the careers of brilliant young men,
especially at Cambridge, helping them in any way he could—and who
would not, in such circumstances, forgive?
Mrs. Luke forgave.
The fact, however, remained that she was now alone, and she couldn’t
think what her life was going to be without Jocelyn. For how, she
wondered, did one live without an object, with no raison d’être of any sort?
How did one live after one has left off being needed?
That year the spring was late and cold. The days dragged along, each one
emptier than the last. There was nothing in them at all; no reason, hardly,
why one should so much as get up every morning and dress for days like
that,—pithless, coreless, dead days. She tried to comfort herself by
remembering that at least she wasn’t any longer beaten down and
humiliated, that she could lift her head and look South Winch in the face,
and look it in the face more proudly than ever before; but even that seemed
to have lost its savour. Still, she mustn’t grumble. This happened to all
mothers sooner or later, this casting loose, this final separation, and to none,
she was sure, had it ever happened more magnificently. She mustn’t
grumble. She must be very thankful. She was very thankful. Like Toussaint
l’Ouverture—Wordsworth, again—she had, she said to herself, sitting
solitary through the chilly spring evenings by her fire after yet another
empty day, great allies; only fortunately of a different kind from poor
Toussaint’s, for however highly one might regard, theoretically, exultations
and agonies and love and man’s unconquerable mind, she, for her part,
preferred the Moulsfords.
But did she?
A bleak little doubt crept into her mind. As the weeks passed, the doubt
grew bleaker. Invisible Moulsfords; Moulsfords delightful and most
friendly when one met them, but whom one never did meet; Moulsfords full
of almost intimacies; Moulsfords who said they were coming to see one
again, and didn’t come; Moulsfords benignant, but somewhere else: were
these in the long run, except as subjects of carefully modest conversation in
South Winch—and South Winch, curiously, while it was plainly awe-struck
by what had happened to Jocelyn yet was also definitely less friendly than it
used to be—were these in the long run as life-giving, as satisfying, as
fundamentally filling as Toussaint’s exultations and agonies?
Ah, one had to feel; feel positively, feel acutely. Anything, anything, any
anger, any pain, any anxiety, any exasperation, anything at all that stabbed
one alive, was better than this awful numbness, this empty, deadly, settled,
stagnant, back-water calm....
And one evening, when it had been raining all day, after a period of
standing at the drawing-room window looking out at the dripping front
garden, where the almond-tree by the gate shivered in the grey twilight like
a frail, half-naked ghost, she turned and went to her writing-table, and sat
down and wrote a little note to Mr. Thorpe, and asked if he would not come
in after his dinner, and chat, and show that they could still be good friends
and neighbours; and when she had finished it, and signed herself Margery,
with no Luke, she rang for the little maid, and bade her take it round to
Abergeldie and bring back an answer.
‘For after all,’ she said to herself while she waited, standing by the fire
and slowly smoothing one cold hand with the other, ‘he has sterling
qualities.’
 THE END

Printed in Great Britain by R. & R. Clark, Limited, Edinburgh.

Typographical error corrected by the etext transcriber:
It it were a=> If it were a {pg 126}
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

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

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com