Introduction to Probability with Mathematica

Second Edition Kevin J. Hastings download

https://ebookgate.com/product/introduction-to-probability-with-
mathematica-second-edition-kevin-j-hastings/

Get Instant Ebook Downloads – Browse at https://ebookgate.com

 Get Your Digital Files Instantly: PDF, ePub, MOBI and More
Quick Digital Downloads: PDF, ePub, MOBI and Other Formats

Introduction to Probability Second Edition George G.

Roussas

https://ebookgate.com/product/introduction-to-probability-second-
edition-george-g-roussas/

An introduction to programming with Mathematica 3rd ed

Edition Paul R Wellin

https://ebookgate.com/product/an-introduction-to-programming-
with-mathematica-3rd-ed-edition-paul-r-wellin/

Introduction to probability and statistics Second

Edition, Revised And Expanded Edition Giri

https://ebookgate.com/product/introduction-to-probability-and-
statistics-second-edition-revised-and-expanded-edition-giri/

Introduction to Algebra Second Edition Peter J. Cameron

https://ebookgate.com/product/introduction-to-algebra-second-
edition-peter-j-cameron/
Introduction to Algebra Second Edition Peter J. Cameron

https://ebookgate.com/product/introduction-to-algebra-second-
edition-peter-j-cameron-2/

CRC Standard Curves and Surfaces with Mathematica

Second Edition David H. Von Seggern

https://ebookgate.com/product/crc-standard-curves-and-surfaces-
with-mathematica-second-edition-david-h-von-seggern/

Introduction to Probability 2nd Edition Dimitri P.

Bertsekas

https://ebookgate.com/product/introduction-to-probability-2nd-
edition-dimitri-p-bertsekas/

Introduction to Probability and Statistics Thirteenth

Edition William Mendenhall

https://ebookgate.com/product/introduction-to-probability-and-
statistics-thirteenth-edition-william-mendenhall/

Introduction to Probability Models 9th Edition Sheldon

M. Ross

https://ebookgate.com/product/introduction-to-probability-
models-9th-edition-sheldon-m-ross/
INTRODUCTION TO
Probability with
Mathematica®
Second Edition
 TEXTBOOKS in MATHEMATICS
Series Editor: Denny Gulick

PUBLISHED TITLES

ABSTRACT ALGEBRA: AN INTERACTIVE APPROACH

William Paulsen
COMPLEX VARIABLES: A PHYSICAL APPROACH WITH APPLICATIONS AND MATLAB®
Steven G. Krantz
ESSENTIALS OF TOPOLOGY WITH APPLICATIONS
Steven G. Krantz
INTRODUCTION TO ABSTRACT ALGEBRA
Jonathan D. H. Smith
INTRODUCTION TO MATHEMATICAL PROOFS: A TRANSITION
Charles E. Roberts, Jr.
INTRODUCTION TO PROBABILITY WITH MATHEMATICA®, SECOND EDITION
Kevin J. Hastings
LINEAR ALBEBRA: A FIRST COURSE WITH APPLICATIONS
Larry E. Knop
MATHEMATICAL AND EXPERIMENTAL MODELING OF PHYSICAL AND BIOLOGICAL PROCESSES
H. T. Banks and H. T. Tran
TEXTBOOKS in MATHEMATICS

INTRODUCTION TO
Probability with
Mathematica®
Second Edition

Kevin J. Hastings
Knox College
Galesburg, Illinois, U.S.A.
Chapman & Hall/CRC
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2010 by Taylor and Francis Group, LLC

Chapman & Hall/CRC is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed in the United States of America on acid-free paper

10 9 8 7 6 5 4 3 2 1

International Standard Book Number: 978-1-4200-7938-8 (Hardback)

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, transmit-
ted, 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.

Library of Congress Cataloging‑in‑Publication Data

Hastings, Kevin J., 1955-

Introduction to probability with Mathematica / Kevin J. Hastings. -- 2nd ed.
p. cm. -- (Textbooks in mathematics)
Includes bibliographical references and index.
ISBN 978-1-4200-7938-8 (hardcover : alk. paper)
1. Probabilities--Data processing--Textbooks. 2. Mathematica (Computer
file)--Textbooks. I. Title. II. Series.

QA273.19.E4H37 2010
519.20285’536--dc22 2009030737

Visit the Taylor & Francis Web site at

http://www.taylorandfrancis.com

and the CRC Press Web site at

http://www.crcpress.com
To my parents, for their wisdom and for helping me to understand the value of doing
your best
 Preface
When Introduction to Probability with Mathematica® was written in the late 1990s, I
had noticed what other probabilists and statisticians had also noticed. Students were having
difficulty understanding the main concepts of probability: event, sample space, random
variable, distribution, and expectation. This weakened their foundation for the study of
statistics and other subjects that relied on probability, such as operations research and
finance. In addition, real problem solving, with ill-specified problems, large disorganized
data sets, and important and intrinsically interesting applications were substantially missing
from mathematical statistics courses. Students would emerge from such courses with some
understanding of theory, a collection of techniques for preprogrammed situations, and not a
great deal of experience in exploring a problem from the model formulation stage completely
through to the conclusion. Good writing was also not especially emphasized. The second
edition of this book does not claim to fully address all of these problems, but its aim is to
make progress on several fronts, as the first edition did.
It is now rather firmly established that technology can be used as a tool to let
students visualize concepts, to make computations easily, and to encourage open-ended
exploration in realistic, non-contrived problems. Students experience an active learning style
in which they can experiment, make conjectures, and trace the consequences of those
conjectures. For an author who is cognizant of these ideas when writing, the end result is a
text that is involving, that develops the habits of going deeper, questioning assumptions, and
wrestling with ambiguities, and in the end leaves students with real understanding of
probabilistic modeling and the analysis of data, and the ability to put that understanding to
use.
One of the main pedagogical tools used in the first edition was simulation. I
recognized then how powerful a tool simulation could be to help students understand
randomness and sampling concepts better. Simulation also was used to gain insight into
probabilistic models that are not analytically easy to solve.
Simulation programs are easy to write in Mathematica and, in addition, with a little
extra support provided by the Mathematica package and notebooks that I have constructed
to accompany this text, students will be able to easily create their own simulations from
templates, as well as use Mathematica in other kinds of problem-solving situations. The
notebook environment, with its excellent symbol processing capability and smooth integra-
tion of graphics, numerics, and text, yields another valuable pedagogical opportunity: to
encourage students to do professional quality mathematical writing of which they can be
legitimately proud. The use of Mathematica provides a consistent platform for the study and
exposition of probability theory.
The time has come for a new edition of Introduction to Probability with
Mathematica® for a few reasons, one of which is very pragmatic. Starting with version 6.0,
Mathematica changed some important functions and syntax and, in addition provided a few
more features that can be exploited. Because of these changes, the Mathematica notebooks
into which the first edition of the book was divided no longer worked seamlessly, and in a
few cases notebooks didn't work at all. An update of the book is necessary to support current
and future users who update their version of the program. Version 7.0 was introduced just
as I was completing work on the second edition, and that also forced a few last-minute
changes.
 But also, years of experience have revealed important ways in which the presentation
could be modified to be even more effective in the new edition.

• Chapter 6 on applications really intends to be a chapter on stochastic

processes and their applications, and it should be billed that way and reinforced. I have
transferred some of the earlier material on Poisson processes to a new section in Chapter 6,
expanded the section on Markov chains to include a study of absorbing chains, and included
a new section on Brownian motion.
• The first edition did not touch in any significant way on order statistics,
which are important in several topic areas in mathematical statistics. A new section in
Chapter 4 has been added.
• Another area of mathematical statistics that was not addressed very well was
multivariate random variables, and transformations of multivariate normal random variables,
particularly with regard to quadratic forms. Such material sets up the study of regression and
analysis of variance in the second course in the usual sequence. So a new section on
transformations of multivariate normal random variables has been included into Chapter 4.
• To give more attention to elementary data analysis and to ground the theory
of the normal distribution in reality better, more example data and the graphical technique of
the normal quantile plot is covered in Section 4.1.
• More attention has been given to conditional expectation in Section 2.6, a
topic which has increased in importance with the rise to prominence of financial mathemat-
ics.
• There are new visualization and animation features in Mathematica version
6.0 and higher that will increase the interactive nature and visual appeal of some of the
existing examples and facilitate new and interesting animations. For example, when a new
probability density is introduced, a user-controlled manipulation showing parameter
dependence can be done.
• Throughout the book, particularly in Chapters 2 and 4 on discrete and
continuous distributions, I have added some well-chosen problems drawn from Actuarial
Exam P. This course does serve that audience, and these problems should give the students a
head start on preparing for that professional credential.
• A new appendix giving a basic introduction to Mathematica has been
included. The end-of-section lists of Mathematica commands used in the section have been
eliminated. They are unnecessary in the presence of the complete index to the commands in
the appendices.
• And finally, it is always a good idea to refresh examples and exercises, and
add a few more. In teaching from the book a number of times, I noticed that a few of the
existing examples, even the technological ones of which I was so fond, didn’t advance
learning especially well, and so those have been retooled or eliminated. New data sets have
been introduced, particularly in the section on the bivariate normal distribution.
A road map through the text is fairly straightforward. The instructor should cover all
of Chapter 1 on discrete probability. Those who prefer a less simulation-oriented approach
may skip lightly over Section 1.3 without major damage, but in my opinion this is the most
interesting and fun part of the book. Chapter 2 is a fairly standard treatment of the common
discrete distributions, but with heavy use of technology. On first reading, the material on
conditional expectation in Section 2.6 can be minimized (students ought to at least know the
definition), but the section will be very helpful later for students going on to study finance.
Chapter 3 is an important transition to continuous probability, but the first section contains
some material on higher level ideas like Σ-algebras that is usually difficult for students at the
first pass, so handle with care. In Chapter 4 on continuous distributions, certainly Sections
4.1, 4.2, 4.3, 4.5, and 4.6 are must-do sections. Section 4.4 on order statistics may be
omitted if time is pressing. The final section of Chapter 4 requires a bit of linear algebra.
Courses that have this as a prerequisite ought not to be afraid to cover this material, which
helps students a great deal when studying regression and analysis of variance in mathemati-
cal statistics. But courses that do not assume linear algebra should probably pass by Section
4.7. Chapter 5 examines some of the history of probability while covering the Laws of Large
Numbers and the Central Limit Theorem, and so it should not be missed. The treatment of
stochastic processes and applications in Chapter 6 is optional; an instructor can include more
of it or less of it depending on time. The sections are nearly independent of one another,
although it is hard to imagine studying queues in Section 6.3 without first doing Poisson
processes in Section 6.2. Among these sections, surely 6.2 is the most important, filling in
the theory of arrival times that was not touched in Section 2.4 where the Poisson process is
first introduced. The instructor may find the Chapter 6 material to be fertile ground for
student projects and/or independent reading. Students intending to go on to operations
research or finance will particularly benefit by reading and doing these sections.
The second edition of Introduction to Probability with Mathematica® , like the first
edition, is available in the traditional print copy and comes with a CD with executable
notebooks, one per text section. It is my intention for the students to really interact with the
electronic book, but I also think that the paper copy stands on its own merits as a nice
treatment of probability. Having the electronic version ought to greatly benefit the instructor
as well in creating class notes, demonstrations, and projects. I continue to believe that the
book is an important contribution to the field, valued particularly by mathematics depart-
ments that are becoming more and more technologically oriented.
I have updated the complete solutions manual to include the new problems and to be
compatible with Mathematica versions 6 and higher. This is available upon request from the
publisher.
This is now my third project with Chapman & Hall/CRC Press and I must thank the
helpful folks there, including Executive Editor Bob Stern, Project Coordinator Jessica
Vakili, and technical assistants and reviewers for their help in getting the book out. I should
also mention Maryka Baraka and others at Wolfram Research for their Mathematica advice.
And, lastly, I continue to marvel at the patience and encouragement given to me by my wife
Gay Lynn and daughter Emily during my authoring experiences. They are the best.

Kevin J. Hastings
 Note on the Electronic Version and the
KnoxProb7Utilities Package
The CD that accompanies this book contains a Mathematica notebook for each text
section. Copyright law forbids you from distributing these notebooks without the express
permission of the publisher, just as with the print version. The print text was made without
color, by publisher specification, but color effects are included into the electronic version.
Otherwise the electronic notebooks are identical to the print version. When a notebook is
first loaded you can select the Evaluate Initialization Cells command under the Evaluation
menu to reproduce the output. The look of the notebooks in the front end will depend on
what style sheet you are using; I have applied the Book/Textbook stylesheet to these files.
Also on the disk are two packages: the KnoxProb7  Utilities package for users of
version 7 of Mathematica, and KnoxProb6`Utilities` for users of version 6. Move the
appropriate folder into the AddOns\ExtraPackages subdirectory of your main Mathematica
directory in order to allow the electronic book to run. The electronic files are written for
version 7, so the Needs commands in the notebooks should be modified to load Knox-
Prob6`Utilities` instead of KnoxProb7`Utilities` for Mathematica 6 users. The main differ-
ences between the two Utilities versions are: (1) the Histogram command that I had written
for the version 6 package is eliminated in the version 7 package because Wolfram Research
included its own suitable Histogram command into the kernel in version 7; and (2) the
ProbabilityHistogram command had to be rewritten for version 7 due to the elimination of
the BarCharts package that it had relied on. Both packages are updates of those in the
KnoxProb`Utilities` package that was available for the first edition of the book, adapting to
the major changes in the structure of Mathematica that happened as it went into version 6.0.
Among the most important changes for our purposes was that many statistical commands
that had previously been contained in external packages were moved into the kernel, so that
these packages no longer needed to be loaded. The electronic book has been subjected to a
lot of testing, and it should operate well.
 Table of Contents
Chapter 1 – Discrete Probability
1.1 The Cast of Characters 1
1.2 Properties of Probability 10
1.3 Simulation 23
1.4 Random Sampling 33
Sampling in Sequence without Replacement 34
Sampling in a Batch without Replacement 38
Other Sampling Situations 42
1.5 Conditional Probability 48
Multiplication Rules 51
Bayes' Formula 56
1.6 Independence 62
Chapter 2 – Discrete Distributions
2.1 Discrete Random Variables, Distributions,
and Expectations 73
Mean, Variance, and Other Moments 79
Two Special Distributions 85
2.2 Bernoulli and Binomial Random Variables 94
Multinomial Distribution 102
2.3 Geometric and Negative Binomial Random Variables 107
2.4 Poisson Distribution 115
Poisson Processes 120
2.5 Joint, Marginal, and Conditional Distributions 124
2.6 More on Expectation 136
Covariance and Correlation 140
Conditional Expectation 148
Chapter 3 – Continuous Probability
3.1 From the Finite to the (Very) Infinite 159
3.2 Continuous Random Variables and Distributions 171
Joint, Marginal, and Conditional Distributions 178
3.3 Continuous Expectation 192
Chapter 4 – Continuous Distributions
4.1 The Normal Distribution 207
4.2 Bivariate Normal Distribution 225
Bivariate Normal Density 229
Marginal and Conditional Distributions 235
 4.3 New Random Variables from Old 246
C.D.F. Technique and Simulation 247
Moment–Generating Functions 252
4.4 Order Statistics 259
4.5 Gamma Distributions 273
Main Properties 273
The Exponential Distribution 277
4.6 Chi–Square, Student’s t, and F–Distributions 281
Chi–Square Distribution 282
Student's t–Distribution 289
F–Distribution 293
4.7 Transformations of Normal Random Variables 300
Linear Transformations 301
Quadratic Forms in Normal Random Vectors 306
Chapter 5 – Asymptotic Theory
5.1 Strong and Weak Laws of Large Numbers 313
5.2 Central Limit Theorem 320
Chapter 6 – Stochastic Processes and Applications
6.1 Markov Chains 329
Short Run Distributions 331
Long Run Distributions 334
Absorbing Chains 338
6.2 Poisson Processes 344
6.3 Queues 352
Long Run Distribution of System Size 353
Waiting Time Distribution 358
Simulating Queues 359
6.4 Brownian Motion 365
Geometric Brownian Motion 373
6.5 Financial Mathematics 379
Optimal Portfolios 380
Option Pricing 385
Appendices
A. Introduction to Mathematica 395
B. Glossary of Mathematica Commands for Probability 423
C. Short Answers to Selected Exercises 431
References 445

Index 447
 CHAPTER 1
DISCRETE PROBABILITY

1.1 The Cast of Characters

Why study probability? All around us is uncertainty and we must understand its
nature in order to conduct ourselves in the world. We join the shortest waiting line in hopes
that it has the best chance of allowing us to be served quickly. We invest our savings in
assets that we believe are most likely to yield high returns. We buy the battery that we think
is most likely to last the longest. The study of probability also lets you understand the
mechanics of games of chance like roulette, poker, and blackjack; what is predictable and
what is not, what to bet on (usually nothing) and what not to bet on (usually everything). In
all of these cases we do not know for certain what will happen before the phenomenon is
observed (we don't know how much time we will wait in line, what our opponent's blackjack
hand will be, etc.). But some common-sense assumptions, background knowledge, or a long
history of observed data from similar experiments can enable us to quantify likelihoods.
Several of these examples have actually given birth to significant subfields of probability or
related areas of study, for example, queueing theory (i.e., the study of waiting lines) and
mathematical finance, which are introduced in Chapter 6. Probability also provides the
theoretical basis for statistics, which is the art of using data obtained randomly to draw
conclusions about social and scientific phenomena.
In addition to all this, probability exhibits many of the aesthetically appealing
properties that attract people to mathematics in general. It proceeds logically from a few
simple axioms to build an abstract structure which is straightforward, yet powerful. The cast
of characters of probability is both foreign and tangible, which implies that you must work a
little to get to know them, but there is hope in succeeding in the acquaintance.
This text will be somewhat different from others in the field. While the main ideas of
probability have not changed much in at least fifty years, the world has, and learning in
mathematics has changed with it. Many things are technologically possible today of which
mathematics education should take advantage. This book is a living, interactive text which
asks you to do activities as you read carefully to help you understand and discover the ideas
of probability in new ways, on your own terms. The complete text is available electronically
in the form of Mathematica notebooks, one for each section. The notebooks are executable,
so that you can reenter existing commands, modify them, or write your own supplemental
commands to take the study in the direction you wish. You will be building probability

1
2 Chapter 1 Discrete Probability

yourself in a kind of workshop, with Mathematica and your own insight and love of
experimentation as your most powerful tools. I caution you against depending too much on
the machine, however. Much of the time, traditional pencil, paper, and brainpower are what
you need most to learn. What the computer does, however, is open up new problems, or let
you get new insights into concepts. So we will be judiciously blending new technology with
traditionally successful pedagogy to enable you to learn effectively. Consistent themes will
be the idea of taking a sample randomly from a larger universe of objects in order to get
information about that universe, and the computer simulation of random phenomena to
observe patterns in many replications that speak to the properties of the phenomena.
The cast of characters in probability whose personalities we will explore in depth in
the coming chapters include the following six principal players: (1) Event; (2) Sample
Space; (3) Probability Measure; (4) Random Variable; (5) Distribution; and (6) Expectation.
The rest of this section will be a brief intuitive introduction to these six.
Most of the time we are interested in assessing likelihoods of events, that is, things
that we might observe as we watch a phenomenon happen. The event that a national chain's
bid to acquire a local grocery store is successful, the event that at least 10 patients arrive to
an emergency room between 6:00 and 7:00 AM, the event that a hand of five poker cards
forms a straight, the event that we hold the winning ticket in the lottery, and the event that
two particular candidates among several are selected for jury duty are just a few examples.
They share the characteristic that there is some uncertain experiment, sampling process, or
phenomenon whose result we do not know "now". But there is a theoretical "later" at which
we can observe the phenomenon taking place, and decide whether the event did or didn't
occur.
An outcome (sometimes called a simple event) is an event that cannot be broken
down into some combination of other events. A composite event is just an event that isn't
simple, that is, it has more than one outcome. For instance the event that we are dealt a poker
hand that makes up a straight is composite because it consists of many outcomes which
completely specify the hand, such as 2 through 6 of hearts, 10 through ace of clubs, etc.
This is where the sample space makes its entry: the sample space is the collection of
all possible outcomes of the experiment or phenomenon, that is all possible indecomposable
results that could happen. For the poker hand example, the sample space would consist of
all possible five-card hands. In specifying sample spaces, we must be clear about assump-
tions; for instance we might assume that the order in which the cards were dealt does not
matter, and we cannot receive the same card twice, so that the cards are dealt without
replacement.

Activity 1 How can you characterize the sample space for the grocery store example?
Make up your own random phenomenon, informally describe the sample space, and give
examples of events.
1.1 The Cast of Characters 3

Activity 2 In the emergency room example above, is the event that was described an
outcome or a composite event? If that event is composite, write down an example of an
outcome relating to that phenomenon.

In random phenomena, we need a way of measuring how likely the events are to
occur. This may be done by some theoretical considerations, or by experience with past
experiments of the same kind. A probability measure assigns a likelihood, that is a number
between 0 and 1, to all events. Though we will have to be more subtle later, for now we can
intuitively consider the extremes of 0 and 1 as representing impossibility and complete
certainty, respectively.
Much of the time in probability problems, probability measures are constructed so
that the probability of an event is the "size" of the event as a proportion of the "size" of the
whole sample space. "Size" may mean different things depending on the context, but the two
most frequent usages are cardinality, that is number of elements of a set, and length (or area
in two dimensions). For example, if you roll a fair die once, since there are six possible faces
that could land on top, the sample space is 1, 2, 3, 4, 5, 6. Since three of the faces are odd,
it seems reasonable that the probability that an odd number is rolled is 3/6. Size means
length in the following scenario. Suppose that a real number is to be randomly selected in
the interval 0, 4. This interval is the sample space for the random phenomenon. Then the
probability that the random real number will be in the interval 1, 3 is 2  4  1  2, which is
the length of the interval 1, 3 divided by the length of the whole sample space interval.
The first two chapters will concentrate on discrete probability, in which counting the
number of elements in sets is very important. We will leave much of the detail for later
sections, except to remind you of the intuitively obvious multiplication principle for
counting. If outcomes of a random experiment consist of two stages, then the total number of
outcomes of the experiment is the number of ways the first stage can occur times the number
of ways that the second can occur. This generalizes to multiple stage experiments. For
instance, if a man has 3 suits, 6 shirts, and 4 ties, and he doesn't care about color coordina-
tion and the like, then he has 3  6  4  72 possible outfits.
It is possible to experimentally estimate a probability by repeating the experiment
over and over again. The probability of the event can be estimated by the number of times
that it occurs among the replications divided by the number of replications. Try this
exercise. In the KnoxProb7`Utilities` package that you should have installed is a Mathemat-
ica command called

DrawIntegerSample[a, b, n]
4 Chapter 1 Discrete Probability

which outputs a sequence of n numbers drawn at random, without reusing numbers previ-
ously in the sequence, from the positive integers in the range a, ... , b. (You should be sure
that the package has been loaded into the ExtraPackages subdirectory within the Mathemat-
ica directory you are using on your computer.) First execute the command to load the
package, then try repeating the experiment of drawing samples of two numbers from
1, ... , 5 a large number of times. I have shown one such sample below. Keep track of how
frequently the number 1 appears in your samples. Empirically, what is a good estimate of the
probability that 1 is in the random sample of size two? Talk with a group of your classmates,
and see if you can set up a good theoretical foundation that supports your conclusion. (Try to
specify the sample space of the experiment.) We will study this type of experiment more
carefully later in this chapter. Incidentally, recall that Mathematica uses braces { } to
delineate sequences as well as unordered sets. We hope that the context will make it clear
which is meant; here the integer random sample is assumed to be in sequence, so that, for
example, 1, 3 is different from 3, 1.

Needs"KnoxProb7`Utilities`"

DrawIntegerSample1, 5, 2

1, 3

We are halfway through our introduction of the cast of characters. Often the outc-
omes of a random phenomenon are not just numbers. They may be sequences of numbers as
in our sampling experiment, or even non-numerical data such as the color of an M&M that
we randomly select from a bag. Yet we may be interested in encoding the simple events by
single numbers, or extracting some single numerical characteristic from more complete
information. For example, if we are dealt a five card hand in poker, an outcome is a full
specification of all five cards, whereas we could be interested only in the number of aces in
the hand, which is a numerical valued function of the outcome. Or, the sample space could
consist of subjects in a medical experiment, and we may be interested in the blood sugar
level of a randomly selected subject. Such a function for transforming outcomes to numerical
values is called a random variable, the fourth member of our cast of characters. Since the
random variable depends on the outcome, and we do not know in advance what will happen,
the value of the random variable is itself uncertain until after the experiment has been
observed.

Activity 3 Use DrawIntegerSample again as below to draw samples of size five from a
population of size 50. Suppose we are interested in the random variable which gives the
minimum among the members of the sample. What values of this random variable do
you observe in several replications?
1.1 The Cast of Characters 5

outcome  DrawIntegerSample1, 50, 5

X  Minoutcome

The cell below contains a command I have called SimulateMinima for observing a
number m of values of the minimum random variable from Activity 3. Read the code
carefully to understand how it works. An integer sample is drawn from the set
{1,2,...,popsize}, and then its minimum is taken. The Table command assembles a list of m
of these minima. To get an idea of the pattern of those sample minima, we plot a graph
called a histogram which tells the proportion of times among the m replications of the
experiment that the sample minimum fell into each of a given number of categories. The
Histogram command is in the kernel in Mathematica 7. Its exact syntax is

Histogram[listofdata, binspecifications, barheightspecifications]

where the latter two arguments are optional. (See the help documentation, the note in the
preface, and the appendix on Mathematica commands for version 6 of Mathematica and the
KnoxProb6`Utilities` package.) The binspecifications argument can be set to be a number
for a desired number of categories, or it can be set to a list d w for a desired category width.
The third argument can be set to "Count," "Probability," or "ProbabilityDensity," respec-
tively, to force the bar heights to be freqencies, relative frequencies, or relative frequencies
divided by category widths. One output cell is shown in Figure 1 using 100 replications, six
histogram rectangles, and using samples of size 5 from {1, ... , 50}. You should reexecute
several times to see whether your replications give similar histograms. Should they be
identical to this one? Why or why not? (If you want to suppress the list of minima and get
only the histogram, put a semicolon between the SimulateMinima and Histogram
commands.)

SimulateMinimam_, popsize_, sampsize_ :

TableMin
DrawIntegerSample1, popsize, sampsize, i, 1, m

mins  SimulateMinima100, 50, 5

Histogrammins, 6, ChartStyle  Gray
6 Chapter 1 Discrete Probability

7, 6, 4, 13, 6, 8, 2, 10, 6, 9, 2, 4, 2, 5, 1, 2, 11, 5,

3, 2, 22, 2, 8, 17, 6, 6, 13, 4, 2, 6, 15, 32, 3, 7, 15,
21, 5, 17, 2, 4, 8, 12, 7, 13, 1, 6, 28, 26, 14, 12, 3,
7, 6, 1, 12, 11, 6, 7, 17, 6, 10, 3, 3, 1, 1, 16, 7, 5,
4, 18, 6, 14, 23, 15, 18, 22, 4, 9, 9, 8, 1, 15, 4, 16,
4, 9, 16, 8, 23, 10, 1, 7, 20, 6, 10, 10, 2, 7, 18, 2

35
30
25
20
15
10
5

5 10 15 20 25 30
Figure 1.1 - Sample histogram of minimum of five values from 1, 2, ... , 50

The histogram for the sample minimum that we just saw, together with your experi-
ments, hints at a very deep idea, which is the fifth member of our cast. A random variable
observed many times will give a list of values that follows a characteristic pattern, with some
random variation. The relative frequencies of occurrence (that is the number of occurrences
divided by the number of replications) of each possible value of the random variable, which
are the histogram rectangle heights, will stabilize around some theoretical probability. Much
as a probability measure can be put on a sample space of simple events, a probability
distribution can be put on the collection of possible values of a random variable. To cite a
very simple example, if 1/3 of the applicants for a job are men and 2/3 are women, if one
applicant is selected randomly, and if we define a random variable
1 if applicant is a man
X 
0 if applicant is a woman

then the two possible values of X are 1 and 0, and the probability distribution for this X gives
probability 1/3 to value 1 and 2/3 to value 0.
The last member of our cast is expectation or expected value of a random variable.
Its purpose is to give a single number which is an average value that the random variable can
take on. But if these possible values of the random variable are not equally likely, it doesn't
make sense to compute the average as a simple arithmetical average. For example, suppose
the number of 911 emergency calls during an hour is a random variable X whose values and
probability distribution are as in the table below.
1.1 The Cast of Characters 7

value of X 2 4 6 8
probability 1  10 2  10 3  10 4  10

What could we mean by the average number of calls per hour? The simple arithmetical
average of the possible values of X is (2+4+6+8)/4 = 5, but 6 and 8 are more likely than 2
and 4, so they deserve more influence. In fact, the logical thing to do is to define the
expected value of X, denoted by E[X], to be the weighted average of the possible values,
using the probabilities as weights:
1 2 3 4 60
EX   2  4  6  8   6
10 10 10 10 10

We will be looking at all of our characters again in a more careful way later, but I
hope that the seeds of intuitive understanding have been planted. You should reread this
introductory section from time to time as you progress to see how the intuition fits with the
formality.

Exercises 1.1
1. Most state lotteries have a version of a game called Pick 3 in which you win an amount,
say $500, if you correctly choose three digits in order. Suppose that the price of a Pick 3
ticket is $1. There are two possible sample spaces that might be used to model the random
experiment of observing the outcome of the Pick 3 experiment, according to whether the
information of the winning number is recorded, or just the amount you won. Describe these
two sample spaces clearly, give examples of outcomes for both, and describe a reasonable
probability measure for both.
2. A grand jury is being selected, and there are two positions left to fill, with candidates A,
B, C, and D remaining who are eligible to fill them. The prosecutor decides to select a pair
of people by making slips of paper with each possible pair written on a slip, and then
drawing one slip randomly from a hat. List out all outcomes in the sample space of this
random phenomenon, describe a probability measure consistent with the stated assumptions,
and find the probability that A will serve on the grand jury.
3. In an effort to predict performance of entering college students in a beginning mathemat-
ics course on the basis of their mathematics ACT score, the following data were obtained.
The entries in the table give the numbers of students among a group of 179 who had the
indicated ACT score and the indicated grade.
Grade
A B C D F
23 2 5 16 31 3
ACT 24 6 8 12 15 2
25 7 10 15 6 1
26 8 15 13 4 0
8 Chapter 1 Discrete Probability

(a) A student is selected randomly from the whole group. What is the probability that this
student has a 25 ACT and received a B? What is the probability that this student received a
C? What is the probability that this student has an ACT score of 23?
(b) A student is selected randomly from the group of students who scored 24 on the ACT.
What is the probability that this student received an A?
(c) A student is selected randomly from the group of students who received B's. What is the
probability that the student scored 23 on the ACT? Is it the same as your answer to the last
question in part (a)?
4. Find the sample space for the following random experiment. Four people labeled A, B, C,
and D are to be separated randomly into a group of 2 people and two groups of one person
each. Assume that the order in which the groups are listed does not matter, only the contents
of the groups.
5. Consider the experiment of dealing two cards from a standard 52 card deck (as in black-
jack), one after another without replacing the first card. Describe the form of all outcomes,
and give two examples of outcomes. What is the probability of the event "King on 1st card?"
What is the probability of the event "Queen on 2nd card?"
6. In the Pick 3 lottery example of Exercise 1, consider the random variable X that gives the
net amount you win if you buy one ticket (subtracting the ticket price). What is the probabil-
ity distribution of X? What is the expected value of X?

a X
b 0
c
d 1
e
f

Exercise 7

7. An abstract sample space has outcomes a, b, c, d, e, and f as shown on the left in the
figure. Each outcome has probability 1/6. A random variable X maps the sample space to the
set {0,1} as in the diagram. What is the probability distribution of X? What is the expected
value of X?
1.1 The Cast of Characters 9

8. Two fair coins are flipped at random and in succession. Write the sample space explicitly,
and describe a reasonable probability measure on it. If X is the random variable that counts
the number of heads, find the probability distribution of X and compute the expected value
of X.
9. Using the data in Exercise 3, give an empirical estimate of the probability distribution of
the random variable X that gives the ACT score of a randomly selected student. What is the
expected value of X, and what is its real meaning in this context?
10. (Mathematica) Use the DrawIntegerSample command introduced in the section to build
a command that draws a desired number of samples, each of a desired size, from the set
1, 2, ..., popsize. For each sample, the command appends the maximum number in the
sample to a list. This list can be used to create a histogram of the distribution of maxima.
Repeat the experiment of taking 50 random samples of size 4 from 1, 2, ... , 7 several
times. Describe in words the most salient features of the histograms of the maximum. Use
one of your replications to give an empirical estimate of the probability distribution of the
maximum.
11. (Mathematica) The command SimulateMysteryX[m] illustrated below will simulate a
desired number m of values of a mystery random variable X. First load the KnoxProb7`Utili-
ties` package, then run the command several times using specific numerical values for m.
Try to estimate the probability distribution of X and its expected value.

SimulateMysteryX 10

12. (Mathematica) Write a command in Mathematica which can take as input a list of pairs
 pi , xi  in which a random variable X takes on the value xi with probability pi and return the
expectation of X. Test your command on the random variable whose probability distribution
is in the table below.
value of X 0 1 2 3
probability 1  4 1  2 1  8 1  8

13. Two flights are selected randomly and in sequence from a list of departing flights
covering a single day at an airport. There were 100 flights in the list, among which 80
departed on time, and 20 did not. Write a description of the sample space of the experiment
of selecting the two flights, and use the multiplication principle to count how many elements
the sample space has. Now let X be the number of flights in the sample of two that were
late. Find the probability distribution of X .
10 Chapter 1 Discrete Probability

1.2 Properties of Probability

We will begin this section with examples of the concrete construction of probability
measures for sample spaces with finitely many outcomes. With our feet on good intuitive
ground, we will then be ready to set up a formal axiom system which permits rigorous proof
of some important properties of probability. This is the way that most mathematical research
actually proceeds: from concrete to abstract rather than the other way around.
Example 1 If, as we said in Section 1.1, a probability measure gives likelihood to events,
then at its most basic level it gives likelihood to simple events, that is, outcomes. As an
example, I recently saw a breakdown of advanced placement scores for Illinois 12th graders
on the subject test in English Language and Composition. Scores range from 1 to 5, and the
numbers of test takers in these categories for the particular year I looked at were: 1: 74, 2:
293, 3: 480, 4: 254, 5: 130, for a total of 1231 students. Consider first the random experi-
ment of drawing a student at random from this group and observing the test score.

1
3
5

Figure 1.2 - Sampling from a discrete population

There are 5 possible outcomes, namely, the scores 1-5. We can think of the experi-
ment of randomly sampling a student score as spinning the arrow on the pie chart of Figure
2, which has been constructed so that the angles of the circular wedges are proportional to
the frequencies in the categories that were listed in the last paragraph.
This example illustrates several important features of the subject of probability.
First, there is an underlying physics that drives the motion and final resting position of the
arrow. But differences in the starting point, the angular momentum imparted to the arrow,
and even slight wind currents and unevenness of the surface of the spinner give rise to a non-
deterministic or random phenomenon. We cannot predict with certainty where the arrow
will land. It could be philosophically argued that if we knew all of these randomizing factors
perfectly, then the laws of physics would perfectly predict the landing point of the arrow, so
that the phenomenon is not really random. Perhaps this is true in most, if not all, phenomena
that we call random. Nevertheless, it is difficult to know these factors, and the ability to
measure likelihoods of events that we model as random is useful to have.
1.2 Properties of Probability 11

Second, this example shows the two main ways of arriving at an assignment of
probabilities. We might make simplifying assumptions which lead to a logical assignment of
probability. Or we might repeat the experiment many times and observe how frequently
events occur. We were fortunate enough here to have the background data from which our
spinner was constructed, which would indicate by simple counting that if the spinner was
truly spun "randomly" we should assign probabilities
74 293 480 254 130
P1  , P2  , P3  , P4  , P5  (1)
1231 1231 1231 1231 1231
to the five outcomes. But step back for a moment and consider what we would have to do
with our spinner if we weren't blessed with the numbers. We would have no recourse but to
perform the experiment repeatedly. If, say, 1000 spins produced 70 1's, then we would
estimate P[1] empirically by 70  1000  .07, and similarly for the other outcomes.
Third, we can attempt to generalize conclusions. This data set itself can be thought
of as a sort of repeated spinning experiment, in which each student represents a selection
from a larger universe of students who could take the test. The probabilities in formula (1),
which would be exact probabilities for the experiment of selecting one student from this
particular batch, become estimates of probabilities for a different experiment; that of
selecting 1231 students from the universe of all students, past and present, who could take
the test. Whether our particular data set is a "random" sample representing that universe
without bias is a very serious statistical question, best left for a statistics course with a good
unit on samples and surveys.
Using the probabilities in (1), how should probability be assigned to composite
events like "1 or 3?" The logical thing to do is to pool together the probabilities of all
outcomes that are contained in the composite event. Then we would have
74 480 554
P1 or 3  P1  P3    (2)
1231 1231 1231
So the probability of an event is the total of all probabilities of the outcomes that make up
the event.
Two implications follow easily from the last observation: the empty event  which
contains no outcomes has probability zero. Also, the event that the spinner lands somewhere
is certain, that is, it has probability 1 (100%). Another way to look at it is that the sample
space, denoted by  say, is the composite event consisting of all outcomes. In this case we
have:
P  P1 or 2 or 3 or 4 or 5
 P1  P2    P5
74 293 130 (3)
    
1231 1231 1231
 1
12 Chapter 1 Discrete Probability

In summary, so far we have learned that outcomes can be given probabilities between
0 and 1 using theoretical assumptions or empirical methods, and probabilities for composite
events are the totals of the outcome probabilities for outcomes in the event. Also P  0
and P  1. Try the following activity, which suggests another property. You will learn
more about that property shortly.

Activity 1 For the spinner experiment we looked at P[1 or 3]. Consider now the two
events "at least 3" and "between 1 and 3 inclusive." What is the probability of the event
"at least 3" or "between 1 and 3 inclusive"? How does this probability relate to the
individual event probabilities? Can you deduce a general principle regarding the
probability of one event or another occurring?

Sometimes we are interested in the set theoretic complement of an event, i.e., the
event that the original event does not occur. Does the probability of the complementary
event relate to the probability of the event? For the spinner, look at the events "at least 3"
and "less than 3," which are complements. We have, using formula (1):
480 254 130 864
Pat least 3  P3 or 4 or 5    
1231 1231 1231 1231
(4)
74 293 367
Pless than 3  P1 or 2   
1231 1231 1231
Notice that the two probabilities add to one, that is for these two events E and Ec :
PE  PEc   1 (5)

This is not a coincidence. Together E and Ec contain all outcomes, and they have no
outcomes in common. The total of their probabilities must therefore equal 1, which is the
probability of the sample space. 
Example 2. Let's look at another random phenomenon. Recall from Section 1.1 the
experiment of selecting a sequence of two numbers at random from the set 1, 2, 3, 4, 5
with no repetition of numbers allowed. Such an ordered selection without replacement is
known as a permutation, in this case a permutation of 2 objects from 5. Mathematica can
display all possible permutations as follows. The command

KPermutations[list, k]

in the KnoxProb7`Utilities` package returns a list of all ordered sequences of length k from
the list given in its first argument, assuming that a list member, once selected, cannot be
selected again.
1.2 Properties of Probability 13

Needs"KnoxProb7`Utilities`"

KPermutations1, 2, 3, 4, 5, 2

1, 2, 2, 1, 1, 3, 3, 1, 1, 4, 4, 1,
1, 5, 5, 1, 2, 3, 3, 2, 2, 4, 4, 2, 2, 5,
5, 2, 3, 4, 4, 3, 3, 5, 5, 3, 4, 5, 5, 4

LengthKPermutations1, 2, 3, 4, 5, 2

20

As the Length output reports, there are 20 such permutations. These 20 are the outcomes that
form the sample space of the experiment. The randomness of the drawing makes it quite
reasonable to take the theoretical approach to assigning probabilities. Each outcome should
have the same likelihood. Since the total sample space has probability 1, the common
likelihood of the simple events, say x, satisfies
x  x    x 20 times  1  20 x  1  x  1  20 (6)

In other words, our probability measure on the sample space is defined on outcomes as:
P1, 2  P2, 1  P1, 3    P5, 4  1  20 (7)

and for composite events, we can again total the probabilities of outcomes contained in them.

Activity 2 In the example of selecting two positive integers from the first five, find the
probability that 1 is in the sample. Explain why your result is intuitively reasonable.

For instance,
Peither 1 or 2 is in the sample
 P1, 2, 2, 1, 1, 3, 3, 1, 1, 4, 4, 1, 1, 5, 5, 1,
2, 3, 3, 2, 2, 4, 4, 2, 2, 5, 5, 2
 14  20

and
P2 is not in sample
 P1, 3, 3, 1, 1, 4, 4, 1, 1, 5, 5, 1, 3, 4, 4, 3, 3, 5,
5, 3, 4, 5, 5, 4
 12  20
14 Chapter 1 Discrete Probability

By counting outcomes you can check that P[2 is in sample] = 8/20, which is complementary
to P[2 is not in sample] = 12/20 from above. Also notice that
8
P1 in sample   P2 in sample
20
however,
14 8 8
P1 in sample or 2 in sample   
20 20 20
We see that the probability of the event "either 1 is in the sample or 2 is in the sample" is not
simply the sum of the individual probability that 1 is in the sample plus the probability that 2
is in it. The reason is that in adding 8  20  8  20 we let the outcomes 1, 2 and 2, 1
contribute twice to the sum. This overstates the true probability, but we now realize that it is
easy to correct for the double count by subtracting 2  20 which is the amount of probability
contributed by the doubly counted outcomes. (See Theorem 2 below.) 

Activity 3 Use Mathematica to display the ordered samples of size 3 without replace-
ment from 1, 2, 3, 4, 5, 6. What is the probability that 1 is in the sample? What is the
probability that neither 5 nor 6 is in the sample?

Armed with the experience of our examples, let us now develop a mathematical
model for probability, events, and sample spaces that proceeds from a minimal set of axioms
to derive several of the most important basic properties of probability. A few other properties
are given in the exercises.
Definition 1. A sample space  of a random experiment is a set whose elements Ω are
called outcomes.
Definition 2 below is provisional we will have to be more careful about what an
event is when we come to continuous probability.
Definition 2. An event A is a subset of the sample space , hence an event is a set of
outcomes. The collection of all events is denoted by .
Suppose for example we consider another sampling experiment in which three letters
are to be selected at random, without replacement and without regard to order from the set of
letters a, b, c, d. The Combinatorica` package in Mathematica (which is loaded automati-
cally when you load KnoxProb7`Utilities`) has the commands

KSubsets[list, k] and Subsets[universe]

which, respectively, return all subsets of size k taken from the given list of objects, and all
subsets of the given universal set. We can use these to find the sample space of the experi-
ment, and the collection of events :
1.2 Properties of Probability 15

  KSubsetsa, b, c, d, 3

a, b, c, a, b, d, a, c, d, b, c, d

  Subsets

, a, b, c, a, b, d, a, c, d,

b, c, d, a, b, c, a, b, d, a, b, c, a, c, d,
a, b, c, b, c, d, a, b, d, a, c, d,
a, b, d, b, c, d, a, c, d, b, c, d,
a, b, c, a, b, d, a, c, d, a, b, c, a, b, d, b, c, d,
a, b, c, a, c, d, b, c, d, a, b, d, a, c, d, b, c, d,
a, b, c, a, b, d, a, c, d, b, c, d

Notice that the empty set  is the first subset of  that is displayed; next are the simple
events a, b, c, a, b, d etc.; next are the two element composite events; and so forth.
In all subsequent work we will continue a practice that we have done so far, which is
to blur the distinction between outcomes Ω, which are members of the sample space, and
simple events Ω which are sets consisting of a single outcome. This will enable us to speak
of the "probability of an outcome" when in fact probability is a function on events, as seen in
the next definition.
Definition 3. A probability measure P is a function taking the family of events  to the real
numbers such that
(i) P  1;
(ii) For all A , P[A] 0;
(iii) If A1 , A2 ,  is a sequence of pairwise disjoint events then
P[A1 A2   = PAi 
i

Properties (i), (ii), and (iii) are called the axioms of probability. Our mathematical
model so far corresponds with our intuition: a random experiment has a set of possible
outcomes called the sample space, an event is a set of these outcomes, and probability
attaches a real number to each event. Axiom (i) requires the entire sample space to have
probability 1, axiom (ii) says that all probabilities are non-negative numbers, and axiom (iii)
says that when combining disjoint events, the probability of the union is the sum of the
individual event probabilities. For finite sample spaces this axiom permits us to just define
probabilities on outcomes, and then give each event a probability equal to the sum of its
(disjoint) outcome probabilities. We will encounter various ways of constructing probability
measures along the way, which we must verify to satisfy the axioms. Once having done so,
other properties of probability listed in the theorems below are ours for free. They follow
from the axioms alone, and not from any particular construction, which of course is the huge
advantage of basing a mathematical model on the smallest possible collection of axioms.
16 Chapter 1 Discrete Probability

Theorem 1. If A is an event with complement Ac , then

PA  PAc   1 (8)

Therefore, PA  1 PAc  and PAc   1 PA.

Proof. The events A and Ac are pairwise disjoint, by definition of complementation. By the
third axiom,
PA Ac   PA  PAc 

Also, A Ac = . By the first axiom, P  1, therefore

1  P  PA Ac   PA  PAc  

Theorem 1 was anticipated earlier. It is generally used in cases where the comple-
ment of an event of interest is easier to handle than the original event. If we have the
probability of one of the two, then we have the other by subtracting from 1.
The next result is based on the idea that the total probability of a union can be found
by adding the individual probabilities and correcting if necessary by subtracting the probabil-
ity in the double-counted intersection region.


A B

A B

Figure 1.3 - The probability of a union is the total of the event probabilities minus the intersection probability

Theorem 2. If A and B are arbitrary events, then

PA B  PA  PB PA B (9)

In particular if A and B are disjoint, then

PA B  PA  PB

Proof. (See Figure 3.) Since A can be expressed as the disjoint union A = (A B)
A Bc , and B can be expressed as the disjoint union B  A B B Ac we have, by
axiom (iii),
PA  PB  PA B  PA Bc   PA B  PB Ac 
 PA Bc   PB Ac   2 PA B

Therefore,
1.2 Properties of Probability 17

PA  PB PA B  PA Bc   PB Ac   PA B (10)

But it is easy to see from Figure 3 that the events (A Bc ), (B Ac ), and A B are
pairwise disjoint and their union is A B. Thus, by axiom (iii) of probability again,
PA Bc   PB Ac   PA B  PA B (11)

Combining (10) and (11) proves the first assertion. The second assertion follows directly as
a specialization of axiom (iii) to two events. Alternatively, if A and B are disjoint, then
A B   and by Exercise 11, P  0. 

Activity 4 Try to devise an analogous thoerem to Theorem 2 for a three set union.
(Hint: if you subtract all paired intersection probabilities, are you doing too much?) See
Exercise 14 for a statement of the general result.

You may be curious that the axioms only require events to have non-negative
probability. It should also be clear that events must have probability less than or equal to 1,
because probability 1 represents certainty. In the exercises you are asked to give a rigorous
proof of this fact, stated below as Theorem 3.
Theorem 3. For any event A, P[A] 1. 
Another intuitively obvious property is that if an event B has all of the outcomes in it
that another event A has, and perhaps more, then B should have higher probability than A.
This result is given as Theorem 4, and the proof is Exercise 12.
Theorem 4. If A and B are events such that A  B, then P[A] P[B]. 
Theorems 3 and 4 help to reassure us that our theoretical model of probability
corresponds to our intuition, and they will be useful in certain proofs. But Theorems 1 and 2
on complements and unions are more useful in computations. We would like to finish this
section by adding a third very useful computational result, usually called the Law of Total
Probability. Basically it states that we can "divide and conquer," that is break up the
computation of a complicated probability P[A] into a sum of easier probabilities of the form
P[A B]. The imposition of condition B in addition to A presumably simplifies the computa-
tion by limiting or structuring the outcomes in the intersection. (See Figure 4.) All of our
computational results will be used regularly in the rest of the text, and the exercise set for
this section lets you begin to practice with them.
18 Chapter 1 Discrete Probability

B1 B2 ... Bn
A

Figure 1.4 - The Law of Total Probability

Theorem 5. Let A be an arbitrary event, and assume that B1 , B2 , B3 , … , Bn is a partition

of the sample space , that is, the sets Bi are pairwise disjoint and ni1 Bi  . Then,
n
PA   PA Bi  (12)
i1

Proof. We will use induction on the number of events Bi beginning with the base case n =
2. In the base case, it is clear that B2 = Bc1 , and therefore A  A B1 A Bc1 and the two
sets in this union are disjoint. By Theorem 2,
PA  PA B1   PA Bc1   PA B1   PA B2 

The anchor step of the induction is complete.

Now assume that for all partitions consisting of n events formula (12) is true. Let
B1 , B2 , B3 , … , Bn1 be a partition of n  1 events; we must show (12) with the upper
index of summation replaced by n  1. Define the events Ci by:
C1  B1 , C2  B2 ,  , Cn 1  Bn 1 , Cn  Bn Bn1

Then the Ci 's are also a partition of  (see Activity 5 below). By the inductive hypothesis
PA  ni1 PA Ci 
 ni11 PA Bi   PA Bn Bn1 
 ni 11 PA Bi   P A Bn A Bn1  (13)
 ni11 PA Bi   PA Bn   PA Bn1 
 n1
i1 PA Bi 

This finishes the inductive step of the proof. 

Activity 5 Why are the Ci 's in the proof of the Law of Total Probability a partition of 
? Why is the fourth line of (13) true?
1.2 Properties of Probability 19

Exercises 1.2
1. Consider the experiment of rolling one red die and two white dice randomly. Observe the
number of the face that lands up on the red die and the sum of the two face up numbers on
the white dice.
(a) Describe the sample space, giving the format of a typical outcome and two specific
examples of outcomes.
(b) Define a probability measure on this sample space. Argue that it satisfies the three
axioms of probability.
red die
1 2 3 4 5 6
2 3b
3 2b 1b
4 1b
5 1b 2b
6 1b
white sum
7 1b HR
8 1b 1b
9 2b 1b
10 1b
11 1b 1b
12 2 b
Exercise 2

2. This is a continuation of Exercise 1. A simulated baseball game is played in such a way

that each batter has a card like the one shown in the figure. One red die and two white dice
are rolled. The outcome of the batter's plate appearance is decided by finding the column
corresponding to the red die and the row corresponding to the sum of the white dice. The
entry in that row and that column could be a base hit (1b), a double (2b), a triple (3b), a
home run (HR) or a blank, which would mean that the batter is out. For the batter shown,
what is the probability that he will get a hit of any kind? If you were designing such a card to
simulate a hitter who historically hits about a home run every 20 plate appearances, where
would you place HR symbols on the card?
3. On a driver's education test there are three questions that are considered most important
and most difficult. The questions are multiple choice, with four alternative answers labeled
a, b, c, and d on each question. Exactly one answer is correct for each question. A test taker
guesses randomly on these three questions.
(a) Describe the sample space, giving the format of a typical outcome and two specific
examples of outcomes.
20 Chapter 1 Discrete Probability

(b) Define a probability measure on this sample space. Argue that it satisfies the three
axioms of probability.
(c) What is the probability that the random guesser gets at least two questions right?
4. (Mathematica) Use Mathematica to display the sample space of all possible random
samples without order and without replacement of three names from the list of names {Al,
Bubba, Celine, Delbert, Erma}. How many events are in the collection of all events  ?
Describe a probability measure for this experiment, and find the probability that either
Bubba or Erma is in the sample.
5. Explain why each of the following attempts P1 and P2 is not a good definition of a
probability measure on the space of outcomes  = {a, b, c, d, e, f, g}.
P1 P2
a .12 .16
b .05 .20
c .21 .20
d .01 .12
e .04 .08
f .28 .15
g .13 .12

6. (Mathematica) Use Mathematica to display the sample space of all possible random
samples in succession and without replacement of two numbers from the set 1, 2, ..., 10.
Describe a probability measure, and find the probability that 10 is not in the sample.
7. A 1999 Internal Revenue Service study showed that 44.9 million U.S. taxpayers reported
earnings of less than $20,000, 20.4 million earned between $20,000 and $30,000, 16.2
million earned between $30,000 and $40,000, 41.9 million earned between $40,000 and
$100,000, and 10.5 million earned more than $100,000.
(a) Write the outcomes for the experiment of sampling one taxpayer at random, and define
a probability measure. In such a random sample of one taxpayer, what is the probability that
the reported earnings is at least $30,000?
(b) Write the outcomes for the experiment of sampling two taxpayers at random and with
replacement, that is, once a taxpayer has been selected that person goes back into the pool
and could be selected again. Define a probability measure. (Hint: for a succession of two
events which do not depend on one another, it makes sense to compute the probability that
they both occur as the product of their probabilities. You can convince yourself of this by
considering a very simple experiment like the flip of two coins.) Find the probability that at
least one of the two earned more than $100,000.
1.2 Properties of Probability 21

x,y
2 1 1 2

Exercise 8

8. Suppose that a point with random x and y coordinates in  2, 2 is selected. By this we

mean that the probability that the random point lies in a region R inside the square is the
proportion of the square's area that R has.
(a) Argue that this means of defining a probability measure on the square  2, 2   2, 2
satisfies the axioms of probability.
(b) Find the probability that the random point will be in the third quadrant.
(c) Find the probability that the random point is not in the circle of radius 2 shown in the
picture.
(d) Find the probability that the random point is either in the third quadrant or not in the
circle of radius 2.
9. If A and B are events such that P[A B] = .24, P[A] = .61, and P[B] = .37, find
P[ A B c ].
10. Prove Theorem 3.
11. Prove that P  0.
12. Prove Theorem 4.
13. List out all events for the sample space  = Ω1 , Ω2 , Ω3 . Find the probability of each
simple event if P[Ω1 , Ω2 ] = .6, P[Ω2 , Ω3 ] = .8, and P[Ω1 , Ω3 ] = .6.
14. Prove that for arbitrary events A, B, and C:
PA B C  PA  PB  PC
PA B  PA C  PB C  PA B C

15. The general result for the probability of a union, which has Theorem 2 and Exercise 14
as special cases, is called the law of inclusion and exclusion. Let A1 , A2 , A3 , ..., An be a
collection of n events. To simplify notation, let Aij stand for a typical intersection of two of
them Ai A j , let Aijk stand for a three-fold intersection Ai Aj Ak , etc. The law states that
22 Chapter 1 Discrete Probability

PA1 A2 A3 ... An 
 PAi  PAij   PAijk  
i i j i  j k
n1
 1 PA1 A2 A3 ... An 

In other words, one alternately adds and subtracts all k-fold intersection probabilities as k
goes from 1 to n. Use this result in the following classical problem. Four men leave their
hats at a hat check station, then on departing take a hat randomly. Find the probability that
at least one man takes his own hat.
16. Recall Exercise 5 of Section 1.1 (the blackjack deal). Use the Law of Total Probability,
Theorem 5, to find P[queen on 2nd] in a well-justified way.
17. There are two groups of experimental subjects, one with 2 men and 3 women and the
other with 3 men and 2 women. A person is selected at random from the first group and
placed into the second group. Then a person is randomly selected from the second group.
Use the Law of Total Probability to find the probability that the person selected from the
second group is a woman.
In this book we will include as appropriate problems taken from the list of sample
questions for the Society of Actuaries/Casualty Actuarial Society qualifying exam P
(Probability). These problems have been obtained with permission from the societies at the
website www.soa.org/files/pdf/P-09-05ques.pdf.
Sample Problems from Actuarial Exam P
18. The probability that a visit to a primary care physician's (PCP) office results in neither
lab work nor referral to a specialist is 35%. Of those coming to a PCP's office, 30% are
referred to specialists and 40% require lab work. Determine the probability that a visit to a
PCP's office results in both lab work and referral to a specialist.
19. An urn contains 10 balls: 4 red and 6 blue. A second urn contains 16 red balls and an
unknown number of blue balls. A single ball is drawn from each urn. The probability that
both balls are the same color is 0.44. Calculate the number of blue balls in the second urn.
20. An auto insurance company has 10,000 policyholders. Each policyholder is classified as:
(i) young or old; (ii) male or female; and (iii) married or single. Of these policyholders, 3000
are young, 4600 are male, and 7000 are married. The policyholders can also be classified as
1320 young males, 3010 married males, and 1400 young married persons. Finally, 600 of
the policyholders are young married males. How many of the company's policyholders are
young, female, and single?
21. An insurer offers a health plan to the employees of a large company. As part of this plan,
the individual employees may choose exactly two of the supplementary coverages A, B, and
C, or they may choose no supplementary coverage. The proportions of the company's
employees that choose coverages A, B, and C are 1/4, 1/3, and 5/12 respectively. Determine
the probability that a randomly chosen employee will choose no supplementary coverage.
1.3 Simulation 23

1.3 Simulation
To simulate means to reproduce the results of a real phenomenon by artificial means.
We are interested here in phenomena that have a random element. What we really simulate
are the outcomes of a sample space and/or the states of a random variable in accordance with
the probability measure or distribution that we are assuming. Computers give a fast and
convenient way of carrying out these sorts of simulations.
There are several reasons to use simulation in problems. Sometimes we do not yet
have a good understanding of a phenomenon, and observing it repeatedly may give us
intuition about a question related to it. In such cases, closed form analytical results might be
derivable, and the simulation mostly serves the purpose of suggesting what to try to prove.
But many times the system under study is complex enough that analysis is either very
difficult or impossible. (Queueing problems  see Section 6.3  are good illustrations of
both situations.) Then simulation is the only means of getting approximate information, and
the tools of probability and statistics also allow us to measure our confidence in our conclu-
sions. Another reason to simulate is somewhat surprising. In some inherently non-probabilis-
tic problems we can use a simulated random target shooting approach to gain information
about physical measurements of size and shape. A simple example of such a problem
solving method (traditionally called Monte Carlo simulation in reference to the European
gaming center) is the example on area of a region in Exercise 8.
I have chosen to give the subject emphasis and early treatment in this book for
another reason. There is more at stake for you intellectually than gaining intuition about
systems by observing simulations. Building an algorithm to solve a problem by simulation
helps to teach you about the subtle concepts of the subject of probability, such as probability
measures and random variables. Thus, here and elsewhere in this book I will be asking you
to carefully study the details of programs I have written, and to write some simulation
programs in Mathematica yourself. You will invest some time and mental effort, but the
payoff will be great.
We have already experienced simulation in the form of selecting ordered samples
without replacement of two numbers from the set 1, 2, 3, 4, 5. In that context we looked
at the empirical histogram of the minimum of the two numbers. We will treat simulations of
samples again in more detail in the next section. For the time being let us see how to use
Mathematica's tools for generating streams of random numbers to solve problems.
Most of the kind of simulations that we will do are adaptations of the simple process
of choosing a random real number in the interval 0, 1. The word "random" here could be
replaced by the more descriptive word "uniform," because the assumption will be that
probability is distributed uniformly through 0, 1, with sets getting large probability only if
their size is large, not because they happen to be in some favored location in 0, 1. More
precisely, we suppose that the sample space of the random experiment is  = 0, 1, the
collection of events  consists of all subsets of 0, 1 for which it is possible to define their
24 Chapter 1 Discrete Probability

length, and the probability measure P gives probability to a set equal to its length. Then for
instance:
1 3 1 1 2 1 1 7
P 0, 1   1, P ,    , P 0,    , 1     ,
2 4 4 4 3 4 3 12
5
P   0
6

Activity 1 Try to describe a uniform distribution of probability on the interval [2,5].

What would be the probability associated to the event [2,3]? [9/4, 13/4]  (4,5]? {2.5}?

In Mathematica's kernel there are several probability distribution objects, including

one for the uniform distribution, and also a means of sampling one or more random numbers
from a distribution. Fittingly,

UniformDistribution[{a,b}]

is the syntax for the uniform distribution on the interval [a,b], and

RandomReal[distribution] and RandomReal[distribution, n]

return, respectively, one or a list of n random numbers from the given distribution. Here are
some examples of their use.

dist  UniformDistribution2, 5;

RandomRealdist
RandomRealdist
RandomRealdist, 5

4.42891

2.00398

2.14133, 2.06261, 3.20962, 3.47772, 3.93109

Notice that new random numbers in 2, 5 are returned each time that RandomReal is called.
Incidentally, RandomReal may be used with an empty argument when the distribution being
sampled from is the uniform distribution on 0, 1, which is what we will usually need.
How do Mathematica and other programs obtain such random numbers? There are
several means that are possible, but the most common type of random number generator is
the linear congruential generator which we describe now. The ironic fact is that our
supposedly random stream of numbers is only simulated in a deterministic fashion in such a
1.3 Simulation 25

way as to be indistinguishable statistically from a list of truly random numbers. Because of

this, the stream of numbers is called pseudo-random. The algorithm is as follows. Begin with
a single non-negative integer called a seed. (Computers can use internal contents such as
collections of digits in their clocks as seeds, which change frequently.) To get the first
pseudo-random number, compute a new value of the seed by:
new value of seed  multiplier  seed  increment mod modulus (1)

where multiplier, increment, and modulus are positive integer constants. Then, using the
newly computed seed, return the number
rand  seed modulus (2)

Subsequent random numbers are found in the same way, updating seed by formula (1) and
dividing by modulus. Notice that because the initial seed is non-negative and all of the
constants are positive, seed will stay non-negative, which makes rand  0 as well. And since
the mod operation returns a value less than modulus, seed will be less than modulus, hence
rand will be less than 1. So each pseudo-random number that we produce will lie in [0, 1).
Now the larger the value of modulus, the more possible values of seed there can be;
hence, the more real numbers that can be returned when formula (2) is applied. In fact, there
would be exactly modulus number of possible seed values. But modulus, multiplier, and
increment must be carefully chosen to take advantage of this, otherwise, the sequence of
pseudo-random numbers could cycle with a very small period. As a simple example, if the
constants were as follows:
new seed  2  seed  2 mod 8

then possible seed values are 0, 1, 2, 3, 4, 5, 6, and 7, but it is easy to check that for these
seeds the new seeds are, respectively, 2, 4, 6, 0, 2, 4, 6, 0, so that after the first value of seed,
subsequent seeds can only take on one of four values. Even worse, should seed ever equal 6,
then new seeds will continue to be 6 forever, because (2 × 6 + 2) mod 8 is again 6.
Therefore, the choice of the constants is critical to making a stream of numbers that
has the appearance of randomness. There is a lot of number theory involved here which we
cannot hope to cover. The book by Rubinstein (see References) gives an introduction. It is
reported there that the values
multiplier  27  1, increment  1, modulus  235
have been successful, but these are not the only choices. We can build our own function
similar to the RandomReal command in Mathematica to generate a list of uniform[0,1]
numbers using these values as follows.
26 Chapter 1 Discrete Probability

MyRandomArrayinitialseed_, n_ :
Moduleseed, rand, thearray, mult, inc, modul, i,
mult  27  1;
inc  1;
modul  235 ;
thearray  ;
seed  initialseed;
Doseed  Modmult  seed  inc, modul;
rand  N seed  modul;
AppendTothearray, rand, i, 1, n;
thearray

The module begins with the list of local variables used in the subsequent lines, and sets the
values above for the multiplier, here called mult, the increment inc, and the modulus modul.
After initializing an empty list of random numbers and setting the initial value of seed
according to the input initialseed, the module enters a loop. In each of the n passes through
the loop, three things are done. The seed is updated by the linear congruence relation (1), the
new random number rand is computed by formula (2), and then rand is appended to the
array of random numbers. Here is a sample run. You should try running the command again
with the same initial seed to see that you get the same sequence of pseudo-random numbers,
then try a different seed.

MyRandomArray224  19, 10

0.0629884, 0.125497, 0.189176, 0.403682, 0.0749859,

0.673181, 0.840295, 0.398108, 0.355958, 0.918557

Activity 2 Revise the random number generating command above using the following
constants, taken from a popular computer science text.

multiplier = 25173 increment = 13849 modulus = 65536

Use the command Histogram[datalist,numrectangles] to plot a histogram of 50 random

numbers using 4 rectangles. Do you see what you expected?

One advantage to the pseudo-random approach is that simulation experiments are

reproducible. If one starts with the same seed, then the same stream of random numbers will
ensue, which will produce the same experimental results. This is advantageous because you
may want to compare the performance of similar systems using the same random inputs to
1.3 Simulation 27

ensure fairness. Mathematica permits this by using the command

SeedRandom[n]

which sets the initial seed value as the integer n. If you use an empty argument to SeedRan-
dom, it resets the seed using the computer's clock time. In the sequence of commands below,
the seed is set to 79, two random numbers are generated, then the seed is reset to 79, and the
same random number .624058 is returned as was returned the first time. We then reset the
seed "randomly" using the computer clock, and the first random number that is returned is
something other than .624058.

SeedRandom79;
RandomReal
RandomReal

0.624058

0.782677

SeedRandom79;
RandomReal

0.624058

SeedRandom;
RandomReal

0.295358

Activity 3 Perform a similar test with SeedRandom using a seed of 591 and requesting
a RandomReal[UniformDistribution[{0,1}], 5]. Reset the seed by computer clock time
and call for another random list of five numbers to see that different random numbers are
returned.

Example 1 It is time to build a simulation model for an actual system. A random walk on
the line is a random process in which at each time instant 0, 1, 2, … an object occupies a
point with integer coordinates shown as circled numbers in Figure 5.
28 Chapter 1 Discrete Probability

1 p 1 p 1 p 1 p 1 p
0 p
1 p
2 p
3 p
4 p
5
Figure 1.5 - Schematic diagram of a random walk

The position on the line cannot be foretold with certainty, but it satisfies the condition that if
the position is n at one time instant, then it will either be n  1 with probability p or n  1
with probability 1  p at the next time instant, as indicated by the annotated arrows in the
figure. Variations are possible, including allowing the object to wait at its current position
with some probability, enforcing boundary behavior such as absorption or reflection to
confine the random walk to a bounded interval, and increasing the dimension of the set on
which the random walk moves to 2, 3, or higher. Random walks have been used as models
for diffusion of heat, the motion of economic quantities such as stock prices, the behavior of
populations, and the status of competitive games among many other applications, and their
fascinating properties have kept probabilists busy for many decades.
For our purposes suppose we are interested in estimating the probability that a
random walk which starts at a point n  0 reaches a point M  n before it reaches 0. This is a
famous problem called the gambler's ruin problem, which really asks how likely it is that a
gambler's wealth that follows a random walk reaches a target level of M before bankruptcy.
We have an initial position n followed by a sequence of random positions X1 , X2 , X3 ,
… . The underlying randomness is such that at each time a simple random experiment
determines whether the move is to the right (probability p) or the left (probability 1  p). So
to simulate the random walk we only need to be able to iteratively simulate the simple
experiment of adding 1 to the current position with probability p and subtracting 1 with
probability 1  p. We will stop when position 0 or position M is reached, and make note of
which of the two was the final position. By running such a simulation many times and
observing the proportion of times that the final state was M, we can estimate the probability
that the final state is M.
To implement the plan in the preceding paragraph, we need a way of randomly
returning +1 with probability p and 1 with probability 1  p. But since we know how to
simulate uniform real random numbers on [0,1], and the probability of an interval is equal to
its length, let's say that if a simulated uniform random number is less than or equal to p then
the result is +1, otherwise 1. Then we have the desired properties of motion, because
Pmove right  P 0, p   length  0, p   p
Pmove left  P p, 1   length p, 1   1  p
Here is a Mathematica function that does the job. A uniform random number is
selected and compared to p. If the  comparison is true, then +1 is returned, else 1 is
returned. So that we can simulate random walks with different p values, we make p an
argument. One function call with p = .5 is displayed; you should try a few others.
1.3 Simulation 29

StepSizep_ : IfRandomReal  p, 1,  1

StepSize.5

1

To simulate the process once, we let the initial position n be given, as well as the
right step probability p and the boundary point M. We repeatedly move by the random
StepSize amount until the random walk reaches 0 or M. The function below returns the
complete list of states that the random walk occupied. Look carefully at the code to see how
it works. The variable statelist records the current list of positions occupied by the random
walk, the variable done becomes true when either the boundary 0 or M is reached, and
variables currstate and nextstate keep track of the current and next position of the random
walk. (The || syntax means logical "or.")

SimulateRandomWalkp_, n_, M0_ :

Modulestatelist, done, currstate, nextstate,
statelist  n;
currstate  n;
done  currstate  0 currstate  M0 ;
WhileNotdone,
nextstate  currstate  StepSizep;
AppendTostatelist, nextstate;
currstate  nextstate;
done  currstate  0 currstate  M0 ;
statelist

Below is an example run of the command. We include a list plot of the sequence of
positions to help your understanding of the motion of the random walk. You should rerun
this command, changing parameters as desired to get more intuition about the behavior of
the random walk. A more systematic investigation is suggested in Exercise 6. We set a
random seed first for reproducibility, but if you are in the electronic version of the book and
want to see different paths of the random walk, just delete the SeedRandom command.

SeedRandom464 758;
thewalk  SimulateRandomWalk0.5, 4, 8
ListPlotthewalk,
PlotStyle  PointSize0.015, Black
30 Chapter 1 Discrete Probability

4, 5, 4, 5, 6, 7, 6, 5, 4, 5, 6, 5, 6,
5, 4, 5, 6, 5, 6, 5, 4, 3, 4, 3, 2, 1, 2, 3,
4, 5, 6, 5, 4, 3, 4, 3, 4, 3, 2, 3, 2, 1, 0

7
6
5
4
3
2
1

10 20 30 40

Figure 1.6 - A random walk on 0, 8

Activity 4 Roughly how would you describe the dependence of the probability that the
random walk reaches M before 0 on the value of p? How does it depend on the value of
n? How could you use SimulateRandomWalk to check your intuition about these
questions?

Continuing the random walk example, to estimate the probability of hitting M before
0 we must repeat the simulation many times, with new streams of random numbers. We
merely take note of the last element of the simulated list of positions, which will be either 0
or M, count the number of times that it is M, and divide by the number of repetitions to
estimate the probability. Here is a command that does this. Any input values we like for p, n,
and M can be used. We also let the number of replications of the simulation be a fourth
argument.

AbsorptionPctAtMp_, n_, M0_, numreps_ :

ModuleabsatM, empprob, poslist, lastpos, i,
absatM  0;
Doposlist  SimulateRandomWalkp, n, M0;
lastpos  Lastposlist;
Iflastpos  M0, absatM  absatM  1, Null,
i, 1, numreps;
empprob  NabsatM  numreps

The variable absatM, initialized to zero, keeps track of how many times so far the walk has
been absorbed at M. The algorithm is simple: we repeat exactly numreps times the actions of
simulating the list of positions (called poslist) of the random walk, then pick out the last
Discovering Diverse Content Through
Random Scribd Documents
history, and far, far more vital than any of the phenomena which
recorded history presents.
Recorded history for this island, and for Northern France and for the
Rhine Valley, is a matter of two thousand years; for the Western
Mediterranean of three; but the things of which I speak are to be
reckoned in tens of thousands of years. Their interest does not lie
only or even chiefly in things that disappeared. It is indeed a great
pleasure to rummage in the earth and find the polished stones of
the men who came so many centuries before us, but of whose blood
we certainly are; and it is a great pleasure to find or to guess that
we find under Canterbury the piles of a lake or marsh dwelling,
proving that Canterbury has been there from all time, and that the
apparently defenceless valley city was once chosen as an
impregnable site when the water-meadows of the Stour were
impassable as marsh, or with difficulty passable as a shallow lagoon.
And it is delightful to stand on the earthwork beyond Chilham and to
say to oneself (as one can say with a fair certitude), "Here was the
British camp defending the south-east; here the tenth legion
charged." All these are pleasant, but more pleasant I think to follow
the thing where it actually survives.
Consider the Track-ways, for instance. How rich England is in these!
No other part of Europe will afford the traveller so permanent and so
fascinating a problem. Elsewhere Rome hardened and straightened
every barbaric trail, but in this distant province of Britain she would
only spend just so much energy as made them a foothold for her
soldiery; and all over England you go if you choose foot by foot
along the ancient roads that were made by the men of your blood
before they had heard of brick, or of stone, or of iron, or of written
laws.
I wonder that more men do not set out to follow, let us say, the
Fosse-way. There it runs right across Western England from the
south-west to the north-east, in a line direct yet sinuous, characters
which are the very essence of a savage trail. It is a modern road for
many miles, let us say; and there you are tramping along the
Cotswold on a hard-metalled modern English highway, with
milestones and notices from the county council telling you that the
culverts will not bear a steam-engine, if so be you were travelling in
one. Then suddenly it comes up against a cross-road and apparently
ceases, making what map draughtsmen call a "T"; but right in the
same line you will see a gate, and beyond it a farm lane, and so you
follow. You come to a spinney where a ride is cut through by the
wood-reeve, and it is all in the same line. The Fosse-way turns into a
little path, but you are still on it; it curls over a marshy brook-valley,
keeping on firm land, and as you go you see old stones put there
Heaven knows how many generations ago—or perhaps yesterday,
for the tradition remains and the countryfolk strengthen their wet
lands as they have strengthened them all these thousands of years;
you climb up out of that depression, you get you over a stile, and
there you are again upon a lane. You follow that lane, and once
more it stops dead. This time there is a field before you. No right of
way, no trace of a path, nothing but grass rounded into those
parallel ridges which mark the decay of the corn lands and pasture
taking the place of agriculture. Now your pleasure comes in casting
about for the trail; you look back along the line of the way; you look
forward in the same line till you find some indication, a boundary
between two parishes, perhaps upon your map, or two or three
quarries set together, or some other sign; and very soon you have
picked up the line again.
So you go on mile after mile, and as you tread that line you feel in
the horizons that you see, in the very nature and feel of the soil
beneath your feet, in the skies of England above you, the ancient
purpose and soul of this kingdom. Up this same line went the Clans
marching when they were called Northward to the Host; and up this
went slow, creaking wagons with the lead of the Mendips or the tin
of Cornwall, or the gold of Wales.
And it is still there; it is still used from place to place as a high-road,
it still lives in modern England. There are some of its peers: as for
instance the Ermine Street, far more continuous, and affording
problems more rarely; others like the ridgeway of the Berkshire
Downs, which Rome hardly touched, and of which the last two
thousand years has therefore made hardly anything. You may spend
a delightful day piecing out exactly where it crossed the Thames,
making your guess at it, and wondering as you sit there by Streatley
Vicarage whether those islands did not form a natural weir below
which lay the ford.
The roads are the most obvious things. There are many more; for
instance, thatch. The same laying of the straw in the same manner,
with the same art, has continued, we may be certain, from a time
long before the beginning of history. See how in the Fen Land they
thatch with reeds, and how upon the Chalk Downs with straw from
the lowlands. I remember once being told of a record in a manor
which held of the church, and which lay upon the southern slope of
the Downs, that so much was entered "for straw from the lowlands";
then years afterwards, when I had to thatch a Bethlehem in an
orchard underneath tall elms—a pleasant place to write in with the
noise of bees in the air—the man who came to thatch said to me:
"We must have straw from the lowlands; this upland straw is no
good for thatching." Then immediately when I heard him say this
there was added to me ten thousand years. And I know another
place in England, far distant from this, where a man said to me that
if I wished to cross in a winter mist, as I had determined to do,
Cross Fell, that great summit of the Pennines, I must watch the drift
of the snow, for there was no other guide to one's direction in such
weather. And I remember another man in a boat in the North Sea,
as we came towards the Foreland, talking to me of the two tides,
and telling me how if one caught the tide all the way up to Long
Nose and then went round it, one made two tides in one day. He
spoke with the same pleasure that silly men show when they talk
about an accumulation of money. He felt wealthy and proud from
the knowledge, for by this knowledge he had two tides in one day.
Now knowledge of this sort is older than ten thousand years; and so
is the knowledge of how birds fly, and of how they call, and of how
the weather changes with the moon.
Very many things a man might add to the list that I am making.
Dew-pans are older than our language or religion; and the finding of
water with a stick; and the catching of that difficult animal, the
mole; and the building of flints into mortar, which if one does it the
old way (as you may see at Pevensey) lasts for ever, and if you do it
the new way does not last ten years; and then there is the
knowledge of planting during the crescent part of the month but not
before the new moon shows; and there is the influence of the moon
on cider, and to a less extent upon the brewing of ale; and talking of
ale, the knowledge of how ale should be drawn from the brewing
just when a man can see his face without mist upon the surface of
the hot brew; and there is the knowledge of how to bank rivers,
which is called "throwing the rives" in the South, but in the Fen Land
by some other name; and how to bank them so they do not silt, but
scour themselves. There are these things and a thousand others. All
are immemorial, but I have no space for any now.
 ON MILTON
The letters of a people reflect its noblest as architecture reflects its
most intimate mind and as its religion (if it has a separate or tribal
religion) reflects its military capacity or incapacity. The word
"noblest" is vague, and nobility must here be defined to mean that
steadiness in the soul by which it is able to express a fixed character
and individuality of its own. Thus a man contradicts himself from
passion or from a variety of experience or from the very ambiguity
and limitation of words, but he himself resides in all he says, and
when this self is clearly and poisedly expressed it is then that we
find him noble.
The poet Milton, according to this conception, has best expressed
the nobility of the English mind, and in doing a work quite different
from any of his peers has marked a sort of standard from which the
ideal of English letters does not depart.
Two things are remarkable with regard to English literature, first that
it came late into the field of European culture, and secondly that it
has proved extraordinarily diversified. The first point is immaterial to
my subject; the second is material to it; for it might be superficially
imagined that such bewildering complexity and, as it were, lawless
exuberance of method and of matter would never find a pole, nor
ever be symbolised by but one aspect of it. Yet Milton has found that
pole, and Milton's work has afforded that symbol.
In any one moment of English literary history you may contrast two
wholly different masterpieces from the end of the Fourteenth to the
end of the Eighteenth Centuries. After the first third of the
Nineteenth, indeed, first-rate work falls into much more
commonplace groove, and it is perceptible that the best verse and
the best prose written in English are narrowing in their vocabulary,
and, in what is far more important, their way of looking at life. The
newspapers have levelled the writers down as with a trowel; you
have not side by side the coarse and the refined, the amazing and
the steadfast, the grotesque and the terrible; but in all those earlier
centuries you had side by side manner and thought so varied that a
remote posterity will wonder how such a wealth could have arisen
upon so small an area of national soil. Piers Plowman and the
Canterbury Tales are two worlds, and a third world separate from
each is the world of those lovely lyrics which are now so nearly
forgotten, but which the populace, spontaneously engendered and
sang throughout the close of the Middle Ages. The Sixteenth Century
was perhaps less modulated, and flowed, especially towards its end,
in one simpler stream, but in the Seventeenth what a growth of
variety from the Jacobean translation of the Bible to Swift. The very
decade in which Paradise Lost was published corresponded with the
first riot of the Restoration.
If we look closely into all this diversity we can find two common
qualities which mark out all English work in a particular manner from
the work of other nations. To qualities of this kind, which are like
colours rather than like measurable things, it is difficult to give a
title; I will hazard, however, these two words, "Adventure" and
"Mystery." There is no English work of any period, especially is there
no English work of any period later than the middle of the Sixteenth
Century, which has not got in it all those emotions which proceed
from the love of Adventure. How notable it is, for instance, that
Landscape appears and reappears in every diverse form of English
verse. Even in Shakespeare you have it now and then as vivid as a
little snapshot, and it runs unceasingly through every current of the
stream; it glows in Gray's Elegy, and it is the binding element of In
Memoriam. It saves the earlier work of Wordsworth, it permeates
the large effect of Byron, and those two poems, which to-day no one
reads, Thalaba and The Curse of Kehama, are alive with it. It is the
very inspiration of Keats and of Coleridge. Now this hunger for
Landscape and this vivid sense of it are but aspects of Adventure;
for the men who thus feel and speak are the men who, desiring to
travel to unknown places, are in a mood for sudden revelations of
sea and land. So a living poet has written—

When all the holy primal part of me

Arises up within me to salute
The glorious vision of the earth and sea
That are the kindred of the destitute....

The note of those four lines is the note of Landscape in English

letters, and that note is the best proof and effect of Adventure. If
any man is too poor to travel (though I cannot imagine any man so
poor), or if he is constrained from travel by the unhappy necessities
of a slavish life, he can always escape through the door of English
letters. Let such a one read the third and fourth books of Paradise
Lost before he falls asleep and he will find next morning that he has
gone on a great journey. Milton by his perpetual and ecstatic delight
in these visions of the world was the normal and the central example
of an English poet.

As when far off at sea a fleet descri'd

Hangs in the clouds....

or, again,

.... Hesperus, that led

The Starry Host, rode brightest 'til the Moon,
Rising in cloudy majesty, at length
Apparent Queen, unveiled....

He everywhere, and in a profusion that is, as it were, rebellious

against his strict discipline of words, sees and expresses the picture
of this world.
If Landscape be the best test of this quality of adventure in English
poets and the Milton as their standard, so the mystic character of
English verse appears in them and in him. No period could be so
formal as to stifle or even to hide this demand of English writers for
Mystery and for emotions communicable only by an art allied to
music. The passion is so strong that many ill-acquainted with foreign
literature will deny such literature any poetic quality because they do
not find in it the unmistakable thrill which the English reader
demands of a poet as he demands it of a musician. As Landscape
might be taken for the best test of Adventure, so of this appetite for
the Mysterious the best measurable test is rhythm. Highly
accentuated rhythm and emphasis are the marks and the
concomitants of that spirit. As powerful a line as any in the language
for suddenly evoking intense feeling by no perceptible artifice is that
line in Lycidas—

Smooth-sliding Mincius, crowned with vocal reeds.

I confess I can never read that line but I remember a certain river of
twenty years ago, nor does revisiting that stream and seeing it again
with my eyes so powerfully recall what once it was to those who
loved it as does this deathless line. It seems as though the magical
power of the poet escaped the effect of time in a way that the
senses cannot, and a man curious in such matters might find the
existence of such gifts to be a proof of human immortality. The pace
at which Milton rides his verse, the strong constraint within which he
binds it, deeply accentuate this power of rhythm and the mystical
effect it bears. Now you would say a trumpet, now a chorus of
human voices, now a flute, now a single distant song. From the
fortieth to the fifty-fifth line of the third book Paradise Lost has all
the power and nature of a solemn chant; the large complaint in it is
the complaint of an organ, and one may say indeed in this
connection that only one thing is lacking in all the tones Milton
commanded; he disdained intensity of grief as most artists will
disdain intensity of terror. But whereas intensity of terror is no fit
subject for man's pen, and has appealed only to the dirtier of our
little modern fellows, intense grief has been from the very beginning
thought a just subject for verse.

Τἡλε δ' ἁπὁ κρατὁς χἑε δἑσματα σιγαλὁεντα

Αμπυκα κεκρὑφαλὁν τ', ἡδἑ πλεκτἡν ἁναδἑσμην
Κρἡδεμνὁν Θ', δ ῥἁ οἱ δὡκε χρυσἑη 'Αφροδἱτη
Ηματι τὩ, ὁτε μιν κορυθαἱολος ἡγἁγεθ' Εκτωρ
'Εκ δὁμου 'Ηετἱωνος, ἑπεἱ πὁρε μυρἱα ἑδνα.

[Greek: Têle d' apo kratos chee desmata sigaloenta

Ampyka kekryphalon t', êde plektên anadesmên
Krêdemnon th', ho rha hoi dôke chryseê Aphroditê
Hêmati tôi, hote min korythaiolos êgageth' Hektôr
Ek domou Êetiônos, epei pore myria hedna.]

Milton will have none of it. It is the absence of that note which has
made so many hesitate before the glorious achievement of Lycidas,
and in this passage which I quote, where Milton comes nearest to
the cry of sorrow, it is still no more than what I have called it, a
solemn chant.
 .... Thus with the year
Seasons return; but not to me returns
Day, or the sweet approach of Ev'n or Morn,
Or sign of vernal bloom, or Summer's Rose,
Or flocks, or herds, or human face divine;
But cloud instead, and ever-during dark
Surrounds me, from the chearful waies of men
Cut off, and, for the Book of knowledge fair,
Presented with a Universal blanc
Of Nature's works, to mee expung'd and ras'd,
And wisdome at one entrance quite shut out.
So much the rather thou, Celestial light,
Shine inward, and the mind through all her powers
Irradiate; there plant eyes, all mist from thence
Purge and disperse, that I may see and tell
Of things invisible to mortal sight.

There is one other character in Milton wherein he stands not so

much for English Letters as for a feature in English nature as a
whole, which is a sort of standing apart of the individual. Where this
may be good and where evil it is not for a short appreciation to
discuss. It is profoundly national and nowhere will you see it more
powerfully than in the verse of this man. Of his life we all know it to
be true, but I say it appears even in his verse. There is a sort of noli
me tangere in it all as though he desired but little friendship and was
not broken by one broken love, and contemplated God and the fate
of his own soul in a lonely manner; of all the things he drew the
thing he could never draw was a collectivity.
 HANS CHRISTIAN ANDERSEN
What a great thing it is in this perplexed, confused, and, if not
unhappy at least unrestful time, to come across a thing which is
cleanly itself! What a pleasure it is amid our entwining controversies
to find straightness, and among our confused noises a chord. Hans
Christian Andersen is a good type of that simplicity; and his own
generation recognised him at once; now, when those
contemporaries who knew him best are for the most part dead, their
recognition is justified. Of men for whom so much and more is said
by their contemporaries, how many can stand the test which his
good work now stands, and stands with a sort of sober triumph?
Contemporary praise has a way of gathering dross. We all know why.
There is the fear of this, the respect for that; there is the genuine
unconscious attachment to a hundred unworthy and ephemeral
things; there is the chance philosophy of the moment over-weighing
the praise-giver. In a word, perhaps not half-a-dozen of the great
men who wrote in the generation before our own would properly
stand this test of a neat and unfringed tradition. It is not to be
pretended that according to that test so must men be judged. Many
of the very greatest, Hugo for instance, and in his line, Huxley (a
master of English); or, again, to go further back, the great Byron,
would not pass the test.
Things have been said about most men, great or little, in our fevered
time, so exaggerated, so local and so lacking balance, whether of
experience or of the fear of posterity, that contemporary opinion
should not be allowed by its misfortunes to weigh them down. But a
man has a quality of his own when he is so made that even his
contemporaries do him justice, and that was the case with Hans
Christian Andersen. I will bargain that if our letters survive five
hundred years, this excellent writer will quietly survive. Even the
French may incorporate him. And next it is the business of one who
praises so much to ask in what the excellence of this writer consists.
It is threefold: in the first place, he always said what he thought; in
the second place, he was full of all sorts of ways of saying it; and, in
the third place, he said only what he had to say.
To say what one thinks, that is, to tell the truth, is so exceedingly
rare that one may almost call it a grace in a man. Just those same
manifold strings which pull contemporary criticism hither and thither,
and which have made me suggest above that contemporary criticism
commonly belittles a man in the long run, just those same strings
pull at every writer to make him conform to what he knows to be
false in his time. But some men—with limitations, it is true, and only
by choosing a particular framework—manage to tell the truth all
their lives; those men, if they have other literary qualities, are secure
of the future.
And this leads me to the second point, which is that Andersen could
not only tell the truth but tell it in twenty different ways, and of a
hundred different things. Now this character has been much
exaggerated among literary men in importance, because literary
men, perceiving it to be the differentiation which marks out the
great writer from the little, think it to be the main criterion of letters.
It is not the main criterion; but it is a permanent necessity in great
writing. There is no great writing without this multiplicity, which is
sometimes called imagination, sometimes experience, and
sometimes judgment, but which is in its essence a proper survey of
the innumerable world. This quality it is which makes the great
writers create what are called "characters"; and whether we
recognise those "characters" as portraits drawn from the real world
(they are such in Balzac), or as figments (they are such in Dickens),
or as heroines and heroes (they are such in Shakespeare and in
Homer, if you will excuse me), yet that they exist and live in the
pages of the writer means that he had in him that quality of
contemplation which corresponds in our limited human nature to the
creative power.
Lastly, I say that Andersen said what he had to say and no more.
This quality in writers is not restraint—a futile word dear to those
who cannot write—it is rather a sort of chastity in the pen. The
writer of this kind is one who unconsciously does not add; if any one
were to ask him why he should not add an ornament or anything
supposititious, he would be bewildered and perhaps might answer:
"Why should I?" The instinct behind it is that which produces all
terseness, all exactitude, and all economy in style.
Andersen, then, had all those three things which make a great
writer, and a very great writer he is.
Note that he chose his framework, or, at any rate, that he was
persuaded to it. He could not have been so complete had he not
addressed himself to children, and it is his glory that he is read in
childhood. There is no child but can read Hans Christian Andersen,
and I at least have come across no man who, having read him in
childhood, does not continue to read him throughout life. He wrote
nothing that was not for the enlivening or the sustenance or the
guiding of the human soul; he wrote nothing that suggested
questions only. If one may speak of him in terms a trifle antiquated
(or rather for the moment old-fashioned), he was instinct with
charity, and therefore he is still full of life.
Having said so much of Andersen in general, something should be
said of him in particular. He was Northern; you always feel as you
read him that if his scene is laid in the open air, the air is fresh and
often frosty; that if he is talking indoors the room is cosy and often
old. Certain passions which the North lacks are lacking in him, both
upon their good and upon their evil side. He is never soldierly, and
he is never revengeful; he is never acute with the desire for life, but,
again, he is never envious. Those who read him and who are also
Northern may well be in love with Denmark. It is a triumph of our
civilisation that this little land, quite outside the limits of the Roman
Empire, not riven by any of the Empires' great vital resurrections,
undisturbed by the vision of the Twelfth and of the Thirteenth
Centuries, spared from the march of Napoleon's armies, should be
so completely European. What could be more European to-day than
that well-organised, contented, peasant State? It is a good irony to
put against the blundering prophecies of barbaric people that
beyond the Germanies this secure and happy State exists. One
might put it in a phrase a little too epigrammatic and say that as one
reads Hans Christian Andersen one remembers Elsinore, and one
recalls the good architecture of Copenhagen. If ever any misfortune
again shall threaten that State, and if barbarism attempts to play the
fool with it, something that really is the conscience of Europe and
not the empty and sham organisation to which that phrase is too
often prostituted will arise and protect the Danes.
 THE CHRISTMAS OF 1808
No British Army in force has capitulated in Europe for many
generations. It is the peculiar historical position of this country. That
historical fact lends to the common history of the schools and
universities an attitude towards military history in general which is
commonly distorted, but it lends to the policy of the country as a
whole a confident tradition, the strength and value of which it is
impossible to exaggerate.
The nearest touch to such a disaster, if we except the sieges, was
passed during the days in which these words are written and read;
the close thing came about in the days just before and just after
Christmas, one hundred and two years ago. I will attempt to
describe as simply as I can the nature of that adventure.
It must first of all be premised that, in the words of Napier, position
determines the fate of armies. No truth is more apparent to the
soldier, none more forgotten by the civilian—and more especially by
the civilian touched with the unmilitary vice of Jingoism. Position
determines the fate of armies, and, armaments being supposedly
equal, he is a great or a fortunate general who, in the critical
moment, has so arranged matters that disposition is upon his side,
or who by some stroke of luck is in that advantage. There are
exceptions to this truth. Certain decisive battles (though very few)
have utterly determined campaigns; and among these battles some,
again, have been won at a drive, and by a sort of impetus, the factor
of position being so simple as to be negligible, or so equally
balanced as to advantage neither side and be eliminated. But, as a
rule, it is true even of decisive actions, that position is the
determining factor. It is necessarily true of the strategy of a
campaign, and it is with this consideration that I return to the
particular crisis of the British Army at the close of December 1808.
Sir John Moore, as every one knows, had raided right into the North
of Spain, with the object of withdrawing the pressure of the French
upon the South of that country. It was in the South that French
ambition had found its first check, and that Napoleon's plan had
been warped by the unexpected and, as it were, impossible
capitulation of Baylen. Close upon twenty thousand of the French
forces had there laid down their arms. The Emperor came in person
to restore the fortunes of his house; it was in the South that
resistance could best be expected; by the occupation of the South
that he might put himself at ease over the whole territory, and from
the South that the English operations were destined to draw him.
On the 21st of December, a Wednesday, Lord Paget, with the Tenth
and Fifteenth Hussars, surprised an advanced body of French
Cavalry at Sahagun. It was the extreme limit of Moore's great raid;
the town was occupied, and all the Thursday, all the Friday, Moore
halted there with his force of some twenty-three thousand and sixty
guns. He was nearly two hundred miles from the port on the sea-
coast, whereto he must retire if he would escape. In front of him
was Soult, against whom it was his business, if he were undisturbed,
to march from Sahagun immediately; but upon his right, nearly as
far off as the sea, though not quite so far, a matter of a hundred and
fifty to a hundred and seventy miles, Napoleon, at Madrid,
commanded the best and the largest of the armies in Spain. Sixty
thousand men, with a hundred and fifty guns, lay at the gates of
Madrid, and during those same hours in which the British Army had
marched into Sahagun, Napoleon's great force began to move
northward over the Guadarrama.
I will not here describe that famous march: I have done so
elsewhere at greater length: but the reader, to appreciate the
conditions of this great duel, must imagine a country denuded and
largely mountainous, deep in snow, and subject throughout those
days to intolerable weather; and the race upon the issue of which
depended so many and such final things was run at a time and in a
place when one would have thought that no man could be abroad.
But the protagonists of the Revolutionary wars were not men like
ourselves.
Christmas Day fell upon the Sunday. Moore had got ahead of his
supplies; they had reached him on the Friday, and on the Saturday,
Christmas Eve, he had intended to go forward and attack the
opponent before him. But on that same Friday when, in the night,
his Infantry were already beginning to march eastward, he heard of
Napoleon's amazing feat; he knew that he had succeeded in drawing
the great commander northward, but he knew also, since that
commander could work miracles, that the distance separating them
would be crossed with a swiftness not to be measured by the old
rules of war, and that the vast force three times his own would, if he
hesitated, be found holding the snow-blocked roads between his
position and the sea. The order to advance was cancelled, the order
to retreat was given. By Christmas Eve Baird and Hope were on the
line of the Esla River; on Christmas Day, Sunday, the troops were
passing that obstacle. On Monday, the 26th, the baggage and the
last of the army, under Moore's own eye, were crossing by the
bridge of Castro Gonzalo before Benevente, and the trick was done.
There was a thick fog, the passage was far slower than the strained
intelligence of the imperilled commander had designed. On that
same day, the 26th, Napoleon was at Tordesillas, one long day's
march away from the Esla River. He had covered in that dash of
three days and a half a hundred and twenty miles, but he was too
late. He was too late by half a day.
In the dark and storm-driven night of that Monday the extreme van
of Napoleon's horse rode up to the bridge of Castro Gonzalo. They
were unsupported, of course, and rode far before the army to
discover; but, though it was not contact in any serious sense, there
is something very worthily dramatic in the appearance of those tall
horsemen suddenly in the night through the blinding snow, come up
just too late to do more than watch the escape of Moore's column.
By the next day the purpose of the British commander was
achieved: Napoleon knew he could no longer intercept: the bridge
was destroyed. The opportunity of recording the envelopment and
destruction of a British force was lost to Napoleon; he abandoned to
Soult the further long pursuit, which is called in history the retreat
upon Corunna.
 ON COMMUNICATIONS
There is nothing more curious in the material change which is
passing so rapidly over the modern world than what I may call the
Romance of Communication.
With the Romance of Discovery every one is thoroughly acquainted.
The modern world is saturated with that form of romance; it has
permeated all our literature and is still the theme of most of our
books of travel. But like all things which have attained a literary
position, the Romance of Discovery already belongs to the past. Not
that nothing remains to be discovered: on the contrary, the modern
world has hardly yet begun to appreciate how it may penetrate from
detail to detail and find perpetually something new in that which it
thinks it knows, but the great broad unknown spaces, the horizons
quite new to Europeans which break upon them for the first time,
are now no longer left to the explorer. With the romance of
communication, luckily for us, there is another, a newer and, in a
certain sense, a much wider field. Many who have travelled largely
have felt this, but it has not yet, I think, been expressed.
What I mean by the Romance of Communication is this: that the
establishment of regular lines for ocean traffic, the building of
railways and, above all, of good roads, have made it possible for a
multitude of men to see those contrasts which travel can afford, and
this development of modern travel has just begun to afford our
generation, and will afford with much greater generosity the
generation to come, an opportunity for feeling physically the
complexity and variety and wonder of the world. This is a good
thing.
Not so long ago it was a difficult matter for a man to go from some
Northern part of Europe, such as England, to so isolated a
community as that which inhabits the Island of Majorca. Now it is
easy for a man and costs him but a few pounds to go from England
to Barcelona, and from Barcelona he can sail with a rapid and
regular service to the port of Palma. When he reaches that port he
cannot but feel the Romance, finding this little isolated State wealthy
and contented in the midst of the sea. Corsica, of which men know
so little, is similarly at hand to-day, and so are the Valleys of the
Pyrenees, especially of the Spanish valleys upon which as yet there
is hardly any Northern literature or experience. In a year or two we
shall have the railway through the Cerdagne, and another line will
take one up the Valley of the Ariège into the middle part of Northern
Spain.
But of all these benefits to the mind which the modern charge is
procuring, I know of none more remarkable than the entry into the
Desert.
That portion of Northern Africa which the French have reclaimed for
Europe, and which was throughout the existence of the Roman
Empire an integral part of European civilisation, consists of the great
tableland buttressed to the north and the south by mountain ranges,
and crossed in its middle part by parallel outcrops of high rock. This
plateau stretches for somewhat more than a thousand miles all
along the southern shore of the Western Mediterranean. If the
reader will take a map he will see jutting out from the general
contour of Africa, an oblong as it were, the eastern end of which is
Tunis, the western end Morocco. All that oblong is the tableland of
which I speak. The coast is warm, fertile, densely cultivated and
populous; full of ports and cities and the coming and going of ships.
The highlands behind and to the south of the coast line are more
arid, very cold in winter, baking in summer, and always dry and
rugged to our Northern eyes. But they are habitable, the population
is spreading upon them, and they contain the past relics of the old
Roman civilisation which prove what man can do with them when
their water supply is stored and their soil is cultivated.
Now this habitable land suddenly ceases, and falls into the Desert of
Sahara. The demarkation is abrupt and is everywhere noticeable to
the eye. It is indeed more noticeable in the eastern than in the
western part. The limit between what Miss Bell has called, in a fine
book of hers, "the Desert and the Sown," is more than a day's march
in width upon the Moroccan frontier; indeed it is several days'
march, and one is not over-sure when one has left the habitable soil
and when one has reached the inhospitable sand at the eastern end.
The limits are not only marked sharply by a differentiation in the
climate and the vegetation, but also by an abrupt escarpment. The
Atlas (as the plateau of Northern Africa is generally called) falls in
huge, precipitous red cliffs right down upon the Sahara. It so
happens that these cliffs, just at the point where they are most
abrupt and most rugged, and most romantic, are cleft by a profound
gorge through which the Wady Biskra runs very clear and cold, filled
with the melting of the snow upon the high mountain of Aures to the
north of it. This gap in the cliffs the Romans knew well. They had a
military station here to guard them against the ravages of the
nomad tribes who afterwards, in the form of the Arabian Invasion,
overran their African province and turned it from an European and a
Christian to an Asiatic and Mohammedan thing. The Romans called
that gorge "The Kick of Hercules," as though the god had here by a
stroke of his foot broken away from the cultivated north toward the
Desert. Through this gap ran their military road, and here, as a
formation of the gorge demanded, they carried that road over the
river, the Wady Biskra, upon the bridge the stones of which still
remain, though renovated and supported by modern work, to recall
the greatness of the empire.
The Arabs, in their turn, have called this astonishing breach "Foum
es Sahara"—the mouth of the Sahara; and, as is always the case
where they found a Roman bridge, they have added the name El
Kantara, the bridge. For it is remarkable that the Arabs were unable
to continue Roman work, especially in masonry, save where they had
a large Roman population to help them after their conquest, and the
bridges which the Romans had built were regarded by them with a
sort of superstitious reverence.
Now this Mouth of the Sahara, this gap in the glaring wall of the
Desert, has, by a coincidence which has its obvious geographical
cause, and which is to be discovered in many another pass
throughout Europe and Northern Africa, served for modern methods
of communication the purpose which it served for the ancients. It is
the nearest approach which the Desert makes to the sea-coast; it is
the approach involving the least engineering effort, the most obvious
and the most natural entry from the northern cultivated land on to
the waterless sandy waste. Therefore, modern civilisation has used
it, and you get here more than anywhere else that romance of
sudden contrast with which, as I have said, modern methods of
travel have gifted the modern world. The French first built down this
track a military road, as hard, excellent and well graded as any that
you will find in Europe. It not only goes through the gorge, but right
on into the sand of the desert, bounded upon either side by
masonry, and it has reached, or very soon will reach, Biskra without
a break.
Some time after this road had been planned, a railway was
constructed along the same track, with certain divergences where
the gradient of the highway was too steep for the rails and where
therefore long curves were necessary. Whether a man goes by the
road or the rail, this is what he sees—and he had best see it in early
spring, or what is with us in England late winter. As the road and the
rail wind downwards from the little plateaux, by great steps as it
were from one level to another, the traveller has about him such
scenery as has accompanied him for the last hundred miles: fields of
cotton, the trees proper to a temperate climate, and rugged, rocky
ridges cropping up from the cultivated soil. There is nothing around
him to remind him of what is called "The East," except the camel
preceding its master up or down the great highway and the
distinctive dress of the natives; for the climate, the crops, and the
temperature, the quality of the sunlight, he might be on one of the
plains of Western America, which indeed this part of Africa most
nearly resembles.
He comes to a clean little inn entirely French in architecture,
surrounded by a cool and quiet garden, and with the river running
behind it. He walks on a few hundred yards through the gorge, and
quite suddenly at the turning of a corner the Desert and all its
horizon breaks upon his eyes. He sees a waste of hot, red, unusable
sand, a brilliant oasis of palm trees, and even the sun, small and
glaring above that plain, seems something different from the familiar
light which he had received but an hour before.
It is the most complete contrast, the most sudden and memorable
revelation which modern travel affords. And if I had to advise any
one who with short leisure desired some experience of modern
travel, at least in the way of landscape, I would advise him to visit
this astounding place. It has already found its way into many English
books, but the great mass of people who could enjoy it do not yet
know how exceedingly easy is its access. For travellers even to so
near a place like to put on an air of mystery. There is no one with a
fortnight to spare and £20 to spend who cannot walk or bicycle or
motor to this place at the right time of the year (for in summer the
heat is insupportable and in winter the snow on the high northern
plains makes travel difficult). Any one who will take that journey
would have a memory to last him all his life.
There are those who say that the popularisation of wonderful things
is the spoiling of them. I have never been able to agree. Places are
not spoiled by the multitude of those who reach them, but by the
character of those who reach them, and no one will journey to this
meeting place of the East and the West, I think, save with a desire
to wonder and to observe, which of itself breeds reverence. Such
men may come in any numbers and the place will be the better for
them. At any rate, I repeat, it ought to be known to any one who
has the sum or the time to spare, that a revelation of such a kind is
quite close at hand, for as yet hundreds of Englishmen with far more
than that leisure and infinitely more than that wealth are ignorant of
the place and its opportunities; and your quickest way is by
Marseilles and Bona, and on your way I beseech you to stop and see
Timgad, which is a dead Roman city lying silent and empty under
the sun upon the Edge of the Desert.
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!

ebookgate.com