Download the full version of the textbook now at textbookfull.

com

A Basic Course in Probability Theory 2nd

Edition Rabi Bhattacharya

https://textbookfull.com/product/a-basic-course-
in-probability-theory-2nd-edition-rabi-
bhattacharya/

Explore and download more textbook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Rabi N Bhattacharya Selected Papers 1st Edition Manfred

Denker

https://textbookfull.com/product/rabi-n-bhattacharya-selected-
papers-1st-edition-manfred-denker/

textbookfull.com

Golosa A Basic Course in Russian Book One 6th Edition

Robin

https://textbookfull.com/product/golosa-a-basic-course-in-russian-
book-one-6th-edition-robin/

textbookfull.com

Foundations of the Theory of Probability 2nd Edition A.N.

Kolmogorov

https://textbookfull.com/product/foundations-of-the-theory-of-
probability-2nd-edition-a-n-kolmogorov/

textbookfull.com

Craft Coffee A Manual Brewing a Better Cup at Home First

Edition Jessica Easto

https://textbookfull.com/product/craft-coffee-a-manual-brewing-a-
better-cup-at-home-first-edition-jessica-easto/

textbookfull.com
Rehearsing Shakespeare Ways of Approaching Shakespeare in
Practice for Actors Directors and Trainers 1st Edition
Leon Rubin
https://textbookfull.com/product/rehearsing-shakespeare-ways-of-
approaching-shakespeare-in-practice-for-actors-directors-and-
trainers-1st-edition-leon-rubin/
textbookfull.com

Sexual Crime and Trauma Helen Swaby

https://textbookfull.com/product/sexual-crime-and-trauma-helen-swaby/

textbookfull.com

Physics of Semiconductor Devices 17th International

Workshop on the Physics of Semiconductor Devices 2013 1st
Edition A. Charris-Hernandez
https://textbookfull.com/product/physics-of-semiconductor-
devices-17th-international-workshop-on-the-physics-of-semiconductor-
devices-2013-1st-edition-a-charris-hernandez/
textbookfull.com

CRITICAL THINKING : understanding and evaluating dental

research. Donald Maxwell Brunette

https://textbookfull.com/product/critical-thinking-understanding-and-
evaluating-dental-research-donald-maxwell-brunette/

textbookfull.com

Guide to the English School in International Studies 1st

Edition Cornelia Navari

https://textbookfull.com/product/guide-to-the-english-school-in-
international-studies-1st-edition-cornelia-navari/

textbookfull.com
Handbook of Biofuels Production Processes and Technologies
2nd Edition Rafael Luque

https://textbookfull.com/product/handbook-of-biofuels-production-
processes-and-technologies-2nd-edition-rafael-luque/

textbookfull.com
Universitext

Rabi Bhattacharya
Edward C. Waymire

A Basic Course
in Probability
Theory
Second Edition
Universitext
Universitext

Series editors
Sheldon Axler
San Francisco State University, San Francisco, CA, USA

Vincenzo Capasso
Università degli Studi di Milano, Milan, Italy

Carles Casacuberta
Universitat de Barcelona, Barcelona, Spain

Angus MacIntyre
Queen Mary University of London, London, UK

Kenneth Ribet
University of California, Berkeley, CA, USA

Claude Sabbah
École Polytechnique, CNRS, Université Paris-Saclay, Palaiseau, France

Endre Süli
University of Oxford, Oxford, UK

Wojbor A. Woyczyński
Case Western Reserve University, Cleveland, OH, USA

Universitext is a series of textbooks that presents material from a wide variety of

mathematical disciplines at master’s level and beyond. The books, often well
class-tested by their author, may have an informal, personal, even experimental
approach to their subject matter. Some of the most successful and established books
in the series have evolved through several editions, always following the evolution
of teaching curricula, into very polished texts.
Thus as research topics trickle down into graduate-level teaching, first textbooks
written for new, cutting-edge courses may make their way into Universitext.

More information about this series at http://www.springer.com/series/223

Rabi Bhattacharya Edward C. Waymire
•

A Basic Course in Probability

Theory
Second Edition

123
Rabi Bhattacharya Edward C. Waymire
Department of Mathematics Department of Mathematics
University of Arizona Oregon State Univeristy
Tucson, AZ Corvallis, OR
USA USA

ISSN 0172-5939 ISSN 2191-6675 (electronic)

Universitext
ISBN 978-3-319-47972-9 ISBN 978-3-319-47974-3 (eBook)
DOI 10.1007/978-3-319-47974-3
Library of Congress Control Number: 2016955325

Mathematics Subject Classification (2010): 60-xx, 60Jxx

1st edition: © Springer Science+Business Media, Inc. 2007

2nd edition: © Springer International Publishing AG 2016
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface to Second Edition

This second edition continues to serve primarily as a text for a lively two-quarter or
one-semester course in probability theory for students from diverse disciplines,
including mathematics and statistics. Exercises have been added and reorganized
(i) for reinforcement in the use of techniques, and (ii) to complement some results.
Sufficient material has been added so that in its entirety the book may also be used
for a two-semester course in basic probability theory. The authors have reorganized
material to make Chapters III–XIII as self-contained as possible. This will aid
instructors of one semester (or two quarter) courses in picking and choosing some
material, while omitting some other. Material from a former chapter on Laplace
transforms has been redistributed to parts of the text where it is used.
The early introduction of conditional expectation and conditional probability
maintains the pedagogic innovation of the first edition. This enables the student to
quickly move to the fundamentally important notions of modern probability besides
independence, namely, martingale dependence and Markov dependence, where
new theory and examples have been added to the text. The former includes Doob’s
upcrossing inequality, the submartingale convergence theorem, and reverse
martingales, and the (reverse) martingale proof of the strong law of large numbers,
while retaining important earlier approaches such as those of Kolmogorov,
Etemadi, and of Marcinkiewicz–Zygmund.
A theorem of Polya is added to the chapter on weak convergence to show that
the convergence to the normal distribution function in the central limit theorem is
uniform.
The Cramér–Chernoff large deviation theory in Chapter V is sharpened by the
addition of a large deviation theorem of Bahadur and Ranga Rao using the Berry–
Esseen convergence rate in the central limit theorem. Also added in Chapter V is a
concentration of measure type inequality due to Hoeffding. The proof of the
aforementioned Berry–Esseen bound is deferred to Chapter VI on Fourier series
and Fourier transform. The Chung–Fuchs transience/recurrence criteria for ran-
dom walk based on Fourier analysis is a new addition to the text.
Special examples of Markov processes such as Brownian motion, and random
walks appear throughout the text to illustrate applications of (i) martingale theory

v
vi Preface to Second Edition

and stopping times in computations of certain important probabilities.

A culmination of the theory developed in the text occurs in Chapters XI and XII on
Brownian motion. This continues to rank among the primary goals attainable for a
course based on the text.
General Markov dependent sequences and their convergence to equilibrium is
the subject matter of the entirely new Chapter XIII. Illustrative examples are pro-
vided, including some of historical importance to the development of the kinetic
theory of matter in physics due to Boltzmann, Einstein, and Smoluchowski. The
treatment centers on describing a prototypical framework, namely Doeblin’s
theorem, for existence and convergence to a unique invariant probability for
Markov processes, together with illustrative examples for students with diverse
interests ranging from mathematics and statistics to contemporary mathematical
finance or biology. Examples include iterated random maps, the Ehrenfest model,
and products of random matrices. The Ornstein–Uhlenbeck process is shown to be
obtained as the unique solution to a stochastic differential equation, namely the
Langevin equation, using Picard iteration. This provides students with a glimpse
into the broad scope and utility of the probability that they have learned, while
motivating continued study of stochastic processes.
Complete references to authors of books cited in footnotes are provided in a
closing list of references. This also includes other textbook resources covering the
same topics and/or further applications.
The authors are grateful to William Faris, University of Arizona, and to Enrique
Thomann, Oregon State University, for providing comments and corrections to an
earlier draft based on their teaching of the course. Partial support from the National
Science Foundation under grants DMS 1406872 and DMS1408947, respectively, is
gratefully acknowledged by the authors.

Tucson, AZ, USA Rabi Bhattacharya

Corvallis, OR, USA Edward C. Waymire
September 2016
Preface to First Edition

In 1937, A.N. Kolmogorov introduced a measure-theoretic mathematical frame-

work for probability theory in response to David Hilbert’s Sixth Problem. This text
provides the basic elements of probability within this framework. It may be used for
a one-semester course in probability, or as a reference to prerequisite material in a
course on stochastic processes. Our pedagogical view is that the subsequent
applications to stochastic processes provide a continued opportunity to motivate
and reinforce these important mathematical foundations. The book is best suited for
students with some prior, or at least concurrent, exposure to measure theory and
analysis. But it also provides a fairly detailed overview, with proofs given in
appendices, of the measure theory and analysis used.
The selection of material presented in this text grew out of our effort to provide a
self-contained reference to foundational material that would facilitate a companion
treatise on stochastic processes that Theory and Applications of Stochastic
Processes we have been developing.1 While there are many excellent textbooks
available that provide the probability background for various continued studies of
stochastic processes, the present treatment was designed with this as an explicit
goal. This led to some unique features from the perspective of the ordering and
selection of material.
We begin with Chapter I on various measure-theoretic concepts and results
required for the proper mathematical formulation of a probability space, random
maps, distributions, and expected values. Standard results from measure theory are
motivated and explained with detailed proofs left to an appendix.
Chapter II is devoted to two of the most fundamental concepts in probability
theory: independence and conditional expectation (and/or conditional probability).
This continues to build upon, reinforce, and motivate basic ideas from real analysis
and measure theory that are regularly employed in probability theory, such as
Carathéodory constructions, the Radon–Nikodym theorem, and the Fubini–Tonelli

1
Bhattacharya, R. and E. Waymire (2007): Theory and Applications of Stochastic Processes,
Springer-Verlag, Graduate Texts in Mathematics.

vii
viii Preface to First Edition

theorem. A careful proof of the Markov property is given for discrete-parameter

random walks on ℝk to illustrate conditional probability calculations in some
generality.
Chapter III provides some basic elements of martingale theory that have evolved
to occupy a significant foundational role in probability theory. In particular,
optional stopping and maximal inequalities are cornerstone elements. This chapter
provides sufficient martingale background, for example, to take up a course in
stochastic differential equations developed in a chapter of our text on stochastic
processes. A more comprehensive treatment of martingale theory is deferred to
stochastic processes with further applications there as well.
The various laws of large numbers and elements of large deviation theory are
developed in Chapter IV. This includes the classical 0–1 laws of Kolmogorov and
Hewitt–Savage. Some emphasis is given to size-biasing in large deviation calcu-
lations which are of contemporary interest.
Chapter V analyzes in detail the topology of weak convergence of probabilities
defined on metric spaces, culminating in the notion of tightness and a proof of
Prohorov’s theorem.
The characteristic function is introduced in Chapter VI via a first principles
development of Fourier series and the Fourier transform. In addition to the oper-
ational calculus and inversion theorem, Herglotz’s theorem, Bochner’s theorem,
and the Cramér–Lévy continuity theorem are given. Probabilistic applications
include the Chung–Fuchs criterion for recurrence of random walks on ℝk, and the
classical central limit theorem for i.i.d. random vectors with finite second moments.
The law of rare events (i.e., Poisson approximation to binomial) is also included as
a simple illustration of the continuity theorem, although simple direct calculations
are also possible.
In Chapter VII, central limit theorems of Lindeberg and Lyapounov are derived.
Although there is some mention of stable and infinitely divisible laws, the full
treatment of infinite divisibility and Lévy–Khinchine representation is more prop-
erly deferred to a study of stochastic processes with independent increments.
The Laplace transform is developed in Chapter VIII with Karamata’s Tauberian
theorem as the main goal. This includes a heavy dose of exponential size-biasing
techniques to go from probabilistic considerations to general Radon measures. The
standard operational calculus for the Laplace transform is developed along the way.
Random series of independent summands are treated in Chapter IX. This
includes the mean square summability criterion and Kolmogorov’s three series
criteria based on Kolmogorov’s maximal inequality. An alternative proof to that
presented in Chapter IV for Kolmogorov’s strong law of large numbers is given,
together with the Marcinkiewicz and Zygmund extension, based on these criteria
and Kronecker’s lemma. The equivalence of a.s. convergence, convergence in
probability, and convergence in distribution for series of independent summands is
also included.
In Chapter X, Kolmogorov’s consistency conditions lead to the construction of
probability measures on the Cartesian product of infinitely many spaces. Applications
include a construction of Gaussian random fields and discrete-parameter Markov
Preface to First Edition ix

processes. The deficiency of Kolmogorov’s construction of a model for Brownian

motion is described, and the Lévy–Ciesielski “wavelet” construction is provided.
Basic properties of Brownian motion are taken up in Chapter XI. Included are
various rescalings and time inversion properties, together with the fine-scale
structure embodied in the law of the iterated logarithm for Brownian motion.
In Chapter XII many of the basic notions introduced in the text are tied together
via further considerations of Brownian motion. In particular, this chapter revisits
conditional probabilities in terms of the Markov and strong Markov properties for
Brownian motion, stopping times, and the optional stopping and/or sampling the-
orems for Brownian motion and related martingales, and leads to weak convergence
of rescaled random walks with finite second moments to Brownian motion, i.e.,
Donsker’s invariance principle or the functional central limit theorem, via the
Skorokhod embedding theorem.
The text is concluded with a historical overview, Chapter XIII, on Brownian
motion and its fundamental role in applications to physics, financial mathematics,
and partial differential equations, which inspired its creation.
Most of the material in this book has been used by us in graduate probability
courses taught at the University of Arizona, Indiana University, and Oregon State
University. The authors are grateful to Virginia Jones for superb word processing
skills that went into the preparation of this text. Also, two Oregon State University
graduate students, Jorge Ramirez and David Wing, did an outstanding job in
uncovering and reporting various bugs in earlier drafts of this text. Thanks go to the
editorial staff at Springer and anonymous referees for their insightful remarks.

March 2007 Rabi Bhattacharya

Edward C. Waymire

NOTE: Some of the first edition chapter numbers have changed in the second
edition. First edition Chapter VII was moved to Chapter IV, and the material in the
first edition Chapter VIII has been redistributed into other chapters. An entirely new
Chapter XIII was added to the second edition.
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
Contents

I Random Maps, Distribution, and Mathematical Expectation . . . . 1

Exercise Set I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
II Independence, Conditional Expectation . . . . . . . . . . . . . . . . . . . . . 25
Exercise Set II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
III Martingales and Stopping Times . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Exercise Set III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
IV Classical Central Limit Theorems. . . . . . . . . . . . . . . . . . . . . . . . . . 75
Exercise Set IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
V Classical Zero–One Laws, Laws of Large Numbers
and Large Deviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Exercise Set V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
VI Fourier Series, Fourier Transform, and Characteristic
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Exercise Set VI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
VII Weak Convergence of Probability Measures on Metric
Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Exercise Set VII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
VIII Random Series of Independent Summands . . . . . . . . . . . . . . . . . . 159
Exercise Set VIII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
IX Kolmogorov’s Extension Theorem and Brownian Motion . . . . . . 167
IX.1 A Wavelet Construction of Brownian Motion:
The Lévy–Ciesielski Construction . . . . . . . . . . . . . . . . . . . . . . 173
Exercise Set IX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
X Brownian Motion: The LIL and Some Fine-Scale Properties . . . . 179
Exercise Set X. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

xi
xii Contents

XI Strong Markov Property, Skorokhod Embedding,

and Donsker’s Invariance Principle . . . . . . . . . . . . . . . . . . . . . . . . 187
Exercise Set XI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
XII A Historical Note on Brownian Motion . . . . . . . . . . . . . . . . . . . . . 207
XIII Some Elements of the Theory of Markov Processes
and Their Convergence to Equilibrium . . . . . . . . . . . . . . . . . . . . . 211
Exercise Set XIII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Appendix A: Measure and Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Appendix B: Topology and Function Spaces. . . . . . . . . . . . . . . . . . . . . . . 241
Appendix C: Hilbert Spaces and Applications in Measure Theory . . . . . 247
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
Symbol Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
Chapter I
Random Maps, Distribution,
and Mathematical Expectation

In the spirit of a refresher, we begin with an overview of the measure–theoretic

framework for probability. Readers for whom this is entirely new material may
wish to consult the appendices for statements and proofs of basic theorems from
analysis. A measure space is a triple (S, S, μ), where S is a nonempty set; S is a
collection of subsets of S, referred to as a σ-field, which includes ∅ and is closed
under complements and countable unions; and ∞μ : S → [0, ∞] satisfies (i) μ(∅) = 0,
(ii) (countable additivity) μ(∪∞ n=1 A n ) = n=1 μ(An ) if A1 , A2 , . . . is a sequence
of disjoint sets in S. Subsets of S belonging to S are called measurable sets. The
pair (S, S) is referred to as a measurable space, and the set function μ is called a
measure. Familiar examples from real analysis are Lebesgue measure μ on S = Rk ,
equipped with a σ-field S containing the class of all k-dimensional  rectangles, say
R = (a1 , b1 ] × · · · × (ak , bk ], of “volume” measure μ(R) = kj=1 (b j − a j ); or
Dirac point mass measure μ = δx at x ∈ S defined by δx (B) = 1 if x ∈ B,
δx (B) = 0 if x ∈ B c , for B ∈ S. Such examples should suffice for the present,
but see Appendix A for constructions of these and related measures based on the
Carathéodory extension theorem. If μ(S) < ∞ then μ is referred to as a finite
measure. If one may write S = ∪∞ n=1 Sn , where each Sn ∈ S(n ≥ 1) and μ(Sn ) <
∞, ∀n, then μ is said to be a σ-finite measure.
A probability space is a triple (Ω, F, P), where Ω is a nonempty set, F is a
σ-field of subsets of Ω, and P is a finite measure on the measurable space (Ω, F) with
P(Ω) = 1. The measure P is referred to as a probability. Intuitively, Ω represents
the set of all possible “outcomes” of a random experiment, real or conceptual, for
some given coding of the results of the experiment. The set Ω is referred to as the
sample space and the elements ω ∈ Ω as sample points or possible outcomes. The
σ-field F comprises “events” A ⊂ Ω whose probability P(A) of occurrence is well
defined.
The finite total probability and countable additivity of a probability have many
important consequences, such as finite additivity, finite and countable subad-
ditivity, inclusion–exclusion, monotonicity, and the formulas for both relative
© Springer International Publishing AG 2016 1
R. Bhattacharya and E.C. Waymire, A Basic Course in Probability Theory,
Universitext, DOI 10.1007/978-3-319-47974-3_I
2 I Random Maps, Distribution, and Mathematical Expectation

complements and universal complements. Proofs of these properties are left to the
reader and included among the exercises.
Example 1 (Finite Sampling of a Fair Coin) Consider m repeated tosses of a fair
coin. Coding the individual outcomes as 1 or 0 (or, say, H, T), the possible outcomes
may be represented as sequences of binary digits of length m. Let Ω = {0, 1}m denote
the set of all such sequences and F = 2Ω , the power set of Ω. The condition that
the coin be fair may be defined by the requirement that P({ω}) is the same for each
sequence ω ∈ Ω. Since Ω has cardinality |Ω| = 2m , it follows from the finite
additivity and total probability requirements that

1 1
P({ω}) = = , ω ∈ Ω.
2 m |Ω|

Using finite additivity this completely and explicitly specifies the model (Ω, F, P)
with
 |A|
P(A) = P({ω}) = , A ⊂ Ω.
ω∈A
|Ω|

The so-called continuity properties also follow from the definition as follows:
A sequence of events An , n ≥ 1, is said to be increasing (respectively, decreasing)
with respect to set inclusion if An ⊂ An+1 , ∀n ≥ 1 (respectively An ⊃ An+1 ∀n ≥ 1).
In the former case one defines limn An := ∪n An , while for decreasing measurable
events limn An := ∩n An . In either case the continuity of a probability, from below
or above, respectively, is the following consequence of countable additivity1 (Exer-
cise 1):
P(lim An ) = lim P(An ). (1.1)
n n

A bit more generally, if {An }∞

n=1 is a sequence of measurable events one defines

lim sup An := ∩n=1 ∪m≥n Am (1.2)

n

and
lim inf An := ∪∞
n=1 ∩m≥n Am . (1.3)
n

The event lim supn An denotes the collection of outcomes ω ∈ Ω that correspond
to the occurrences of An for infinitely many n; i.e., the events An occur infinitely
often This event is also commonly denoted by [An i.o.] := lim supn An . On the other
hand, lim inf n An is the set of outcomes ω that belong to An for all but finitely many
n. Note that [An i.o.]c is the event that the complementary event Acn occurs for all
but finitely many n and equals lim inf n Acn .

1 Withthe exception of properties for “complements” and “continuity from above,” these and the
aforementioned consequences can be checked to hold for any measure.
I Random Maps, Distribution, and Mathematical Expectation 3

Lemma 1 (Borel–Cantelli
 I) Let (Ω, F, P) be a probability space and An ∈ F, n =
1, 2, . . . . If ∞
n=1 P(A n ) < ∞ then P(An i.o.) = 0.
Proof Apply (1.1) to the decreasing sequence of events

∪∞ ∞
m=1 Am ⊃ ∪m=2 Am ⊃ · · · ,

followed by subadditivity of the probability to get

∞

P(lim sup An ) = lim P(∪∞
m=n Am ) ≤ lim P(Am ) = 0.
n n→∞ n→∞
m=n


A partial converse (Borel–Cantelli II) will be given in the next chapter.
Example 2 Suppose that T1 , T2 , . . . is a sequence of positive random variables
defined on a probability space (Ω, F, P) such that for some constant λ > 0, P(Tn > t)
= e−λt , t ≥ 0, for n = 1, 2, . . . . Then P(Tn > n i.o.) = 0. In fact, P(Tn > θ log
n i.o.) = 0 for any value of θ > λ1 . This may also be expressed as P(Tn ≤ θ log n
eventually for all n) = 1 if θ > λ1 .
Example 3 (Infinite Sampling of a Fair Coin) The possible outcomes of nontermi-
nated repeated coin tosses can be coded as infinite binary sequences of 1’s and 0’s.
Thus the sample space is the infinite product space Ω = {0, 1}∞ . Observe that a
sequence ω ∈ Ω may be viewed as the digits in a binary expansion of a number x in
the unit interval. The binary expansion x = ∞ −n
n=1 ωn (x)2 , where ωn (x) ∈ {0, 1},
is not unique for binary rationals, e.g., 2 = .1000000 . . . = .011111 . . .. However
1

it may be made unique by requiring that infinitely many 0’s occur in the expansion.
Thus, up to a subset of probability zero, Ω and [0, 1) may be put in one-to-one
correspondence. Observe that for a given specification εn ∈ {0, 1}, n = 1, . . . , m,
of the first m tosses, the event A = {ω = (ω 1 , ω2 , . . . ) ∈ Ω : ωn = εn , n ≤ m} cor-
−n −n
responds to the subinterval [ m n=1 ε n 2 , m
n=1 εn 2 + 2−m ) of [0, 1) of length
−m
(Lebesgue measure) 2 . Again modeling the repeated tosses of a fair coin by
the requirement that for each fixed m, P(A) not depend on the specified values
εn ∈ {0, 1}, 1 ≤ n ≤ m, it follows from finite additivity and total probability one
that P(A) = 2−m = |A|, where |A| denotes the one-dimensional Lebesgue measure
of A. Based on these considerations, one may use Lebesgue measure on [0, 1) to
define a probability model for infinitely many tosses of a fair coin. As we will see
below, this is an essentially unique choice. For now, let us exploit the model with
an illustration of the Borel–Cantelli Lemma 1. Fix a nondecreasing sequence rn of
positive integers and let An = {x ∈ [0, 1) : ωk (x) = 1, k = n, n + 1, . . . , n +rn − 1}
denote the event that a run of 1’s occurs of length at least rn starting at the nth toss.
Note
∞ that this set is a union of length 2−rn . Thus, if rn increases so quickly that
−rn
n=1 2 < ∞ then the Borel–Cantelli Lemma 1 yields that P(An i.o.) = 0. For a
concrete illustration, let rn = [θ log2 n], for fixed θ > 0, with [·] denoting the integer
part. Then P(An i.o.) = 0 for θ > 1. Analysis of the case 0 < θ ≤ 1 requires more
4 I Random Maps, Distribution, and Mathematical Expectation

detailed consideration of the fundamental notion of “statistical independence” of the

outcomes of the individual tosses implicit to this model. This concept is among the
most important in all of probability theory and will be precisely defined in the next
chapter.

Remark 1.1 A detailed construction of Lebesgue measure is given in Example 1

of Appendix A. The existence of Lebsegue measure on [0, 1) plays a fundamental
role in providing the probability space for repeated unending tosses of a fair coin in
the previous example. The existence of probability models corresponding to infinite
sequences of experiments is as fundamentally important to probability as existence
of Lebesgue measure is to analysis. A general existence theorem will be given in
Chapter IX that will cover the theory developed in the chapters leading up to it. For
now we generally take such existence theory for granted.

For a given collection C of subsets of Ω, the smallest σ-field that contains all of the
events in C is called the σ-field generated by C and is denoted by σ(C); if G is any σ-
field containing C then σ(C) ⊂ G. Note that, in general, if Fλ , λ ∈ Λ, is an arbitrary
collection of σ-fields of subsets of  Ω, then λ∈Λ Fλ := {F ⊂ Ω : F ∈ Fλ ∀λ ∈ Λ}
is a σ-field. On the other hand λ∈Λ Fλ := {F ⊂ Ω : F ∈ F λ for some λ ∈ Λ}
is not generally a σ-field.
 Define the join σ-field, denoted by λ∈Λ Fλ , to be the
σ-field generated by λ∈Λ Fλ .
It is not uncommon that F = σ(C) for a collection C closed under finite
intersections; such a collection C is called a π-system, e.g., Ω = (−∞, ∞),
C = {(a, b] : −∞ ≤ a ≤ b < ∞}, or infinite sequence space Ω = R∞ , and
C = {(a1 , b1 ] × · · · × (ak , bk ] × R∞ : −∞ ≤ ai ≤ bi < ∞, i = 1, . . . , k, k ≥ 1}.
A λ-system is a collection L of subsets of Ω such that (i) Ω ∈ L, (ii) If A ∈ L then
Ac ∈ L, (iii) If An ∈ L, An ∩ Am = ∅, n = m, n, m = 1, 2, . . . , then ∪n An ∈ L. A
σ-field is clearly also a λ-system. The following π-λ theorem provides a very useful
tool for checking measurability.

Theorem 1.1 (Dynkin’s π-λ Theorem) If L is a λ-system containing a π-system C,

then σ(C) ⊂ L.

Proof Let L(C) = ∩F, where the intersection is over all λ-systems F containing
C. We will prove the theorem by showing (i) L(C) is a π-system, and (ii) L(C)
is a λ-system. For then L(C) is a σ-field (see Exercise 15), and by its definition
σ(C) ⊂ L(C) ⊂ L. Now (ii) is simple to check. For clearly Ω ∈ F for all F,
and hence Ω ∈ L(C). If A ∈ L(C), then A ∈ F for all F, and since every F is a
λ-system, Ac ∈ F for every F. Thus Ac ∈ L(C). If An ∈ L(C), n ≥ 1, is a disjoint
sequence, then for each F, An ∈ F, for all n and A ≡ ∪n An ∈ F for all F. Since
this is true for every λ-system F, one has A ∈ L(C). It remains to prove (i). For
each set A, define the class L A := {B : A ∩ B ∈ L(C)}. It suffices to check that
L A ⊃ L(C) for all A ∈ L(C). First note that if A ∈ L(C), then L A is a λ-system,
by arguments along the line of (ii) above (Exercise 15). In particular, if A ∈ C, then
A ∩ B ∈ C for all B ∈ C, since C is closed under finite intersections. Thus L A ⊃ C.
This implies, in turn that L(C) ⊂ L A . This says that A ∩ B ∈ L(C) for all A ∈ C and
I Random Maps, Distribution, and Mathematical Expectation 5

for all B ∈ L(C). Thus, if we fix B ∈ L(C), then L B ≡ {A : B ∩ A ∈ L(C)} ⊃ C.

Therefore L B ⊃ L(C). In other words, for every B ∈ L(C) and A ∈ L(C), one has
A ∩ B ∈ L(C). 
In view of the additivity properties of a probability, the following is an immediate
and important corollary to the π-λ theorem.
Corollary 1.2 (Uniqueness) If P1 , P2 are two probability measures such that
P1 (C) = P2 (C) for all events C belonging to a π-system C, then P1 = P2 on
all of F = σ(C).
Proof Check that {A ∈ F : P1 (A) = P2 (A)} ⊃ C is a λ-system. 
Remark 1.2 It is rather simple to construct examples of generating collections of
sets C and probability measures P1 , P2 such that P1 = P2 on C, but P1 = P2
on σ(C). For example take Ω = {1, 2, 3, 4}, C = {{1, 2, 3}, {2, 3, 4}}. Then
σ(C) = {{1, 2, 3}, {2, 3, 4}, {1}, {4}, {1, 4}, {2, 3}, Ω, ∅}. Let P1 ({1}) = P2 ({1}) =
P1 ({4}) = P2 ({4}) = 1/8, but P1 ({2}) = P2 ({3}) = 1/8, P1 ({3}) = P2 ({2}) = 5/8.
For a related application suppose that (S, ρ) is a metric space. The Borel σ-field
of S, denoted by B(S), is defined as the σ-field generated by the collection C = T of
open subsets of S, the collection T being referred to as the topology on S specified by
the metric ρ. More generally, one may specify a topology for a set S by a collection
T of subsets of S that includes both ∅ and S, and is closed under arbitrary unions
and finite intersections. Then (S, T ) is called a topological space and members of
T define the open subsets of S. The topology is said to be metrizable when it may
be specified by a metric ρ as above. In any case, one defines the Borel σ-field by
B(S) := σ(T ).
Definition 1.1 A class C ⊂ B(S) is said to be measure-determining if for any two
finite measures μ, ν such that μ(C) = ν(C) ∀C ∈ C, it follows that μ = ν on B(S).
One may directly apply the π-λ theorem, noting that S is both open and closed, to
see that the class T of all open sets is measure-determining, as is the class K of all
closed sets.
If (Si , Si ), i = 1, 2, is a pair of measurable spaces then a function f : S1 → S2
is said to be a measurable map if f −1 (B) := {x ∈ S1 : f (x) ∈ B} ∈ S1 for
all B ∈ S2 . In usual mathematical discourse the σ-fields required for this defini-
tion may not be explicitly mentioned and will need to be inferred from the context.
For example, if (S, S) is a measurable space, by a Borel-measurable function
f : S → R is meant measurability when R is given its Borel σ-field. A ran-
dom variable, or a random map, X is a measurable map on a probability space
(Ω, F, P) into a measurable space (S, S). Measurability of X means that each event2
[X ∈ B] := X −1 (B) belongs to F ∀ B ∈ S. The σ-field generated by X, denoted
σ(X ), is the smallest σ-field of subsets of Ω for which X : Ω → S is measurable. In

2 Throughout, this square-bracket notation will be used to denote events defined by inverse images.
6 I Random Maps, Distribution, and Mathematical Expectation

particular, therefore, σ(X ) = {[X ∈ A] : A ∈ S} (Exercise 11). The term random

variable is most often used to denote a real-valued random variable, i.e., where
S = R, S = B(R). When S = Rk , S = B(Rk ), k > 1, one uses the term random
vector.
A common alternative to the use of a metric to define a metric space topology, is
to indirectly characterize the topology by specifying what it means for a sequence to
converge in the metric. That is, if T is a topology on S, then a sequence {xn }∞
n=1 in S
converges to x ∈ S with respect to the topology T if for arbitrary U ∈ T such that
x ∈ U , there is an N such that xn ∈ U for all n ≥ N . A topological space (S, T ),
or a topology T , is said to be metrizable if T coincides with the class of open sets
defined by a metric ρ on S. Alternatively, by specifying the meaning of convergence
in the metric, one has that closed sets, and therefore open sets via complements,
can also be defined. Using this notion, other commonly occurring measurable image
spaces may be described as follows: (i) S = R∞ —the space of all sequences of
reals with the (metrizable) topology of pointwise convergence, and S = B(R∞ ),
(ii) S = C[0, 1]—the space of all real-valued continuous functions on the interval
[0, 1] with the (metrizable) topology of uniform convergence, and S = B(C[0, 1]),
and (iii) S = C([0, ∞) : Rk )—the space of all continuous functions on [0, ∞) into
Rk , with the (metrizable) topology of uniform convergence on compact subsets of
[0, ∞), S = B(S) (see Exercise 10).
The relevant quantities for a random map X on a probability space (Ω, F, P) are
the probabilities with which X takes sets of values. In this regard, P determines the
most important aspect of X , namely, its distribution Q ≡ P ◦ X −1 defined on the
image space (S, S) by

Q(B) := P(X −1 (B)) ≡ P(X ∈ B), B ∈ S. (1.4)

The distribution is sometimes referred to as the induced measure of X under P.

For random vectors X = (X 1 , . . . , X k ) with values in Rk , it is often convenient to
restrict consideration to the (multivariate) distribution function defined by F(x) =
P(X ≤ x) ≡ P(X 1 ≤ x1 , . . . , X k ≤ xk ), x = (x1 , . . . , xk ) ∈ Rk ; see Exercise 16. A
familiar and important special case is that of an absolutely continuous distribution
function given by
 xk  x1
F(x) = ··· g(u)du, x ∈ Rk ,
−∞ −∞

for a nonnegative density function g with respect to Lebesgue measure on Rk ; here

we have used the convention of representing Lebesgue measure as du. In an abuse of
terminology, a random variable with an absolutely continuous distribution is often
referred to as a continuous random variable.
If a real-valued random variable X has the distribution function F, then P(X ∈
(a, b]) = P(a < X ≤ b) = F(b)− F(a). Moreover, P(X ∈ (a, b)) = P(a < X < b)
= F(b− ) − F(a). Since the collection C of all open intervals (a, b), −∞ < a ≤ b <
I Random Maps, Distribution, and Mathematical Expectation 7

∞ is closed under finite intersections, and every open subset of R can be expressed
as a countable disjoint union of sets in C, the collection C is measure-determining (cf.
Exercise 19). One may similarly check that the (multivariate) distribution function of
a probability Q on the Borel sigma-field of Rk uniquely determines Q; cf. Exercise
19.
In general, let us also note that given any probability measure Q on a measurable
space (S, S) one can construct a probability space (Ω, F, P) and a random map X on
(Ω, F) with distribution Q. The simplest such construction is given by letting Ω = S,
F = S, P = Q, and X the identity map, X (ω) = ω, ω ∈ S. This is often called a
canonical construction, and (S, S, Q) with the identity map X is called a canonical
model. Note that any canonical model for X will generally be a noncanonical model
for a function of X . So it would not be prudent to restrict the theoretical development
to canonical models alone!
Before proceeding, it is of value to review the manner in which abstract Lebesgue
integration and, more specifically, mathematical expectation is defined. Throughout
1 A denotes the indicator
 function of the set A, i.e., 1 A (x) = 1 if x ∈ A, and is zero
otherwise. If X = mj=1 a j 1 A j , A j ∈ F, Ai ∩ A j = ∅(i = j), is a discrete random
variable
m or, equivalently, a simple random variable, then EX ≡ Ω X d P :=
j=1 ja P(A j ). If X : Ω → [0, ∞) is a random variable, then EX , expected
is defined by the “simple function approximation” EX ≡ Ω X d P := sup{EY :
0 ≤ Y ≤ X, Y simple}. In particular, one may apply the standard simple function
approximations X = limn→∞ X n given by the nondecreasing sequence


n2 n
−1
j
X n := 1[ j2−n ≤X <( j+1)2−n ] + n1[X ≥n] , n = 1, 2, . . . , (1.5)
j=0
2n

to write
⎧ n ⎫
⎨n2−1
j ⎬
EX = lim EX n = lim −n −n
P( j2 ≤ X < ( j + 1)2 ) + n P(X ≥ n) .
n→∞ n→∞ ⎩ 2 n ⎭
j=0
(1.6)
Note that if EX < ∞, then n P(X > n) → 0 as n → ∞ (Exercise 30). Now, more
generally, if X is a real-valued random variable, then the expected value (or, mean,
first moment) of X is defined as

E(X ) ≡ X d P := EX + − EX − , (1.7)
Ω

provided at least one of E(X + ) and E(X − ) is finite, where X + = X 1[X ≥0] and X − =
−X 1[X ≤0] . If both EX + < ∞ and EX − < ∞, or equivalently, E|X | = EX + +
EX − < ∞, then X is said to be integrable with respect to the probability P. Note
that if X is bounded a.s., then applying (1.5) to X + and X − , one obtains a sequence
8 I Random Maps, Distribution, and Mathematical Expectation

X n (n ≥ 1) of simple functions that converge uniformly to X , outside a P-null set

(Exercise 1.5(i)).
If X is a random variable with values in (S, S) and if h is a real-valued Borel-
measurable function on S, then using simple function approximations to h, one may
obtain the following basic change of variables formula
 
E(h(X )) ≡ h(X (ω))P(dω) = h(x)Q(d x), (1.8)
Ω S

where Q is the distribution of X , provided one of the two indicated integrals may be
shown to exist.
For arbitrary p ≥ 1, the order p-moment of a random variable X on (Ω, F, P)
having distribution Q is defined by
 
μ p := EX = p
X (ω)P(dω) =
p
x p Q(d x), (1.9)
Ω R

provided that X p is integrable, or nonnegative. Moments of lower order p than one,

including negative order ( p < 0) moments, may be defined similarly so long as X p
is real-valued random variable. Moments of absolute values |X | are referred to as
absolute moments of X . Let us record a useful formula for the moments of a random
variable derived from the Fubini–Tonelli theorem before proceeding. Namely,

Proposition 1.3 If X is a random variable on (Ω, F, P), then for any p > 0,
 ∞
E|X | p = p y p−1 P(|X | > y)dy. (1.10)
0

x
Proof For x ≥ 0, simply use x p = p 0 y p−1 dy in the formula
    |X (ω)| 
E|X | =
p
|X (ω)| P(dω) =
p
p y p−1
dy P(dω)
Ω Ω 0

and apply the Tonelli part (a) to reverse the order of integration. The assertion
follows. 

Example 4 As a generalization of Example 2, suppose that X 1 , X 2 , . . . is a sequence

of positive random variables, each having distribution Q, with a finite moment of
order p > 0. Then an application of Borel–Cantelli I together with Proposition 1.3
1 p
shows that P(X n > n p i.o.) = 0 (Use Exercise 29 applied to X 1 .).

If X = (X 1 , X 2 , . . . , X k ) is a random vector whose components are integrable

real-valued random variables, then define E(X ) = (E(X 1 ), . . . , E(X k )). Similarly
for complex valued random variables X = U + i V , where U, V are integrable real-
valued random variables, one defines EX = EU + iEV . In particular, √ for complex
valued random variables EX exists if and only if E|X | = E U 2 + V 2 < ∞;
Exercise 36.
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
I Random Maps, Distribution, and Mathematical Expectation 9

This definition of expectation as an integral in the sense of Lebesgue is precisely

the same as that used in real analysis to define S f (x)μ(d x) for a real-valued Borel-
measurable function f on an arbitrary measure space (S, S, μ); see Appendix A.
Almost sure convergence of a sequence of random maps X n , n ≥ 1, to X , each
defined on (Ω, F, P), is defined by X n (ω) → X (ω) as n → ∞ for all ω ∈ Ω up to
a subset of probability zero; i.e., convergence almost everywhere with respect to P.
One may exploit standard tools of real analysis (see Appendices A and C), such as
Lebesgue’s dominated convergence theorem, Lebesgue’s monotone convergence
theorem, Fatou’s lemma, Fubini–Tonelli theorem, Radon–Nykodym theorem,
for estimates and computations involving expected values.
The following lemma and proposition illustrate the often used exchange in the
order of integration.

Lemma 2 (Integration by parts) Let μ1 , μ2 be signed measures on R, which are

finite on finite intervals. Let

Fi (y) = μi (0, y], i = 1, 2, −∞ < y < ∞.

Then for any −∞ < a < b < ∞, one has

 
F1 (y)μ2 (dy) = F1 (b)F2 (b) − F1 (a)F2 (a) − F2 (y−)μ1 (dy).
(a,b] (a,b]

Proof Since a signed measure may be expressed as the difference of two measures,
without loss of generality it is sufficient to let both μ1 and μ2 be measures that are
finite on finite intervals. Then, using the Fubini–Tonelli theorem, one has
 
μ1 (du)μ2 (dv) = [F1 (v) − F1 (a)]μ2 (dv)
a<u≤v,a<v≤b a<v≤b

= −F1 (a)[F2 (b) − F2 (a)] + F1 (v)μ2 (dv).
a<v≤b

Also,
 
μ1 (du)μ2 (dv) = [F2 (b) − F2 (u−)]μ1 (du)
a<u≤v,a<v≤b a<u≤b

= F2 (b)[F1 (b) − F1 (a)] − F2 (u−)μ1 (du).
a<u≤b

Comparing these two iterations yields the asserted formula. 

Remark 1.3 The “distribution functions” Fi , i = 1, 2, can be defined as Fi (y) =

μi ((c, y]), i = 1, 2, y ∈ R, for any real number c in place of zero, and the lemma
still holds. This formula has special utility when applied to a nondecreasing function,
or more generally a function of bounded variation, as an integrand.
10 I Random Maps, Distribution, and Mathematical Expectation

The following is a useful version for expected values. A clever application is given
in Theorem 3.4.

Proposition 1.4 Let μ1 be an arbitrary measure on (0, ∞] which is finite on finite

intervals, and such that μ1 ({0}) = 0. Suppose that μ2 is a probability measure on
[0, ∞), and let Y be a random variable with distribution μ2 . Then, with Fi (y) =
μi ((0, y]), i = 1, 2, y ≥ 0, one has

EF1 (Y ) = P(Y ≥ y)μ1 (dy)
[0,∞)

Proof By the lemma, for any b > 0 one has

 
F1 (y)μ2 (dy) = F1 (b)F2 (b) − F1 (0)F2 (0) − F2 (y−)μ1 (dy)
(0,b] (0,b]

= F1 (b)F2 (b) − F2 (y−)μ1 (dy)
(0,b]

= [F2 (b) − F2 (y−)]μ1 (dy). (1.11)
(0,b]

The assertion follows by letting b ↑ ∞ on both sides, and using Lebesgue’s monotone
convergence theorem to obtain
 
F1 (y)μ2 (dy) = [1 − F2 (y−)]μ1 (dy).
(0,∞) (0,∞)

Definition 1.2 A sequence {X n }∞ n=1 of random variables on a probability space

(Ω, F, P) is said to converge in probability to a random variable X if for each
ε > 0, limn→∞ P(|X n − X | > ε) = 0. The convergence is said to be almost sure
(a.s.) if the event [X n → X ] ≡ {ω ∈ Ω : X n (ω) → X (ω)} has P-measure zero.

Convergence in probability is referred to as “convergence in measure” in analysis;

see Appendix A. Note that almost sure convergence always implies convergence in
probability, since for arbitrary ε > 0 one has 0 = P(∩∞ ∞
n=1 ∪m=n [|X m − X | >
∞
ε]) = limn→∞ P(∪m=n [|X m − X | > ε]) ≥ lim supn→∞ P(|X n − X | > ε); also see
Exercise 5. An equivalent formulation of convergence in probability can be cast in
terms of almost sure convergence as follows.

Proposition 1.5 A sequence of random variables {X n }∞ n=1 on (Ω, F, P) converges

in probability to a random variable X on (Ω, F, P) if and only if every subsequence
has an a.s. convergent subsequence to X .

Proof Suppose that X n → X in probability as n → ∞. Let {X n k }∞ k=1 be a subse-

quence, and for each m ≥ 1 recursively choose n k(0) = 1, n k(m) = min{n k > n k(m−1) :
I Random Maps, Distribution, and Mathematical Expectation 11

P(|X n k − X | > 1/m) ≤ 2−m }. Then it follows from the Borel–Cantelli lemma (Part
I) that X n k(m) → X a.s. as m → ∞. For the converse suppose that X n does not
converge to X in probability. Then there exists ε > 0 and a sequence n 1 , n 2 , . . . such
that limk P(|X n k − X | > ε) = α > 0. Since a.s. convergence implies convergence
in probability (see Appendix A, Proposition 2.4), there cannot be an a.s. convergent
subsequence of {X n k }∞
k=1 . 

The utility of Proposition 1.5 can be seen, for example, in demonstrating that if
a sequence of random variables X n , n ≥ 1, say, converges in probability to X , then
X n2 will converge in probability to X 2 by virture of continuity of the map x → x 2 ,
and considerations of almost sure convergence; see Exercise 6.
The notion of measure-determining classes of sets extends to classes of functions
as follows. Let μ, ν be arbitrary finite measures on the Borel σ-field of a metric space
S. A class Γ of real-valued bounded Borel-measurable functions on S is measure-
determining if S g dμ = S g dν ∀g ∈ Γ implies μ = ν.

Proposition 1.6 The class Cb (S) of real-valued bounded continuous functions on S

is measure-determining.

Proof To prove this, it is enough to show that for each (closed) F ∈ K there exists
a sequence of nonnegative functions { f n } ⊂ Cb (S) such that f n ↓ 1 F as n ↑
∞. Since F is closed, one may view x ∈ F in terms of the equivalent condition
that ρ(x, F) = 0, where ρ(x, F) := inf{ρ(x, y) : y ∈ F}. Let h n (r ) = 1 − nr
for 0 ≤ r ≤ 1/n, h n (r ) = 0 for r ≥ 1/n. Then take f n (x) = h n (ρ(x, F)). In
particular, 1 F (x) = limn f n (x), x ∈ S, and Lebesgue’s dominated convergence
theorem applies. 

Note that the functions f n in the proof of Proposition 1.6 are uniformly continu-
ous, since | f n (x) − f n (y)| ≤ (nρ(x, y)) ∧ (2 supx | f n (x)|). It follows that the set
U Cb (S) of bounded uniformly continuous real-valued functions on S is measure-
determining. Measure-determining classes of functions are generally actually quite
extensive. For example, since the Borel σ-field on R can be generated by classes
of open intervals, closed intervals, half-lines etc., each of the corresponding class
of indicator functions 1(a,b) , −∞ ≤ a < b < ∞, 1[a,b] , −∞ < a < b < ∞,
1(−∞,x] , x ∈ R, is measure-determining (see Exercises 9, 16).
Consider the L p -space L p (Ω, F, P) of (real-valued) random variables X such
that E|X | p < ∞. When random variables that differ only on a P-null set are iden-
tified, then for p ≥ 1, it follows from Theorem 1.7(e) below that L p (Ω, F, P) is a
1
normed linear space with norm X  p := ( Ω |X | p d P) p ) ≡ (E|X | p ) p . It is in this
sense that elements of L p (Ω, F, P) are, strictly speaking, represented by equiva-
lence classes of random variables that are equal almost surely. It may be shown that
with this norm (and distance X − Y  p ), it is a complete metric space, and therefore
a Banach space (Exercise 35). In particular, L 2 (Ω, F, P) is a Hilbert space with
inner product (see Appendix C)
12 I Random Maps, Distribution, and Mathematical Expectation


1
X, Y  = EX Y ≡ X Y d P, ||X ||2 = X, X  2 . (1.12)
Ω

The L 2 (S, S, μ) spaces are the only Hilbert spaces that are required in this text,
where (S, S, μ) is a σ-finite measure space; see Appendix C for an exposition of
the essential structure of such spaces. Note that by taking S to be a countable set
with counting measure μ, this includes the l 2 sequence space. Unlike the case of a
measure space (Ω, F, μ) with an infinite measure μ, for finite measures it is always
true that
L r (Ω, F, P) ⊂ L s (Ω, F, P) if r > s ≥ 1, (1.13)

as can be checked using |x|s < |x|r for |x| > 1. The basic inequalities in the
following Theorem 1.7 are consequences of convexity at some level. So let us be
precise about this notion.
Definition 1.3 A function ϕ defined on an open interval J is said to be a convex
function if ϕ(ta + (1 − t)b) ≤ tϕ(a) + (1 − t)ϕ(b), for all a, b ∈ J , 0 ≤ t ≤ 1.
If the function ϕ is sufficiently smooth, one may use calculus to check convexity, see
Exercise 24. The following lemma is required to establish a geometrically obvious
“line of support property” of convex functions.

Lemma 3 (Line of Support) Suppose ϕ is convex on an interval J . (a) If J is open,

then (i) the left-hand and right-hand derivatives ϕ− and ϕ+ exist and are finite and
nondecreasing on J , and ϕ− ≤ ϕ+ . Also (ii) for each x0 ∈ J there is a constant
m = m(x0 ) such that ϕ(x) ≥ ϕ(x0 ) + m(x − x0 ), ∀x ∈ J . (b) If J has a left (or right)
endpoint and the right-hand (left-hand) derivative is finite, then the line of support
property holds at this endpoint x0 .

Proof (a) In the definition of convexity, one may take a < b, 0 < t < 1. Thus
convexity is equivalent to the following inequality with the identification a = x,
b = z, t = (z − y)/(z − x): For any x, y, z ∈ J with x < y < z,

ϕ(y) − ϕ(x) ϕ(z) − ϕ(y)

≤ . (1.14)
y−x z−y

More generally, use the definition of convexity to analyze monotonicity and bounds
on the Newton quotients (slopes of secant lines) from the right and left to see that
(1.14) implies ϕ(y)−ϕ(x)
y−x
≤ ϕ(z)−ϕ(x)
z−x
≤ ϕ(z)−ϕ(y)
z−y
(use the fact that c/d ≤ e/ f for
d, f > 0 implies c/d ≤ (c + e)/(d + f ) ≤ e/ f ). The first of these inequalities
shows that ϕ(y)−ϕ(x)
y−x
decreases as y decreases, so that the right-hand derivative ϕ+ (x)
exists and ϕ(y)−ϕ(x)
y−x
≥ ϕ+ (x). Letting z ↓ y in (1.14), one gets ϕ(y)−ϕ(x)
y−x
≤ ϕ+ (y)
for all y ∈ J . Hence ϕ+ is finite and nondecreasing on J . Now fix x0 ∈ J . By taking
x = x0 and y = x0 in turn in these two inequalities for ϕ+ , it follows that ϕ(y) −
ϕ(x0 ) ≥ ϕ+ (x0 )(y − x0 ) for all y ≥ x0 , and ϕ(x0 ) − ϕ(x) ≤ ϕ+ (x0 )(x0 − x) for all
I Random Maps, Distribution, and Mathematical Expectation 13

x ≤ x0 . Thus the “line of support” property holds with m = ϕ+ (x0 ). (b) If J has
a left (right) endpoint x0 , and ϕ+ (x0 ) (ϕ− (x0 )) is finite, then the above argument
remains valid with m = ϕ+ (x0 ) (ϕ− (x0 )).
A similar proof applies to the left-hand derivative ϕ− (x) (Exercise 24). On letting
x ↑ y and z ↓ y in (1.14), one obtains ϕ− (y) ≤ ϕ+ (y) for all y. In particular, the
line of support property now follows for ϕ− (x0 ) ≤ m ≤ ϕ+ (x0 ). 

Theorem 1.7 (Basic Inequalities) Let X, Y be random variables on (Ω, F, P).

(a) (Jensen’s Inequality) If ϕ is a convex function on the interval J and P(X ∈ J ) = 1,
then ϕ(EX ) ≤ E(ϕ(X )) provided that the indicated expectations exist. More-
over, if ϕ is strictly convex, then equality holds if and only if X is a.s. constant.
1 1
(b) (Lyapounov Inequality) If 0 < r < s then (E|X |r ) r ≤ (E|X |s ) s .
(c) (Hölder Inequality) Let p ≥ 1. If X ∈ L p , Y ∈ L q , 1p + q1 = 1, then X Y ∈ L 1
1 1
and E|X Y | ≤ (E|X | p ) p (E|Y |q ) q .
(d) (Cauchy–Schwarz
√ √ Inequality) If X, Y ∈ L 2 then X Y ∈ L 1 and one has |E(X Y )|
≤ EX EY .
2 2

(e) (Minkowski Triangle Inequality) Let p ≥ 1. If X, Y ∈ L p then X + Y  p ≤

X  p + Y  p .
(f) (Markov and Chebyshev-type Inequalities) Let p ≥ 1. If X ∈ L p then P(|X | ≥
E(|X | p 1[|X |≥λ] )
λ) ≤ λp
≤ E|X
λp
|p
, λ > 0. More generally, if h is a nonnegative
increasing function on an interval containing the range of X , then P(X ≥ λ) ≤
E(h(X )1[X ≥λ] )/ h(λ).

Proof The proof of Jensen’s inequality hinges on the line of support property of
convex functions in Lemma 3 by taking x = X (ω), ω ∈ Ω, x0 = EX . The Lya-
s
pounov inequality follows from Jensen’s inequality by writing |X |s = (|X |r ) r , for
s
0 < r < s, since ϕ(x) = x r is convex on [0, ∞). For the Hölder inequality, let
p, q > 1 be conjugate exponents in the sense that 1p + q1 = 1. Using convex-
ity of the function exp(x) one sees that |ab| = exp(ln(|a| p )/ p + ln(|b|q )/q)) ≤
1
p
|a| p + q1 |b|q . Applying this to a = X|X|p , b = Y|Y|q and integrating, it fol-
1 1
lows that E|X Y | ≤ (E|X | p ) p (E|Y |q ) q . The Cauchy–Schwarz inequality is the
Hölder inequality with p = q = 2. For the proof of Minkowski’s inequality, first
use the inequality (1.27) to see that |X + Y | p is integrable from the integrabil-
ity of |X | p and |Y | p . Applying Hölder’s inequality to each term of the expansion
E(|X | + |Y |) p = E|X |(|X | + |Y |) p−1 + E|Y |(|X | + |Y |) p−1 , and solving the result-
ing inequality for E(|X | + |Y |) p (using conjugacy of exponents), it follows that
X + Y  p ≤ X  p + Y  p . Finally, for the Markov and Chebyshev-type inequali-
|X | p 1{|X |≥λ} )
≤ |Xλ p| on Ω, taking expectations
p
ties simply observe that since 1{|X |≥λ} ≤ λ p
E(|X | p 1[|X |≥λ] )
yields P(|X | ≥ λ) ≤ λp
≤ E|X
λp
|p
, λ > 0. More generally, for increasing
h with h(λ) > 0, one has E(h(X )1[X ≥λ] ) ≥ h(λ)P(X ≥ λ). 

Remark 1.4 One may note that the same proof may be used to check that correspond-
ing formulations of both the Hölder and the Minkowski inequality for functions on
arbitrary measure spaces can be verified with the same proof as above.
Random documents with unrelated
content Scribd suggests to you:
free fingers I made short work of the bonds on my ankles and
leaped up, just as he began to curse me, and cry out that I had
fooled him.
“Come!” I said, and pushed him toward the window.
He reeled against it and I sprang upon the sill. The night was black
before me and the city seemed full of noises, cries, and discordant
sounds, and above, in the sky, there was that luminous pallor that
precedes the dawn.
“Where’s your drink?” the tippler mumbled, trying to pull me back,
but I knocked him aside.
“Yonder!” I said; “go to the river, fool, and drink it dry,” and I
dropped out of the window and left him cursing, too far gone to
follow me, though I think he tried.
I felt my way along by the wall of the house; I was in the court-yard,
a cul-de-sac between the huts, and it was like pitch, but at last I
came to a gate and, fumbling at the latch, got it open and went out
into a street—a street that I did not know—in the Zemlianui-gorod,
and beset on all sides with dangers, seen and unseen. I went swiftly
forward, guided only by the dimly seen towers of the Kremlin, until
the light above grew fuller and I saw the crosses gleam on the
cathedrals, and so kept steadily on.
The wild night—full of its horrors—was spent; another day was
breaking, and here and there I stumbled on a man lying in the
street, either dead or in a drunken stupor, and once or twice I
turned aside to avoid a group of tipsy ruffians, but, in places, there
was quiet, the quiet of fear or worse, and I went on. The thought
that I might be too late drove me well-nigh to madness and winged
my feet, yet the way seemed endless.
But at last I came in sight of Kurakin’s house, and beyond it saw Le
Bastien’s, the windows still shuttered. There were no signs of
outrage, the street was quiet, the house closed and silent. I hurried
to the door and tried it and found it barred within. My heart beat
high with hope, and I made for the rear door and, crossing the
court, tried my key in the lock, but here, too, there were bars within.
Day dawned, a ghastly whiteness shone on the scene, even the sky
was white rather than blue. I beat upon the door. Silence. Then I
heard a step within, and beat upon the door and shouted. At my
voice the bars were lifted, the door opened softly, and Maluta’s white
face and great ears appeared.
“Where is the Princess Daria?” I cried, pushing past him.
He fell to trembling, his teeth chattered, he clutched at my knees.
“Be not angry, O my master,” he cried shrilly; “as the saints live, I
know not—she is not here!”
 XXIII: A SPRIG OF RUE
I HAD entered the hall, and the dwarf closed the door and secured it
with shaking hands. On the floor a lanthorn burned low, casting a
dim light, and the house—with its tight-closed shutters—was as dark
as pitch. There was a settle in the hall and, overcome with fasting
and exertion, I walked over to it and sat down in pure weariness and
dejection. Maluta meanwhile shied off from me and cowered behind
the lanthorn; for some reason the creature seemed to expect a
beating whenever I was displeased or disappointed. But I regarded
him as little at the time as a toad; I looked at him with dull eyes.
The event had only justified my fears.
“Where has the princess gone?” I asked slowly; “do you even know
that?”
“No,” he answered, shaking his head forlornly. “I came here last
night and found the house barred and silent, and yet I saw a light
under one of the shutters. I knocked and scratched there, and at
last the Princess Daria came and spoke to me; she thought you had
returned; she had been alone an hour or more, and she had heard
cries in the stable-yard.” He stopped and looked at me anxiously.
I nodded. “Go on,” I said briefly.
“She let me in,” he continued more quietly, seeing that I was not
angry, “and I gave her the message from her father.”
“Ah! you saw the prince?” I interrupted.
“Yes, I saw him and delivered your message,” he replied meekly.
“How did you escape the rioters in the palace?” I asked sharply. “I
thought you were surely trailing them to your death.”
He smiled, his little face puckering. “No,” he rejoined, “I ran down
the gallery—I am light of foot. I outran them and leaped out of a
window and left them to bay at the air. Then I found Prince Voronin
hidden in the Church of Saint Basil the Blessed, and I told him of his
daughter, but”—Maluta hesitated—“but not of you, O my master!”
“Your wit is as long as your ears,” I said approvingly; “go on, you
little rogue.”
“He bade me tell her to meet him at Troïtsa; he dared not leave his
hiding-place till night.”
“And he did not think of her peril?” I asked drily.
The dwarf shrugged his shoulders. “She is only a daughter,
excellency; the prince divorced her mother because she bore him no
sons, and he has wedded since, and will wed again.”
“Is that four times?” I asked absently, my mind dwelling on the
princess.
“Nay,” said Maluta, “four times can no Russian marry; even if he
presents himself before the priest he receives punishment, he——”
“And you found the princess here?” I said, cutting him short. “What
then? There has been no struggle here, no housebreaking—where is
she?”
Maluta’s face showed white in the dim light of the lanthorn.
“I gave her the message and she sent me to the stable to look for a
trace of you. I was gone scarcely twenty minutes, and when I
returned she was gone—gone as completely as though the earth
had swallowed her.”
I sat staring at the feeble flame of the lanthorn; burning low. She
had gone, then—as I had felt from the first—of her own free will, but
where and how? She had seemed to shrink from me; she probably
distrusted me, and she had fled—but whither? The wild disorder of
the city, the perils that beset her on every side, would have stayed
the most intrepid, and how could she go, alone, in the darkness of
that night? The atmosphere of the house suddenly grew choking. I
rose and, opening a shutter, let in the light of morning and the chill
spring air. Then I fell to pacing the floor, pondering upon it, and
taking no notice of the dwarf. I was, indeed, half frantic at the
thought of the girl alone and unprotected, fleeing from me into such
awful danger. It might be that she had friends near at hand, but it
was unlikely; the Russian girl—kept behind “the twenty-eight bolts”
of the terem—made few friends, and though I knew the Princess
Daria had enjoyed a greater freedom than others and that old-time
rules were everywhere relaxing, it was not at a time like this. Even
while I meditated I heard the roll of drums, the shouts of soldiers,
and a man lay in his blood in the very stable, and yet she had gone
alone!
I went into the room where I had seen her last, weeping so bitterly,
and I opened a shutter there and looked about me for some sign of
her. The room was just as she had left it, and lacking her presence it
was bare and sombre: a candle had burned out and guttered in the
candlestick upon the table; beside it stood her chair, pushed aside,
and the whole aspect of the place spoke of her presence. I stood
and looked at it with a heavy heart. She despised me, and I was still
her husband. The miracle of it made my brain swim, but,
nevertheless, my heart was very heavy. I sat down in her chair and
leaned upon the table, thinking, thinking of her, and of what fate
had befallen her, and as I sat there I saw something lying on the
table—a little sprig of rue, broken, doubtless, from her wedding
crown and caught in her hair to fall here. I took it up sadly and was
looking at it, when Maluta came sidling into the room and,
approaching me cautiously, laid a curious buckle on the table and
looked up at me with a wrinkled forehead. I glanced at the buckle
and then at him with some impatience.
“Well,” I said sharply, too weary and heartsore for trifles, “what ails
it?”
“’Tis such as the guards at the palace wear,” said he, an expression
of rare intelligence in his sharp eyes.
“Where did you find it?” I asked quickly.
“In the lobby, O my master,” he replied.
I rose and looked about me for some weapons, but as I did it I
turned dizzy, and remembered that, in twenty-four hours, I had
eaten nothing but a piece of bread. The dwarf had been watching
me and, with his usual singular penetration, he divined my need and
ran toward the kitchen. I stood leaning on the window, glad of the
air, and pondering on many things, while I put that bit of rue into my
bosom, a keepsake—and a sad one. I was still standing there, in
strange perplexity, when Maluta came back with some rice bread
and a little cold meat and wine, all that he could find in the desolate
larder, which he had doubtless visited heavily on his own account.
He set the food on the table and I ate hurriedly and in silence, for I
was sad enough. If the palace guards had been there, if Sophia had
entrapped the princess, what fate might not befall her? Yet, the
mystery of her disappearance was wholly unsolved. I could not
believe that she would go willingly with the czarevna’s emissaries,
and there could scarcely have been a forced entrance where no
trace of violence appeared.
At last I spoke my thought aloud.
“I must go to Sophia Alexeievna,” I said; “she shall tell me where the
princess is.”
I had scarcely spoken before Maluta fell on his knees beside me and
clasped my ankles, with every sign of terror.
“Never!” he cried; “oh, never go to her, my master; she would hang
you. She never forgets or forgives—not she!”
“But the Princess Daria?” I said sharply.
“There are other ways,” he pleaded; “let your servant find her, let
——”
I rose, casting him aside. “Let you find her,” I cried, with passion;
“you little fool, do you take me for a man of stone?”
And I went, searching for a sword and pistol, to another room, and
had found them and fastened them at my belt, and chosen a long
cloak that would muffle me, when I found him again on his knees at
my feet.
“You little rogue,” I said kindly, “why beset me? Do you think a man
fears to risk his neck?”
He shook his head and, laying both hands over his heart, besought
me with mute gestures.
“You will kill her also,” he protested, “the Princess Daria!”
At this I paused; the little creature’s wisdom had been almost equal
to his devotion. I could not afford to spurn either the one or the
other.
“What, then, can I do?”
“Let me go,” he pleaded.
“Ay, go, by all means,” I said promptly, “but I also must be stirring.
But content you, I will not go to the czarevna until all else fails, but I
will find the princess if she is still on the face of this old earth of ours
—that I swear, on the honour of a French gentleman!”
His face lighted a little at the opening of my speech, but clouded at
the end, and I saw that he feared that I would blunder, and, touched
by the creature’s devotion, I laid my hand kindly on his head and
thanked him for his courage and his wit in saving us in the palace
and for his willingness to serve me. It pleased him; a strange light
flickered in his eyes, he touched my coat hurriedly, and kissed his
own hand afterwards.
“You saved my life,” he said; “I am yours, O my master!”
And after that we went out together into the streets, where the
glare of day shone horridly upon a scene of death and turmoil, and
those things that night had cloaked with charity glared at noonday
and sent a shudder even to a strong man’s heart.
 XXIV: GALITSYN
WHEN I left our quarters I had made no plans, and was only
determined to find the Princess Daria, and to save her from the
dangers that encompassed her at every step. A wild enough design
for one man—and a foreigner at that—to form at such a time in
Moscow, and the aspect of the deserted streets smote me with a
sharp sense of the desperate nature of my enterprise. Fear crept
here, behind the close-shuttered windows, the grim, double-bolted
doors—fear and silence. Once or twice, as I passed, a grille opened
and I saw a white face peep out and vanish again at the sound of
my footstep, and here and there, notably among the houses of the
better sort, doors and windows gaped wide, and a heap of refuse, of
broken furniture and torn clothing, lay piled in the court-yard,
showing that this house and that had been gutted by the mob, and
across more than one threshold lay something that was neither
furniture, nor clothing, nor a sack of meal, though it lay as helpless,
still for evermore. Where I walked it was silent, so silent that my
tread woke the echoes, but the city was not so; it was full of
confused noises, of shouting and crying, of drum and musket, and
now and then, of the deep voices of the bells of Moscow’s many
churches and cathedrals. Riot and murder and robbery were loose
there, and as I drew nearer to the Kremlin I saw ever more bodies
and more bodies, lying in the sunlight with upturned faces and
helpless hands that would fight no more forever. I turned aside
thrice to avoid parties of rioters in pursuit of some wretched victim,
but through all my devious turns I kept on toward the palace. If the
princess lived, if my wife lived, she was there, of that I was
convinced. But before I entered the Red Place I heard the trampling
of horses, the shouts of men approaching, and stepped back into the
shadow of a friendly doorway and waited to see who passed that
way with such an escort. And presently, at the end of the street,
appeared a band of serfs, running ahead of a carriage, as they
always ran before a great nobleman; they came swiftly toward me,
two and two, clad in long crimson tunics, with collars and belts of
white and high green caps; and as they advanced I counted twenty
before the horses and behind the carriage there were twenty more,
and with all their splendid dress their feet were bare, as had been
the feet of the Prince Voronin’s slaves. The horses, three splendid
creatures, were hung with fox-tails that floated as they moved, and
in the carriage sat a noble figure, in a magnificent dress of gold and
silver brocade; his handsome face was but slightly concealed by his
high collar, and jewels flashed on his breast. It was Prince Basil
Galitsyn himself, the rising star of the new order of things, the
favourite of Sophia, and the lover of the Princess Daria. And at his
feet sat Maître le Bastien. I stepped out of my concealment and
called aloud for them to stop. The driver, a fierce Cossack, cracked
his whip and would have driven over me, but for Le Bastien, who
saw me and cried out to Prince Galitsyn. At a nod from him the
whole procession halted, the horses plunging and rearing on their
haunches, and the serfs crowding about me, as if they waited orders
to seize and carry me away. But I walked up to the carriage itself,
and demanded speech with his excellency the Boyar Prince Galitsyn,
aware, all the while, of the master goldsmith’s perplexed
amazement, but the prince was bent on benevolence. It was an
hour, indeed, when he had need of all his diplomacy and suavity to
hold his supremacy, and he bade me follow to his palace, where I
should have an audience. His air of patronage stuck in my throat,
but reflecting that he knew me only as the French goldsmith, I
swallowed my pride and followed at a distance to his house.
Galitsyn was noted for his magnificence and his foreign tastes, and
his palace was furnished more in the style of Europe than any other
in Moscow. I had been there before as Maître le Bastien’s apprentice
and knew it well. The carriage of the prince and his attendants
arriving in advance, I found the court-yard thronged with his scarlet
tunics, and the wide doors of the great house stood open as I
approached, and in the lobby, too, were serfs in scarlet, and then I
saw the prince very gracefully and graciously offer Maître le Bastien
the bread and salt, and when the goldsmith had tasted both, they
were extended to me and, happily, I took them also. Then his
excellency led the way, and Maître le Bastien and I followed into a
grand salon, where the prince seated himself in a chair, carved with
his arms and covered with cloth of gold, and signing to the master
goldsmith to sit in a lower and humbler seat, he turned to me and
asked my business, while a slave brought in the salver laden with
vodka and caviare—the zakuska.
I was in no mood to mince matters, and, despite various frowns and
grimaces from Maître le Bastien, I came bluntly to the point,
speaking in Russ; for the prince knew only his own language and
Latin.
“Can your excellency tell where the Czarevna Sophia has hidden the
Princess Daria?” I asked, fixing my eyes sternly on him, for I was not
without suspicions of the man himself, but my doubts were instantly
dispelled by the change that swept over his face.
The prince was a proud man, haughty and reserved, as all these
Muscovite aristocrats were, but he could not disguise his
discomfiture at the mention of those two names together.
“The Princess Daria!” he repeated blankly. “I was told that she and
her cousin, Vassalissa, were safe in Troïtsa yesterday morning.”
I could not doubt that he spoke the truth; his manner was full of a
noble sincerity, and, indeed, I think the man’s worst fault was the
common one of a not over-scrupulous ambition.
“The Princess Daria was in the palace yesterday,” I said deliberately,
“and imprisoned there by the Czarevna Sophia, who would have
forced her to marry the Boyar Kurakin.”
Galitsyn sprang from his chair; his face was as white as ashes, but
with wrath rather than dismay. He turned fiercely on Maître le
Bastien.
“This is your man,” he said thickly, for something seemed to choke
him; “does he speak the truth?”
The goldsmith caught my eye and understood my gesture; he rose
with a dignified composure that became him well.
“My prince,” he said gravely, “you deserve my confidence and my
service; I will disguise nothing from you. This man is not an
apprentice, but a patron of mine, a French nobleman, M. le Marquis
de Cernay, and his honour and your excellency’s are one.”
The prince bowed gracefully; but the strenuous expression of his
face did not relax. He asked my Christian name of Maître le Bastien,
and addressed me after the manner of the Russians.
“Ivan Feodorvitch,” he said, which was my name, being translated;
“did you see the princess in the—the power of Sophia Alexeievna?”
he stammered, in spite of himself; one woman he loved, the other
he courted for ambition’s sake, and I have seen this a hundred times
in Paris.
I looked down the long room, and across the end of it stood a
double row of scarlet tunics, thirty-seven, I thought, and before the
only other entrance stood the major-domo, leaning on a wand of
ivory and gold. I counted the cost and smiled.
“M. le Prince,” I said quietly, “I saw and heard the Czarevna Sophia
threaten and compel the Princess Daria Voronin to wed Kurakin, in
the private chapel by the painted gallery.”
He drew a deep breath, his eyes blazed, his whole figure seemed to
dilate with passion. Maître le Bastien leaned forward, listening
eagerly; I even caught a flicker on the face of the old steward, who
was otherwise as motionless as stone.
“Did she marry Kurakin?” the prince demanded, in a low voice, but in
a tone that might well strike terror to a weak heart; “did Sophia
force her to that?”
“No,” I replied, “no, your excellency, for I prevented it——”
He broke in upon me with a kind of fierce joy.
“You prevented it—and how, sir?”
“I married the Princess Daria myself,” I said.
A pause followed, a pause so deadly that I heard Maître le Bastien
breathing like a man spent with running. I believe that Galitsyn
thought me mad; he looked at me as if he doubted his own senses,
and that doubt alone stayed his hand. I think, on the first impulse,
he would have struck me dead—if he could. He wore the look men
wear when they strike to kill; I saw just such a look on the face of
M. d’Argenson that morning in Easter week, on the Place Royale.
And Galitsyn meant to kill me, but after a moment, a moment of
sharp suspense, he laughed harshly.
“I think you jest, sir,” he said, with bitter pleasantry, “but ’tis a
dangerous jest—here.”
Saint Denis! who could doubt it, with that fierce eye of his upon me
and that row of scowling savage faces below me in the hall? It was
like to be a sorry jest indeed. But I cared neither for him nor for his
menials, now that I was sure that the princess was not in his power,
yet I meant to let him know that she was mine—and mine she
should be—in spite of him.
“I jest so little, M. le Prince,” I said, “that he who dares to contradict
my statement will do it at his peril. The Princess Daria is my wife.”
“Your wife,” he replied bitterly, measuring me with a fierce eye, “by
Saint Nikolas of Mojaïsk, she shall be your widow, then!” and he
raised his right hand sharply, but Maître le Bastien flung himself
upon that arm and held it.
“Hold your hand, Prince Galitsyn!” he cried hoarsely, “and remember
the sacred bond of hospitality—the marquis has eaten of your bread
and salt.”
The prince paused; his breast heaving with passion, his eyes kindling
with a savage triumph, his face deeply flushed, Maître le Bastien
holding his right arm by main force. Below, at the end of the hall, I
heard the serfs stirring restlessly and the clash of swords. I folded
my arms on my breast and waited. I had never been more
indifferent; let the barbarian do his worst!
“The bond of bread and salt, M. le Prince, remember it,” said the
goldsmith gravely. “The last of our Valois kings permitted his guest
to be murdered, and he also fell by the murderer’s hand. As you
sow, so will you reap!”
Galitsyn looked at me with eyes that devoured every detail of my
face and figure; scorn and rage were mingled on his countenance.
“‘Her husband,’ he calls himself,” he said, with a bitter laugh. “This
foreigner, this soldier of fortune—the mate for one of Russia’s
noblest born—for the Princess Daria. Why, master goldsmith, the
man is mad!”
“You forget to whom you speak, M. le Prince,” I retorted hotly. “A
gentleman of France, and the head of one of its noble houses, is
your equal—ay, and something more,” I added, as truculent as he,
though I saw Le Bastien’s warning gesture.
“‘And something more’—I thank you, sir,” he said, with bitter disdain,
“and the czarevna—did she plan this marriage also?”
“No,” I answered promptly, “a thousand times no—she did not know
of the exchange of bridegrooms——”
He interrupted me at this; he shook off the goldsmith and came
nearer to me.
“Did the Princess Daria choose you—instead of Kurakin?” he asked in
a deep voice, and I saw that the man was shaken to the soul.
“What it that to you?” I retorted scornfully; “the princess is my wife,
and hark ye, M. le Prince, mine she shall be—against the world, and
no man shall put asunder.”
Again he half raised his hand to deliver me to his slaves, and again
he desisted, but his face was distorted with contending passions,
and he pointed to the door with a quivering finger.
“Go, sir!” he said hoarsely; “go, before I violate the bond of host and
guest—but, beyond my gates, look to yourself!”
“Nay, M. le Prince,” I said courteously. “Beyond your gates I am at
your service. In France we settle these matters on the field of
honour. I should be happy, monsieur, I——”
But the goldsmith had me by the arm.
“Saint Denis, man!” he cried in French, “tempt him no more, unless
you would imperil the princess as well as your own head and mine!”
And I yielded to this reasoning the more easily because I saw that
Galitsyn did not heed my challenge; his ideas of settling the matter
differed from mine, as vastly as the customs of his country and
mine. He stood there, pointing steadily at the door, and his face was
so distorted with passion that I marvelled to see so great a change
wrought in a handsome countenance.
I bowed profoundly.
“M. le Prince will find me ready when and where he pleases,” I said
pleasantly, “and he will remember that the Princess Daria is my
wife.”
I could say no more, for Maître le Bastien was dragging me away by
main force, and the serfs, parting to let us go through, closed up
behind us like a wall and eyed us so viciously that I saw the
goldsmith wipe his brow twice before we reached the gate of the
court-yard. The good man was nothing of a fighter, and yet too stout
to run.
 XXV: MICHAUD’S REPENTANCE
WHEN we reached the street Maître le Bastien stopped, panting and
wiping the cold perspiration from his forehead.
“Lack-a-day, M. le Marquis, you have undone us now!” he cried,
between his gasps. “These Russians—holy Virgin! to tell the prince,
to his face, that you had married this Princess Daria, and to defy
him, too! And he’s on the crest of the wave, and just declared the
head of the Department of Foreign Affairs, in the place of Larion
Ivanof!”
“Precisely, monsieur,” I replied, smiling, “and let him desert the
Czarevna Sophia for the pursuit of the Princess Daria,—of my wife,—
and he will find that his house is of cards and falls at a breath.”
The force of this argument struck even the goldsmith, and he was
silent for a moment, but the next his equilibrium was again
destroyed by the sudden appearance of a band of Streltsi at the
further end of the street. He seized me by the arm, without
ceremony, and hurried me into a deserted garden close at hand, and
there, drawing me into the shadow of the stone wall, he began to
recount the horrors of the preceding day, many of which I had
witnessed, and to argue from this point that I was powerless to save
the Princess Daria, and that my marriage was of as little moment as
the cooing of two pigeons on the stone arch over our heads.
“The power of the Naryshkins has all crumbled to pieces; Matveief is
dead, and Dolgoruky. Athanasius Naryshkin, the czarina’s brother,
was betrayed by the dwarf, Homyak, and cut to pieces in the
Cathedral of the Assumption. Even the patriarch barely escaped;
they are determined to have Ivan Naryshkin and the Jew Von Gaden.
They killed the privy councillor Ivanof, and his son, and two
lieutenant-colonels on the portico, between the banqueting-hall and
the Cathedral of the Annunciation; the Boyar Ramodanofsky is
butchered, the Boyars Soltykof, father and son; Peter Naryshkin, and
—this morning—Kirillof and Dr. Gutemensch, and there are others—
by the score. They declare to-day that the serfs shall be free, and
now ’tis thought that many slaves will betray their masters; it was a
servant who betrayed the elder Soltykof. ’Tis just as I predicted: the
devils are let loose and no man can curb them, and Sophia
Alexeievna is the only one in the royal family equal to facing the
crisis; she alone can plead with and influence these beasts, and yet
you think to get the Princess Daria from her!” The good man threw
up his hands with an expression of despair.
Outside the garden wall I heard the Streltsi screaming and singing
as they passed—a tipsy crowd. About us the weeds and plants grew
thickly and almost blocked the doors and windows of a deserted
house. One great clump of bushes opposite choked the entrance to
the cellar.
“I have seen enough, and been through enough, to realise my
difficulties,” I said calmly, “but what is a man’s right arm for, Maître
le Bastien, if it be not to fight in a good cause? I will not allow even
Sophia to snatch my wife away, and I will not leave her in the hands
of her enemies. But tell me, how fares it with you? I would not
imperil you.”
“I am safe enough,” he replied at once. “Galitsyn will protect me;
against me he has no quarrel, and has sent some Streltsi already to
keep my house from the mob. I fear nothing for myself, but you,
monsieur, you are lost!”
I laughed, although my reflections were grim enough.
“Nay,” I replied, “was I not lost in Paris? Did I not flee from the Place
Royale, with the provost-marshal at my heels? Tush, monsieur, I am
inured to perils; fear not for me. But now, I would give much for a
single clew of her whereabouts; whether she went of her own will or
not, and who she went with?”
Deeply flushed with embarrassment, Maître le Bastien hesitated, and
then ventured to speak his mind.
“Pardon me, monsieur,” he said; “did you not marry her against her
will, and almost—well, yes, M. le Marquis—almost, we might say, by
force?”
I assented gloomily, my eyes on the bushes opposite, that swayed,
though there was no wind.
“Then—may she not have fled—of her free will—monsieur?”
suggested the goldsmith, avoiding my eye.
“’Tis not impossible,” I replied, “yet I think it improbable. She had
less to fear from me than from others, and I thought her at the last
inclined to trust me.”
“‘Souvent femme varie,’” quoted Maître le Bastien softly.
I did not reply; instead, I went across the garden and, diving my
arm into the bushes, drew forth—by the nape of the neck—that
rogue, Michaud.
At the sight of him the master goldsmith’s face flushed with
mortification; he felt that I had borne much from his apprentice; and
he was more severe now than I, who knew, by the knave’s looks and
his chattering teeth, that he had hidden there in an agony of terror
and from no sinister motive.
“What are you doing here, you rogue?” demanded his master
harshly; “have you not done harm enough already, without playing
also the part of eavesdropper?”
The scenes that he had witnessed, and the near approach of violent
death, had cowed the fellow completely, and he hung his head,
sullen and wretched to the last degree.
“I hid here to escape these savages,” he said, “and had no thought
of seeing you, Maître le Bastien.”
“That may be!” retorted the goldsmith sharply, “but when you saw
only M. de Cernay and me here, why did you lurk like a snake in the
bushes!”
Michaud turned deeply red and glanced aside at me.
“I thought that you despised me,” he said bluntly, “and could well
dispense with the sight of me.”
But Le Bastien was not appeased, and would have said more but
that I interposed.
“We cannot judge him too severely, monsieur,” I said, laughing; “only
last night I hid in a wine-butt!”
The goldsmith, who knew me for a choleric man and a fighter,
smiled in spite of himself at this, and Michaud cast a look of
something akin to gratitude at me. The fellow loved his master. I
myself had once just such a follower, devoted to jealousy, and full of
sullen fits and the changeful moods of a woman.
“You speak truly, M. le Marquis,” said Le Bastien thoughtfully; “we
cannot judge too harshly; but where have you been, Michaud?”
“I got out of the palace in the tumult, monsieur,” replied the
apprentice, “and I have been hiding in one place and another, ever
since; only an hour ago I got out of the Kremlin through the
kindness of a soldier, whom I knew, and who got me through the
guard at the Gate of Saint Nikolas, for a rouble—all I had.”
“Ah, then money still has its virtue, has it?” I exclaimed, seeing a
gleam in the thick cloud of trouble.
Michaud shuffled his feet in the weeds and stood looking down
without answering, and then suddenly he lifted his head and looked
at me—squarely—for the first time since he had released the fat
chamberlain.
“M. le Marquis, I have somewhat to tell you,” he said slowly. “Last
night, as I hid under the portico of the banqueting-hall, I saw the
palace guards under that man—the stout man—you know whom I
mean?” I nodded. “He who was imprisoned in Maître le Bastien’s
house. They came by, bringing a lady, veiled and muffled, and
afterwards one of their number, who speaks a little French,—the
man I bribed—told me that she was taken from our quarters and—
was the Princess Daria.”
I drew a deep breath, and then I took two roubles and put them into
Michaud’s hand.
“To repay your loss,” I said, “and to thank you; your good deed
outweighs your evil.”
But Maître le Bastien shook his head. “’Tis ill news, M. le Marquis,”
he said ponderously, “and an ill wind that will blow nobody good.”
“No news can be worse than bad news, monsieur,” I replied, as
lightly as I could, though, I confess, my heart sank. “At least I know
now where to look.”
He regarded me in despair. “You cannot dream, monsieur, of going
there?” he exclaimed, pointing toward the Kremlin with a shudder of
repulsion.
I nodded gaily, as if I thought it a light matter.
“Ay, and at once,” I said.
He held up his hands. “’Tis madness,” he cried, “sheer madness! The
dead lie there to warn you!”
“And she is there!” I retorted, and drawing my pistol from my belt, I
primed it.
“Tut, tut, monsieur!” said Maître le Bastien, “you dream! Why rush to
death? The Church accounts suicide sin, and what is this you
contemplate but suicide? Come, monsieur, come home with me,” and
he plucked at my sleeve, in honest consternation.
I thanked him pleasantly. “You are a good friend, Maître le Bastien,”
I said, “and I am glad to feel that you will be safe. Cut loose from
me, monsieur, however, for I will be henceforth a marked man—the
Princess Daria’s husband,” and I smiled bitterly.
He shook his head despondently, knowing me too well to interfere
further.
“Michaud,” I said, looking to my sword, “where is your Streltsi? I
must get into the Kremlin, instead of out of it.”
But at these words the apprentice turned white as paper.
“I cannot go back there, monsieur,” he protested; “I dare not.”
“Are you a woman?” I asked scornfully, eyeing him in a way that
brought the blood to his face, but he stood sullenly silent.
“Bah!” I said; “where are your petticoats?”
But Maître le Bastien was inclined to support him, if only to thwart
me, and my temper was rising when—happily for all—the dwarf,
Maluta, suddenly appeared in the gateway, and, at the sight of me,
came pattering in, and ’twas he, as usual, who found a way to cut
the knot. In fact, in those five bloody days of the insurrection, it was
these little creatures—these playthings of the court—who wrought
much good and evil. Homyak, the dwarf, betrayed the czarina’s
brother to his death; Komar, the dwarf, saved the son of the
Chancellor Matveief and Feodor Naryshkin; and Maluta, the dwarf,
was my true friend and ally. Yet they were the veriest waifs of
fortune, the most miserable toys of tyrants, the outcasts of society—
and its spies. Misery and secret power, degradation and triumph,
merriment and despair, these things made up the sum of their short
lives, but love and happiness and honour passed them by and left
them to perish by the wayside.
 XXVI: MALUTA BUYS TWO SOULS
NO sooner was Michaud’s information explained to the dwarf than
the latter fell into a reverie, rubbing his chin with his forefinger and
puckering up his forehead in that monkey-fashion which never failed
him in time of thought, when it always seemed to me that his ears
moved also. For my part, I was not inclined to reflections, but to
action; if the Princess Daria was in the Kremlin, I would be there
too. But there was a difficulty to be surmounted first; my Streltsi
captors had relieved me of all the money I had upon my person, and
I had only put three roubles and a few kopecks in my pocket when I
visited our quarters; it would, therefore, be necessary for me to go
there, not only for more money, but for a stock of ammunition,
before I attempted to pass the gates of the Kremlin. Meanwhile
Maluta had evidently reached a conclusion, and asked me the name
of the soldier who knew Michaud, for the dwarf and the apprentice
could only communicate by signs, neither understanding the other’s
language. I repeated the question to our knave and he shrugged his
shoulders sullenly.
“I know not, monsieur,” he replied; “’tis one of their villainous names
that make a man sneeze and then forget them.”
This I did not translate, but told Maluta that he did not know the
Streltsi’s name.
“Ask him if it is Grotsky?” said the dwarf shrewdly.
Michaud was not sure, but thought it might be. He described the
man, however, as tall, black-haired, and with a cast in the left eye.
Maluta nodded when this was translated to him; and held out his
small, claw-like hand.
“Give me money, O excellency,” he said in his shrill voice, “for ’tis for
money that men sell their souls—and for drink, their understanding.”
I put the rouble and the few kopecks that I had left in his hand, but
he was by no means satisfied.
“Prebavit!” he cried shrilly, stamping his foot, “prebavit! Can thy
servant buy a soul for a rouble or twenty kopecks?”
“Some souls would be dear at that, Maluta,” I said drily, “and you
have drained my purse.”
But he stamped the more. “’Tis not enough, O my master,” he said,
“prebavit, prebavit!”
“Maître le Bastien, for the nonce, I must even borrow of you,” I said.
Whereupon the goldsmith gave me ten roubles more, though he was
open in his disapproval.
“All this will lead to nothing but misfortune,” he said, “and, after all,
’twould have been wiser to let you fall into the hand of his majesty’s
provost-marshal.”
“Ah, monsieur,” I said, “you do not know the Princess Daria.”
Le Bastien shook his head despairingly. “No,” he replied, “and ’tis
well I do not, since she can turn a sane man’s head so completely.”
While he was speaking I was giving the money to the dwarf, and
trying to fathom his plans, but to no purpose, he would tell nothing
—only looked at me, in an elfish fashion.
“Let be, O excellency,” he said; “I will go to the Kremlin and by
nightfall I will return to you, then we will go together, and we will
find the Princess Daria, if she is still in the fortress.”
My faith in his acuteness was growing to be almost a superstition,
but it was too much to ask me to wait from noon until nightfall, and
I told him so.
“You must wait,” he replied wisely. “You cannot go—as I can—in and
out of the palaces and the cathedrals; you are big—a tall man,” he
measured me with his eye, as though he thought me a giant. “Such
a man cannot hide—any more than the Tower of Ivan Veliki can hide
among the churches! I must find where the lady is, and then I must
come for my master. Nay, fear not—she is safe to-day and to-
morrow and the day after; not even Sophia Alexeievna rules to-day,
but only the Streltsi. They have mobbed the Department of Justice
and Serfage; they have thrown the papers in the street, they have
threatened the Danish Resident, they would have killed the
patriarch. Sophia will do nothing yet to the princess, for she has
hidden her, and to-night we will find her.”
“Nay, now we will find her,” I said, determined to go with him.
But he fell on his knees. “Have I failed you at all, O excellency?” he
cried.
I was forced to admit that he had, on the contrary, saved my life and
hers.
“Then give me two hours—two hours and eleven roubles, O my
master,” he said, “and surely I may buy a man’s soul and also his
body—for there is vodka yet in the cellar—where, they know not—
and red wine.”
“The little fellow talks sense, monsieur,” said the goldsmith; “I pray
you be ruled; give him, at least, two hours. It is madness for you to
attempt the Kremlin now. Until they find the poor Jew, Von Gaden,
they will suspect every man in foreign clothes. Come, therefore, to
our quarters; two hours of rest, and food, and more money will but
prepare you for this desperate adventure.”
I knew there was truth in his argument, and my need of money was
absolute if I was to take the princess out of Moscow. So, hard as it
was to wait, I yielded, at last, to their entreaties, and went back to
our quarters for two hours’ rest, of which I stood in great need, and
for money and another pistol.
Prince Galitsyn had kept his promise and sent Maître le Bastien a
guard of Streltsi; a score of villainous-looking knaves who were as
ready to rob us as to watch us, and sat about the court-yard making
frequent demands for brandy and meat, and their presence proving
a cause for fresh alarms to the goldsmith, who kept the shutters up
and the doors bolted and busied himself hiding his valuables in the
earth of the cellar, assisted by Michaud. Meanwhile, stretched myself
on my bed, in my clothes, resolved to rest but a few moments;
however, we are only human, and my body was worn out with
continued exertions and loss of sleep, and I had hardly touched the
pillow before I dropped off into the heavy slumber of sheer
exhaustion.
My room being in the second story, the shutter of the one window
was open and the red glow of sunset shone full in my face when I
awoke, and, for a moment, kept me from seeing Maluta, who was
standing beside me, watching my slightest movement. With the
return of consciousness came back the memory of my situation, and
I started up, and began to snatch up my weapons. The sweet air of
a May evening came softly in at the casement, and with it far-off
sounds, harsh cries, and discordant music.
“Where is she?” I cried. “Why have I slept like a brute? Where is the
Princess Daria?”
“She is safe,” replied the dwarf gravely, “and ’tis well that my master
slept, for we must work to-night and travel to-night, if we would get
her away, from the czarevna and from Kurakin.”
I had, by this time, arranged my disordered clothing and belted on
my weapons, and now I hid a bag of gold upon my person.
“Was the money sufficient?” I asked eagerly.
He nodded gravely. “The souls were cheap this day,” he said
solemnly. “I bought two for nine roubles and two kopecks, and the
rest of the money I put into good liquor, and the drug that they keep
in the Gostinnoi Dvor, in the street of sweet perfumes and spices and
myrrh and aloes.”
By this he referred to the divisions of the Gostinnoi Dvor, or great
bazaar, where every commodity had a separate street, and a man
could find only silk merchants in one avenue, and dealers in jewels—
opals and pearls and the great amethysts of Russia—in another.
“Come, then,” I said with impatience, “let us go!”
He nodded his assent, and without more delay we descended the
stairs together, and I went to bid Maître le Bastien farewell. I found
him burrowing, like a mole, in the cellar.
“The Prince Galitsyn means well to me,” the good man said
earnestly, “but these Streltsi have already demanded a rouble
apiece, a flagon of good brandy, and ten loaves of rice bread,
besides the pickled mushrooms and the fine sterlet, that I would
have kept for my dinner! Moreover, Michaud and I are spent with
hard labour, but, praised be the saints! everything is buried now but
the great vase, and that being for Sophia, I think they will leave it.”
“Unless they sell it to satisfy their stomachs,” I replied; “but adieu,
Maître le Bastien, I go on a perilous enterprise, and may not soon
see you again.”
The goldsmith wiped his eyes surreptitiously, and then shook my
hand heartily.
“Farewell,” he said; “you go against my counsel and my will, but yet
—I suppose, if I were young, I would do likewise! But, I pray you,
be cautious, if you can, and if I can do aught——”
“But you cannot, monsieur,” I interrupted heartily, “nor do I wish you
in this tangle,” and I bade him a hasty farewell and hurried away,
lest he should delay me by more arguments.
It took us a few moments more to get past our own Streltsi, and
then Maluta and I struck off into the lanes that led through some
byways to the Kremlin, avoiding the short route which would have
taken us into the crowded quarter.
The city still wore its desolate air and, by contrast with the sky, yet
glowing with sunset, it seemed dark and haunted with the grimmest
of shadows. Close-shuttered windows and barred doors surrounded
us; not a woman looked out, not a child ran in the court-yards, and
if we chanced on a peaceful citizen, he skurried away at the sight of
an armed man. As we neared the Gate of Saint Nikolas, I saw that it
was occupied by a guard of Streltsi, and the dwarf signed to me to
advance slowly, while he began to whistle softly, holding his hands
over his mouth. Then I saw one of the guards at the gate lounge
toward us, and began to suspect Maluta’s designs. When the soldier
had placed a few yards between himself and his comrades, he
quickened his steps and the dwarf whispered to me to keep close to
the wall of the house, in front of which we had halted, while he
sidled forward, crab-fashion, to meet the Streltsi; and I saw that he
must have saved some money, for it changed hands before my eyes.
Then he called to me in a low voice:
“We will go forward now, excellency,” he said. “Way here, for the
ambassador to their czarish majesties!” he added, in his shrillest
tone, clapping his hands.
The soldier bowed low and saluted, and then walked before me,
with Maluta, crying:
“Way for the ambassador to the czar and czarevitch!”
The Streltsi at the gate stared curiously; they had been well enough
supplied with liquor to be happy, but they could not let me pass until
one of their number, who knew Von Gaden, was called.
“Is this the Jew doctor?” they demanded, “for, by Saint Nikolas of
Mojaïsk, he shall not escape us!”
But the fellow, happily, shook his head. “This is not the Jew!” he
declared.
And my Streltsi, all the while, cried out:
“’Tis an ambassador from the French King to their czarish majesties;
the okolnitchy bade me pass him. Way there, or your beards will be
plucked out by the roots, and ye will have the pravezh!”
But I found a speedier remedy, and quietly drew some money from
my wallet and cast a handful of it among them, and in the fierce
scramble for it, I slipped through the gate and hurried away towards
the Red Place, at the top of my speed, with Maluta at my heels. I
who, but yesterday, had struggled so hard to get out of the Kremlin,
had now paid high to be in again, and rejoiced at my good fortune. I
clapped my hand on the dwarf’s shoulder.
“Maluta,” I cried, “you deserve to be a prime minister, I——”
But he put his finger on his lips, and catching hold of the skirt of my
coat, hurried along under the shadow of the Cathedral of Saint Basil
the Blessed. At last he stopped and pointed at a narrow door in the
lower part of the Palace of Facets.
“Yonder,” he whispered, “beyond that door, is a room with no
windows, and in it is the Princess Daria.”
I caught my breath sharply. “Let us go in,” I cried.
But he seized my sleeve. “Stay,” he whispered, and his thin, three-
cornered face showed white in the dusk.
“In the entry there is a soldier, but he drinks good wine,” he said,
“wine that thy servant bought, and also the stuff from the Gostinnoi
Dvor, and at this hour he troubles us but little. But one false step—
one outcry—one whisper—and——”
He put his hand to his throat with a significant gesture.
“It is death,” he said, “for you—and for her, the shaved head!”
And that was a living death for a Russian woman.
 XXVII: “IS IT THOU?”
WE stood in the parvis of the cathedral, in the deep shadows that
fall at dusk, but the light still streamed broadly over the Red Place—
the Palace Beautiful, as the Russians love to call it—and across the
square I caught the gleam of golden crosses against the sky. It was
silent—with the silence of terror. A guard of Streltsi sat on the Red
Staircase, throwing dice, but their voices were subdued. Some dark
figures flitted softly across the portico, between the cathedral and
the banqueting-hall. Lights shone fitfully in the windows of the
palace. An officer of the guard lighted a pine torch, and it flared up
in a vivid tongue of flame, throwing his burly figure and dark face
into relief. He was scarcely twenty yards away from the spot where
Maluta and I stood waiting. I pressed my hand heavily on the
dwarf’s shoulder.
“Whether it is death or not, I must enter that prison,” I said, in a low
tone.
He nodded, motioning to me to be silent, and pointing. I followed
the direction of his finger with my eyes and saw a short, stout
female figure on the portico, accompanied by two men. The three
stood a while talking together, in subdued voices, and then they
entered the palace and I drew my breath more easily. At least, they
had not entered by that low door that would lead them to the
princess. Meanwhile the officer with the torch had walked away and
the soldiers were busy with their dice. Maluta stripped off a short
cloak he wore and drew his dagger.
“I am going in,” he said, “and when I whistle, the door will be
unbarred. Then it will rest with you, O excellency!”
“Go, then,” I exclaimed, with impatience, “and make haste, in the
name of your patron saint!”
He stooped low and scudded across the space that intervened
between the cathedral and that low door, and I watched eagerly to
see how he intended to enter, for I purposed following without more
delay. But a few moments sufficed to show me the futility of such a
design, for the dwarf did not go to the door; he made straight for a
window ten feet away—to the left—and swinging himself up on the
ledges of the stone copings, as I had seen him do that first day in
Kurakin’s court-yard, he reached the window-sill, and here I thought
to see him balked, for I could plainly perceive the iron bars across it.
However, I was destined to another surprise, for he stopped, and,
hanging on the ledge, began to work at the farthest bar on the left-
hand side, and in a moment he had it out and began to wriggle
through the opening, which was scarcely large enough, to my eyes,
to admit a monkey, and, small as the dwarf was, he had much ado
to squeeze and twist his body through, but finally he disappeared
within, and left me writhing with impatience without. The creature’s
cleverness had, by this time, so far impressed me that I was given to
expecting marvels from him, and I knew that he had a liberty about
the palace—in common with the other dwarfs—that no one else had,
or even dreamed of. So I forced myself to await his signal and to use
all the caution that I could, reflecting that a misstep now would ruin
the adventure, and, perhaps, separate me from the princess. In this
frame of mind, certainly not a happy one, I waited in the ever
increasing shadow, and watched and listened for ten minutes, which
I took—in that mood—for two hours, though sober reflection
afterwards showed me that the time was short, as men count it. I
thought of Maluta’s possible failure, of the death or removal of the
prisoner, while we had delayed, of a change of guard, of the arrival
of Sophia, of a dozen things, in fact, that might wreck my happiness
for all time.
And then I heard Maluta’s shrill whistle.
I drew my sword, trusting the silent weapon rather than the noisy
one, and in a moment I was at the door and had pushed it inward.
So far all was well: I stepped in and closed it quickly behind me.
Before me I saw a long narrow passage lighted by a single lanthorn
swung on the wall. Under it stood Maluta, and farther on I saw the
figure of a man sitting at a table with his back toward me. The light
shone on me, and Maluta saw me grasp my weapon more tightly,
and he laughed—his shrill little laugh.
“He sleeps, O my master,” he said, wagging his head; “he sleeps well
—come and look at him.”
I thought the rogue had killed him, but a glance told me that the
man slept heavily, as a drunken or a drugged person does. His head
hung forward on his breast and his mouth was open, while he
breathed deeply, and his arms hung limp at his sides. I drew a
bunch of keys from his belt.
“Is that the door?” I asked Maluta, pointing at the end of the
passage.
He nodded and I went swiftly forward. It was strongly secured with
both locks and bars, but I removed the latter and found a key, in the
guard’s bunch, to undo the former. Then I tapped upon the door and
called the princess by name. At first there was no answer, nor indeed
any sound within, though there was a grille in the upper half,
doubtless to admit the air, and I was on the point of opening it when
I heard a soft step on the other side, and then silence as though she
listened.
“Princess,” I called gently, “Princess Daria—open the door, I pray
you.”
I heard a little cry—quivering and soft and, I thought, joyful—and
the door was opened.
“Is it thou?” she exclaimed eagerly.
A sudden joy filled my heart, I took a step forward flushed,
expectant, forgetful of all but her and my love for her. But her cell
was lighted, too, by a lanthorn, and as its rays fell on my face she
retreated—not angrily, or even coldly—but with a sudden timidity
that chilled me. After all, the welcome was not for me, but for her
deliverer.