Experience Seamless Full Ebook Downloads for Every Genre at textbookfull.

com

Probability and Computing Randomization and

Probabilistic Techniques in Algorithms and Data
Analysis 2nd Edition Michael Mitzenmacher

https://textbookfull.com/product/probability-and-computing-
randomization-and-probabilistic-techniques-in-algorithms-
and-data-analysis-2nd-edition-michael-mitzenmacher/

OR CLICK BUTTON

DOWNLOAD NOW

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

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

Probability and Computing Randomization and Probabilistic

Techniques in Algorithms and Data Analysis 2 Edition
Edition Michael Mitzenmacher
https://textbookfull.com/product/probability-and-computing-
randomization-and-probabilistic-techniques-in-algorithms-and-data-
analysis-2-edition-edition-michael-mitzenmacher/
textboxfull.com

Probabilistic data structures and algorithms for big data

applications Gakhov

https://textbookfull.com/product/probabilistic-data-structures-and-
algorithms-for-big-data-applications-gakhov/

textboxfull.com

Inequalities in Analysis and Probability 2nd Edition Odile

Pons

https://textbookfull.com/product/inequalities-in-analysis-and-
probability-2nd-edition-odile-pons/

textboxfull.com

Algorithms Design Techniques and Analysis M H Alsuwaiyel

https://textbookfull.com/product/algorithms-design-techniques-and-
analysis-m-h-alsuwaiyel/

textboxfull.com
Randomization in clinical trials : theory and practice 2nd
Edition Lachin

https://textbookfull.com/product/randomization-in-clinical-trials-
theory-and-practice-2nd-edition-lachin/

textboxfull.com

Analysis for Computer Scientists Foundations Methods and

Algorithms Michael Oberguggenberger

https://textbookfull.com/product/analysis-for-computer-scientists-
foundations-methods-and-algorithms-michael-oberguggenberger/

textboxfull.com

Analysis Techniques for Racecar Data Acquisition 2nd

Edition Jorge Segers

https://textbookfull.com/product/analysis-techniques-for-racecar-data-
acquisition-2nd-edition-jorge-segers/

textboxfull.com

An Introduction to Categorical Data Analysis 3rd Edition

Wiley Series in Probability and Statistics Agresti

https://textbookfull.com/product/an-introduction-to-categorical-data-
analysis-3rd-edition-wiley-series-in-probability-and-statistics-
agresti/
textboxfull.com

Scientific Computing and Algorithms in Industrial

Simulations Projects and Products of Fraunhofer SCAI 1st
Edition Michael Griebel
https://textbookfull.com/product/scientific-computing-and-algorithms-
in-industrial-simulations-projects-and-products-of-fraunhofer-
scai-1st-edition-michael-griebel/
textboxfull.com
 Probability and Computing

Randomization and probabilistic techniques play an important role in modern computer

science, with applications ranging from combinatorial optimization and machine learning
to communication networks and secure protocols.
This textbook provides an indispensable teaching tool to accompany a one- or two-
semester course for advanced undergraduate or beginning graduate students in computer
science and applied mathematics. It offers a comprehensive introduction to the role of ran-
domization and probabilistic techniques in modern computer science, in particular to tech-
niques and paradigms used in the development and probabilistic analysis of algorithms
and for data analyses. It assumes only an elementary background in discrete mathematics
and gives a rigorous yet accessible treatment of the material, with numerous examples and
applications.
The first half of the book covers core material, including random sampling, expecta-
tions, Markov’s inequality, Chebyshev’s inequality, Chernoff bounds, balls-and-bins mod-
els, the probabilistic method, and Markov chains. In the second half, the authors delve
into more advanced topics such as continuous probability, applications of limited indepen-
dence, entropy, Markov chain Monte Carlo methods, coupling, martingales, and balanced
allocations.
This greatly expanded new edition includes several newly added chapters and sec-
tions, covering topics including normal distributions, sample complexity, VC dimension,
Rademacher complexity, power laws and related distributions, cuckoo hashing, and appli-
cations of the Lovász Local Lemma. New material relevant to machine learning and big
data analysis enables students to learn up-to-date techniques and applications. Among the
many new exercises and examples are programming-related exercises that provide students
with practical experience and training related to the theoretical concepts covered in the text.

Michael Mitzenmacher is a Professor of Computer Science in the School of Engineering

and Applied Sciences at Harvard University, where he was also the Area Dean for Com-
puter Science from 2010 to 2013. Michael has authored or co-authored over 200 confer-
ence and journal publications on a variety of topics, including algorithms for the Internet,
efficient hash-based data structures, erasure and error-correcting codes, power laws, and
compression. His work on low-density parity-check codes shared the 2002 IEEE Informa-
tion Theory Society Best Paper Award and won the 2009 ACM SIGCOMM Test of Time
Award. He is an ACM Fellow, and was elected as the Chair of the ACM Special Interest
Group on Algorithms and Computation Theory in 2015.
Eli Upfal is a Professor of Computer Science at Brown University, where he was also the
department chair from 2002 to 2007. Prior to joining Brown in 1998, he was a researcher and
project manager at the IBM Almaden Research Center, and a Professor of Applied Math-
ematics and Computer Science at the Weizmann Institute of Science. His main research
interests are randomized algorithms, probabilistic analysis of algorithms, and computa-
tional statistics, with applications ranging from combinatorial and stochastic optimization,
computational biology, and computational finance. He is a Fellow of both the IEEE and the
ACM.
Probability and Computing
Randomization and Probabilistic
Techniques in Algorithms and
Data Analysis

Second Edition

Michael Mitzenmacher Eli Upfal

 University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi - 110002, India
79 Anson Road, #06-04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781107154889
10.1017/9781316651124
© Michael Mitzenmacher and Eli Upfal 2017
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2017
Printed in the United States of America by Sheridan Books, Inc.
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging in Publication Data
Names: Mitzenmacher, Michael, 1969– author. | Upfal, Eli, 1954– author.
Title: Probability and computing / Michael Mitzenmacher Eli Upfal.
Description: Second edition. | Cambridge, United Kingdom ;
New York, NY, USA : Cambridge University Press, [2017] |
Includes bibliographical references and index.
Identifiers: LCCN 2016041654 | ISBN 9781107154889
Subjects: LCSH: Algorithms. | Probabilities. | Stochastic analysis.
Classification: LCC QA274.M574 2017 | DDC 518/.1 – dc23
LC record available at https://lccn.loc.gov/2016041654
ISBN 978-1-107-15488-9 Hardback
Additional resources for this publication at www.cambridge.org/Mitzenmacher.
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party Internet Web sites referred to in this publication
and does not guarantee that any content on such Web sites is, or will remain,
accurate or appropriate.
 To

Stephanie, Michaela, Jacqueline, and Chloe

M.M.

Liane, Tamara, and Ilan

E.U.
 Contents

Preface to the Second Edition page xv

Preface to the First Edition xvii

1 Events and Probability 1

1.1 Application: Verifying Polynomial Identities 1
1.2 Axioms of Probability 3
1.3 Application: Verifying Matrix Multiplication 8
1.4 Application: Naïve Bayesian Classifier 12
1.5 Application: A Randomized Min-Cut Algorithm 15
1.6 Exercises 17

2 Discrete Random Variables and Expectation 23

2.1 Random Variables and Expectation 23
2.1.1 Linearity of Expectations 25
2.1.2 Jensen’s Inequality 26
2.2 The Bernoulli and Binomial Random Variables 27
2.3 Conditional Expectation 29
2.4 The Geometric Distribution 33
2.4.1 Example: Coupon Collector’s Problem 35
2.5 Application: The Expected Run-Time of Quicksort 37
2.6 Exercises 40

3 Moments and Deviations 47

3.1 Markov’s Inequality 47
3.2 Variance and Moments of a Random Variable 48
3.2.1 Example: Variance of a Binomial Random Variable 51

vii
 contents

3.3 Chebyshev’s Inequality 51

3.3.1 Example: Coupon Collector’s Problem 53
3.4 Median and Mean 55
3.5 Application: A Randomized Algorithm for Computing the Median 57
3.5.1 The Algorithm 58
3.5.2 Analysis of the Algorithm 59
3.6 Exercises 62

4 Chernoff and Hoeffding Bounds 66

4.1 Moment Generating Functions 66
4.2 Deriving and Applying Chernoff Bounds 68
4.2.1 Chernoff Bounds for the Sum of Poisson Trials 68
4.2.2 Example: Coin Flips 72
4.2.3 Application: Estimating a Parameter 72
4.3 Better Bounds for Some Special Cases 73
4.4 Application: Set Balancing 76
4.5 The Hoeffding Bound 77
4.6∗ Application: Packet Routing in Sparse Networks 79
4.6.1 Permutation Routing on the Hypercube 80
4.6.2 Permutation Routing on the Butterfly 85
4.7 Exercises 90

5 Balls, Bins, and Random Graphs 97

5.1 Example: The Birthday Paradox 97
5.2 Balls into Bins 99
5.2.1 The Balls-and-Bins Model 99
5.2.2 Application: Bucket Sort 101
5.3 The Poisson Distribution 101
5.3.1 Limit of the Binomial Distribution 105
5.4 The Poisson Approximation 107
5.4.1∗ Example: Coupon Collector’s Problem, Revisited 111
5.5 Application: Hashing 113
5.5.1 Chain Hashing 113
5.5.2 Hashing: Bit Strings 114
5.5.3 Bloom Filters 116
5.5.4 Breaking Symmetry 118
5.6 Random Graphs 119
5.6.1 Random Graph Models 119
5.6.2 Application: Hamiltonian Cycles in Random Graphs 121
5.7 Exercises 127
5.8 An Exploratory Assignment 133

6 The Probabilistic Method 135

6.1 The Basic Counting Argument 135

viii
 contents

6.2 The Expectation Argument 137

6.2.1 Application: Finding a Large Cut 138
6.2.2 Application: Maximum Satisfiability 139
6.3 Derandomization Using Conditional Expectations 140
6.4 Sample and Modify 142
6.4.1 Application: Independent Sets 142
6.4.2 Application: Graphs with Large Girth 143
6.5 The Second Moment Method 143
6.5.1 Application: Threshold Behavior in Random Graphs 144
6.6 The Conditional Expectation Inequality 145
6.7 The Lovász Local Lemma 147
6.7.1 Application: Edge-Disjoint Paths 150
6.7.2 Application: Satisfiability 151
6.8∗ Explicit Constructions Using the Local Lemma 152
6.8.1 Application: A Satisfiability Algorithm 152
6.9 Lovász Local Lemma: The General Case 155
6.10∗ The Algorithmic Lovász Local Lemma 158
6.11 Exercises 162

7 Markov Chains and Random Walks 168

7.1 Markov Chains: Definitions and Representations 168
7.1.1 Application: A Randomized Algorithm for 2-Satisfiability 171
7.1.2 Application: A Randomized Algorithm for 3-Satisfiability 174
7.2 Classification of States 178
7.2.1 Example: The Gambler’s Ruin 181
7.3 Stationary Distributions 182
7.3.1 Example: A Simple Queue 188
7.4 Random Walks on Undirected Graphs 189
7.4.1 Application: An s–t Connectivity Algorithm 192
7.5 Parrondo’s Paradox 193
7.6 Exercises 198

8 Continuous Distributions and the Poisson Process 205

8.1 Continuous Random Variables 205
8.1.1 Probability Distributions in R 205
8.1.2 Joint Distributions and Conditional Probability 208
8.2 The Uniform Distribution 210
8.2.1 Additional Properties of the Uniform Distribution 211
8.3 The Exponential Distribution 213
8.3.1 Additional Properties of the Exponential Distribution 214
8.3.2∗ Example: Balls and Bins with Feedback 216
8.4 The Poisson Process 218
8.4.1 Interarrival Distribution 221
ix
 contents

8.4.2 Combining and Splitting Poisson Processes 222

8.4.3 Conditional Arrival Time Distribution 224
8.5 Continuous Time Markov Processes 226
8.6 Example: Markovian Queues 229
8.6.1 M/M/1 Queue in Equilibrium 230
8.6.2 M/M/1/K Queue in Equilibrium 233
8.6.3 The Number of Customers in an M/M/∞ Queue 233
8.7 Exercises 236

9 The Normal Distribution 242

9.1 The Normal Distribution 242
9.1.1 The Standard Normal Distribution 242
9.1.2 The General Univariate Normal Distribution 243
9.1.3 The Moment Generating Function 246
9.2∗ Limit of the Binomial Distribution 247
9.3 The Central Limit Theorem 249
9.4∗ Multivariate Normal Distributions 252
9.4.1 Properties of the Multivariate Normal Distribution 255
9.5 Application: Generating Normally Distributed Random Values 256
9.6 Maximum Likelihood Point Estimates 258
9.7 Application: EM Algorithm For a Mixture of Gaussians 261
9.8 Exercises 265

10 Entropy, Randomness, and Information 269

10.1 The Entropy Function 269
10.2 Entropy and Binomial Coefficients 272
10.3 Entropy: A Measure of Randomness 274
10.4 Compression 278
10.5∗ Coding: Shannon’s Theorem 281
10.6 Exercises 290

11 The Monte Carlo Method 297

11.1 The Monte Carlo Method 297
11.2 Application: The DNF Counting Problem 300
11.2.1 The Naïve Approach 300
11.2.2 A Fully Polynomial Randomized Scheme for DNF Counting 302
11.3 From Approximate Sampling to Approximate Counting 304
11.4 The Markov Chain Monte Carlo Method 308
11.4.1 The Metropolis Algorithm 310
11.5 Exercises 312
11.6 An Exploratory Assignment on Minimum Spanning Trees 315

x
 contents

12 Coupling of Markov Chains 317

12.1 Variation Distance and Mixing Time 317
12.2 Coupling 320
12.2.1 Example: Shuffling Cards 321
12.2.2 Example: Random Walks on the Hypercube 322
12.2.3 Example: Independent Sets of Fixed Size 323
12.3 Application: Variation Distance Is Nonincreasing 324
12.4 Geometric Convergence 327
12.5 Application: Approximately Sampling Proper
Colorings 328
12.6 Path Coupling 332
12.7 Exercises 336

13 Martingales 341
13.1 Martingales 341
13.2 Stopping Times 343
13.2.1 Example: A Ballot Theorem 345
13.3 Wald’s Equation 346
13.4 Tail Inequalities for Martingales 349
13.5 Applications of the Azuma–Hoeffding Inequality 351
13.5.1 General Formalization 351
13.5.2 Application: Pattern Matching 353
13.5.3 Application: Balls and Bins 354
13.5.4 Application: Chromatic Number 355
13.6 Exercises 355

14 Sample Complexity, VC Dimension, and Rademacher

Complexity 361
14.1 The Learning Setting 362
14.2 VC Dimension 363
14.2.1 Additional Examples of VC Dimension 365
14.2.2 Growth Function 366
14.2.3 VC dimension component bounds 368
14.2.4 -nets and -samples 369
14.3 The -net Theorem 370
14.4 Application: PAC Learning 374
14.5 The -sample Theorem 377
14.5.1 Application: Agnostic Learning 379
14.5.2 Application: Data Mining 380
14.6 Rademacher Complexity 382
14.6.1 Rademacher Complexity and Sample Error 385

xi
 contents

14.6.2 Estimating the Rademacher Complexity 387

14.6.3 Application: Agnostic Learning of a Binary Classification 388
14.7 Exercises 389

15 Pairwise Independence and Universal Hash Functions 392

15.1 Pairwise Independence 392
15.1.1 Example: A Construction of Pairwise Independent Bits 393
15.1.2 Application: Derandomizing an Algorithm for Large Cuts 394
15.1.3 Example: Constructing Pairwise Independent Values Modulo
a Prime 395
15.2 Chebyshev’s Inequality for Pairwise Independent Variables 396
15.2.1 Application: Sampling Using Fewer Random Bits 397
15.3 Universal Families of Hash Functions 399
15.3.1 Example: A 2-Universal Family of Hash Functions 401
15.3.2 Example: A Strongly 2-Universal Family of Hash Functions 402
15.3.3 Application: Perfect Hashing 404
15.4 Application: Finding Heavy Hitters in Data Streams 407
15.5 Exercises 411

16 Power Laws and Related Distributions 415

16.1 Power Law Distributions: Basic Definitions and Properties 416
16.2 Power Laws in Language 418
16.2.1 Zipf’s Law and Other Examples 418
16.2.2 Languages via Optimization 419
16.2.3 Monkeys Typing Randomly 419
16.3 Preferential Attachment 420
16.3.1 A Formal Version 422
16.4 Using the Power Law in Algorithm Analysis 425
16.5 Other Related Distributions 427
16.5.1 Lognormal Distributions 427
16.5.2 Power Law with Exponential Cutoff 428
16.6 Exercises 429

17 Balanced Allocations and Cuckoo Hashing 433

17.1 The Power of Two Choices 433
17.1.1 The Upper Bound 433
17.2 Two Choices: The Lower Bound 438
17.3 Applications of the Power of Two Choices 441
17.3.1 Hashing 441
17.3.2 Dynamic Resource Allocation 442
17.4 Cuckoo Hashing 442
17.5 Extending Cuckoo Hashing 452
17.5.1 Cuckoo Hashing with Deletions 452

xii
 contents

17.5.2 Handling Failures 453

17.5.3 More Choices and Bigger Bins 454
17.6 Exercises 456

Further Reading 463

Index 464

Note: Asterisks indicate advanced material for this chapter.

xiii
 Preface to the Second Edition

In the ten years since the publication of the first edition of this book, probabilistic
methods have become even more central to computer science, rising with the growing
importance of massive data analysis, machine learning, and data mining. Many of the
successful applications of these areas rely on algorithms and heuristics that build on
sophisticated probabilistic and statistical insights. Judicious use of these tools requires
a thorough understanding of the underlying mathematical concepts. Most of the new
material in this second edition focuses on these concepts.
The ability in recent years to create, collect, and store massive data sets, such as
the World Wide Web, social networks, and genome data, lead to new challenges in
modeling and analyzing such structures. A good foundation for models and analysis
comes from understanding some standard distributions. Our new chapter on the nor-
mal distribution (also known as the Gaussian distribution) covers the most common
statistical distribution, as usual with an emphasis on how it is used in settings in com-
puter science, such as for tail bounds. However, an interesting phenomenon is that in
many modern data sets, including social networks and the World Wide Web, we do not
see normal distributions, but instead we see distributions with very different proper-
ties, most notably unusually heavy tails. For example, some pages in the World Wide
Web have an unusually large number of pages that link to them, orders of magnitude
larger than the average. The new chapter on power laws and related distributions covers
specific distributions that are important for modeling and understanding these kinds of
modern data sets.
Machine learning is one of the great successes of computer science in recent years,
providing efficient tools for modeling, understanding, and making predictions based on
large data sets. A question that is often overlooked in practical applications of machine
learning is the accuracy of the predictions, and in particular the relation between accu-
racy and the sample size. A rigorous introduction to approaches to these important
questions is presented in a new chapter on sample complexity, VC dimension, and
Rademacher averages.

xv
 preface to the second edition

We have also used the new edition to enhance some of our previous material. For
example, we present some of the recent advances on algorithmic variations of the pow-
erful Lovász local lemma, and we have a new section covering the wonderfully named
and increasingly useful hashing approach known as cuckoo hashing. Finally, in addi-
tion to all of this new material, the new edition includes updates and corrections, and
many new exercises.
We thank the many readers who sent us corrections over the years – unfortunately,
too many to list here!

xvi
 Preface to the First Edition

Why Randomness?

Why should computer scientists study and use randomness? Computers appear to
behave far too unpredictably as it is! Adding randomness would seemingly be a dis-
advantage, adding further complications to the already challenging task of efficiently
utilizing computers.
Science has learned in the last century to accept randomness as an essential com-
ponent in modeling and analyzing nature. In physics, for example, Newton’s laws led
people to believe that the universe was a deterministic place; given a big enough calcu-
lator and the appropriate initial conditions, one could determine the location of planets
years from now. The development of quantum theory suggests a rather different view;
the universe still behaves according to laws, but the backbone of these laws is proba-
bilistic. “God does not play dice with the universe” was Einstein’s anecdotal objection
to modern quantum mechanics. Nevertheless, the prevailing theory today for subparti-
cle physics is based on random behavior and statistical laws, and randomness plays a
significant role in almost every other field of science ranging from genetics and evolu-
tion in biology to modeling price fluctuations in a free-market economy.
Computer science is no exception. From the highly theoretical notion of probabilis-
tic theorem proving to the very practical design of PC Ethernet cards, randomness
and probabilistic methods play a key role in modern computer science. The last two
decades have witnessed a tremendous growth in the use of probability theory in comput-
ing. Increasingly more advanced and sophisticated probabilistic techniques have been
developed for use within broader and more challenging computer science applications.
In this book, we study the fundamental ways in which randomness comes to bear on
computer science: randomized algorithms and the probabilistic analysis of algorithms.
Randomized algorithms: Randomized algorithms are algorithms that make random
choices during their execution. In practice, a randomized program would use values
generated by a random number generator to decide the next step at several branches
of its execution. For example, the protocol implemented in an Ethernet card uses ran-
dom numbers to decide when it next tries to access the shared Ethernet communication
xvii
 preface to the first edition

medium. The randomness is useful for breaking symmetry, preventing different cards
from repeatedly accessing the medium at the same time. Other commonly used applica-
tions of randomized algorithms include Monte Carlo simulations and primality testing
in cryptography. In these and many other important applications, randomized algo-
rithms are significantly more efficient than the best known deterministic solutions.
Furthermore, in most cases the randomized algorithms are also simpler and easier to
program.
These gains come at a price; the answer may have some probability of being incor-
rect, or the efficiency is guaranteed only with some probability. Although it may seem
unusual to design an algorithm that may be incorrect, if the probability of error is suf-
ficiently small then the improvement in speed or memory requirements may well be
worthwhile.
Probabilistic analysis of algorithms: Complexity theory tries to classify computa-
tion problems according to their computational complexity, in particular distinguishing
between easy and hard problems. For example, complexity theory shows that the Trav-
eling Salesman problem is NP-hard. It is therefore very unlikely that we will ever know
an algorithm that can solve any instance of the Traveling Salesman problem in time that
is subexponential in the number of cities. An embarrassing phenomenon for the clas-
sical worst-case complexity theory is that the problems it classifies as hard to compute
are often easy to solve in practice. Probabilistic analysis gives a theoretical explanation
for this phenomenon. Although these problems may be hard to solve on some set of
pathological inputs, on most inputs (in particular, those that occur in real-life applica-
tions) the problem is actually easy to solve. More precisely, if we think of the input as
being randomly selected according to some probability distribution on the collection of
all possible inputs, we are very likely to obtain a problem instance that is easy to solve,
and instances that are hard to solve appear with relatively small probability. Probabilis-
tic analysis of algorithms is the method of studying how algorithms perform when the
input is taken from a well-defined probabilistic space. As we will see, even NP-hard
problems might have algorithms that are extremely efficient on almost all inputs.

The Book

This textbook is designed to accompany one- or two-semester courses for advanced

undergraduate or beginning graduate students in computer science and applied math-
ematics. The study of randomized and probabilistic techniques in most leading uni-
versities has moved from being the subject of an advanced graduate seminar meant
for theoreticians to being a regular course geared generally to advanced undergraduate
and beginning graduate students. There are a number of excellent advanced, research-
oriented books on this subject, but there is a clear need for an introductory textbook.
We hope that our book satisfies this need.
The textbook has developed from courses on probabilistic methods in computer sci-
ence taught at Brown (CS 155) and Harvard (CS 223) in recent years. The emphasis in
these courses and in this textbook is on the probabilistic techniques and paradigms, not
on particular applications. Each chapter of the book is devoted to one such method or
xviii
 preface to the first edition

technique. Techniques are clarified though examples based on analyzing randomized

algorithms or developing probabilistic analysis of algorithms on random inputs. Many
of these examples are derived from problems in networking, reflecting a prominent
trend in the networking field (and the taste of the authors).
The book contains fourteen chapters. We may view the book as being divided into
two parts, where the first part (Chapters 1–7) comprises what we believe is core mate-
rial. The book assumes only a basic familiarity with probability theory, equivalent to
what is covered in a standard course on discrete mathematics for computer scientists.
Chapters 1–3 review this elementary probability theory while introducing some inter-
esting applications. Topics covered include random sampling, expectation, Markov’s
inequality, variance, and Chebyshev’s inequality. If the class has sufficient background
in probability, then these chapters can be taught quickly. We do not suggest skipping
them, however, because they introduce the concepts of randomized algorithms and
probabilistic analysis of algorithms and also contain several examples that are used
throughout the text.
Chapters 4–7 cover more advanced topics, including Chernoff bounds, balls-and-
bins models, the probabilistic method, and Markov chains. The material in these chap-
ters is more challenging than in the initial chapters. Sections that are particularly chal-
lenging (and hence that the instructor may want to consider skipping) are marked with
an asterisk. The core material in the first seven chapters may constitute the bulk of a
quarter- or semester-long course, depending on the pace.
The second part of the book (Chapters 8–17) covers additional advanced material
that can be used either to fill out the basic course as necessary or for a more advanced
second course. These chapters are largely self-contained, so the instructor can choose
the topics best suited to the class. The chapters on continuous probability and entropy
are perhaps the most appropriate for incorporating into the basic course. Our intro-
duction to continuous probability (Chapter 8) focuses on uniform and exponential
distributions, including examples from queueing theory. Our examination of entropy
(Chapter 10) shows how randomness can be measured and how entropy arises naturally
in the context of randomness extraction, compression, and coding.
Chapters 11 and 12 cover the Monte Carlo method and coupling, respectively; these
chapters are closely related and are best taught together. Chapter 13, on martingales,
covers important issues on dealing with dependent random variables, a theme that con-
tinues in a different vein in Chapter 15 is the development of pairwise independence
and derandomization. Finally, the chapter on balanced allocations (Chapter 17) covers
a topic close to the authors’ hearts and ties in nicely with Chapter 5 concerning analysis
of balls-and-bins problems.
The order of the subjects, especially in the first part of the book, corresponds to
their relative importance in the algorithmic literature. Thus, for example, the study
of Chernoff bounds precedes more fundamental probability concepts such as Markov
chains. However, instructors may choose to teach the chapters in a different order. A
course with more emphasis on general stochastic processes, for example, may teach
Markov chains (Chapter 7) immediately after Chapters 1–3, following with the chapter
on balls, bins, and random graphs (Chapter 5, omitting the Hamiltonian cycle exam-
ple). Chapter 6 on the probabilistic method could then be skipped, following instead
xix
 preface to the first edition

with continuous probability and the Poisson process (Chapter 8). The material from
Chapter 4 on Chernoff bounds, however, is needed for most of the remaining material.
Most of the exercises in the book are theoretical, but we have included some pro-
gramming exercises – including two more extensive exploratory assignments that
require some programming. We have found that occasional programming exercises are
often helpful in reinforcing the book’s ideas and in adding some variety to the course.
We have decided to restrict the material in this book to methods and techniques based
on rigorous mathematical analysis; with few exceptions, all claims in this book are fol-
lowed by full proofs. Obviously, many extremely useful probabilistic methods do not
fall within this strict category. For example, in the important area of Monte Carlo meth-
ods, most practical solutions are heuristics that have been demonstrated to be effective
and efficient by experimental evaluation rather than by rigorous mathematical analy-
sis. We have taken the view that, in order to best apply and understand the strengths
and weaknesses of heuristic methods, a firm grasp of underlying probability theory and
rigorous techniques – as we present in this book – is necessary. We hope that students
will appreciate this point of view by the end of the course.

Acknowledgments

Our first thanks go to the many probabilists and computer scientists who developed
the beautiful material covered in this book. We chose not to overload the textbook
with numerous references to the original papers. Instead, we provide a reference list
that includes a number of excellent books giving background material as well as more
advanced discussion of the topics covered here.
The book owes a great deal to the comments and feedback of students and teaching
assistants who took the courses CS 155 at Brown and CS 223 at Harvard. In particular
we wish to thank Aris Anagnostopoulos, Eden Hochbaum, Rob Hunter, and Adam
Kirsch, all of whom read and commented on early drafts of the book.
Special thanks to Dick Karp, who used a draft of the book in teaching CS 174 at
Berkeley during fall 2003. His early comments and corrections were most valuable in
improving the manuscript. Peter Bartlett taught CS 174 at Berkeley in spring 2004, also
providing many corrections and useful comments.
We thank our colleagues who carefully read parts of the manuscript, pointed out
many errors, and suggested important improvements in content and presentation: Artur
Czumaj, Alan Frieze, Claire Kenyon, Joe Marks, Salil Vadhan, Eric Vigoda, and the
anonymous reviewers who read the manuscript for the publisher.
We also thank Rajeev Motwani and Prabhakar Raghavan for allowing us to use some
of the exercises in their excellent book Randomized Algorithms.
We are grateful to Lauren Cowles of Cambridge University Press for her editorial
help and advice in preparing and organizing the manuscript.
Writing of this book was supported in part by NSF ITR Grant no. CCR-0121154.

xx
 chapter one
Events and Probability

This chapter introduces the notion of randomized algorithms and reviews some basic
concepts of probability theory in the context of analyzing the performance of simple
randomized algorithms for verifying algebraic identities and finding a minimum cut-set
in a graph.

1.1. Application: Verifying Polynomial Identities

Computers can sometimes make mistakes, due for example to incorrect programming
or hardware failure. It would be useful to have simple ways to double-check the results
of computations. For some problems, we can use randomness to efficiently verify the
correctness of an output.
Suppose we have a program that multiplies together monomials. Consider the prob-
lem of verifying the following identity, which might be output by our program:
?
(x + 1)(x − 2)(x + 3)(x − 4)(x + 5)(x − 6) ≡ x6 − 7x3 + 25.
There is an easy way to verify whether the identity is correct: multiply together the
terms on the left-hand side and see if the resulting polynomial matches the right-hand
side. In this example, when we multiply all the constant terms on the left, the result
does not match the constant term on the right, so the identity cannot be valid. More
generally, given two polynomials F (x) and G(x), we can verify the identity
?
F (x) ≡ G(x)
 d i

by converting the two polynomials to their canonical forms i=0 ci x ; two polynomi-
als are equivalent if and only if all the coefficients in their canonical forms are equal.
From thisdpoint on let us assume that, as in our example, F (x) is given as a product
F (x) = i=1 (x − ai ) and G(x) is given in its canonical form. Transforming F (x) to
its canonical form by consecutively multiplying the ith monomial with the product of

1
 events and probability

the first i − 1 monomials requires (d 2 ) multiplications of coefficients. We assume in

what follows that each multiplication can be performed in constant time, although if
the products of the coefficients grow large then it could conceivably require more than
constant time to add and multiply numbers together.
So far, we have not said anything particularly interesting. To check whether the
computer program has multiplied monomials together correctly, we have suggested
multiplying the monomials together again to check the result. Our approach for check-
ing the program is to write another program that does essentially the same thing we
expect the first program to do. This is certainly one way to double-check a program:
write a second program that does the same thing, and make sure they agree. There
are at least two problems with this approach, both stemming from the idea that there
should be a difference between checking a given answer and recomputing it. First, if
there is a bug in the program that multiplies monomials, the same bug may occur in
the checking program. (Suppose that the checking program was written by the same
person who wrote the original program!) Second, it stands to reason that we would like
to check the answer in less time than it takes to try to solve the original problem all over
again.
Let us instead utilize randomness to obtain a faster method to verify the identity. We
informally explain the algorithm and then set up the formal mathematical framework
for analyzing the algorithm.
Assume that the maximum degree, or the largest exponent of x, in F (x) and G(x) is
d. The algorithm chooses an integer r uniformly at random in the range {1, . . . , 100d},
where by “uniformly at random” we mean that all integers are equally likely to be
chosen. The algorithm then computes the values F (r) and G(r). If F (r) = G(r) the
algorithm decides that the two polynomials are not equivalent, and if F (r) = G(r) the
algorithm decides that the two polynomials are equivalent.
Suppose that in one computation step the algorithm can generate an integer chosen
uniformly at random in the range {1, . . . , 100d}. Computing the values of F (r) and
G(r) can be done in O(d) time, which is faster than computing the canonical form of
F (r). The randomized algorithm, however, may give a wrong answer.
How can the algorithm give the wrong answer?
If F (x) ≡ G(x), then the algorithm gives the correct answer, since it will find that
F (r) = G(r) for any value of r.
If F (x) ≡ G(x) and F (r) = G(r), then the algorithm gives the correct answer since
it has found a case where F (x) and G(x) disagree. Thus, when the algorithm decides
that the two polynomials are not the same, the answer is always correct.
If F (x) ≡ G(x) and F (r) = G(r), the algorithm gives the wrong answer. In other
words, it is possible that the algorithm decides that the two polynomials are the
same when they are not. For this error to occur, r must be a root of the equation
F (x) − G(x) = 0. The degree of the polynomial F (x) − G(x) is no larger than d and,
by the fundamental theorem of algebra, a polynomial of degree up to d has no more
than d roots. Thus, if F (x) ≡ G(x), then there are no more than d values in the
range {1, . . . , 100d} for which F (r) = G(r). Since there are 100d values in the range
{1, . . . , 100d}, the chance that the algorithm chooses such a value and returns a wrong
answer is no more than 1/100.
2
 1.2 axioms of probability

1.2. Axioms of Probability

We turn now to a formal mathematical setting for analyzing the randomized algorithm.
Any probabilistic statement must refer to the underlying probability space.
Definition 1.1: A probability space has three components:
1. a sample space , which is the set of all possible outcomes of the random process
modeled by the probability space;
2. a family of sets F representing the allowable events, where each set in F is a subset1
of the sample space ; and
3. a probability function Pr : F → R satisfying Definition 1.2.
An element of  is called a simple or elementary event.
In the randomized algorithm for verifying polynomial identities, the sample space
is the set of integers {1, . . . , 100d}. Each choice of an integer r in this range is a simple
event.
Definition 1.2: A probability function is any function Pr : F → R that satisfies the
following conditions:
1. for any event E, 0 ≤ Pr(E ) ≤ 1;
2. Pr() = 1; and
3. for any finite or countably infinite sequence of pairwise mutually disjoint events
E1 , E2 , E3 , . . . ,
 

Pr Ei = Pr(Ei ).
i≥1 i≥1

In most of this book we will use discrete probability spaces. In a discrete probability
space the sample space  is finite or countably infinite, and the family F of allow-
able events consists of all subsets of . In a discrete probability space, the probability
function is uniquely defined by the probabilities of the simple events.
Again, in the randomized algorithm for verifying polynomial identities, each choice
of an integer r is a simple event. Since the algorithm chooses the integer uniformly at
random, all simple events have equal probability. The sample space has 100d simple
events, and the sum of the probabilities of all simple events must be 1. Therefore each
simple event has probability 1/100d.
Because events are sets, we use standard set theory notation to express combinations
of events. We write E1 ∩ E2 for the occurrence of both E1 and E2 and write E1 ∪ E2 for
the occurrence of either E1 or E2 (or both). For example, suppose we roll two dice. If
E1 is the event that the first die is a 1 and E2 is the event that the second die is a 1, then
E1 ∩ E2 denotes the event that both dice are 1 while E1 ∪ E2 denotes the event that at
least one of the two dice lands on 1. Similarly, we write E1 − E2 for the occurrence

1 In a discrete probability space F = 2 . Otherwise, and introductory readers may skip this point, since the events
need to be measurable, F must include the empty set and be closed under complement and union and intersection
of countably many sets (a σ -algebra).

3
 events and probability

of an event that is in E1 but not in E2 . With the same dice example, E1 − E2 consists
of the event where the first die is a 1 and the second die is not. We use the notation Ē
as shorthand for  − E; for example, if E is the event that we obtain an even number
when rolling a die, then Ē is the event that we obtain an odd number.
Definition 1.2 yields the following obvious lemma.

Lemma 1.1: For any two events E1 and E2 ,

Pr(E1 ∪ E2 ) = Pr(E1 ) + Pr(E2 ) − Pr(E1 ∩ E2 ).

Proof: From the definition,

Pr(E1 ) = Pr(E1 − (E1 ∩ E2 )) + Pr(E1 ∩ E2 ),

Pr(E2 ) = Pr(E2 − (E1 ∩ E2 )) + Pr(E1 ∩ E2 ),
Pr(E1 ∪ E2 ) = Pr(E1 − (E1 ∩ E2 )) + Pr(E2 − (E1 ∩ E2 )) + Pr(E1 ∩ E2 ).

The lemma easily follows. 

A consequence of Definition 1.2 is known as the union bound. Although it is very

simple, it is tremendously useful.

Lemma 1.2: For any finite or countably infinite sequence of events E1 , E2 , . . . ,

 

Pr Ei ≤ Pr(Ei ).
i≥1 i≥1

Notice that Lemma 1.2 differs from the third part of Definition 1.2 in that Definition
1.2 is an equality and requires the events to be pairwise mutually disjoint.
Lemma 1.1 can be generalized to the following equality, often referred to as the
inclusion–exclusion principle.

Lemma 1.3: Let E1 , . . . , En be any n events. Then

 n 
 n
Pr Ei = Pr(Ei ) − Pr(Ei ∩ E j ) + Pr(Ei ∩ E j ∩ Ek )
i=1 i=1 i< j i< j<k
 

+1
− · · · + (−1) Pr Eir + ··· .
i1 <i2 <···<i r=1

The proof of the inclusion–exclusion principle is left as Exercise 1.7.

We showed before that the only case in which the algorithm may fail to give the
correct answer is when the two input polynomials F (x) and G(x) are not equivalent;
the algorithm then gives an incorrect answer if the random number it chooses is a root
of the polynomial F (x) − G(x). Let E represent the event that the algorithm failed to
give the correct answer. The elements of the set corresponding to E are the roots of
the polynomial F (x) − G(x) that are in the set of integers {1, . . . , 100d}. Since the
polynomial has no more than d roots it follows that the event E includes no more than
4
 1.2 axioms of probability

d simple events, and therefore

d 1
Pr(algorithm fails) = Pr(E ) ≤ = .
100d 100
It may seem unusual to have an algorithm that can return the wrong answer. It may
help to think of the correctness of an algorithm as a goal that we seek to optimize in
conjunction with other goals. In designing an algorithm, we generally seek to minimize
the number of computational steps and the memory required. Sometimes there is a
trade-off; there may be a faster algorithm that uses more memory or a slower algorithm
that uses less memory. The randomized algorithm we have presented gives a trade-off
between correctness and speed. Allowing algorithms that may give an incorrect answer
(but in a systematic way) expands the trade-off space available in designing algorithms.
Rest assured, however, that not all randomized algorithms give incorrect answers, as
we shall see.
For the algorithm just described, the algorithm gives the correct answer 99% of
the time even when the polynomials are not equivalent. Can we improve this prob-
ability? One way is to choose the random number r from a larger range of integers.
If our sample space is the set of integers {1, . . . , 1000d}, then the probability of a
wrong answer is at most 1/1000. At some point, however, the range of values we
can use is limited by the precision available on the machine on which we run the
algorithm.
Another approach is to repeat the algorithm multiple times, using different random
values to test the identity. The property we use here is that the algorithm has a one-sided
error. The algorithm may be wrong only when it outputs that the two polynomials are
equivalent. If any run yields a number r such that F (r) = G(r), then the polynomials are
not equivalent. Thus, if we repeat the algorithm a number of times and find F (r) = G(r)
in at least one round of the algorithm, we know that F (x) and G(x) are not equivalent.
The algorithm outputs that the two polynomials are equivalent only if there is equality
for all runs.
In repeating the algorithm we repeatedly choose a random number in the range
{1, . . . , 100d}. Repeatedly choosing random numbers according to a given distribution
is generally referred to as sampling. In this case, we can repeatedly choose random
numbers in the range {1, . . . , 100d} in two ways: we can sample either with replace-
ment or without replacement. Sampling with replacement means that we do not remem-
ber which numbers we have already tested; each time we run the algorithm, we choose
a number uniformly at random from the range {1, . . . , 100d} regardless of previous
choices, so there is some chance we will choose an r that we have chosen on a previous
run. Sampling without replacement means that, once we have chosen a number r, we
do not allow the number to be chosen on subsequent runs; the number chosen at a given
iteration is uniform over all previously unselected numbers.
Let us first consider the case where sampling is done with replacement. Assume
that we repeat the algorithm k times, and that the input polynomials are not equiva-
lent. What is the probability that in all k iterations our random sampling from the set
{1, . . . , 100d} yields roots of the polynomial F (x) − G(x), resulting in a wrong output
by the algorithm? If k = 1, we know that this probability is at most d/100d = 1/100.
5
 events and probability

If k = 2, it seems that the probability that the first iteration finds a root is at most 1/100
and the probability that the second iteration finds a root is at most 1/100, so the prob-
ability that both iterations find a root is at most (1/100)2 . Generalizing, for any k, the
probability of choosing roots for k iterations would be at most (1/100)k .
To formalize this, we introduce the notion of independence.
Definition 1.3: Two events E and F are independent if and only if
Pr(E ∩ F ) = Pr(E ) · Pr(F ).
More generally, events E1 , E2 , . . . , Ek are mutually independent if and only if, for any
subset I ⊆ [1, k],
 
Pr Ei = Pr(Ei ).
i∈I i∈I

If our algorithm samples with replacement then in each iteration the algorithm chooses
a random number uniformly at random from the set {1, . . . , 100d}, and thus the choice
in one iteration is independent of the choices in previous iterations. For the case where
the polynomials are not equivalent, let Ei be the event that, on the ith run of the algo-
rithm, we choose a root ri such that F (ri ) − G(ri ) = 0. The probability that the algo-
rithm returns the wrong answer is given by
Pr(E1 ∩ E2 ∩ · · · ∩ Ek ).
Since Pr(Ei ) is at most d/100d and since the events E1 , E2 , . . . , Ek are independent,
the probability that the algorithm gives the wrong answer after k iterations is
k k k
d 1
Pr(E1 ∩ E2 ∩ · · · ∩ Ek ) = Pr(Ei ) ≤ = .
i=1 i=1
100d 100

The probability of making an error is therefore at most exponentially small in the num-
ber of trials.
Now let us consider the case where sampling is done without replacement. In this
case the probability of choosing a given number is conditioned on the events of the
previous iterations.
Definition 1.4: The conditional probability that event E occurs given that event F
occurs is
Pr(E ∩ F )
Pr(E | F ) = .
Pr(F )
The conditional probability is well-defined only if Pr(F ) > 0.
Intuitively, we are looking for the probability of E ∩ F within the set of events defined
by F. Because F defines our restricted sample space, we normalize the probabilities
by dividing by Pr(F ), so that the sum of the probabilities of all events is 1. When
Pr(F ) > 0, the definition can also be written in the useful form
Pr(E | F ) Pr(F ) = Pr(E ∩ F ).

6
 1.2 axioms of probability

Notice that, when E and F are independent and Pr(F ) = 0, we have

Pr(E ∩ F ) Pr(E ) Pr(F )
Pr(E | F ) = = = Pr(E ).
Pr(F ) Pr(F )
This is a property that conditional probability should have; intuitively, if two events are
independent, then information about one event should not affect the probability of the
second event.
Again assume that we repeat the algorithm k times and that the input polynomials are
not equivalent. What is the probability that in all the k iterations our random sampling
from the set {1, . . . , 100d} yields roots of the polynomial F (x) − G(x), resulting in a
wrong output by the algorithm?
As in the analysis with replacement, we let Ei be the event that the random num-
ber ri chosen in the ith iteration of the algorithm is a root of F (x) − G(x); again, the
probability that the algorithm returns the wrong answer is given by
Pr(E1 ∩ E2 ∩ · · · ∩ Ek ).
Applying the definition of conditional probability, we obtain
Pr(E1 ∩ E2 ∩ · · · ∩ Ek ) = Pr(Ek | E1 ∩ E2 ∩ · · · ∩ Ek−1 ) · Pr(E1 ∩ E2 ∩ · · · ∩ Ek−1 ),
and repeating this argument gives
Pr(E1 ∩ E2 ∩ · · · ∩ Ek )
= Pr(E1 ) · Pr(E2 | E1 ) · Pr(E3 | E1 ∩ E2 ) · · · Pr(Ek | E1 ∩ E2 ∩ · · · ∩ Ek−1 ).
Can we bound Pr(E j | E1 ∩ E2 ∩ · · · ∩ E j−1 )? Recall that there are at most d values
r for which F (r) − G(r) = 0; if trials 1 through j − 1 < d have found j − 1 of them,
then when sampling without replacement there are only d − ( j − 1) values out of the
100d − ( j − 1) remaining choices for which F (r) − G(r) = 0. Hence
d − ( j − 1)
Pr(E j | E1 ∩ E2 ∩ · · · ∩ E j−1 ) ≤ ,
100d − ( j − 1)
and the probability that the algorithm gives the wrong answer after k ≤ d iterations is
bounded by
k k
d − ( j − 1) 1
Pr(E1 ∩ E2 ∩ · · · ∩ Ek ) ≤ ≤ .
j=1
100d − ( j − 1) 100

Because (d − ( j − 1))/(100d − ( j − 1)) < d/100d when j > 1, our bounds on the
probability of making an error are actually slightly better without replacement. You
may also notice that, if we take d + 1 samples without replacement and the two poly-
nomials are not equivalent, then we are guaranteed to find an r such that F (r) − G(r) =
0. Thus, in d + 1 iterations we are guaranteed to output the correct answer. However,
computing the value of the polynomial at d + 1 points takes (d 2 ) time using the stan-
dard approach, which is no faster than finding the canonical form deterministically.
Since sampling without replacement appears to give better bounds on the probability
of error, why would we ever want to consider sampling with replacement? In some
cases, sampling with replacement is significantly easier to analyze, so it may be worth
7
Exploring the Variety of Random
Documents with Different Content
She either did not believe in its truth, or lulled herself into security
by depending upon the fidelity of her friends. Unmoved by the
danger that threatened her, she concealed from her husband the
information she had received; for which, when it was too late to
retrace her steps, he afterwards severely censured her. Ostermann,
who was early made aware of the proceedings of the conspirators,
warned the regent of her danger, and entreated her to take some
decisive measures to avert it: and the British ambassador, detecting,
probably, the insidious hand of France, predicted her destruction in
vain. Her facile nature still lingered inactive, until at last she received
an anonymous letter, in which she was strongly admonished of the
perils by which she was surrounded. A more energetic mind would
have acted unhesitatingly upon these repeated proofs of the
approaching insurrection; but Anna, still clinging to the side of
mercy, instead of seizing upon the ringleaders, who were known to
her, and quieting at once the apprehensions of her advisers, read the
whole contents of the letter in open court in the presence of
Elizabeth, and stated the nature of the reports that had reached her.
Elizabeth, of course, protested her ignorance of the whole business,
burst into a flood of tears, and asserted her innocence with such a
show of sincerity that the regent was perfectly satisfied, and took no
further notice of the matter.
This occurred on the 4th of December, 1741. Lestocq had
previously appointed the day of the consecration of the waters, the
6th of January, 1742, for Elizabeth to make her public appearance at
the head of the guards, to issue declarations setting forth her claims
upon the throne, and to cause herself to be proclaimed. But the
proceeding that had taken place in the court determined him to
hasten his plans. Now that the vigilance of the court was awakened,
he knew that his motions would be watched, and that the affair did
not admit of any further delay. He applied himself, accordingly, with
redoubled vigilance, to the business of collecting and organising the
partisans of the princess; continued to bribe them with French gold;
and, when everything was prepared, he again impressed upon his
mistress the urgent necessity of decision. He pointed out to her that
the guards, upon whose assistance she chiefly relied, were under
orders to march for Sweden, and that in a short time all would be
lost. She was still, however, timid and doubtful of the result, when
the artful Lestocq drew a card from his pocket, which represented
her on one side in the habit of a nun, and on the other with a crown
upon her head—asking her which fate she preferred; adding that the
choice depended upon herself, and upon the promptitude with which
she employed the passing moment. This argument succeeded; she
consented to place herself in his hands; and, remembering the
success that had attended the midnight revolution that consigned
Biron to banishment, he appointed the following night, the 5th of
December, for the execution of his plan—undertaking the principal
part himself, in the hope of the honours that were to be heaped
upon him in the event of success.
When the hour arrived Elizabeth again betrayed irresolution, but
Lestocq overcame her fears; and after having made a solemn vow
before the crucifix that no blood should be shed in the attempt, she
put on the order of St. Catherine, and placing herself in a sledge,
attended by Lestocq and her chamberlain, she drove to the barracks
of the Preobrajenski guards. When she arrived at this point, she
advanced towards the soldiers on foot, holding the cross in her
hand; and, addressing them in a speech of some length, justified the
grounds on which she advanced her claims to the throne; reminded
them that she was the daughter of Peter the Great; that she had
been illegally deprived of the succession; that a foreign child wielded
the imperial sceptre; and that foreigners were advanced, to the
exclusion of native Russians, to the highest offices in the state. A
considerable number of the guards had been previously prepared for
this proceeding by bribes and promises, and inflammatory liquors
were distributed amongst them to heighten their zeal. With the
exception of a few, who would not violate their duty and who were,
in consequence, manacled by the remainder, the whole body
responded to the address with enthusiasm.
They now proceeded to the palace of the emperor and his
parents, pressing into their train everybody they met on the way, to
prevent their object from being betrayed; and, forcing the sentries at
the gates, obtained easy admittance to the sleeping apartments of
the regent and the duke, whom they dragged, unceremoniously, and
without affording them time to dress, out of their beds, and
conveyed to the palace of Elizabeth, where they confined them
under a strong guard. The infant Ivan, unconscious of the misery
that awaited him, was enjoying a gentle slumber during this scene of
violence; and when he awoke he was carried, in a similar manner, to
the place where his unhappy parents were immured. On the same
night the principal persons connected with the government were
seized in the same way, and thrown into prison. Amongst them were
Lewis Ernest of Brunswick, the brother of the duke, Ostermann, and
Munich.
This revolution was as rapid and complete as that which deprived
Biron of the regency, and was effected by a similar stealthy
proceeding in the silence of the night. Early on the following
morning, the inhabitants were called upon to take the oath of fealty
to Elizabeth. But they were accustomed to these sudden movements
in the palace; and before the day was concluded the shouts of the
intoxicated soldiery announced that the people had confirmed, by
the usual attestation of allegiance, the authority of the empress.[49]
A manifesto was immediately issued, which contained the following
statement:
The empress Anna having nominated the grandson of her sister, a
child born into the world only a few weeks before the empress’
death, as successor to the throne; during the minority of whom
various persons had conducted the administration of the empire in a
manner highly iniquitous, whence disturbances had arisen both
within the country and out of it, and probably in time still greater
might arise; therefore all the faithful subjects of Elizabeth, both in
spiritual and temporal stations, particularly the regiments of the life-
guards, had unanimously invited her, for the prevention of all the
mischievous consequences to be apprehended, to take possession of
the throne of her father as nearest by right of birth; and that she
had accordingly resolved to yield to this universal request of her
faithful subjects, by taking possession of her inheritance derived
from her parents, the emperor Peter I and the empress Catherine.
Shortly after this another manifesto appeared, in which Elizabeth
grounded her legitimacy on the will of Catherine I. As the
statements in this document respecting the right of inheritance are
singular in themselves, and as they illustrate in a very remarkable
degree the irregularity with which the question of the succession
was suffered to be treated, the passage touching upon those points
appears to be worthy of preservation. It will be seen, upon reference
to previous facts, that these statements are highly coloured to suit
the demands of the occasion. After some preliminaries, the
manifesto proceeds to observe, that on the demise of Peter II,
whom she (Elizabeth) ought to have succeeded, Anna was elected
through the machinations of Ostermann; and afterwards, when the
sovereign was attacked by a mortal distemper, the same Ostermann
appointed as successor the son of Prince Antony Ulrich of Brunswick
and the princess of Mecklenburg, a child only two months old, who
had not the slightest claim by inheritance to the Russian throne;
and, not content with this, he added, to the prejudice of Elizabeth,
that after Ivan’s death the princes afterwards born of the said prince
of Brunswick and the princess of Mecklenburg should succeed to the
Russian throne; whereas even the parents themselves had not the
slightest right to that throne. That Ivan was, therefore, by the
machinations of Ostermann and Munich, confirmed emperor in
October, 1740; and because the several regiments of guards, as well
as the marching regiments, were under the command of Munich and
the father of Ivan, and consequently the whole force of the empire
was in the hands of those two persons, the subjects were compelled
to take the oath of allegiance to Ivan. That Antony Ulrich and his
spouse had afterwards broken this ordinance, to which they
themselves had sworn; had forcibly seized upon the administration
of the empire; and Anna had resolved, even in the lifetime of her
son Ivan, to place herself upon the throne as empress. That, in
order, then, to prevent all dangerous consequences from these
proceedings, Elizabeth had ascended the throne, and of her own
imperial grace had ordered the princess with her son and daughter
to set out for their native country.
Such were the arguments upon which Elizabeth attempted to
justify her seizure of the throne. With what sincerity she fulfilled the
act of grace towards the regent and her family, expressed in the last
sentence, will be seen hereafter.

ELIZABETH PETROVNA (1741-1762 A.D.)

The revolution which elevated Elizabeth to the throne and the

circumstances which preceded that elevation were in every respect
remarkable. She had no claim to the dignity, either by birth or by the
regulation in regard to the succession introduced by the innovating
Peter. Elizabeth was the younger daughter of Peter: Anna, who had
been married to the duke of Holstein, was the elder; and though this
princess was dead, she left a son, the representative of her rights,
who, as we shall hereafter perceive, did ultimately reign as Peter III.
The right of primogeniture, indeed, had, in the regulation to which
we have alluded, been set aside, and the choice, pure and simple, of
the reigning potentate substituted; but the infant Peter had the
additional claim of being expressly indicated in the will of Catherine
I. These claims, however, had been utterly disregarded when Anna,
duchess of Courland and daughter of Ivan, brother of Peter I, had
been raised by a faction to the throne. On the death of this empress
without issue, Peter, as we have seen, was again overlooked,
through the ambition rather of an individual than of a faction—the
bloodthirsty Biron.
Ivan, the son of Anna, had been preferred to his mother, who had
been married to Prince Antony Ulrich of Brunswick; and no doubt
could be entertained that the object of Biron, in prevailing on the
empress to nominate the child, was to retain the supreme power in
his own hands as regent. We have seen by what means his ruin was
effected; what circumstances accompanied the regency of the
duchess Anna, mother of the youthful emperor; and how, by a
similar revolution, Anna herself was replaced by the princess
Elizabeth.
That Ivan had no other right to the throne than that conferred by
the will of the empress Anna, was one of the pretexts which
Elizabeth employed to prove the validity of her own title. That will, in
the manifesto published three days after the revolution, was
insinuated—probably with great truth—to have been irregularly
obtained; but in either case it was of no validity, since the right of
Elizabeth was asserted to be superior even to that of the former
empress. But the instrument was a tissue of sophistry. Though she
had been placed on the throne by about three hundred soldiers, she
did not hesitate to affirm that the revolution had been effected at
the demand of all her subjects. In ostentatiously displaying her
clemency, in proclaiming that she had sent back the parents of Ivan
to their own country, with all the honours due to their station, she
was equally insincere. Both passed their lives in captivity, and were
transferred from one fortress to another, according to her caprice or
jealousy. Until his eighth year Ivan was permitted to remain with
them; but, apprehensive lest his mind should be taught ambition, he
was consigned to solitary confinement first in the fortress of
Oranienburg, next in that of Schlüsselburg. In one respect his fate
was worse than that of his parents: they died in the course of
nature[50]; he, as we shall hereafter perceive, perished by violence.
One of Elizabeth’s first cares was to punish the men who had,
during the former reigns, kept her from the throne—those especially
who had assisted the regent Anna in overturning the power of Biron,
and had instigated her afterwards to seize the throne. All were
condemned to death; but the new empress was not a woman of
blood, and the sentence was commuted into perpetual banishment.
Ostermann, Munich, Golovkin, Mengden, Lövenwold, driven from a
power scarcely less than supreme and from riches almost
inexhaustible, were forced to earn their own subsistence in the wilds
of Siberia. Munich opened a school. The hand which had conquered
the Turks, which had given a
king to Poland, was employed in
tracing mathematical figures for
children.
If Elizabeth could punish, she
could also reward. The surgeon,
Lestocq, was made head
physician of the court, president
of the college of the faculty, and
privy councillor, with a
magnificent income. The
company of grenadiers who had
raised her to the throne were all
declared noble; and the
common soldiers ranked in Elizabeth Petrovna
future as lieutenants. But under
a despotic government there is (1709-1762)
little security for the great, least
of all for those whom capricious
favour has exalted. Presuming on his services, the ambition of
Lestocq urged him to demand higher preferment, and he had the
mortification to be refused. Nor was this all: by his arrogance he
offended the most powerful favourites of Elizabeth, especially the
grand chancellor Bestuzhev, who had been the minister of Anna;
and, in seven years after the revolution, he was exiled to a fortress
in the government of Archangel. Exile, in short, was perpetual in this
reign. The empress vowed that no culprit should suffer death; but
death would often have been preferable to the punishments which
were inflicted. Torture, the knout, slitting of the tongue, and other
chastisements—so cruel that the sufferer frequently died in
consequence—were not spared even females.
Soon after her accession a conspiracy was
[1743 a.d.] discovered, the object of which was the
restoration of young Ivan. The conspirators,
who were encouraged by a foreign minister, were seized, severely
chastised, and sent into exile. Among them was a court beauty,
whose charms had long given umbrage to the czarina, and we may
easily conceive that the revenge was doubly sweet which could at
once destroy the rebel and the rival. But the number of these victims
was small, compared with that which was consigned to unknown
dungeons, and doomed to pass the rest of life in hopeless
despondency. With all her humanity, Elizabeth suffered that most
inquisitorial court, the secret chancery, to subsist; and the
denunciations which were laid before it were received as implicitly as
the clearest evidence in other tribunals.

Foreign Affairs (1743-1757 A.D.)

In her foreign policy this empress seems scarcely to have had an

object. Averse to business, and fond of pleasure, she allowed her
ministers, especially Bestuzhev, to direct the operations of the wars
in which she was engaged, and to conduct at will the diplomacy of
the empire. Her first enemy was Sweden. That power demanded the
restitution of Finland, and was refused; hostilities which, indeed, had
commenced at the instigation of France during the last reign, were
resumed, but they were prosecuted with little vigour by the Swedes.
The valour of the nation appeared to have died with their hero,
Charles XII. So unfortunate were their arms that, by the Treaty of
Nystad, in 1721, and that of Åbo, in 1743, Livonia, Esthonia, Karelia,
Ingermanland, Viborg, and Kexholm passed under the domination of
Russia.
Still worse than the loss of their possessions was the influence
thenceforward exercised over the court of Stockholm by that of St.
Petersburg. In vain did Sweden endeavour to moderate the
exactions of the empress by electing the duke of Holstein, her
nephew, successor to the throne of the Goths: the Treaty of Åbo was
not the less severe. It is, indeed, true that the intelligence of this
election did not reach St. Petersburg until Elizabeth herself, who was
resolved never to marry,[51] had already nominated Duke Peter as
her own successor; but she ought to have received in a better spirit
a step designed as an act of homage to herself.
Had Elizabeth known her own interests, she would never have
engaged in the celebrated war which during so many years shook all
Europe to its centre. But, in the first place, she affected much
commiseration for the Polish king, whose Saxon dominions were
invaded by the Prussians, and whom she called her ally. In the
second, she was evidently actuated by a personal antipathy to
Frederick, and whoever were his enemies were sure to be her allies.
It would, however, be wrong to suppose that personal feeling alone
was her sole motive for interfering in a foreign war. There can be no
doubt that even at this early period, and indeed long before this
period, the ministers of Russia had cast a longing eye on the
possessions of Poland.
Courland and Semigallia, though nominally
[1757 a.d.] dependent on the Polish crown, were in reality
provinces of Russia. They had been lost to
Poland through the marriage of Anna, niece of Peter I, to Kettler,
sovereign of the duchy. Though she had no issue; though Ferdinand,
the successor of Kettler, was also childless; though the Polish diet
contended, with justice, that the fief was revertible to the republic,
Anna was resolved that its future destiny should be changed. Under
the pretext of certain pecuniary claims, the Russian troops overran
the territory; and the states were compelled to elect Biron, the
parent of the empress, to the vacant dignity. After the fall of that
unprincipled adventurer, the states, disgusted with Russian
preponderance, had ventured to unite their suffrages in favour of
Charles, son of Frederick Augustus III king of Poland; but Frederick
durst not sanction the election until he had obtained the permission
of the empress Elizabeth. She could, for once, well afford to be
generous; and Duke Charles was suffered to take possession of the
dignity. And, while on this subject, we may so far anticipate events
as to add that Peter III, successor of Elizabeth, refused to admit the
rights of Charles, whom he expelled from the duchy; and that
Catherine II incorporated it with her dominions. That Elizabeth
herself had the ambitious views of her father, in reference not only
to Courland but to other provinces, is certain; and, as we have
already observed, one of her motives for engaging in the great
European contest was the prospect of ulterior advantages. The
pretext of succouring an ally was sufficient to justify, in the eyes of
Europe, the march of her armies. In this respect, her policy was
macchiavellian enough. But to her the war was an imprudent one;
whatever her views, the time was not yet arrived when they could
be fully executed. Nor were the events always honourable to the
military glory of the empire. The reason is generally and, perhaps,
justly assigned to the partiality of the grand duke Peter, the heir
presumptive, for the Prussian monarch—a partiality so great as to be
inexplicable. The Russian generals, however anxious to win the
favour of their sovereign, still more the honours of successful
warfare, were yet loth to incur the dislike of Peter: hence the
operations were indecisive; and success, when gained, was not
pursued.

Antecedents of the Future Peter III

Charles Peter Ulrich, duke of Holstein Gottorp, whom Elizabeth

had nominated her successor, who had embraced the Greek religion,
and who, at his baptism, had received the name of Peter
Fedorovitch, had arrived at St. Petersburg immediately after her
accession. He was then in his fourteenth year. The education of this
unfortunate prince was neglected; and the cause must be attributed
alike to his own aversion to study and to the indifference of the
empress. Military exercises were the only occupation for which he
had any relish, and in them he was indulged. At the palace of
Oranienbaum, with which his aunt had presented him, he passed the
months of his absence from court—a period of freedom for which he
always sighed. As his recollections were German, so also were his
affections. He had little respect for those over whom he was one day
to reign: instead of native, he surrounded himself with young
German officers. His addiction to such exercises became a passion,
and was doubtless one of the causes that so strongly indisposed him
to more serious and more important pursuits.
But it was not the only cause. In his native province he had
probably learned to admire another propensity, common enough in
his time—that of hard drinking; and it was not likely to be much
impaired in such a country as Russia. His potations, which were
frequent and long, were encouraged by his companions; and, in a
few years, he became a complete bacchanalian. If we add that both
he and they indulged in gratifications still more criminal—in
licentious amours—we shall not hesitate to believe the charge of
profligacy with which he has been assailed. Whether the empress
was for some time privy to his excesses has been disputed; but
probability affirms that she was, and that, by conniving at these
ignoble pursuits, her policy was to keep him at a distance from the
affairs of state. In this base purpose she was, from motives
sufficiently obvious, zealously assisted by her ministers, especially by
Bestuzhev. Profligate as was the grand duke, he was displeased with
this state of restraint; and he sometimes complained of it with a
bitterness that was sure to be exaggerated by the spies whom they
had placed near him.

The Future Catherine II Appears

The empress paid little attention to the reports concerning him.

Her purpose was to disqualify him for governing, to render him too
contemptible to be dreaded; nor was she much offended with his
murmurs. That purpose was gained; for Peter had the reputation of
being at once ignorant, vicious, and contemptible. In a country so
fertile in revolutions, where unprincipled adventurers were ever
ready to encourage the discontent of anyone likely to disturb the
existing order of things, this reputation was one of the surest
safeguards of Elizabeth’s throne. She no longer feared that he would
be made the tool of the designing, and she secretly exulted in the
success of a policy which Macchiavelli himself would have admired.
Nor did she prove herself unworthy of that great master in the
refined hypocrisy which made her represent her nephew as a prince
of hopeful talents. But even she blushed at some of his irregularities;
and, in the view of justifying him, had furnished him with a wife. Her
choice was unfortunate; it was Sophia Augusta, daughter of the
prince of Anhalt-Zerbst, who, on her conversion to the Greek faith—a
necessary preliminary to her marriage—had received the baptismal
name of Catherine.
This union was entitled to the more attention as in its
consequences it powerfully affected not only the whole of Russia but
the whole of Europe. Shortly before its completion Peter was seized
with the smallpox, which left hideous traces on his countenance. The
sight of him is said so far to have so affected Catherine that she
fainted away. But, though she was only in her sixteenth year,
ambition had already over her more influence than the tender
passion, and she smothered her repugnance. Unfortunately, the
personal qualities of the husband were not of a kind to remove the ill
impression; if he bore her any affection, which appears doubtful, his
manners were rude, even vulgar; and she blushed for him whenever
they met in general society. What was still worse, she soon learned
to despise his understanding; and it required little penetration to
foresee that, whatever might be his title after Elizabeth’s death, the
power must rest with Catherine. Hence the courtiers in general were
more assiduous in their attentions to her than to him—a
circumstance which did not much dispose him for the better. Finding
no charms in his new domestic circle, he naturally turned to his boon
companions; his orgies became frequent, and Catherine was
completely neglected. Hence her indifference was exchanged into
absolute dislike.
The contrast between their characters exhibited itself in their
conduct. While he was thus earning contempt for himself, she was
assiduously strengthening her party. She had the advantage—we
should rather say the curse—of being directed by a wily mother, who
had accompanied her into Russia, and whose political intrigues were
so notorious that at length she was ordered by the empress to
return into Germany. The grand duchess, however, had been too
well tutored to suffer much by her mother’s departure; and she
prosecuted her purpose with an ardour that would have done
honour to a better cause.
So long as the German princess remained at court, the conduct of
Catherine was outwardly decorous; but now less restraint was
observable in her behaviour. She was little deterred by the fear of
worldly censure, in a court where the empress herself was anything
but a model of chastity; and her marital fidelity soon came to be
more than doubtful.

Court Intrigues; the Death of Elizabeth (1762 A.D.)

That, in concert with several Russian nobles,

[1762 a.d.] of whom Bestuzhev was the chief, Catherine
meditated the exclusion of her husband from
the throne and the elevation of herself as regent during the minority
of her son Paul, is a fact that can no longer be disputed. Hence the
criminal condescension of the chancellor to the views of Catherine;
hence his efforts to prevail on the empress to nominate the infant
Paul as her successor. The indiscretion of the grand duke, who was
no favourite with anybody; his frequent complaints of the tutelage in
which he was held; his bursts of indignation at his exclusion from
the councils of the empire—were carefully related to his aunt, with
such exaggeration as were most likely to destroy the last traces of
the lingering regard she bore him. All, indeed, who had been the
friends of Catherine, all who had shared in the confidence of the
minister, might well contemplate with alarm the succession of one
that had vowed revenge against the partisans of both. Besides, the
contempt which Peter felt, and which he seldom hesitated to
express, for the Russian people, rendered his succession far from
agreeable to them.
Thus, when, in 1757, Apraxin, field marshal of the Russian forces,
invaded Prussia, took Memel, and, near Jägerndorf, obtained a
brilliant victory over the troops of Frederick, yet, as if defeated,
instantly fell back upon Courland, the cause was something more
than the fear of offending Peter. This retrograde movement
surprising, as well it might, both the empress and her people,
Apraxin was placed under arrest, and the command of the army
bestowed on another general. He was tried for the crime, but
absolved—a result still more surprising to men who regarded merely
the surface of things. The reason was that the grand-chancellor,
Bestuzhev, had secretly ordered the marshal to retreat, and was, of
course, his protector in the trial. It was not to please the heir-
presumptive of the crown, whose blind adoration of the Prussian
king was so well known, that Bestuzhev despatched the secret order
for Apraxin to retreat: it was that the chiefs of the army, of whom
many were his creatures, might be ready to join in effecting the
revolution which was meditated. But the ambitious minister,
presuming on the distaste which his imperial mistress generally
showed for affairs, and still more on her bodily indisposition, which
at this time placed her life in danger, proceeded too rapidly. His
intrigues were discovered; his letter to the marshal was produced;
he was deprived of all his power; and Peter had the joy of seeing
him exiled.
The general who succeeded Apraxin obtained advantages over the
Russian monarch, which had never been contemplated by his
predecessor. But though he took Königsberg, placed most of Prussia
under contribution, and defeated the Prussian army in a decisive
engagement, he, too, was unwilling to irritate beyond forgiveness
the heir of the empire, especially as the reports which daily reached
him of Elizabeth’s health convinced him that the succession was not
far distant. Under the pretext of illness, he demanded leave to retire.
His successor, Soltikov (not, we may be sure, the favourite of that
name), was still more successful. Frederick was defeated in one of
the best contested battles of this famous war; Berlin was taken, and
Kolberg reduced after a vigorous siege. The news of this last success
reached the empress, but she was no longer capable of deriving
satisfaction from it. Much to her honour, she withstood all the
solicitations of the intriguers who wished to exclude her nephew and
to place Paul on the throne, under the regency of his mother. She
died on the 5th day of January, 1762.b

Spread of Art, Literature, and Education under Elizabeth

The empress Elizabeth had a passion for building; Peter the

Great’s summer palace and even the empress Anna’s winter palace
appeared to her small and confined. Upon the site of the latter she
began to build the present edifices; during her reign was also built
the vast, elegant, and beautiful palace at Tsarskoi Selo; the palace of
Oranienbaum was reconstructed, and the fine churches of the
Smolni convent, of Vladimirskaia and of Nicholas Morskoi (in St.
Petersburg) were also erected. Some handsome private houses were
built by Elizabeth’s noblemen, and in general St. Petersburg, which
had not long before been a desert place, consisting chiefly of
wooden houses, became greatly embellished; the palace quay, as
may be seen from drawings and engravings of the time, already
showed a continuous row of huge stone edifices.
Of course all these buildings cost enormous sums which led
private persons into debt and the government into superfluous
expenditure, but it is impossible not to observe that there was to be
seen in this luxury an artistic quality which had never before existed.
The finest edifices of that period form a special style, which after
temporary neglect is now beginning to be imitated; the creator of
this style in Russia was Count Rastrelli—a foreigner, of whom,
however, Russia has the right to speak. The palaces and churches
built by Rastrelli merit description, and although painting at that time
did not represent a very high standard, yet the ceilings painted in
accordance with the fashion of the day, with bouquets of flowers and
mythological goddesses, even now attract the attention of artists.
The grandees gave high prices for pictures by foreign masters; their
houses became distinguished not only for their handsome façades
but also for the comfort of their interior arrangements; it would
hardly be possible, for instance, to imagine anything more nobly
elegant than the house of the chancellor Vorontzov (now the corps
des Pages).
All these beautiful architectural productions, and likewise those of
music and painting, were for the greater part the work of foreign
artists—visitors to Russia; but under their influence Russian artists
were formed and taste developed. The church of Nicholas Morskoi
was built by a pupil of Rastrelli. The almost daily theatrical
representations produced at court gave rise to the idea of organising
similar representations at the corps des Cadets. The empress took a
lively interest in them; she often assisted at them and lent her
diamonds for the women’s costumes. In their turn these
representations could not but assist the development of a taste for
the stage, for dramatic art and literature in general and from
amongst the number of cadet actors not a few became well known
writers, as for instance Beketov, Kheraskov, and Sumarokov.
We must dwell for a few moments on Sumarokov—a man who in
his time enjoyed an extensive literary reputation and secured for
himself the appellation of Father of the Russian Stage. The love of
literature, and especially of the stage, was already developed in
Sumarokov when he was in the corps des Cadets; when he was
afterwards made aide-de-camp to Razumovski, he could almost daily
assist at operas and ballets. At that period he read with avidity the
dramatic authors then in fashion: Corneille, Racine, Voltaire, and
Molière became his idols; he decided to try to imitate them in his
own native language then very undeveloped, and in 1747 he wrote a
tragedy, the Chorists.
It was not the merits of this work, which were very insignificant,
but the unwontedness of the appearance of an original Russian
tragedy, and besides that the fact of its being in verse, that so
astounded and enraptured his contemporaries that they proclaimed
Sumarokov the “Russian Racine”; encouraged by such a success he
wrote a second and yet, a third tragedy; he took up comedy (for
which he had hardly any more vocation) and in fact wrote a whole
repertory; there were, however, no actors; because neither in St.
Petersburg nor in Moscow did there any longer exist such company
and such theatres as were begun in the time of Peter.
Meanwhile, far away from both capitals, in Iaroslav there was
formed, almost of itself without any commands or even any
encouragement being given, a Russian dramatic company which is
indissolubly bound up with the name of Volkov. Theodore Volkov was
the son of a merchant and had been educated in the Iaroslav
seminary, where, following the example of the Academy of Kiev, and
others, representations of a spiritual or religious character were
given. They produced a great impression upon the young merchant;
when later on he managed to get to St. Petersburg and saw on the
stage of the corps des Cadets a dramatic representation given with
scenery, lighting, and mechanical contrivances, Volkov was stupefied
with rapture and astonishment. Being to the highest degree sensitive
to every artistic impression, being a painter, a musician, and a
sculptor—all self-taught—Volkov was also endued with that
constancy and patience without which even gifted natures do not
attain to any results. Volkov studied the material side of scenic art to
the smallest details—that is, the arrangement of the machinery, of
the scenes, etc.; when he returned to Iaroslav he asked his parents,
with whom he lived, to let him have an empty tanner’s shed; there
he arranged a pit and a stage, and making up a company of young
merchants like himself, sons of citizens and clerks, gave
representations which aroused the enthusiasm of all the spectators.
The intelligent and practical Volkov, seeing how the population of
Iaroslav flocked to his representations, named a price for them—a
five kopeck piece for the first rows—and thus little by little he
amassed a sum with which in 1752 he was able to build a general
public theatre with room for one thousand spectators.
The taste for the stage had meanwhile greatly spread in St.
Petersburg; in various private houses dramatic representations were
given at evening parties; it was therefore not surprising that the
Iaroslav theatre soon began to be talked of. The empress invited
Volkov to come to St. Petersburg with his company, as she wished to
see his representations given on the stage of the court theatre. She
was remarkably pleased with them, and four years later issued an
ukase for the establishment of a public theatre. The first director of
this theatre and almost the only dramatic writer was Sumarokov;
according to the testimony of contemporaries Volkov was one of its
most talented actors and his friend and fellow worker Dmitrievski a
great artist.
We must here speak of another still more remarkable Russian
native genius—Lomonosov. It is well known how, when he was a
youth of sixteen, devoured by a thirst for knowledge, he secretly left
the paternal roof and made his way on foot from Kholmogori to
Moscow. How unattractive must life and learning have appeared to
him in those early days! “Having only one altyn (a three-kopeck
piece) a day for salary, it was impossible for him to spend more on
food than a halfpenny a day for bread and a halfpenny worth of
kvass (a kind of beer or mead); the rest had to go for paper, books,
and other necessities.” Thus he described his life in the
Zaikonospaskvi Ecclesiastical Academy to Ivan Shuvalov and
concluded with the following words: “I lived thus for five years and
did not abandon science!” Theodore Prokopovitch, when he was
already an old man, visited the Moscow academy a few years before
his death; he noticed Lomonosov there and praised him for his
laboriousness and learning. In 1737 Lomonosov was sent abroad to
perfect himself and placed himself under the surveillance of the then
famous scholar, Wolff, who, while despising him for his disorderly
life, spoke with respect of his capacities and success in study.
Lomonosov followed the lectures of the German professors and
amused himself with the German students. The news of Minikh’s
great victories and the taking of Khotin reached him; his patriotic
feelings were aroused, and he wrote an ode. When the verses were
received in St. Petersburg everyone was struck with their harmony;
and when Lomonosov returned from Germany in the beginning of
Elizabeth’s reign his reputation as a poet had already preceded him
—the more he wrote the greater his fame became. Poetry, however,
was not Lomonosov’s strongest point, and verses do not occupy a
quarter of his entire works. His mind worked even more than his
imagination, and his scholarly writings are striking in their variety. He
composed a grammar of the Russian language from which several
generations have learned; he laid down rules of versification, the
foundation of which are even now recognised by everyone; he wrote
on chemistry, physics, astronomy, metallurgy, geology; he composed
a Russian history, wrote a hypothesis concerning the great learned
expeditions and memoranda bearing on questions of the state (as
for instance measures for increasing and maintaining the population
in Russia): in fact, Lomonosov’s extraordinary intellect seemed to
touch upon every branch of mental activity. He was made a member
of the St. Petersburg Academy of Sciences, but there the German
element reigned supreme and Lomonosov was one of those who,
while venerating the work of Peter the Great and the European
learning introduced by him, yet was oppressed by foreign tutorage
and took offence when the Germans put forward their own
countrymen to the detriment of meritorious Russians. Continual
disputes and quarrels arose between Lomonosov and his fellow
members; nor, being of a very impetuous and obstinate nature, was
Lomonosov always in the right. His rough and sharp measures
frequently led him into quarrels even outside the academy, for
instance with his literary brethren, Frediakovski and Sumarokov. All
this might greatly have injured Lomonosov, but fortunately for him
he possessed powerful protectors in the persons of Count Worontzov
and Count Razumovski, who liked to show favour to the first Russian
scholar and poet.
But the strongest, truest, and most constant of his protectors was
Ivan Shuvalov. Shuvalov had many defects—his character was weak,
lazy, and careless; but he nevertheless represented one of the most
consolatory types of his epoch: strong, energetic types were not
uncommon in the first half of the eighteenth century, but gentle,
benevolent, indulgent natures were rarely to be met with. Shuvalov
was not captivated by clamorous deeds, like the men of Peter’s time,
but by the peaceful progress of science and art. Therefore if the
weakness of his character made him an instrument for the ambitious
designs of his cousin, his heartfelt sympathies drew him towards
Lomonosov, of whom he naturally learned much and—what is of
more importance—with whom he devised means for the spread of
education in Russia. The result of these deliberations was a vast plan
for the establishment of schools throughout the governments, and
finally of a university in Moscow. The establishment of a university
seemed of the first necessity, as it was to furnish Russia with
teachers; this had been Peter’s intention with regard to the
academy: but it had not been fulfilled. In his report to the senate
upon this subject, Shuvalov wrote that it would be desirable to
appoint a “sufficient number of worthy men of the Russian
nationality, acquainted with the sciences, to spread education in
distant parts among the common people, so that thus superstition,
dissent, and other like heresies proceeding from ignorance might be
destroyed.” The senate approved Shuvalov’s proposition and in 1755
the University of Moscow was founded.
We have given as just and complete a picture of the period of the
empress Elizabeth as is possible in view of the scarcity of information
obtainable concerning many circumstances of that time. Elizabeth
left behind her if not a great memory yet, broadly speaking, a good
one. Her administration may be reproached with much: in its foreign
policy it was not sufficiently independent; it was not sufficiently
watchful in interior affairs, where oversights occasioned special evils;
moreover examples of unlawful enrichment attained huge
dimensions. But her reign may be said to have led Russia out of
bondage to the Germans, while the level of education was not in the
smallest degree lowered, but on the contrary considerably raised.
Much that brought forth such brilliant fruits under Catherine II was
sown under Elizabeth.d

Bain’s Estimate of Elizabeth

It is the peculiar glory of Elizabeth Petrovna that she consulted

once for all the life work of her illustrious father. During the first
fifteen years after the death of the great political regenerator, his
stupendous creation, Russia, (before him we only hear of Muscovy,)
was frequently in danger. The reactionary boyars who misruled the
infant empire under Peter II would have sacrificed both the new
capital and the new fleet, the twin pivots upon which the glory and
the prosperity of the new state may be said to have turned; the
German domination under the empress Anna, directly contrary as it
was to the golden rule of Peter, “Russia for the Russians,” threatened
the nation with a western yoke far more galling than the eastern or
Tatar yoke of ruder times. From this reaction, from this yoke the
daughter of Peter the Great set the nation free, and beneath her
beneficent sceptre Russia may be said to have possessed itself
again. All the highest offices of state were once more entrusted to
natives and to natives only, and whenever a foreigner was proposed
for the next highest, Elizabeth, before confirming the appointment,
invariably inquired: “Is there then no capable Russian who would do
as well?” Moreover she inherited from her father the sovereign gift
of choosing and using able councillors, and not only did she summon
to power a new generation of native statesmen and administrators,
but she constrained them to work harmoniously together despite
their mutual jealousies and conflicting ambitions. She herself had
advantageously passed through the bitter but salutary school of
adversity. With all manner of dangers haunting her path from her
youth upwards, she had learnt the necessity of circumspection,
deliberation, self-control; she had acquired the precious faculty of
living in the midst of people intent on jostling each other, without in
any way jostling them; and these great qualities she brought with
her to the throne without losing anything of that infinite good-
nature, that radiant affability, that patriarchal simplicity which so
endeared her to her subjects and made her, deservedly, the most
popular of all the Russian monarchs. As regards her foreign policy, it
may be safely affirmed she laid down the deep and durable
foundations upon which Catherine II was to build magnificently
indeed, but too often, alas! so flimsily. The diplomacy of Elizabeth,
on the whole, was not so confident or so daring as the diplomacy of
her brilliant successor; but, on the other hand, it was more correct,
equally dignified and left far less to chance. It must also be borne in
mind that the energy and firmness of Elizabeth considerably
facilitated the task of Catherine by rendering Prussia, Russia’s most
dangerous neighbour, practically harmless to her for the remainder
of the century. This of itself was a political legacy of inestimable
value, and it was not the only one. All the great captains, all the
great diplomatists of the “ever victorious Catherine,” men like
Rumiantsev, Suvarov, Riepnin, Besborodko, the Panins and the
Galitzins, were brought up in the school of Elizabeth. Excellent was
the use which the adroit and audacious Catherine made of these
instruments of government, these pioneers of empire, but it should
never be forgotten that she received them all from the hands of the
daughter of Peter the Great.g

PETER III (1762 A.D.)

As Elizabeth, on her death-bed, had confirmed the rights of Peter

III; and as the conspirators, deprived of Bestuzhev their guide, were
unable to act with energy, the new emperor encountered no
opposition. On the contrary, he was immediately recognised by the
military; and the archbishop of Novgorod, in the sermon preached
on the occasion, thanked heaven that a prince so likely to imitate his
illustrious grandfather was vouchsafed to Russia. Catherine was
present. She wore a peculiar dress to conceal her pregnancy, and
her countenance exhibited some indication of the anxious feeling
which she was obliged to repress. Compelled to defer the execution
of her ambitious purposes, and uncertain what vengeance the czar
might exert for her numerous infidelities, she might well be
apprehensive.
But she had no real foundation for the fear. Of all the sovereigns
of that or any age, Peter was among the most clement. Whether he
thought that clemency might bind to his interests one whose talents
he had learned to respect, or that her adherents were too numerous
and powerful to allow of her being punished—whether, in short, he
had some return of affection for her, or his own conscience told him
that she had nearly as much to forgive as he could have, we will not
decide. One thing only is certain—that, in about three months after
his accession, he invested her with the domains held by the late
empress. Certainly his was a mind incapable of long continued
resentment. His heart was better than his head. Resolved to
signalise his elevation by making others happy, he recalled all whom
his predecessor had exiled, except Bestuzhev. Many he restored to
their former honours and possessions. Thus the aged Munich was
made governor-general of Siberia, restored to his military command;
while Biron, who certainly deserved no favour, was reinvested with
the duchy of Courland. He did more: he restored the prisoners made
by the generals of Elizabeth, and gave them money to defray their
passage home. And, as Frederick had always been the object of his
idolatry, the world expected the armistice which he published, and
which was preparatory to a peace between the two countries.
That declaration was an extraordinary document. In it the
emperor declares that, his first duty being the welfare of his people,
that welfare could not be consulted so long as hostilities were
continued; that the war, which had raged six years, had produced no
advantage to either party, but done incredible harm to both; that he
would no longer sanction the wanton destruction of his species; that,
in conformity with the divine injunction relative to the preservation
of the people committed to his charge, he would put an end to the
unnatural, impious strife; and that he was resolved to restore the
conquests made by his troops. In this case he had been praised, and
with great justice, for his moderation. We fear, however, he does not
merit so high a degree of praise of humanity as many writers have
asserted. At this moment, while proclaiming so loudly his
repugnance to war, he was sending troops into his native principality
of Holstein, with the intention of wresting from the king of Denmark
the duchy of Schleswig, which he considered the rightful inheritance
of his family. He even declared that he would never rest until he had
sent that prince to Malabar.
Nor must we omit to add that from the enemy he became the ally
of Frederick; that his troops joined with the Prussians to expel the
Austrians from the kingdom. His humanity only changed sides; if it
spared the blood of Prussians, it had little respect for that of
 Austrians. We may add, too,
that there was something like
madness in his enthusiastic
regard for Frederick. He
corresponded with that
monarch, whom he proclaimed
his master, whose uniform he
wore, and in whose armies he
obtained the rank of major-
general. Had he been capable of
improvement, his intercourse
with that far-sighted prince
might have benefited him.
Frederick advised him to
Peter III celebrate at Moscow his
(1728-1762) coronation—a rite of
superstitious importance in the
eyes of the multitude. He was
advised, too, not to engage in the Danish war, not to leave the
empire. But advice was lost on him.
In some other respects, Peter deserves more credit than the
admirers of Catherine are willing to allow him. (1) Not only did he
pardon his personal enemies—not only did the emperor forget the
wrongs of the grand duke—but on several he bestowed the most
signal favours. He suppressed that abominable inquisitorial court,
the secret chancery, which had consigned so many victims to
everlasting bondage, which had received delations from the most
obscure and vicious of men, which had made every respectable
master of a family tremble lest his very domestics should render him
amenable to that terrible tribunal. Had this been the only benefit of
his reign, well would he have been entitled to the gratitude of
Russia. (2) He emancipated the nobles from the slavish dependence
on the crown, so characteristic of that barbarous people. Previous to
his reign, no boyar could enter on any profession, or forsake it when
once embraced, or retire from public to private life, or dispose of his
property, or travel into any foreign country, without the permission
of the czar. By breaking their chains at one blow, he began the
career of social emancipation. (3) The military discipline of the
nation loudly demanded reform, and he obeyed the call. He rescued
the officers from the degrading punishments previously inflicted; he
introduced a better system of tactics; and he gave more
independence to the profession. He did not, however, exempt the
common soldier from the corporal punishment which at any moment
his superior officers might inflict. (4) He instituted a useful court to
take cognisance of all offences committed against the public peace,
and to chastise the delinquencies of the men entrusted with the
general police of the empire. (5) He encouraged commerce, by
lessening the duties on certain imports, and by abolishing them on
certain exports. (6) In all his measures, all his steps, he proved
himself the protector of the poor. In fact, one reason for the dislike
with which he was regarded by the nobles arose from the preference
which he always gave to the low over the high.

Impolitic Acts of Peter III

But if impartial history must thus eulogise many of this monarch’s

acts, the same authority must condemn more. He exhibited
everywhere great contempt for the people whom he was called to
govern. He had no indulgence for their prejudices, however
indifferent, however inveterate. Thus, in commanding that the
secular clergy should no longer wear long beards, and should wear
the same garb as the clergy of other countries, he offended his
subjects to a degree almost inconceivable to us. In ordering the
images to be removed from the churches—he was still a Lutheran, if
anything—he did not lessen the odium which his other acts had
produced. The archbishop of Novgorod flatly refused to obey him,
and was in consequence exiled; but the murmurs of the populace
compelled the czar to recall him. Still more censurable were his
efforts to render the church wholly dependent on the state—to
destroy everything like independence in its ministers; to make
religion a mere engine in the hands of arbitrary power for the
attainment of any object. His purpose, in fact, was to seize all the
demesnes of the church—its extensive estates, its numerous serfs—
and to pension the clergy like other functionaries.
In the ukase published on this occasion, he expressed a desire to
relieve ecclesiastics of the temporal cares so prejudicial to their
ghostly utility; to see that they indeed renounced the world, and free
from the burden of perishing treasures, applied their whole attention
to the welfare of souls. He therefore decreed that the property of
the church should in future be managed by imperial officers; and
that the clergy should receive, from the fund thus accumulated,
certain annual pensions, corresponding to their stations. Thus the
archbishops of Novgorod, Moscow, and St. Petersburg were to have
each 2,500 rubles; and the same sum was to be allowed for the
support of their households, of their capitular clergy, and for the
sustentation of the sacred edifices. But the twenty-three other
archbishops and bishops were to have only 3,000 rubles for both
purposes. The salaries of the other ecclesiastics were carefully
graduated. The inferior were divided into three classes—individuals
of the first to receive 500, of the second 300, of the third 150 rubles
per annum. The surplus funds were to be applied to the foundation
of hospitals, to the endowment of colleges, and to the general
purposes of the state.
Peter attempted these and other innovations in virtue of the two-
fold character which, from the time of his grandfather, the czars had
been anxious to assume, as supreme heads alike of religion and of
the state. Not even the grand lama of Thibet ever arrogated a higher
degree of theocratic authority. Indeed, our only surprise is that in
addition to their other functions they did not assume that of bishops;
that they did not array themselves in pontificals, and celebrate mass
at the altar. But they certainly laid something like a claim to the
sacerdotal character. Thus, on the death of the patriarch, Peter I
opposed the election of another supreme head of the church; and
when he found that the synod durst not venture on so far irritating
the people as to dispense with the dignity, he insisted on being
elected himself. If the sultan of Constantinople combined with
himself the two-fold character, why should it be refused to him? The
reign of Peter was too short to permit his designs of spoliation to be
carried into effect; but, by confirming the dangerous precedent of
his grandfather, he had done enough, and his successor Catherine
was enabled to complete the robbery which he commenced.
But the most impolitic measure of Peter—that which rendered
those who might have defended him indifferent to his fate—was his
conduct towards the imperial guards. Two regiments he ordered to
be in readiness for the Danish war. This was contrary to custom. In
the faith of remaining near the court, most of the soldiers had
embraced the military life; and they were as indignant as they were
surprised when told that they must exchange the dissipations of a
metropolis for the fatigues and privations attending a distant
campaign. They were offended, too, with the introduction of the
Prussian discipline, which they found by experience to be far more
rigid than that to which they had hitherto been subject; and they
patriotically condemned the innovation as prejudicial to the military
fame of the empire. Still more irritating was the preference which he
everywhere gave to the German over the native troops. His most
intimate friends were Germans; the officers around his person were
of the same nation; Germans directed the manœuvres not only of
his household but of all his regiments; and a German—Prince George
of Holstein, his uncle—was placed at the head of all the imperial
armies.
Couple these acts of imprudence with others of which he was
hourly guilty. In his palace of Oranienbaum he constructed a
Lutheran chapel; and though he appears to have been indifferent to
every form of religion, he held this in much more respect than the
Greek form, which in fact, he delighted to ridicule. If churchmen
became his enemies, the people in general were not likely to
become his friends when they heard of a boast—probably a true one
—that in the last war he had acquainted the Prussian monarch with
the secrets of the imperial cabinet. Lastly, he insulted men of honour
by making them the jest of his buffoons.
 Circumstances much less numerous and much less cogent than
these would have sufficed so ambitious, able, and unprincipled a
woman as Catherine to organise a powerful conspiracy against the
czar. But he was accused of many other things of which he was
perfectly innocent. In fact, no effort seems to have been spared to
invent and propagate stories to his disadvantage. In some instances,
it is scarcely possible to separate the true from the false. Whether,
for example, he, from the day of his accession, resolved to divorce
his wife, to marry his mistress, to set aside Paul from succession,
and to adopt Ivan, still confined in the fortress of Schlüsselburg, can
never be known with certainty. That he secretly visited that unhappy
prince seems undoubted; but we have little evidence for the
existence of the design attributed to him. If, in fact, he sincerely
contemplated raising the daughter of Count Vorontzov to the
imperial throne, he would scarcely have adopted Ivan, unless he felt
assured that no issue would arise from the second marriage. He
could not, however, entertain any regard for a consort who had so
grievously injured him, and little for a boy whom he knew was not
his own. And, as there is generally some foundation for every report,
there seems to be no doubt that Peter had promised to marry his
mistress if she survived his wife. The report was enough for
Catherine: on it she built her own story that her life was in danger;
and that if her son were not designed for a similar fate, he would at
least have that of Ivan.

Catherine Plots against the Czar

The anxiety of the empress to secure adherents was continually

active; and as her husband passed so much time in drunkenness,
her motions were not so closely scrutinised as they should have
been. Gregory Orlov, her criminal favourite, was the man in whom
she placed the most reliance. Gregory had four brothers—all men of
enterprise, of courage, of desperation; and none of them restricted
by the least moral principle. Potemkin, afterwards so celebrated, was
the sixth. This man was, perhaps, the most useful of the
conspirators, as by means of his acquaintance with the priests of the
metropolis he was able to enlist that formidable body in the cause.
They were not slow to proclaim the impiety of the czar, his contempt
of the orthodox faith, his resolution “to banish the fear of the Lord”
from the Russian court, to convert churches into hospitals and
barracks, to seize on all revenues of the church, and to end by
compelling the most orthodox of countries to embrace the errors of
Luther. The archimandrites received these reports from the parish
priests, the bishops from the archimandrites; nor was there much
difficulty in obtaining an entrance for them into the recesses of the
neighbouring monasteries. The hetman of the Cossacks, an officer of
great authority and of great riches, was next gained. Not less
effectual than he was the princess Dashkov, who, though the sister
of Peter’s mistress, was the most ardent of the conspirators: perhaps
the threatened exaltation of that sister, by rendering her jealous,
only strengthened her attachment to the czarina. Through the
instrumentality of this woman, Count Panin, the foreign minister and
the governor of the grand duke Paul, was gained over. Whether the
argument employed was, as one writer asserts, the sacrifice of her
sister, or whether, as another affirms, she was the daughter of the
count, who notoriously intrigued with her mother, is of no moment.
What is certain is, that the count was exceedingly fond of her; and
one authority expressly asserts that he became acquainted with the
details of the conspiracy before her, and admitted her into the plot.
This, however, is less probable than the relation we have given; for
the princess had long been the friend of Catherine.
Her activity was unceasing. A Piedmontese adventurer, Odart by
name, being forced to leave his native country for some crime, and
having tried in vain to obtain a subsistence in the neighbouring
capitals, wisely resolved to try his fortune in St. Petersburg—a city
where guilt might reside with impunity, and where it had only to be
successful to win the applause of mankind. As he had a considerable
knowledge of the fine arts, especially of music and painting, he had
little difficulty in obtaining an introduction to the princess Dashkov.
She, who had a shrewd insight into human character, soon perceived
that this supple, crafty, active, sober, intriguing, unprincipled
foreigner was just the man that was required to act as spy and
confidential agent. He was introduced to Catherine, whose opinion
confirmed that of her favourite. No choice could, indeed, have been
better. Little cared he in what service he was employed. If a partisan
were to be gained, no man could be more insinuating: if an enemy
were to be removed, he had his pistols and his dirk, without which
he never appeared in the street. His penetration soon enabled him
to secure the aid of two other bravos—the one, Possik, a lieutenant
in the guards; the other, Globov, a lawyer in the employment of the
senate. Of the character of these men, some notion may be formed
from the fact that Possik offered to stab the emperor in the midst of
the court. He knew how to ally duplicity with desperation; he was at
once the hypocritical intriguer and the remorseless bravo.
Through the same Princess Dashkov, Volkonski, major-general of
the guards, was won; and by Potemkin, or his ghostly allies, the
archbishop of Novgorod was soon in the secret. The hetman of the
Cossacks went further. Great as was the danger of entrusting that
secret to many, he assembled the officers who served under him,
assured them that he had heard of a conspiracy to dethrone the
emperor, too irresistible to be appeased; and exhorted them to seize
the favourable moment of propitiating the favour of the czarina,
rather than, by remaining hostile or inactive, to bring down
vengeance on their own heads. His advice had all the success that
he could desire.
While these most vicious and in every way most worthless of men
were thus employed in her behalf, Catherine was no less active. She
knew that Count Panin espoused the cause of her son—less,
perhaps, from affection to his charge, than from the hope of
exercising more power under an infant emperor than under one of
the mother’s enterprising character. Her promise, that his influence
should be second only to her own, made him her willing instrument.
His defection constrained the rest of the conspirators: there was no
more heard of a regency; and Catherine was to be proclaimed
autocratrix of all the Russias.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

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

Let us accompany you on the journey of exploring knowledge and

personal growth!

textbookfull.com