Visit https://ebookfinal.

com to download the full version and

explore more ebook

Introduction to Combinatorial Designs Second Edition

W.D. Wallis

_____ Click the link below to download _____

https://ebookfinal.com/download/introduction-to-
combinatorial-designs-second-edition-w-d-wallis/

Explore and download more ebook at ebookfinal.com

Here are some recommended products that might interest you.
You can download now and explore!

Introduction to Metaphysics Second Edition Martin

Heidegger

https://ebookfinal.com/download/introduction-to-metaphysics-second-
edition-martin-heidegger/

ebookfinal.com

Introduction to Green Chemistry Second Edition Matlack

https://ebookfinal.com/download/introduction-to-green-chemistry-
second-edition-matlack/

ebookfinal.com

Introduction to Credit Risk Modeling Second Edition Bluhm

https://ebookfinal.com/download/introduction-to-credit-risk-modeling-
second-edition-bluhm/

ebookfinal.com

Introduction to Chemical Reactor Analysis Second Edition

Hayes

https://ebookfinal.com/download/introduction-to-chemical-reactor-
analysis-second-edition-hayes/

ebookfinal.com
Introduction to Survey Sampling Second Edition Graham
Kalton

https://ebookfinal.com/download/introduction-to-survey-sampling-
second-edition-graham-kalton/

ebookfinal.com

Introduction to water in California Second Edition Carle

https://ebookfinal.com/download/introduction-to-water-in-california-
second-edition-carle/

ebookfinal.com

An Introduction to Buddhism Second Edition Peter Harvey

https://ebookfinal.com/download/an-introduction-to-buddhism-second-
edition-peter-harvey/

ebookfinal.com

How to Count An Introduction to Combinatorics Second

Edition R.B.J.T. Allenby

https://ebookfinal.com/download/how-to-count-an-introduction-to-
combinatorics-second-edition-r-b-j-t-allenby/

ebookfinal.com

An Introduction to Japanese Society Second Edition Yoshio

Sugimoto

https://ebookfinal.com/download/an-introduction-to-japanese-society-
second-edition-yoshio-sugimoto/

ebookfinal.com
Introduction to Combinatorial Designs Second Edition
W.D. Wallis Digital Instant Download
Author(s): W.D. Wallis
ISBN(s): 9781584888390, 1584888393
Edition: 2
File Details: PDF, 4.48 MB
Year: 2007
Language: english
 DISCRETE MATHEMATICS AND ITS APPLICATIONS
Series Editor KENNETH H. ROSEN

IntroductIon to

combInatorIal desIgns
Second edition
 DISCRETE
MATHEMATICS
and
ITS APPLICATIONS
Series Editor
Kenneth H. Rosen, Ph.D.

Juergen Bierbrauer, Introduction to Coding Theory

Kun-Mao Chao and Bang Ye Wu, Spanning Trees and Optimization Problems
Charalambos A. Charalambides, Enumerative Combinatorics
Henri Cohen, Gerhard Frey, et al., Handbook of Elliptic and Hyperelliptic Curve Cryptography
Charles J. Colbourn and Jeffrey H. Dinitz, Handbook of Combinatorial Designs, Second Edition
Steven Furino, Ying Miao, and Jianxing Yin, Frames and Resolvable Designs: Uses,
Constructions, and Existence
Randy Goldberg and Lance Riek, A Practical Handbook of Speech Coders
Jacob E. Goodman and Joseph O’Rourke, Handbook of Discrete and Computational Geometry,
Second Edition
Jonathan L. Gross and Jay Yellen, Graph Theory and Its Applications, Second Edition
Jonathan L. Gross and Jay Yellen, Handbook of Graph Theory
Darrel R. Hankerson, Greg A. Harris, and Peter D. Johnson, Introduction to Information
Theory and Data Compression, Second Edition
Daryl D. Harms, Miroslav Kraetzl, Charles J. Colbourn, and John S. Devitt, Network Reliability:
Experiments with a Symbolic Algebra Environment
Leslie Hogben, Handbook of Linear Algebra
Derek F. Holt with Bettina Eick and Eamonn A. O’Brien, Handbook of Computational Group Theory
David M. Jackson and Terry I. Visentin, An Atlas of Smaller Maps in Orientable and
Nonorientable Surfaces
Richard E. Klima, Neil P . Sigmon, and Ernest L. Stitzinger, Applications of Abstract Algebra
with Maple™ and MATLAB®, Second Edition
Patrick Knupp and Kambiz Salari, Verification of Computer Codes in Computational Science
and Engineering
William Kocay and Donald L. Kreher, Graphs, Algorithms, and Optimization
Donald L. Kreher and Douglas R. Stinson, Combinatorial Algorithms: Generation Enumeration
and Search
 Continued Titles
Charles C. Lindner and Christopher A. Rodgers, Design Theory
Hang T. Lau, A Java Library of Graph Algorithms and Optimization
Alfred J. Menezes, Paul C. van Oorschot, and Scott A. Vanstone, Handbook of Applied
Cryptography
Richard A. Mollin, Algebraic Number Theory
Richard A. Mollin, Codes: The Guide to Secrecy from Ancient to Modern Times
Richard A. Mollin, Fundamental Number Theory with Applications
Richard A. Mollin, An Introduction to Cryptography, Second Edition
Richard A. Mollin, Quadratics
Richard A. Mollin, RSA and Public-Key Cryptography
Carlos J. Moreno and Samuel S. Wagstaff, Jr., Sums of Squares of Integers
Dingyi Pei, Authentication Codes and Combinatorial Designs
Kenneth H. Rosen, Handbook of Discrete and Combinatorial Mathematics
Douglas R. Shier and K.T. Wallenius, Applied Mathematical Modeling: A Multidisciplinary
Approach
Jörn Steuding, Diophantine Analysis
Douglas R. Stinson, Cryptography: Theory and Practice, Third Edition
Roberto Togneri and Christopher J. deSilva, Fundamentals of Information Theory and
Coding Design
W. D. Wallis, Introduction to Combinatorial Designs, Second Edition
Lawrence C. Washington, Elliptic Curves: Number Theory and Cryptography
 DISCRETE MATHEMATICS AND ITS APPLICATIONS
Series Editor KENNETH H. ROSEN

IntroductIon to

combInatorIal desIgns
Second edition

W.D. Wall i s
Front cover art: Shallow Waters by Burghilde Gruber

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2007 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Version Date: 20110725

International Standard Book Number-13: 978-1-58488-839-0 (eBook - PDF)

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

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

For permission to photocopy or use material electronically from this work, please access www.copyright.
com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood
Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and
registration for a variety of users. For organizations that have been granted a photocopy license by the CCC,
a separate system of payment has been arranged.

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

and the CRC Press Web site at

http://www.crcpress.com
This book is dedicated to the coffee growers of the world,
without whom many of its theorems would not exist.

(Paul Erdös said, “A mathematician is a machine

for turning coffee into theorems.”)
Contents

List of Figures xiii

Preface xv

1 Basic Concepts 1
1.1 Combinatorial Designs . . . . . . . . . . . . . . . . . . . . . 1
1.2 Some Examples of Designs . . . . . . . . . . . . . . . . . . . 4
1.3 Block Designs . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Systems of Distinct Representatives . . . . . . . . . . . . . . 11

2 Balanced Designs 15
2.1 Pairwise Balanced Designs . . . . . . . . . . . . . . . . . . . 15
2.2 Balanced Incomplete Block Designs . . . . . . . . . . . . . . 20
2.3 Another Proof of Fisher’s Inequality . . . . . . . . . . . . . . 29
2.4 t-Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Finite Geometries 35
3.1 Finite Affine Planes . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Finite Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Construction of Finite Affine Geometries . . . . . . . . . . . 45
3.4 Finite Projective Geometries . . . . . . . . . . . . . . . . . . 49

ix
x

4 Some Properties of Finite Geometries 55

4.1 Ovals in Projective Planes . . . . . . . . . . . . . . . . . . . 55
4.2 The Desargues Configuration . . . . . . . . . . . . . . . . . . 58

5 Difference Sets and Difference Methods 63

5.1 Difference Sets . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 Construction of Difference Sets . . . . . . . . . . . . . . . . . 65
5.3 Properties of Difference Sets . . . . . . . . . . . . . . . . . . 71
5.4 General Difference Methods . . . . . . . . . . . . . . . . . . . 73
5.5 Singer Difference Sets . . . . . . . . . . . . . . . . . . . . . . 79

6 More about Block Designs 85

6.1 Residual and Derived Designs . . . . . . . . . . . . . . . . . 85
6.2 Resolvability . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

7 The Main Existence Theorem 97

7.1 Sums of Squares . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.2 The Bruck-Ryser-Chowla Theorem . . . . . . . . . . . . . . . 105
7.3 Another Proof . . . . . . . . . . . . . . . . . . . . . . . . . . 112

8 Latin Squares 117

8.1 Latin Squares and Subsquares . . . . . . . . . . . . . . . . . 117
8.2 Orthogonality . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.3 Idempotent Latin Squares . . . . . . . . . . . . . . . . . . . . 126
8.4 Transversal Designs . . . . . . . . . . . . . . . . . . . . . . . 130

9 More about Orthogonality 137

9.1 Spouse-Avoiding Mixed Doubles Tournaments . . . . . . . . 137
9.2 Three Orthogonal Latin Squares . . . . . . . . . . . . . . . . 140
9.3 Bachelor Squares . . . . . . . . . . . . . . . . . . . . . . . . . 146

10 One-Factorizations 153
10.1 Basic Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
10.2 The Variability of One-Factorizations . . . . . . . . . . . . . 158
 xi

10.3 Starters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

11 Applications of One-Factorizations 167

11.1 An Application to Finite Projective Planes . . . . . . . . . . 167
11.2 Tournament Applications of One-Factorizations . . . . . . . 169
11.3 Tournaments Balanced for Carryover . . . . . . . . . . . . . 173

12 Steiner Triple Systems 179

12.1 Construction of Triple Systems . . . . . . . . . . . . . . . . . 179
12.2 Subsystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
12.3 Simple Triple Systems . . . . . . . . . . . . . . . . . . . . . . 189
12.4 Cyclic Triple Systems . . . . . . . . . . . . . . . . . . . . . . 191
12.5 Large Sets and Related Designs . . . . . . . . . . . . . . . . 195

13 Kirkman Triple Systems and Generalizations 201

13.1 Kirkman Triple Systems . . . . . . . . . . . . . . . . . . . . 201
13.2 Kirkman Packings and Coverings . . . . . . . . . . . . . . . 209

14 Hadamard Matrices 217

14.1 Basic Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
14.2 Hadamard Matrices and Block Designs . . . . . . . . . . . . 222
14.3 Further Hadamard Matrix Constructions . . . . . . . . . . . 226
14.4 Regular Hadamard Matrices . . . . . . . . . . . . . . . . . . 230
14.5 Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

15 Room Squares 243

15.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
15.2 Starter Constructions . . . . . . . . . . . . . . . . . . . . . . 246
15.3 Subsquare Constructions . . . . . . . . . . . . . . . . . . . . 251
15.4 The Existence Theorem . . . . . . . . . . . . . . . . . . . . . 255
15.5 Howell Rotations . . . . . . . . . . . . . . . . . . . . . . . . . 260

16 Further Applications of Design Theory 265

16.1 Statistical Applications . . . . . . . . . . . . . . . . . . . . . 265
xii

16.2 Information and Cryptography . . . . . . . . . . . . . . . . . 273

16.3 Golf Designs . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

References 281

Answers and Solutions 293

Index 307
List of Figures

2.1 A design and its complement. . . . . . . . . . . . . . . . . . . 27

3.1 Addition table for GF (32 ). . . . . . . . . . . . . . . . . . . . 43

2
3.2 Multiplication table for GF (3 ). . . . . . . . . . . . . . . . . 43

4.1 The Desargues Configuration. . . . . . . . . . . . . . . . . . . 58

5.1 Some difference sets. . . . . . . . . . . . . . . . . . . . . . . . 64

5.2 Supplementary difference sets. . . . . . . . . . . . . . . . . . . 73
5.3 Initial blocks: general examples. . . . . . . . . . . . . . . . . 76

6.1 Residual and derived designs. . . . . . . . . . . . . . . . . . . 86

7.1 Small parameters ruled out by Theorem 7.13. . . . . . . . . . 111

7.2 Small undecided parameters for a BIBD. . . . . . . . . . . . . 111

8.1 Result of substituting for three blocks. . . . . . . . . . . . . . 129

8.2 Final arrays. . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.3 For Theorem 8.12: v = 36 + v1 , t = 12 + t1 . . . . . . . . . . . 135
8.4 Self-orthogonal Latin squares of sides 10 and 14. . . . . . . . 135

9.1 Initial rows and columns for Theorem 9.6. . . . . . . . . . . . 144

9.2 Exceptions to be checked. . . . . . . . . . . . . . . . . . . . . 145
9.3 Bachelor square A of order 11 (× stands for 10). . . . . . . . 150

10.1 A regular graph without a one-factorization. . . . . . . . . . . 154

10.2 Graph for Exercise 10.1.2 . . . . . . . . . . . . . . . . . . . . 157
10.3 One-factorizations of K8 . . . . . . . . . . . . . . . . . . . . . 160
10.4 Divisions in one-factorizations of K8 . . . . . . . . . . . . . . . 161

xiii
xiv

10.5 The patterned starter . . . . . . . . . . . . . . . . . . . . . . 165

11.1 The one-factorizations of 2K6 , properly oriented. . . . . . . . 172

11.2 A schedule balanced for carryover. . . . . . . . . . . . . . . . 175

12.1 Skew transversal squares of sides 4 and 9. . . . . . . . . . . . 189

12.2 Some small difference partitions. . . . . . . . . . . . . . . . . 193
12.3 Families of difference partitions. . . . . . . . . . . . . . . . . . 194
12.4 All triad designs on seven treatments. . . . . . . . . . . . . . 200

13.1 Small Kirkman triple systems. . . . . . . . . . . . . . . . . . . 202

13.2 Construction of KP (12, 4). . . . . . . . . . . . . . . . . . . . 210

14.1 Incidence matrix and Hadamard matrix. . . . . . . . . . . . . 224

14.2 Initial rows for matrices used in Williamson’s construction. . 230
14.3 Construction of a (36, 15, 6)-design. . . . . . . . . . . . . . . . 233
14.4 Representative Hadamard matrices of order 16. . . . . . . . . 237

15.1 A Room square of side 7. . . . . . . . . . . . . . . . . . . . . 243

15.2 Standardized Room squares of side 7. . . . . . . . . . . . . . 245
15.3 Room square from a starter and adder. . . . . . . . . . . . . . 247
15.4 The array R23 of Theorem 15.9. . . . . . . . . . . . . . . . . 259
15.5 Another array to illustrate Theorem 15.9. . . . . . . . . . . . 260
15.6 Room square and Howell rotation. . . . . . . . . . . . . . . . 263

16.1 A golf design for seven clubs. . . . . . . . . . . . . . . . . . . 277

16.2 Array to generate a golf design for 17 teams. . . . . . . . . . 278
Preface

This book is a revised edition of my 1988 Combinatorial Designs, after eigh-

teen years.
I have deleted some things from the first edition that some readers (myself
included) thought were a little too specialized. In their place, I have added
some material that reflects recent results, and a very broad outline of some
applications. I have also significantly reorganized the development, with the
necessary algebra and number theory background being incorporated into the
chapters where they are first used.
It has been suggested that combinatorial theory is the fastest-growing area
of modern mathematics. The largest part of the subject is graph theory, as
evidenced by the huge number of research publications in that part of the
discipline. Accordingly, there have appeared a good many textbooks devoted
to graph theory, suitable for use at junior, senior, and graduate levels. A
second major part of combinatorial mathematics is enumeration. The third—
the one we treat here—is the theory of combinatorial designs.
Design theory began with Euler’s invention of the Latin square 200 years
ago. The next area of particular interest was the balanced incomplete block
design and the special case of finite projective and affine planes. This interest
was fanned by the statistical interpretation of designs and by the geometri-
cal nature of planes. Many textbooks are written as though these, what we
might call “classical designs,” are the only combinatorial designs. However,
over the last 25 years, combinatorial researchers have discussed a wider range
of designs: one-factorizations, Room squares, and other designs based on un-
ordered pairs, various tournament designs, nested designs, and so on. This
interest has been spurred by the discovery of numerous applications. These in-
clude coding theory and cryptography, both theoretical and applied computer
science, design of enterprise networks and other aspects of communications
theory, industrial engineering, chemistry and biology.
In writing the book my aims were:
• to enable instructors to present combinatorial designs to undergraduate
and beginning graduate students in more than a cursory fashion;
• to give a good groundwork in the classical areas of design theory: block
designs, finite geometries, and Latin squares; and then to introduce some
modern extensions of design theory;

xv
xvi

• to lead students toward the current boundaries of the subject—I hope

they will be capable of reading current research papers after studying
the book;
• to present a well-rounded textbook, providing motivation, instructive
examples, theorems, and exercises for all its topics.
After an overview we introduce balanced designs. There are two chapters on
finite geometries; Chapters 5 and 6 go further into balanced incomplete block
designs (difference methods, residual and derived designs, resolvability), and
Chapter 7 is on the existence theorem of Bruck, Ryser and Chowla. There are
two chapters each on Latin squares, one-factorizations and triple systems, and
chapters on Hadamard matrices and Room squares. We finish with a rough
sketch of some statistical and other applications of designs. The applications
are outside our scope, but a number of references for further reading are given
in that chapter.
I have included exercises with each section, and selected answers are given.
Exercises with answers or (complete or partial) solutions are indicated by an
asterisk after the question number. More complete solutions will appear on
the web at
http://www.math.siu.edu/Wallis/designs
and errata, comments, updates and extensions will appear on that site.
I would like to thank Ebad Mahmoodian and Katherine Heinrich for their
extensive comments on the first edition, and Elizabeth Billington for her care-
ful reading and helpful suggestions on this edition.

W. D. Wallis
Department of Mathematics
Southern Illinois University
Carbondale, IL 62901-4408
Chapter 1
Basic Concepts

1.1 Combinatorial Designs

Informally, one may define a combinatorial design to be a way of selecting
subsets from a finite set in such a way that some specified conditions are
satisfied. Although we shall not define the type of condition precisely, they
generally involve incidence: set membership, set intersection, and so on.
As an example, say it is required to select 3-sets from the seven objects
{0, 1, 2, 3, 4, 5, 6} in such a way that each object occurs in three of the 3-sets
and every intersection of two 3-sets has precisely one member. The solution
to this problem—the way of selecting the 3-sets—is a combinatorial design.
One possible solution is

{012, 034, 056, 135, 146, 236, 245}

where 012 represents {0, 1, 2} and so on. (We shall often omit set brackets,
when no confusion can arise.)
The example above does not use the arithmetical properties of the numbers
0, 1, 2, 3, 4, 5, 6. In fact, we do not even know what these symbols stand for.
They could be the corresponding integers; they could be the integers modulo
7, or modulo some larger base; or they could be labels attached to any seven
objects. The elements of the set are not viewed from an arithmetical viewpoint
(although arithmetic will sometimes be very useful in showing that a design
exists or in solving other problems).
Our example involved showing that a certain design exists. Other problems
can arise. For example, in our problem say there are n 3-sets. Then they
have between them 3n entries. But there are seven objects, and each occurs
in three of the subsets. So when they are listed, there are 21 entries. Thus
3n = 21 and n = 7. In other words, the number of subsets is determined; we
can prove that, if a solution exists, it must contain precisely seven subsets.
So one can establish properties of possible solutions.
We now prove the uniqueness of the given solution of our example, in the
following sense. If we have any one solution, we can obviously obtain another

1
2 Introduction to Combinatorial Designs

by relabeling the objects in some way. For example, the relabeling of “1” as
“2” and “2” as “1” in the given solution leads to the different-looking answer

{012, 034, 056, 235, 246, 136, 145}.

But this is essentially the same as the original. We shall consider two solu-
tions to be isomorphic if the list of 3-sets in one can be obtained from the
list of 3-sets of the other by carrying out some permutation of the symbols
consistently. What we now show is that the solution of the problem is unique,
up to isomorphism.
Assume that we have a solution to the problem. There must be three 3-sets
that contain the object 0. Suppose they are 0ab, 0cd, and 0ef, where a, b, c,
d, e and f are chosen from the numbers 1 through 6. Since the intersection
of any two sets is to have one entry, and since 0 is common to 0ab and 0cd,
we must have
a
= c, a
= d, b
= c, b
= d.
Moreover, since 0ab is to have three members, we know that

a
= b, a
= 0, b
= 0.

By similar observations we see that a, b, c, d, e, and f must all be different

from each other and must all be nonzero. So

{a, b, c, d, e, f} = {1, 2, 3, 4, 5, 6}.

Whichever ordering is used, we may permute the symbols at will, so we may

take
a = 1, b = 2, c = 3, d = 4, e = 5, f = 6
up to isomorphism, and the design contains the sets

012, 034, 056.

Now 1 belongs to three sets. One of them is 012. Suppose 1gh is another.
Since two sets have one common member, we have

g
= 0, g
= 2, h
= 0, h
= 2

(consider the two sets 012 and 1gh). Also,

|{g, h} ∩ {3, 4}| = 1.

If h is the common member, we can swap the labels g and h; if 4 is the

common member, we can swap the labels 3 and 4. So up to isomorphism, we
can assume that
g = 3, h
= 3, h
= 4.
 Basic Concepts 3

If h = 6, we can exchange labels 5 and 6. So we can assume that h = 5.

Therefore, two of the sets containing 1 are 012 and 135. It is easy to see that
the third set cannot contain 0, 2, 3, or 5, so it is 146.
We next consider the sets containing 2. In the same way we see that they
must be
012, 236, 245.
So the design is uniquely determined, up to isomorphism.
We could also show that our problem has no solution unless the number of
objects is precisely seven. More generally, suppose we replaced “3” with “k”:
Each object is to belong to precisely k of the k-sets. Then we could prove
that exactly k 2 − k + 1 objects are needed. But this necessary condition is
not sufficient: Even with the correct number of elements, there is no solution
for k = 7. This will be proven in Section 7.2.
Finally, our example design has the following property. If you select any
two of the seven objects, there is exactly one 3-set that contains both of them.
We say that the design is balanced; this important property will be explored
in Section 1.3.

Exercises 1.1

1.1.1∗ Show that there exists a combinatorial design on the nine symbols
{0, 1, 2, 3, 4, 5, 6, 7, 8}, that consists of 12 3-sets with the property that
any two symbols occur together in exactly one 3-set. Show that the
design is uniquely defined up to isomorphism.

1.1.2∗ A combinatorial design consists of the selection of 2-sets and 3-sets from
the set {0, 1, 2, 3, 4, 5}, such that:
(a) Every pair of symbols occur together in exactly one subset;
(b) There are at least three 3-sets.
Prove that no two 3-sets can be disjoint. Show that the number of 3-
sets can be either three or four, but no larger; and show that there are
precisely two nonisomorphic solutions, one with three and one with four
3-sets.

1.1.3 A tennis club has 20 members. One day the members schedule exactly
14 singles (that is, two-player) matches among themselves, with each
member playing in at least one game. Prove that the schedule includes
a set of six games that are disjoint—that is, no two of them contain a
common player—but no set of seven disjoint matches.

1.1.4 Can problem 1.1.3 be generalized to n players and a schedule of m

matches?
4 Introduction to Combinatorial Designs

1.1.5 It is required to select 3-sets from the seven objects {0, 1, 2, 3, 4, 5, 6} in

such a way that each object occurs in three of the 3-sets and every pair
of objects belongs to precisely one 3-set. Show that every intersection
of two 3-sets has precisely one member.

1.2 Some Examples of Designs

Many examples of combinatorial designs have been discussed in the litera-
ture. We shall define some of these, to provide examples of the concept. Some
of these ideas will be restated when we consider the concepts again in later
chapters.
Linked Design. A linked design is a way of selecting subsets from a set
in such a way that any two subsets have intersection size, µ, where µ is a
constant for the design. Such a constant is called a parameter of the design;
one might say that a linked design on v objects, in which every pair of blocks
has intersection size µ, is a “linked design with parameters (v, µ).” The designs
discussed in Section 1.1 were linked designs with parameters (7, 1), but with
certain other conditions (constant subset size, and every element belonging to
the same number of subsets).
Balanced Design. A way of selecting subsets from a set is called balanced
if any two elements occur together in precisely λ of the subsets, where λ is a
constant. Balanced designs are similar to linked designs, and we shall see that
they are in a sense equivalent classes of objects. However, balanced designs
have significant applications, particularly in statistical theory, and will be
studied extensively in this book. Of special interest are balanced incomplete
block designs, which have constant subset size; see below, Section 1.3.
One-Factorization. A one-factorization on a set S of n symbols is a way of
organizing the n(n − 1)/2 unordered pairs of members of S into classes called
one-factors, or simply factors, such that every pair belongs to precisely one
factors and such that every symbol belongs to precisely one member of each
class. A one-factorization is a combinatorial design; one may interpret the
phrase “way of selecting subsets” in the definition to refer to the selection of
(all n(n−1)/2) unordered pairs from the set S, and interpret the arrangement
of these pairs into classes as a “specified condition”; alternatively, one could
say that the universal “finite set” is the set of unordered pairs on S. A one-
factorization has only one parameter, the number n, the size of the symbol
set S.
Latin Square. A Latin square L of side n is an n × n array whose entries all
come from some n-set S, with the property that each row and each column is
 Basic Concepts 5

a permutation of S. One can either interpret the rows of L as subsets (of size
n) that are ordered, or interpret the entries of L as subsets (of size 1) that are
subjected to a two-dimensional positioning rule. Examples of Latin squares
are
132 132
12
321 213
21
213 321

Graph. A graph G = G(V, E) consists of a finite set V of objects called ver-

tices (or points), and a set E of unordered pairs of members of V called edges.
This definition is consistent with the conventions that the edges of a graph
are not directed, that the two vertices constituting an edge (its “endpoints”)
are distinct, and that two vertices belong together to no edge or to one edge
(“are adjacent ”) but to no more. However, the reader should be aware that
the word “graph” is sometimes used for objects with directed edges, with
loops on vertices, with two or more edges with the same endpoints (“multiple
edges”) or with infinite vertex sets. If this more general terminology is used,
“graphs” without loops or multiple edges are called simple. For our part, we
use the word multigraph for the case where multiple edges are allowed.
Graph theory is a large area of study in combinatorial mathematics, and the
reader has probably already encountered it. It is not our intention to claim
that graph theory is a part of design theory, and the methods of approach of a
graph theorist and a design theorist are often (but not always) quite different.
However, it should be noted that a graph is a form of design. Moreover, the
terminology of graph theory is occasionally useful to us; see Chapter 10 in
particular.

Exercises 1.2

1.2.1 Prove that the parameter of a one-factorization must be even.

1.2.2∗ Construct a one-factorization on the sets {1, 2, 3, 4}.

1.2.3 Construct a one-factorization on the set {1, 2, 3, 4, 5, 6}.

Hint: You might as well start with factor {12 34 56}. (Factorizations
starting {12 35 46}, for example, can be transformed by exchanging all
4’s and 5’s throughout.)

1.2.4 Prove that there are precisely two Latin squares of side 3 with first row

1 3 2

namely, those shown in this section.

1.2.5 Exhibit a Latin square of side 4.

6 Introduction to Combinatorial Designs

1.2.6 Graphs G and H are isomorphic if there is a bijection φ from the vertex
set of G to the vertex set of H such that, when {x, y} is an edge of G,
then {xφyφ} is an edge of H.

1.2.7∗ How many different graphs are there with vertex set {1, 2, 3}? How
many isomorphism classes of these graphs are there?

1.2.8 What is the smallest value of v such that there exist nonisomorphic
graphs on v vertices that have the same number of edges?

1.3 Block Designs

The members of the universal set S in a combinatorial design are usually
called treatments, or varieties, and the subsets chosen are called blocks. S
is called the support of the design. We shall say that a design of a certain
type is a block design if in the definition of the type of design, the blocks are
simply unordered sets of treatments and there is no structural ordering or
pattern in the blocks. This vague definition (containing the undefined phrase
“structural ordering or pattern”) is really a statement of intention: If the
blocks are ordered internally (as in a Latin square) or arranged in a pattern
(as in a one-factorization), we shall usually study them in a different way from
a linked design, for example.
There is no requirement that all the blocks in a design should be different.
If two blocks have the same set of elements, we say there is a “repeated block.”
A design that has no repeated block is called simple.
We now define an important class of block design. We say a design is
regular if every treatment occurs equally often in the design. Formally, a
regular design based on a v-set S is a collection of k-sets from S such that
every member of S belongs to r of the k-sets. It is usual to write b for the
number of blocks in a design. So a regular design has four parameters: v, b,
r, and k. However, they are not independent.

THEOREM 1.1
In any regular design,
bk = vr. (1.1)

Proof. One counts, in two different ways, all the ordered pairs (x, y) such that
treatment x belongs to block y. Since every treatment belongs to r blocks,
there are r ordered pairs for each treatment, so the number is vr. Similarly,
 Basic Concepts 7

each block contributes k ordered pairs, so the summation yields bk. Therefore
bk = vr.

The number of blocks that contain a given treatment is called the replication
number or frequency of that treatment. So the defining characteristics of a
regular design are that all elements have the same replication number and
that all blocks have the same size.
If all v treatments occur in a block of a design, that block is called complete.
If a regular design has that property, then k = v and obviously r = b. Such a
design is a complete design, and it has very little interest unless some further
structure is imposed (as, for example, in a Latin square). We say that a design
is incomplete if at least one block is incomplete. If v = b, the design is called
symmetric.
If x and y are any two different treatments in an incomplete design, we
shall refer to the number of blocks that contain both x and y as the covalency
of x and y, and write it as λxy . Many important properties of block designs
are concerned with this covalency function. The one that has most frequently
been studied is the property of balance: a balanced incomplete block design, or
BIBD, is a regular incomplete design in which λxy is a constant, independent
of the choice of x and y. These designs were defined (essentially as a puzzle) by
Woolhouse [181], in the annual Lady’s and Gentleman’s Diary, a “Collection of
mathematical puzzles and aenigmas” that he edited; Yates [188] first studied
them from a statistical viewpoint.
It is usual to write λ for the constant covalency value, in a balanced incom-
plete block design, λ is sometimes called the index of the design. One often
refers to a balanced incomplete block design by using the five parameters
(v, b, r, k, λ); for example, we say “a (9, 12, 4, 3, 1)-design” or “a (9, 12, 4, 3, 1)-
BIBD” to mean a balanced incomplete block design with 12 blocks of size 3,
based on 9 treatments, with replication number 4 and index 1.
A balanced design with λ = 0, or a null design, is often called “trivial,”
and so is a complete design. We shall demand that a balanced incomplete
block design be not trivial; since completeness is already outlawed, the added
restriction is that λ > 0.

THEOREM 1.2
In a (v, b, r, k, λ)-BIBD,

r(k − 1) = λ(v − 1). (1.2)

Proof. Consider the blocks of the design that contain a given treatment, x
say. There are r such blocks. Because of the balance property, every treatment
other than x must occur in λ of them. So if we list all entries of the blocks,
we list x r times and we list every other treatment λ times.
8 Introduction to Combinatorial Designs

The list contains rk entries. So

rk = r + λ(v − 1),
r(k − 1) = λ(v − 1).

The word incidence is used to describe the relationship between blocks and
treatments in a design. One can say block B is incident with treatment t,
or treatment t is incident with block B, to mean that t is a member of B.
A block design may be specified by its incidence matrix: If the design has b
blocks B1 , B2 , . . . , Bb and v treatments t1 , t2 , . . . , tv , define a v × b matrix A
with (i, j) entry aij as follows:


1 if ti ∈ Bj ;
aij =
0 otherwise.

This matrix A is called the incidence matrix of the design. The definition
means that each block corresponds to a column of the incidence matrix, and
each treatment corresponds to a row. For example, the regular design with
treatments t1 , t2 , t3 , t4 and blocks B1 = {t1 , t2 }, B2 = {t3 , t4 }, B3 = {t1 , t4 },
B4 = {t2 , t3 }, has incidence matrix
 
1010
1 0 0 1
A= 0 1 0 1.


0110

As another example, observe that the following blocks form a (6, 10, 5, 3, 2)-
design:
123, 124, 135, 146, 156,
236, 245, 256, 345, 346.
This design has incidence matrix
 
11111 00000
1 1 0 0 0
 1 1 1 0 0

1 0 1 0 0
 1 0 0 1 1
.
0 1 0 1 0
 0 1 0 1 1

0 0 1 0 1 0 1 1 1 0
00011 10101

It is often useful to discuss designs that lie within other designs. The
generic term “subdesign” is used. However, in some cases it is not clear what
the appropriate definition should be: For example, does one require that each
block be a subset of a block in the original design, or that the set of blocks be a
subset of the original set? Usually, we take the latter view, that a subdesign of
 Basic Concepts 9

a block design is the block design formed from a subset of the block set and a
(sufficiently large) subset of the treatment set. However, different definitions
may be used for different types of designs. Observe that, for example, a
subdesign of a balanced incomplete block design will not necessarily be a
balanced incomplete block design.
As was pointed out in Section 1.1, we do not wish to consider two designs to
be different if it is possible to relabel the treatments and blocks of one in such
a way that the other design is obtained. To formalize the earlier discussion,
block designs D and E are defined to be isomorphic if there is a one-to-one
map from the set of treatments of D onto the set of treatments of E such that
when the mapping is applied to a block of D, the result is a block of E, and
such that every block of E can be derived from some block of D in this way.
A map with this property is called an isomorphism.
Alternatively, one could define an isomorphism to consist of two one-to-one
maps, one mapping the set of treatments of D onto the set of treatments of E,
and the other mapping the set of blocks of D onto the set of blocks of E, with
the property that if treatment t maps to treatment tφ and block B maps to
block Bφ, then t belongs to B if and only if tφ belongs to Bφ:

t ∈ B ⇔ tφ ∈ Bφ.

This property is described by saying that “isomorphisms preserve incidence.”

In terms of incidence matrices, two block designs are isomorphic if the
incidence matrix of one can be transformed into the incidence matrix of the
other by row and column permutations. As an example of this, consider
the design that was discussed in Section 1.1. It is easy to verify that it is
a balanced incomplete block design with parameters (7, 7, 3, 3, 1). The first
representation of the design had blocks

{012, 034, 056, 135, 146, 236, 245}.

The corresponding incidence matrix is

 
11100 00
1 0 0 1 1
 0 0

1 0 0 0 0
 1 1

A= 0 1 0 1 0 1 0
.
0 1 0 0 1
 0 1

0 0 1 1 0 0 1
00101 10

When we interchanged treatments 1 and 2 we obtained blocks

{012, 034, 056, 235, 246, 136, 145},

with incidence matrix

10 Introduction to Combinatorial Designs

 
1110 000
1 0 0 0
 0 1 1

1 0 0 1
 1 0 0

B=
0 1 0 1 0 1 0
.
0 1 0 0
 1 0 1

0 0 1 1 0 0 1
0010 110

Obviously, B can be converted into the original matrix A by interchanging the

second and third rows (which correspond to treatments 1 and 2, respectively).
There are, in fact, other row and column permutations that have the same
effect: If the last two rows of B are interchanged and then the permutation
(46)(57) is carried out on the columns of B, we obtain A.

Exercises 1.3
1.3.1∗ Prove that there are exactly three nonisomorphic regular designs with
v = 4, b = 6, r = 3 and k = 2. (Some will have repeated blocks.)
Hint: Subdivide the problem by asking, “How many solutions are there
in which some block occurs n times but none occurs more than n times?”
1.3.2 Prove that there is a (7, 7, 4, 4, 2)-design and that it is unique up to
isomorphism.
1.3.3 Prove that every (6, 10, 5, 3, 2)-design is simple.
1.3.4∗ Find a (7, 14, 6, 3, 2)-design that is not simple.
1.3.5 Show that there is a (7, 7, 3, 3, 1)-design whose blocks include 124, 235
and 347; find its other blocks. What is its incidence matrix?
1.3.6 A balanced incomplete block design has eight treatments and blocks of
size four. Prove that the design has at least 14 blocks.
1.3.7 Write down the incidence matrices of the three designs you found in
Exercise 1.1.1.
1.3.8∗ A design is constructed in the following way. The names of 16 treat-
ments are written in a 4 × 4 array: for example,
1 2 3 4
5 6 7 8
.
9 10 11 12
13 14 15 16
There are 16 blocks: To form a block, one chooses a row and a column
of the array, deletes the common treatment, and takes the other six
elements. (For example, row 1 and column 2 yield the block {1, 3, 4, 6, 10,
14}; treatment 2 is deleted.)
 Basic Concepts 11

(i) Prove that the design is a balanced incomplete block design. What
are its parameters?
(ii) Does the construction above work if the array is of a size different
from 4 × 4?

1.3.9∗ In each row of the following table, fill in the blanks so that the param-
eters are possible parameters for a balanced incomplete block design, or
else show that this is impossible.

v b r k λ
35 3 1
6 4 2
14 7 4
13 6 1
21 28 3
17 8 5
21 30 7
17 7 1

1.3.10 Suppose a balanced incomplete block design satisfies v = b, and λ is

odd. Show that v cannot be even.
Hint: Use (1.2).

1.4 Systems of Distinct Representatives

Suppose D is a block design based on S = {1, 2, . . . , v} with blocks B1 , B2 ,
. . . , Bb . We define a system of distinct representatives (SDR) for D to be a
way of selecting a member xi from each block Bi such that x1 , x2 , . . . are all
different.
As an example, consider the blocks

124, 124, 134, 235, 246, 1256.

One system of distinct representatives for them is

1, 2, 3, 5, 4, 6
12 Introduction to Combinatorial Designs

(where the representatives are listed in the same order as the blocks). There
are several others. On the other hand, the blocks

124, 124, 134, 23, 24, 1256

have no SDR.
If the design D is to have an SDR, it is clearly necessary that D have at
least as many treatments as blocks. The example above shows that this is not
sufficient: If we consider the first five blocks of the design, they constitute a
design with b = 5 and v = 4. If there were an SDR for the six-block design,
then the first five elements would be an SDR for the five-block subdesign,
which is impossible. The implied necessary condition, that D contain no
subdesign with v < b, is in fact sufficient, as we now prove.

THEOREM 1.3 [64]

A block design D has a system of distinct representatives if and only if it never
occurs that some n blocks contain between them fewer than n treatments.

Proof. We proceed by induction on the number of blocks. If D has one block,

the result is obvious. Assume the theorem to be true for all designs with fewer
than b blocks. Suppose D has b blocks B1 , B2 , . . . , Bb and the v treatments
{1, 2, . . . , v}, and suppose D satisfies the hypothesis that the union of any n
blocks has size at least n, for 1 ≤ n ≤ b. By induction, any set of n blocks
has an SDR, provided that n < b. We distinguish two cases.
(i) Suppose no set of n blocks contains less than n + 1 treatments in its
union, for n < b. Select any element x1 ∈ B1 , and write Bi∗ = Bi \{xi }, for
i ∈ {1, 2, . . . , b}. Then the union of any n of the Bi∗ has at least n elements, for
n = 1, 2, . . . , b−1. By the induction hypothesis, there is an SDR x2 , x3 , . . . , xb
for B2 , B3 , . . . , Bb , so x1 , x2 , x3 , . . . , xb is an SDR for the original design D.
(ii) Suppose there is a set of n blocks whose union has precisely n ele-
ments, for some n less than b. Without loss of generality, take these blocks
as B1 , B2 , . . . , Bn . For i > n, write Bi∗ to mean Bi with all members of
B1 ∪ B2 ∪ . . . ∪ Bn deleted. It is easy to see that the design with blocks
∗ ∗
Bn+1 , Bn+2 , . . . , Bb∗ satisfies the conditions of the theorem (if Bn+i ∗
1
∗
, Bn+i 2
,
∗
. . . , Bn+ik were k blocks whose union has less than k elements, then B1 , B2 ,
∗ ∗ ∗
. . . , Bn , Bn+i 1
, Bn+i 2
, . . . , Bn+i k
would be n + k blocks of D whose union has
less than n + k elements, which is impossible). From the induction hypothesis
both sets have SDRs, and clearly they are disjoint, so together they comprise
an SDR for D.

We give two important applications of this result. Another important ap-

plication occurs later, as Theorem 8.1.
 Basic Concepts 13

THEOREM 1.4
A set of tasks are to be performed by a set of workers; no task may be assigned
to a worker who is not qualified to perform it, and no worker may be required
to perform more than one task. A necessary and sufficient condition that all
tasks can be assigned is that for any set of n tasks, there are at least n workers
capable of performing at least one of them each.

Proof. Number the tasks; for task i, write Bi for the set of workers qualified
to perform it. Then what is required is an SDR for the Bi , and the necessary
and sufficient condition is precisely that of Theorem 1.3.

THEOREM 1.5
If M is a square matrix of zeros and ones with the property that every row and
every column sums to r, then there are permutation matrices P1 , P2 , . . . , Pr
such that
M = P1 + P2 + . . . + Pr .

Proof. We proceed by induction on r. If r = 1, the theorem is clearly true.

Assume it holds for r < k; let M be a v × v matrix of zeros and ones with
constant row and column sums k. Define

Bi = {j : mij = 1}.

These Bi form a design with blocks of constant size k.

We show that the union of any n blocks contains at least n elements. With-
out loss of generality, say they are blocks B1 , B2 , . . . , Bn . Consider

{(i, j) : j ∈ Bi , 1 ≤ i ≤ n}.

This set has kn elements. But no integer j can occur as right-hand element
in more than k of the pairs (since column j has only k nonzero entries), so at
least n different values of j must be represented.
Now the blocks Bi have a system of distinct representatives. If xi is the
representative of Bi , write Pk for the permutation matrix with (i, j) entry 1
if and only if j = xi . Then M − Pk is a zero-one matrix. Moreover, it has
constant row-sum and column-sum k − 1. So by the induction hypothesis
there are permutation matrices P1 , P2 , . . . , Pk−1 such that

M − Pk = P1 + P2 + . . . + Pk−1,

giving the result.

14 Introduction to Combinatorial Designs

Exercises 1.4

1.4.1 Do there exist SDRs for the following blocks?

(i) 12, 145, 12, 123
(ii) 12, 145, 12, 13, 23.

1.4.2∗ Show that the blocks

123, 124, 134, 235, 246, 1256

have precisely 17 SDRs.

1.4.3∗ Count the SDRs of the following sets of blocks:

(i) 123, 234, 345, 451, 512;
(ii) 123, 124, 125, 345, 345;
(iii) 123, 124, 135, 245, 345.

1.4.4 Suppose the design D has an SDR, and suppose each of the b blocks has
at least t elements. Prove that if t ≤ b, then D has at least t! SDRs.

1.4.5 The set Si has i + 1 members. Show that the system {S1 , S2 , . . . , Sn }
has at least 2n SDRs. Construct such a system with precisely 2n SDRs.

1.4.6 S and T are sets of v elements each. P is a set of ordered pairs(s, t),
where s is in S and t is in T . Define

A(s) = {t : t ∈ T, (s, t) ∈ P }
B(t) = {s : s ∈ S, (s, t) ∈ P }.

Prove that the following two properties are equivalent:

(a) For every k-subset X of S, 1 ≤ k ≤ v, ∪s∈X A(s) has at least k
elements;
(b) For every k-subset Y of T , 1 ≤ k ≤ v, ∪s∈Y B(t) has at least k
elements.
Chapter 2
Balanced Designs

2.1 Pairwise Balanced Designs

A pairwise balanced design of index λ is a way of selecting blocks from a set

of treatments (support set) such that any two treatments have covalency λ.
If there are v treatments and if every block size is a member of some set K of
positive integers, the design is designated a P B(v; K; λ). So the parameters
consist of two positive integers and one set of positive integers. To avoid
trivial differences between designs, block size 1 is not allowed.
The number of blocks is not normally treated as a parameter; one can have
two pairwise balanced designs with the same parameters but with different
numbers of blocks. Both

123, 145, 24, 25, 34, 35

and
123, 14, 15, 24, 25, 34, 35, 45

are P B(5; {3, 2}; 1)s, but they have six and eight blocks, respectively.
It should be observed that we do not require every member of K to be a
block size. For instance, the two examples are also PB(5; {4, 3, 2}; 1)s. More
generally, any P B(v; K; λ) is also a P B(v; L; λ) whenever K is a subset of L.
Various easy results may be proven about pairwise balanced designs. For
example, two copies of a P B(v; K; λ), taken together, constitute a P B(v; K;
2λ). Just as obviously, there is a P B(v; {v}; λ) for all positive integers v and
λ (it is the complete design with λ sets, each containing all v elements). The
following theorem is more interesting.

THEOREM 2.1
Suppose there exists a P B(v; K; λ), and for every element k of K there exists
a P B(k; L; µ). Then there exists a P B(v; L; λµ).

15
16 Introduction to Combinatorial Designs

Proof. Suppose we have a P B(v; K; λ) based on a v-set V . Suppose its blocks

are B1 , B2 , . . . , Bn , where Bi has ki elements. We replace each block by a new
collection of blocks. Given Bi , form a P B(ki ; L; µ), but instead of taking the
numbers 1, 2, . . . , ki as treatments, use the elements of Bi . This is done for
every i. If x and y are any two treatments, then {x, y} was contained in λ of
the Bi , and in each case Bi has been replaced by a collection of blocks, µ of
which contain {x, y}. So x and y occur together λµ times in total. Therefore,
the total collection of new blocks is a pairwise balanced design of index λµ,
based on V . The size of any block of the new design is a member of L, so the
design is a P B(v; L; λµ).

As an example, we construct a P B(15; {4, 3}; 3) from the P B(15; {5, 3}; 1)
with blocks

01234 56789 ABCDE

05A 06C 07E 08B 09D
16B 17D 18A 19C 15E
27C 2SE 29B 25D 26A
38D 39A 35C 36E 37B
49E 45B 46D 47A 48C.

We need a P B(5; {4, 3}; 3) and a P B(3; {4, 3}; 3). The former has blocks

1234 1235 1245 1345 2345 (2.1)

and the latter is the complete design

123 123 123.

The block 01234 is replaced by a copy of the set of blocks (2.1) in which
the treatments have been relabeled using the mapping

(1, 2, 3, 4, 5) → (0, 1, 2, 3, 4).

The new blocks are
0123 0124 0134 0234 1234. (2.2)
Similarly, 56789 and ABCDE are replaced by

5678 5679 5689 5789 6789 (2.3)

and
ABCD ABCE ABDE ACDE BCDE, (2.4)
respectively. The block 05A is replaced by

05A 05A 05A

 Balanced Designs 17

and similarly for every other 3-block. So the required design consists of the
blocks listed in (2.2), (2.3) and (2.4) and three copies of every 3-set in the
original design.
The P B(15; {5, 3}; 1) that we used above is an example of a useful class of
designs. We now give a more general construction.

THEOREM 2.2
There exists a P B(3k; {3, k}; 1) whenever k is odd.

Proof. We construct a design with treatment set {a1 , a2 , . . . , ak , b1 , b2 , . . . ,

bk , c1 , c2 , . . . , ck }. There are three blocks of size k (which we call “big blocks”),
namely
{a1 a2 . . . ak }, {b1 b2 . . . bk }, {c1 c2 . . . ck }.
The blocks of size 3 are the blocks

{ai bi+x ci−2x : 1 ≤ i ≤ k, 1 ≤ x ≤ k}

where subscripts that exceed k are reduced modulo k.

The pairs {ai , aj }, {bi , bj }, and {ci , cj } occur once each, in the big blocks.
The pair {ai , bj } occurs in {ai bi+x ci+2x } if and only if j ≡ i + x ( mod k); this
will happen for only one value of x, either x = j −i (if j > i) or x = k −j −i (if
j ≤ i). So only one block contains {ai , bi }. A similar remark applies to {bi , cj }.
If {ai , cj } occurs in {ai bi+x ci+2x } then i − 2x ≡ j (mod k), and this also
uniquely defines x; since k is odd, the solution is x ≡ 12 ·(k − 1)(j − 1) ( mod k).

The construction above does not generalize to even values of k, because

1
2
(k − 1)(j − i) is not necessarily an integer in that case. However, we shall
see in Chapter 8 that another construction is available for even k.
We now prove an interesting theorem concerning the number of blocks in a
pairwise balanced design. We start with an easy remark.

LEMMA 2.3
In a pairwise balanced design with λ = 1, no two blocks have two common
elements.

Proof. Suppose B1 = {x, y, . . .} and B2 = {x, y, . . .}. Then{x, y} is a subset

of B1 , and also of B2 . So λxy ≥ 2. But this contradicts the property that
λ = 1.

18 Introduction to Combinatorial Designs

THEOREM 2.4 [27]

Suppose there is a P B(v; K; 1) with b blocks, where b > 1. Then b ≥ v. If
b = v, then either the P B(v; K; 1) has one block of size v − 1 and the rest of
size 2, or else b = v = k 2 − k + 1 for some integer k and all the blocks have
size k.

Proof. Suppose a P B(v; K; 1) has treatments t1 , t2 , . . . , tv and b blocks

B1 , B2 , . . . , Bb . Say ki is the number of elements in Bi , and that tj belongs
to rj blocks. (We call rj the frequency or replication number of tj .) Then,
counting all elements of all blocks,


v 
b
rj = ki . (2.5)
j=1 i=1

If tj does not belong to Bi , then tj must belong to at least ki blocks: For

every element x of Bi there is a block that contains tj and x, and these blocks
are all disjoint because of Lemma 2.3. So

tj ∈
/ Bi ⇒ ki ≤ rj . (2.6)

The blocks are incomplete, and there are no blocks of size 1. So

1 < ki < v for 1 ≤ i ≤ b. (2.7)

There must be some treatment whose replication number is minimal. Say

it is tv , and write rv = m. Relabel the blocks so that those containing tv
are B1 , B2 , . . . , Bm . We can select an element of each block, other than tv ,
and clearly all these elements are different. Suppose (after relabeling) that
ti ∈ Bi , ti
= tv . Then if 1 ≤ i ≤ m, 1 ≤ j ≤ m, and i
= j, tj ∈ / Bi ; in
particular,
t1 ∈/ B2 , t2 ∈
/ B3 , . . . , tm−1 ∈
/ Bm , tm ∈
/ B1
so from (2.6)
k2 ≤ r1 , k3 ≤ r2 , . . . , km ≤ rm−1 , k1 ≤ rm , (2.8)
whence

m 
m
ki ≤ rj (2.9)
i=1 j=1

Also, tv ∈
/ Bi for i > m, so ki ≤ rv for i > m, so


b 
b
ki ≤ rv . (2.10)
i=m+1 j=m+1
 Balanced Designs 19

Adding (2.9) and (2.10) and comparing with (2.5), we obtain


v 
b
rj = ki
j=1 i=1


m 
b
= ki + ki
i=1 i=m+1


m 
b 
b
≤ rj + rv ≤ rj
j=1 j=m+1 j=1

(since rv ≤ rj for all j), and this is impossible if b < v because the ri are all
positive.
In particular, suppose b = v. Then each of the inequalities in (2.8) must
be an equality, and also ki = r − i for all i > m. If we relabel the treat-
ments t1 , t2 , . . . , tm , we obtain ri = ki , all i ∈ {1 . . . v} for some ordering of
treatments and blocks. Moreover, tv is unchanged. Let us further relabel the
treatments (and simultaneously, the blocks) so that
r1 ≥ r2 ≥ . . . ≥ rv .
(Since tv had minimum frequency, it has still not been disturbed.)
We consider the various possibilities.
(i) Suppose r1 > r2 . Then r1 > rj for all j ≥ 2. So k1 = r1 > rj (j ≥ 2).
From (2.6), tj ∈ B1 for all j > 1. Of course, tr ∈
/ B1 . So
B1 = {t2 , t3 , . . . , tv },
and the other blocks must be
{tl , t2 }, {t1, t3 }, . . . , {t2 , tv }.

(ii) Suppose r1 = r2 = . . . = rj−1 > rj , where j > 2. Then tj ∈ B1 and

tj ∈ B2 (from (2.6)); since tv ∈ B1 ∩ B2 , the only possibility (according to
Lemma 2.3) is tj = tv and j = v. So we consider that case. Since rv < rv−1 =
kv−1 < v (from (2.7)) there are at least two blocks not containing tv . One
might be Bv , but suppose the other is Bx , where x
= v. Then from (2.6) we
have rx = kx ≤ rv , a contradiction.
(iii) Finally, suppose r1 = r2 = . . . = rv . We have constant block size and
constant frequency, a balanced incomplete block design. From (1.1) and (1.2)
we immediately deduce that b = v = k 2 − k + 1, where k is the common block
size.

We could generalize pairwise balanced designs by requiring that every set

of t treatments occurs in a fixed number of blocks. This is called a t-wise
balanced design.
20 Introduction to Combinatorial Designs

Exercises 2.1

2.1.1 Assume there exists a P B(v; {k, 3}; 1) with v ≡ 2 (mod 3). Prove that
k ≡ 2 (mod 3).
2.1.2∗ Assume a P B(7; {5, 4, 3}; 1) exists. Prove that the number of blocks of
size 5 is divisible by 3, and that consequently no such blocks exist. Can
there be any blocks of size 4?
2.1.3∗ Prove that no P B(8; {4, 3}; 1) can exist.
2.1.4 Prove1 that there is no P B(v; {4, 3}; 1) when v ≡ 2 (mod 3).
2.1.5 Suppose there is a P B(7; {5, 4, 3}; 1). Prove that the number of blocks
of size 5 is divisible by 3, and that consequently no such blocks exist.
Can there be any blocks of size 4?
2.1.6 Does there exist any P B(5; {3, 2}; 1) not isomorphic to the one given at
the beginning of this section?
2.1.7 Use Theorems 2.1 and 2.2 and the design exhibited in Section 1.1 to
prove that there is a balanced incomplete block design on 7 × 3t treat-
ments with k = 3 and λ = 1 for every integer t ≥ 0.

2.2 Balanced Incomplete Block Designs

One can consider a balanced incomplete block design as a kind of pairwise
balanced design, so all the results of the preceding section apply to them.
In particular, Theorem 2.1 can be applied. If the set L in that theorem has
only one element, the resulting design is regular (as defined in Section 1.3).
So we have the following corollary to Theorem 2.1.

THEOREM 2.5
Suppose there exists a P B(v; K; λ), and for every member k of K there is
a balanced incomplete block design on k treatments with block size l and bal-
ance parameter µ. Then there exists a balanced incomplete block design with
parameters
λµv(v − 1) λµ(v − 1)
v, , , l, λµ .
l(l − 1) l(l − 1)

1 See also exercises 12.1.8 and 13.1.2.

 Balanced Designs 21

In a balanced incomplete block design all the sets chosen are the same size,
every treatment occurs equally often and the design is pairwise balanced.
It is possible to relax the condition of pairwise balance but keep the other
two conditions, and the result is a regular block design, which we discussed
in Chapter 1. It is also possible to retain balance and retain the constant
number of replications but have more than one block size; such designs are
called (r, λ)-designs, where r and λ are the appropriate parameters. However,
we now prove that there are no examples where the restriction on the number
of replications is the only one that is dropped.

THEOREM 2.6
A pairwise balanced design in which every set has the same size is a balanced
incomplete block design.

Proof. If treatment x has frequency rx , then

(k − 1)rx = λ(v − 1)

as in Theorem 1.2. But this means that all treatments have the same frequency
r, where
v−1
r=λ .
k−1

If one treatment is deleted from a balanced incomplete block design, the

result is a pairwise balanced design with two block sizes, k and k − 1. These
pairwise balanced designs will have constant replication number (unless k =
2, and the blocks of size 1 that arise are then deleted). This gives rise to
many examples of (r, λ)-designs, but there are other examples D that are not
constructed in this way.
We shall now discuss the incidence matrices of balanced incomplete block
designs. Incidence matrices belong to the class of (0, 1)-matrices, matrices
each of whose entries is 0 or 1. It is interesting to observe that two equa-
tions characterize the incidence matrices of balanced incomplete block designs
among the wider class of (0, 1)-matrices. To express these equations, we use
two notations. An n × n identity matrix is denoted In , and an m × n matrix
with every entry 1 is denoted Jm×n . (If m = n, we simply write Jm for Jm×n ;
and in any event, we usually omit the subscripts whenever possible.)

THEOREM 2.7
If A is the incidence matrix of a (v, b, r, k, λ) design, then

AAT = (r − λ)Iv + λJv (2.11)

22 Introduction to Combinatorial Designs

and
Jv A = kJv×b . (2.12)

Conversely, if there is a v×b (0, 1)-matrix A that satisfies (2.11) and (2.12),
then
r(k − 1)
v= + 1,
λ
vr
b= ,
k
and provided that k < v, A is the incidence matrix of a (v, b, r, k, λ)-design.

Proof. First, suppose A is the incidence matrix of a (v, b, r, k, λ)-design. As

aij = 1 if and only if treatment i belongs to block j, the number of entries 1
in column j equals the number of members of block j, that is, k. All other
entries of A are zero, so the number of entries 1 in a column equals the sum
of the entries in the column. But each entry of column j of JA equals the
sum of the entries in column j of A. Therefore, (2.12) holds.
The (i, j) entry of AAT is

b
ain aj n. (2.13)
n=1
Now ain ajn equals 1 when both treatments i and j belong to block n, and 0
otherwise. So the sum (2.13) equals the number of blocks that contain both
ti and tj ; this is r when i = j and λ when i
= j. Therefore,
AAT = rI + λ(J − I)
= (r − λ)I + λJ. (2.11)

Conversely, suppose A is a (0, 1) matrix satisfying (2.11) and (2.12). Define

a block design with treatments t1 , t2 , . . . , tv and blocks B1 , B2 , . . . , Bb by ti ∈
Bj if and only if aij = 1.
One sees easily that this is a balanced incomplete block design (provided
that k < v). Consequently, v and b must satisfy the given equations.

It is useful to know the determinant of the matrix AAT . We prove a slightly

more general result, in that we do not assume the parameters r and λ to be
positive integers.

LEMMA 2.8
The determinant of the v × v matrix
M = (r − λ)I + λJ
 Balanced Designs 23

is (r − λ)v−1 [r + (v − 1)λ].

Proof. The matrix M is transformed as follows. First, subtract column 1

from every other column; second, add each of rows 2, 3, . . . , n to row 1.
The process is

   
r λ λ ... λ r λ −r λ −r ... λ− r
 λ r λ ... λ   λ r−λ 0 ... 0 
   
M = λ λ r ... λ  →  λ r − λ ... 0 
   
 . . . ... .   . . . ... . 
λ λ λ ... r λ 0 0 ... r −λ
 
r + (v − 1)λ 0 0 ... 0

 λ r − λ 0 ... 0  
→ 
 λ 0 r − λ ... 0  .
 . . . ... . 
λ 0 0 ... r − λ

These transformations do not change the determinants; since the final matrix
is zero above the diagonal, the determinant equals the product of the diagonal
elements, which is (r − λ)v−1 [r + (v − 1)λ].

THEOREM 2.9 [53]

In a balanced incomplete block design, b ≥ v.

Proof. Let A be the incidence matrix of a (v, b, r, k, λ)-BIBD. Since k < v,

the equation λ(v − 1) = r(k − 1) implies that r > λ. So the determinant of
AAT , which equals (r − λ)v−1 [r + (v − 1)λ], is nonzero. Therefore, AAT has
rank v, and v = rank (AAT ) ≤ rank (A) ≤ min(v, b). So v ≤ b.

Theorem 2.9 is called Fisher’s inequality. The original proof did not involve
matrices, but rather used an ingenious method that takes its inspiration from
statistics. (That proof is reproduced in the next section.)
If A is the incidence matrix of a regular design with parameters v, b, r
and k, then AT is the incidence matrix of a regular design with parameters
b, v, k and r. The design is easy to construct. If the original design had
blocks B1 , B2 , . . . , Bb and treatments t1 , t2 , . . . , tv , the new design has blocks
C1 , C2 , . . . , Cv and treatments u1 , u2 , . . . , ub , and ui belongs to Cj if and only
if tj belongs to Bi . The new design is called the dual of the original.
24 Introduction to Combinatorial Designs

It is clear that the dual of an incomplete design is an incomplete design.

We shall, however, prove that balance is preserved if and only if v = b. We
would need to show that a v × v matrix of zeros and ones that satisfies

AAT = (k − λ)I + λJ (2.14)

JA = kJ (2.15)

[the results of substituting r = k in (2.11) and (2.12)] will also satisfy

AT A = (k − λ)I + λJ (2.16)
JAT = kJ. (2.17)

In fact, we shall prove the following, stronger theorem.

THEOREM 2.10 [122]

If A is a nonsingular matrix of side v, which satisfies one of (2.14), (2.16)
and one of (2.15), (2.17), then it satisfies all four equations, and

k(k − 1) = λ(v − 1).

Proof. All matrices in the proof are v × v. Since A is nonsingular, both AAT
and AT A have nonzero determinant. So either (2.14) or (2.16) implies that
(k − λ)I + λJ has nonzero determinant; from Lemma 2.8, we have

k − λ
= 0, λ(v − 1) + k
= 0.

First suppose (2.14) and (2.15) hold and A is nonsingular. From (2.14) we
deduce that
JAAT = (k − λ)J + λJ 2
and substituting kJ for JA (from (2.15)) and vJ for J 2 yields

kJAT = (k − λ)J + λvJ.

So

kJAT J = k − λ + λv)J 2
KJ(JA)T = (k − λ + λv)J 2
kJ(kJ)T = (k − λ + λv)J 2
k 2 J 2 = (k − λ + λv)J 2

whence k 2 = k − λ + λv and k(k − 1) = λ(v − 1). Also, from

kJAT = (k − λ + λv)J
Exploring the Variety of Random
Documents with Different Content
 THE
BLACK RAT.

This animal is the old genuine English House-Rat, no other being

known in and about houses, for many ages, in this country, and
differs in many particulars from the Norway Rat, last described, for
they do not burrow and run into shores as the others do, but chiefly
lie in the cielings and wainscots in houses, and in out-houses they lie
under the ridge-tiles and behind the rafters, and run along the side-
plates; but their numbers are greatly diminished to what they were
formerly, not many of them being now left, for the Norway Rats
always drive them out, and kill them wherever they can come at
them; as a proof of which I was once exercising my employment at a
gentleman’s house, and when the night came that I appointed to
catch, I set all my traps going as usual, and in the lower part of the
house, in the cellars, I caught the Norway Rats, but in the upper part
of the house I took nothing but the black Rats; I then put them
together into the great cage, to keep them alive till the morning, that
the gentleman might see them, when the Norway Rats killed the
black ones immediately, and devoured them in my presence.
These vermin are not near so bold, nor will feed so freely as the
Norway Rat, and when you are troubled with any of them you must
observe where they use, which you may do very easily, for if you go
into any place and look up at the side-plates, you will perceive they
will be quite black where they run along them, and likewise along the
cross beams; on these cross beams place one of the traps you set for
the Norway Rats, and put some of the same feed in them, but mix
more bread in it than you did for the Norway Rat, and scent the trap
in the same manner, and put some corn in of any sort: set a trap at
each place where they use, and set them all to feed, as you do the
others for the Norway Rat; and when you perceive that they come to
feed boldly, then take them; but this must be done in the night, for
they do not move by day-light, as the other rats will. By this method I
catch them alive, but I have likewise taken them on the side-plates
and beams, where they run, in wires and in snares, so that they swing
off the beams; they are not near so savage as the Norway Rats, for
they seldom kill any chickens or any thing of that kind, unless
extremely hungry and sharp set indeed, but they are sly thieves for
cheese, bacon, or any kind of eatables they can get at and pilfer.
 THE
WATER RAT.

These Rats lie always by the waterside, and feed chiefly on grass and
vegetables, eating neither flesh or corn, but they will devour green
peas, when in season, if there are any growing near the ditches where
they lie; these vermin do not much mischief, except in making holes
in banks and spoiling fences; they make kennels or holes by the side
of ditches, very much resembling those of the otter, for they have one
way out in the water, and another way out by land, in order to escape
from their enemies, which ever way pursued. The method of taking
them is by setting little hutch traps by the side of the ditch or pond
where they lie, with wings made with bushes or raised with mould,
running aslant from the trap as a guide for them to go in, such as
they have in warrens. Feed them with any thing green, such as the
hard part or stalk cut out of a cabbage, or cabbage-leaves, which they
will eat; but do not set your traps going till they feed boldly, and give
them some leaves regularly, as you do the other Rats their proper
feed; tie some of these leaves in the trap, then you will be a judge of
what comes to feed, else one will convey away as much as ten will
eat, which occasions you to conclude there are a great many of them.
You may take them another way; get some small steel traps and put
them in their runs, even with the surface of the ground, and covered
over very nicely with the mould. These vermin are something like the
Norway Rat, but smaller, their noses and tails are shorter, and their
heads rounder, or what is commonly termed muffheaded.
 THE
SQUIRREL.

These little animals are admired for their beauty, and for the sport
they afford in hunting them among the trees; they are about the size
of the stoat or cain, with a tail as large as their body, which they
always turn up when they sit still, to keep their bodies warm; they are
of a reddish colour, except under the belly, which is white; they make
themselves nests or lodging places, commonly called drays, in a very
neat manner, with small sticks, leaves, and moss, in the tops of the
trees, where they hoard up nuts to serve them in the winter; but
besides these, they make a reserve of nuts, acorns, and other things
of a similar nature in retired and bye places, which they know where
to find, in case the other should be taken away, which often happens;
they are chiefly hurtful in destroying wall-fruit, for they will run
along the top of the wall, taking the first choice of the fruit, whether
nectarines, peaches, apricots, &c. and do prodigious mischief; I have
caught them on the wall with a small steel trap, covered nicely, and
in wood-walks I have seen them take the eggs out of birds-nests and
break them, by that means destroying their nests. Their flesh is
reckoned a great dainty by some persons, and is said to be superior
to venison in flavour.
 THE
MOLE.

This animal is in some places, but chiefly in the North of England,

called a Want, and, contrary to most other vermin, lives chiefly under
ground; its skin is of a fine black jet colour, very smooth and soft,
and has short legs, with which they, with incredible swiftness, will
dig themselves into the earth when they apprehend any danger; they
are supposed by some people to be entirely blind, it being in many
places proverbial to say as blind as a mole, but this is a mistake, for
they have eyes as well as other vermin, but very small, appearing
scarcely larger than the head of a common pin, but sufficient to serve
their purposes, their residence being, as was before observed,
generally under the earth.
These animals do great mischief in gardens and grounds, and if
you find they come observe the outsides, for their angle, or run; or, if
there is a path in a field, it is very probable but they have a run across
the path, or at a gate-way, they will frequently have one; these are
what we call the main runs, and about two or three inches under the
earth, and may very easily be found by the heaving up of the mould
or earth, along which they will run ten times in a day. When you have
discovered one of these runs, you must tread in the earth tight, and
when you come that way again, see whether it is as you left it, and if
you perceive the Mole has been along, then set a trap, by which
means I have caught several of them in an afternoon; these being
their main roads out of one part of the ground to the other, for it will
be of little signification to set a trap in any other angles or runs, and
if you should, it may possibly remain a great while before a Mole
comes; for in the spring, when they run near the surface of the earth,
they make a great many different angles in search of the worms, on
which and chaffers their chief food consists.
 Now if they make hills in your fields or gardens, take notice of the
places before mentioned, and set a trap in the following manner:
Take a piece of board half an inch thick, four inches and a half long,
by two and a half wide, then put a small hoop or bow at each end,
with just room for the Mole to go through; then in the centre, at each
side, put two small pegs, in order to keep them in the trap; for
sometimes one that is shy, when he finds the peg before him that
springs the trap, will turn out at the side and spring the trap, and not
be taken: it is necessary therefore to use these small pegs, which will
keep them in the straight road, placed as before directed. In the next
place, get two strong horse-hairs, or pieces of small wire, then in the
centre or middle of the bow, at each end, make a hole to put the hair
or wire through double, then open the hair or wire just to fit, and lie
close inside of the bow, like a noose, get some fine mould, make it
moist like paste, and work some of it with your finger and thumb all
round in the inside of the bow, so that the horse-hair or wire may not
be perceivable. Through the hole in the centre of the trap let a short
bit of string come. Put a forked peg tight in the hole, that may keep
the string from slipping through, till the Mole, by going through,
pushes it out; then the string slips up. When you have thus prepared
the trap, then open one of the runs, exactly the length of the trap,
and put it down in the run quite level, and make it all smooth, that
there may be no light discovered. Then take three good strong
hooked pegs, two on one side, and one on the other, and stick them
down tight, then take a stiff stick, about four feet long, stick one end
in the ground tight, bring the other end down to the trap and hitch in
a loop, that comes from the hair or wire, and then it is set, and when
the Mole comes he pushes out this little peg, then the string draws
out, the bow-string flies up, and the Mole is caught; in the spring
time, when you catch a she Mole, rub her back part about the bows
and the inside of the trap, by which method I have caught eight he
Moles successively; observe, when you have caught all that you
perceive to move, you need only look round the outside of your
fields, and keep some traps constantly going there, and they will lay
hold of them as they come in and out; another expedient you may
make use of, by putting some dead Moles in the runs where they
come, which will prevent their coming, and keep your ground quiet
and free from these trouble-vermin.
 THE
HOUSE-MOUSE.

These are a nasty little vermin, well known, and are very nauseous,
for wherever they come, whether in the pantry or larder, you may
easily discover it by the disagreeable smell they leave behind them.
They are very troublesome in dairies, where they will skim the milk,
and will get into bird-cages after the seed, and kill the birds; and are
mischievous in other instances, of which no mention needs here to
be made. Many persons would doubtless be glad to be informed of a
method to kill these vermin, which is done very easily; and to
accomplish it nothing is so good as nux vomica; take a quart of the
feed above prescribed for the rat, before there is any bread mixed
with it, then take four figs of nux vomica, and rasp them very fine, or
else they will pick the feed from it, upon account of the bitter taste;
rub it well together, and it will be their certain bane. My method of
laying it is after this manner; if they come into the larder at night,
put the bread into a pan, and take the other eatables out of the way;
then lay some of the feed on a piece of paper, and do this in two or
three other places, which will not be attended with any danger,
however, for fear of any accident, what is not eaten take away in the
morning, and at night lay it down again, and so keep on as long as
any come; in the dairy do the same, or in any other place where they
appear; I have found several lay dead on the paper at one time, but
remember always to lay the ingredients on paper, then you may take
away what is left.
 THE
FIELD MOUSE.

This is a larger Mouse than the former, being the largest of the
Mouse kind, with a reddish back and a good deal of white under the
belly. They abound in gardens, and frequently swarm in prodigious
numbers in the fields: they will eat either corn or flesh just as it
comes in their way. I have seen five or six of them by moon-light
attack a large toad and kill it, which they do often; and when I have
been trailing for a fox in the night, I have observed, by turning the
light round, several of these Field-Mice following the trail. They are
more difficult to destroy than the house-mice, on account that they
are shy in taking any bait; however, sometimes, when I have been
going to catch a fox, I have been sometimes obliged to put some of
the feed (so often before mentioned) in the shrape, in order to kill
the Mice, before I could get any to lie for the fox. In winter they will
likewise come into houses at the sink-holes, and get into the pantries,
larders and dairies, being driven from the fields by extremity of
hunger, and prove very troublesome and pernicious; and in spring
they will get into the gardens, and eat whole rows or drills of peas,
after they are set: I can give no better method to destroy these
vermin, than to lay some of the above feed for them, as you do for the
other, on a tile, and shelter it with some pieces of old boards that it
may not get wet: lay it out at night and take it in again in the
morning, and all that eat of it will certainly die.
 THE
RED SHREW-MOUSE.

This species of the Mouse-kind has a nose like a mole, and is in

general a very harmless, inoffensive little animal, in regard to its
preying on any thing: it is much smaller than the house-mouse. The
chief mischief done by these vermin is, that in harvest-time they will
come home with the corn, remain there and breed. I have known
great numbers taken out of a rick; and I once remember half a bushel
of these and other Mice caught at one time out of a rick or stand,
which is often owing to the farmer’s own negligence and inattention;
for you may often see a rick-stand made a nursery, occasioned by
putting under it harrows, ploughs, hurdles and props: what avails it
then to have a stand to fill the ricks in this manner, and assist the
vermin up? You cannot indeed prevent their coming home with the
corn, or when rats hop up and down the stand, which they do very
easily: the best method, therefore, would be to nail some tin to the
props, which would keep the Mice from getting up them; and when
once you find they have got to a corn-rick, the best way is to take it in
as soon as you can, for these Mice, in one particular, are worse than
the rats among the corn, and do a great deal more mischief; for this
reason, that they will live a great while without water, only by licking
the ends of the straws, while any moisture remains therein, and all
the while they remain in the ricks they do infinite prejudice. If it is
not convenient to remove the rick, observe the following method;
take a quarter of a pound of the best nux vomica, put it into an old
saucepan, with three quarts of water, boil it till it comes to two
quarts, and put two pounds of treacle to it, in order to overcome the
bitter taste of the nux vomica; then take some small earthen pans,
into which pour some of this mixture, and set the pans in different
places under the eaves of the ricks; the Mice being in want of water,
will greedily drink of the mixture, which will kill them; and this
method you must continue till they are destroyed.
Before I conclude this account, I cannot help taking notice of a
gross blunder, which a certain writer has made in his description of
this little vermin, who asserts that it is as large as a rat, of the colour
of a weasel, and very mischievous to cattle, and that it will get upon a
beast’s back, and make it lame in the chine; and in consequence of
the bite, the beast will swell to the heart and die; all which particulars
are entirely erroneous; for its size is less than that of the common
house-mouse, its colour like the squirrel; and as to the miraculous
affair of its killing any beast or cattle, it is of a piece with the rest. On
the contrary, I am convinced from long observation, that it feeds on
roots, vegetables, and corn, and will not touch flesh on any account.
 THE
BLACK SHREW-MOUSE.

This animal has a nose like a mole, and much resembles the red
shrew-mouse, last described, except in colour; they are very harmless
little creatures, for they seldom come to barns or ricks, their
residence being amongst the grass, by the sides of ditches which they
feed on. You may often hear them as you are walking along,
squeaking by the path’s side, and making a noise something like the
grasshopper; but they do not the least mischief that I know of; I
thought proper however to mention them, that the reader might
know that such a little animal really exists.
 THE
GRASS MOUSE.

This Mouse is rather larger than the common Mouse, with a muff-
head like a water rat, and a short club tail, looking as if the top was
cut off, and has a great deal of the nature and disposition of that
animal, for it eats no corn at all, but lives on grass and vegetables;
being of a very harmless nature, it is a prey to other vermin, as the
water rat is, (such as stoats and weasels) which are the worst
enemies it has.
 THE
DORMOUSE.

This is a small red Mouse, with a tail like a squirrel, and is the most
harmless and inoffensive of all the mouse-tribe. It is a very beautiful
tame little animal, and is kept by many persons in small boxes or
cages, and fed with crumbs of bread, &c. They have one singular and
peculiar quality, which is that of sleeping for several weeks together,
without receiving the least sustenance, and appearing all the while as
if dead. You may oftentimes find them in dry banks, or in wood, laid
up very warm in a nest, in the ground, made with oak-leaves, and
always asleep when you discover them.
 THE
BAT.

This little animal is half mouse, half bird, having wings something
resembling leather, and a head very much like that of the mouse,
except that their nose is flatter, and their mouth wider; they keep
close in holes all the day and never fly abroad till the evening, and
that in the summer time, for which reason they are called in many
places, the winged mouse; they live chiefly on flies and other small
insects, and dwell in holes under the roofs of houses, churches,
barns, old ruinated buildings, and in the holes of hollow trees, where
in winter time I have sometimes found several scores of them closely
adhering together, in large clusters, and in a torpid state, in which
condition they doubtless remain till vivified again by the enlivening
warmth of the sun, on the approach of the summer. I have been the
more particular in my account and description of the different kinds
of Mice, as perhaps many of my readers might be unacquainted, that
there are so many distinct species of the mouse-kind.
 THE
LARGE BLACK EAGLE
BUZZARD.

Having finished my account of the four-footed vermin, I now come to

describe the winged-tribe of vermin, inhabitants of the air, which,
notwithstanding, often descend to earth, and do much mischief; I
shall begin with the Black Eagle-Buzzard, which is one of the largest
and fiercest birds of the hawk kind, that breeds in England.
This bird frequents parks and warrens, and often catches leverets,
young rabbits and pheasants, or any thing else that moves, as soon as
the morning light approaches, for they are a very early bird; in winter
time I have caught them in a steel trap, the same as is set in a cube
for a dog; bait it with the guts of rabbits or fowls, or some pieces of
rabbits that you find dead by accident; in spring I have caught them
in steel traps, with the skin of a young rabbit stuffed and tied to the
bridge, but after the young rabbits begin to be able to run about, they
sometimes are shy and will not strike at a rabbit, unless they see him
run; the steel traps that you set for this vermin should strike seven or
eight inches high, in order to clear the bait as it ought to be,
otherwise the jaws might only catch the bait, and miss the vermin; I
have sometimes caught them with a rat in the following manner: set
two traps in the manner as you see in plate VI. fig. 1. head to head,
covered very nicely with moss, then round the circle as in the same
plate, put some small pieces of fern, or a bush or two, quite low, that
the bait may be visible and open; put the bait, whatever it may be, in
the middle, between the two traps, then he must of necessity go over
one of them, in order to get the bait, and by this means I have often
caught him. Sometimes you may observe him to haunt a place in a
warren or park for a week or fortnight together, then set your traps
in the same method as before laid down, and you will be almost
certain of taking him. Let your bait be a dead rat, if you can procure
one.
 THE
GREY BOB-TAILED
BUZZARD,
OR
PUTTOCK.

This bird is much of the same nature as the former, in haunting

parks and warrens, for the destruction of rabbits and pheasants,
going in search of them when the morning light appears. In this
respect, however, it differs from the Eagle-Buzzard, that it is smaller.
You may catch them with the steel trap above described, baited with
the entrails of fowls or rabbits, or with some of the pieces of rabbits
which may have been accidentally killed. They are to be caught
easiest in winter time, for then I have known them haunt one place
for a month together, in a warren, where you may easily catch them,
as at that season they are short and destitute of food.
 THE
LARGE FORKED TAIL
KITE.

This is the largest and heaviest bird of the hawk kind in England, but
not near so fierce as the Eagle-Buzzard, being rather of a sluggish
and indolent disposition, not caring much to hunt after prey; but
when the other kites and hawks have killed any birds, they come
upon them, and beat them away, and then devour the birds
themselves; they chiefly dwell in woods and desart places, and
frequent the sides of rivers and brooks, being fond of fish, and often
eat the tails of the fish, which the otters have left behind them. You
may catch them by setting two traps, in the same manner as you do
for the buzzard; bait them with a piece of fish, if you have it, or with a
rat, or the guts of fowls or rabbits, and when once you discover the
places they haunt, you may catch them with the greatest facility.
 THE
LARGE BLUE SHERARD
KITE.

This kind of Kite frequents forests, heaths, and other lonesome

places, but especially bogs and marshy grounds, where they destroy
great numbers of snipes, to which they are the worst enemy of any
bird of prey of the hawk kind; they beat all over the bog with the
greatest regularity and exactness, till they find them lying: for the
nature of the snipe is, if they perceive their enemy in the pursuit,
they lie as close to the ground as possible, when the Kite pounces
upon them directly and takes them: I have observed one of these
vermin beating at one end of a bog or marshy piece of ground; and at
the other end I have stooped down till he has come up close to the
place where I was, he looking stedfastly down for his prey, regardless
of any thing else; and when he came near enough I have shot him;
sometimes they will have a beat cross a warren, where I have caught
them in traps, baited with what is termed a stall, which is a young
rabbit-skin stuffed.
These vermin are very remarkable for one particular circumstance,
for at any time when you observe any of them come along in the
morning, you will be certain of seeing them return the same way
back again in the afternoon, and three or four succeeding mornings
they will have the same beat, if they are not frightened or disturbed;
whenever, therefore, you see one go along in the morning, get a trap
set ready against his return, baited with a rabbit’s skin stuffed and
put on the bridge of the trap, as before directed, but remember to
cover the trap nicely with moss; if you set it in a green sward, then
cover it with grass, and in this manner I have often taken them; I
have likewise caught them by means of two steel traps, set in the
same manner as for the buzzard; but always mind when you set traps
in this manner, to fasten down the bait between them with a peg,
otherwise they will frequently take it away, and not strike the trap;
but when the bait is fixed they cannot get it away, but are constrained
to stop, when by hopping and picking about, they spring the trap and
are caught.
 THE
BLUE SHERARD
HAWK.

This bird, in many respects, resembles the blue kite last described, in
haunting heaths, forests, and lonely places; especially quags and
marshy grounds, doing much mischief, particularly among the
snipes. You will always be sure of seeing him return home by the
same path, in the afternoon, by which he pursued his flight in the
morning, but he is not so large, and preys in much the same manner,
but does not, however, frequent warrens so often as the other; you
may catch them by the same method as the Sherard Kite, with a
rabbit’s skin stuffed and put on the bridge of the trap, remembering
to cover the trap nicely with moss or grass, observing the directions
before given.
 THE LARGE
BROWN, WHITE ARSE,
RING-TAILED
HAWK,

Is the most pernicious and mischievous Hawk that flies; but

especially in destroying that admired game the partridge, which it
takes, not so much by swift flying, as by the following manner or
stratagem: viz. when they find a covey and spring them, they fly after
them as fast as they are able, and mark where they pitch down again,
and then spring them again; and thus keep following them till they
catch their prey; for partridges, when they are driven hard for three
flights, become so tired and languid that they are easily taken; I have
sometimes observed one of these Hawks in full pursuit after a poor
partridge, almost spent, when the Hawk has suddenly stopped, and
alighted on a tree in the next hedge, to which I imagined the
partridge had taken, as they frequently do, when they are almost
tired or down. On this I got near enough the Hawk, being intent and
earnest after his prey, shot him dead, and on examining, to my great
surprize, found that the partridge, being driven by the Hawk, had
taken refuge in a hutchtrap, that stood on the other side of the hedge,
where the Hawk had marked it, and sat watching for its coming out
again; these vermin will keep their walk or beat for some time very
regularly. I have caught them in a trap, baited with a rabbit-skin
stuffed, as before observed.
 THE
SMALL RING-TAILED
HAWK

Is the largest of the Sparrow-Hawk kind, and is a very fierce and

pernicious bird, destroying young ducks and chickens, in yards, &c.
about farm-houses, in the fields, they kill black-birds, thrushes, and
in the winter-season fieldfares and other small birds; I have driven
one from a blackbird, and immediately set down a steel trap, covered
quite smoothly, placing the same bird on the bridge of the trap, and
have taken the Hawk, which returned soon after for its prey. They
will likewise frequently come to the chicken-gardens and
pheasantries, when you must set a steel trap, baited with a dead
chicken or any other of the birds before mentioned; which they
delight to feed on, and in all probability you will lay hold of them,
and if you find they have been made shy, by reason of the trap
striking and not catching them, which will sometimes, though very
rarely, happen, you may destroy them after the following manner:
take a live chicken, tie it to a small stake drove in the ground, by one
of its legs, laying some bread or corn for it to feed on; then get out of
sight, with your gun, and the hawk will come and strike at the
chicken, when you may easily shoot him, and this method I have
often practised with good success.
 THE
SMALL SPARROW
HAWK.

Birds which fly wild in the fields, are the chief prey of this little
Hawk. It very seldom destroys chickens, or any other kind of poultry,
unless driven by the greatest extremity of hunger; I have driven them
from a bird, and caught them in a steel trap as I have the others,
having been allowed so much per head for all the winged vermin, as
well as others that I destroyed, which made me very anxious and
assiduous in trying and inventing various methods to destroy them.
 THE
SMALL BLACK-HEADED
SPARROW-HAWK.

I never caught this Hawk in a trap, but have shot several of them; it
is a very beautiful bird, and has a black head, with pointed wings,
and is very quick and sharp after its prey, which are small birds of
every kind. It is more timid than the last mentioned Hawk, and is
seldom seen, there being but few of them in England. On which
account we need not dwell minutely upon him.
 THE
WIND-HOVER
HAWK.

You may often see this little brown Hawk, the smallest of the species,
hovering in the air, then they steadily look down on the ground, and
if they chance to see a mouse, they will dart with incredible swiftness,
like an arrow out of a bow, and catch it; sometimes they will catch a
mole, and I have frequently taken this kind of Hawk in a steel trap,
by laying one of these animals on the bridge; and with this I conclude
my account of the birds of the Hawk kind. The three last mentioned
indeed do not much damage or mischief, but if I had not mentioned
them the catalogue would have been incomplete; besides, I flatter
myself it will be acceptable to my readers, some of whom might,
possibly, not be acquainted that there were so many different kinds
of Hawks which breed in England.
 THE
BROWN
WOOD-OWL.

Few persons perhaps imagine this to be a pernicious animal, yet in

my opinion, it is the very worst winged vermin that exists; they seize
their prey with great fierceness, and break it up exactly as the cat
does, by beginning at the fore part. It is very happy for the farmer,
and others who keep poultry, that the ducks and chickens are gone to
roost before the Owl comes to prey in the night, otherwise there
would be few, if any of them left; however I now proceed to make
good my assertion, and describe the infinite mischief done by this
sort of Owls. When the evening approaches, many animals such as
leverets, young rabbits, young pheasants, young partridges, who had
lain quiet in the day-time, come out to feed in the stillness and
silence of the evening, when there is nothing else to disturb them, the
Owl then comes, when they are exposed, and takes away what he
pleases, and if they happen to have young ones at the same time, it is
impossible to conceive what numbers will be taken away in a night’s
time to satisfy his voracious appetite; and this I can affirm from my
own observation, as I used to make it my frequent practice to go out
and hunt the Owls all round the outsides of the warren. The method
was this, two of us used to go with a gun, one on one side of the
hedge, and the other on the other side; a third person accompanying
us with a long pole or stick, who therewith beat the hollow and ivy
trees, when the Owls would fly out, and we have shot sometimes a
dozen in a morning. But one morning particularly I found an Owl’s
nest, and shot the hen; and, on examining the nest, I found two
young ones, with several pieces of young rabbits, leverets, &c.
therein, but we could not meet with the old cock; on which I took out
one of the young ones, leaving the other in order to entice him home
to the nest again; the next morning I repaired to the place, when he
flew out, and I shot him dead, and took away the other young one
which had been left behind. But, what is remarkable, the old Owl had
carried no less than three young rabbits to the single young one
remaining, in one night’s time, which is an evident proof what havoc
they make among the game. Another particular of these vermin
deserves mentioning, which is, that they seldom eat more than the
fore-quarters of any animal they kill, as the cat does, and always
leaves the hind quarters untouched, which are found in the nest. I
have likewise caught them in moon-light evenings in steel traps,
baited with a rabbit-skin stuffed, and have often killed many of them
with a gun, by keeping close under the hedge, and imitating the
squeaking of a mouse, when the Owl has come directly to the place,
and I have shot him dead on the spot.
 THE
LIGHT COLOURED BARN,
OR,
SCREECH OWL.

This kind of Owl is a very beautiful bird, of a palish brown colour, a

little spotted and white on the breast, doing more good than harm,
for it seldom preys on any thing but mice, which it will watch for in
the barns like a cat, and destroy great numbers of that little vermin.
On this account these Owls are encouraged by the farmers. If you
should see one of them, mimick the squeaking of the mouse, and he
will come to you directly, and you may easily shoot him, as was
before observed of the Wood-Owl.
 THE
HORNED OWL

Has its name from the feathers which stand up on each side of his
head, resembling two horns, and making a very beautiful
appearance. There are many people, who do not believe there are any
Owls of this species bred amongst us; the large brown horned Owl,
(which is bigger than our largest birds of prey, whether kite or
buzzard) is indeed from foreign parts, and is esteemed a very great
curiosity, but never breeds here, which this little Horned Owl does,
and much resembles the large one, only in miniature. I have shot
several of them on Windsor-Forest: they do no mischief that I could
ever learn, however I thought it proper to mention them, that the
reader may be assured there is such a bird existing.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

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

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com