Download the full version of the textbook now at textbookfull.

com

Probabilistic Deep Learning with Python 1st

Edition Oliver Duerr

https://textbookfull.com/product/probabilistic-
deep-learning-with-python-1st-edition-oliver-
duerr/

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

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

Deep Learning with Python 1st Edition François Chollet

https://textbookfull.com/product/deep-learning-with-python-1st-
edition-francois-chollet/

textbookfull.com

Deep Learning with Python MEAP 2nd Edition Francois

Chollet

https://textbookfull.com/product/deep-learning-with-python-meap-2nd-
edition-francois-chollet/

textbookfull.com

Deep Learning with Python Learn Best Practices of Deep

Learning Models with PyTorch 2nd Edition Nikhil Ketkar
Jojo Moolayil
https://textbookfull.com/product/deep-learning-with-python-learn-best-
practices-of-deep-learning-models-with-pytorch-2nd-edition-nikhil-
ketkar-jojo-moolayil/
textbookfull.com

Popular Woodworking s Arts Crafts Furniture Projects 42

Designs for Every Room in Your Home 2 ed 2nd Edition Lang
Robert W. (Ed.)
https://textbookfull.com/product/popular-woodworking-s-arts-crafts-
furniture-projects-42-designs-for-every-room-in-your-home-2-ed-2nd-
edition-lang-robert-w-ed/
textbookfull.com
Economics in America : An Immigrant Economist Explores the
Land of Inequality 1st Edition Angus Deaton

https://textbookfull.com/product/economics-in-america-an-immigrant-
economist-explores-the-land-of-inequality-1st-edition-angus-deaton/

textbookfull.com

The social and political philosophy of Mary Wollstonecraft

1st Edition Berges

https://textbookfull.com/product/the-social-and-political-philosophy-
of-mary-wollstonecraft-1st-edition-berges/

textbookfull.com

The Oedipus Plays of Sophocles: Philosophical Perspectives

Paul Woodruff

https://textbookfull.com/product/the-oedipus-plays-of-sophocles-
philosophical-perspectives-paul-woodruff/

textbookfull.com

Yours to Take Hunter s Prey 1 1st Edition Grace Cooper

[Cooper

https://textbookfull.com/product/yours-to-take-hunter-s-prey-1-1st-
edition-grace-cooper-cooper/

textbookfull.com

Drug Abuse Sourcebook Laura Larsen

https://textbookfull.com/product/drug-abuse-sourcebook-laura-larsen/

textbookfull.com
An American Quilt May

https://textbookfull.com/product/an-american-quilt-may/

textbookfull.com
With Python, Keras,
and TensorFlow Probability

Oliver Dürr
Beate Sick
with Elvis Murina

MANNING
 RNN1
CNN2 ZIP GAUS1 GAUS2
CNN1
FCNN1

Network Shelf Prob. Distribution Shelf

DATA

Like
Max

FCNN2 POIS
M

xL
a

ik
e

Data modeling with probabilistic DL. The network determines the parameters of a
probability distribution. Fit the model using the MaxLike principle. In the example shown,
the outcome is count data. Here it’s modeled by a Poisson distribution, where NN is used
to control its rate parameter λ (see the chosen last plate with one output node).
Probabilistic Deep Learning
 Probabilistic
Deep Learning
WITH PYTHON, KERAS, AND
TENSORFLOW PROBABILITY

OLIVER DÜRR
BEATE SICK
WITH ELVIS MURINA

MANNING
SHELTER ISLAND
For online information and ordering of this and other Manning books, please visit
www.manning.com. The publisher offers discounts on this book when ordered in quantity.
For more information, please contact
Special Sales Department
Manning Publications Co.
20 Baldwin Road
PO Box 761
Shelter Island, NY 11964
Email: [email protected]

©2020 by Manning Publications Co. All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in

any form or by means electronic, mechanical, photocopying, or otherwise, without prior written
permission of the publisher.

Many of the designations used by manufacturers and sellers to distinguish their products are
claimed as trademarks. Where those designations appear in the book, and Manning Publications
was aware of a trademark claim, the designations have been printed in initial caps or all caps.

Recognizing the importance of preserving what has been written, it is Manning’s policy to have
the books we publish printed on acid-free paper, and we exert our best efforts to that end.
Recognizing also our responsibility to conserve the resources of our planet, Manning books
are printed on paper that is at least 15 percent recycled and processed without the use of
elemental chlorine.

Development editor: Marina Michaels

Technical development editors: Michiel Trimpe and
Arthur Zubarev
Manning Publications Co. Review editor: Aleksandar Dragosavljević
20 Baldwin Road Production editor: Deirdre S. Hiam
PO Box 761 Copy editor: Frances Buran
Shelter Island, NY 11964 Proofreader: Keri Hales
Technical proofreader: Al Krinker
Typesetter: Dennis Dalinnik
Cover designer: Marija Tudor

ISBN: 9781617296079
Printed in the United States of America
 brief contents
PART 1 BASICS OF DEEP LEARNING .............................................1
1 ■ Introduction to probabilistic deep learning 3
2 ■ Neural network architectures 25
3 ■ Principles of curve fitting 62

PART 2 MAXIMUM LIKELIHOOD APPROACHES FOR

PROBABILISTIC DL MODELS .........................................91
4 ■ Building loss functions with the likelihood approach 93
5 ■ Probabilistic deep learning models with TensorFlow
Probability 128
6 ■ Probabilistic deep learning models in the wild 157

PART 3 BAYESIAN APPROACHES FOR PROBABILISTIC

DL MODELS. ..............................................................195
7 ■ Bayesian learning 197
8 ■ Bayesian neural networks 229

v
 contents
preface xi
acknowledgments xii
about this book xiv
about the authors xvii
about the cover illustration xviii

PART 1 BASICS OF DEEP LEARNING ...................................1

1 Introduction to probabilistic deep learning

1.1 A first look at probabilistic models 4
3

1.2 A first brief look at deep learning (DL) 6

A success story 8
1.3 Classification 8
Traditional approach to image classification 9 Deep learning
■

approach to image classification 12 Non-probabilistic

■

classification 14 Probabilistic classification 14

■

Bayesian probabilistic classification 16

1.4 Curve fitting 16
Non-probabilistic curve fitting 17 Probabilistic curve
■

fitting 18 Bayesian probabilistic curve fitting 20

■

vii
viii CONTENTS

1.5 When to use and when not to use DL? 21

When not to use DL 21 When to use DL 22
■ ■
When to use
and when not to use probabilistic models? 22
1.6 What you’ll learn in this book 23

2 Neural network architectures

2.1
25
Fully connected neural networks (fcNNs) 26
The biology that inspired the design of artificial NNs 26
Getting started with implementing an NN 28 Using a fully ■

connected NN (fcNN) to classify images 38

2.2 Convolutional NNs for image-like data 44
Main ideas in a CNN architecture 44 A minimal CNN for edge
■

lovers 47 Biological inspiration for a CNN architecture 50

■

Building and understanding a CNN 52

2.3 One-dimensional CNNs for ordered data 56
Format of time-ordered data 57 What’s special about ordered
■

data? 58 Architectures for time-ordered data 59

■

3 Principles of curve fitting

3.1
62
“Hello world” in curve fitting 63
Fitting a linear regression model based on a loss function 65
3.2 Gradient descent method 69
Loss with one free model parameter 69 ■ Loss with two free
model parameters 73
3.3 Special DL sauce 78
Mini-batch gradient descent 78 Using SGD variants to speed up
■

the learning 79 Automatic differentiation 79

■

3.4 Backpropagation in DL frameworks 80

Static graph frameworks 81 ■ Dynamic graph frameworks 88

PART 2 MAXIMUM LIKELIHOOD APPROACHES FOR

PROBABILISTIC DL MODELS ...............................91

4 Building loss functions with the likelihood approach

4.1 Introduction to the MaxLike principle: The mother
of all loss functions 94
93
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
 CONTENTS ix

4.2 Deriving a loss function for a classification problem 99

Binary classification problem 99 Classification problems with
■

more than two classes 105 Relationship between NLL, cross

■

entropy, and Kullback-Leibler divergence 109

4.3 Deriving a loss function for regression problems 111
Using a NN without hidden layers and one output neuron for
modeling a linear relationship between input and output 111
Using a NN with hidden layers to model non-linear relationships
between input and output 119 Using an NN with additional
■

output for regression tasks with nonconstant variance 121

5 Probabilistic deep learning models with TensorFlow

Probability 128
5.1 Evaluating and comparing different probabilistic
prediction models 130
5.2 Introducing TensorFlow Probability (TFP) 132
5.3 Modeling continuous data with TFP 135
Fitting and evaluating a linear regression model with constant
variance 136 Fitting and evaluating a linear regression model
■

with a nonconstant standard deviation 140

5.4 Modeling count data with TensorFlow Probability 145
The Poisson distribution for count data 148 Extending the
■

Poisson distribution to a zero-inflated Poisson (ZIP)

distribution 153

6 Probabilistic deep learning models in the wild

6.1 Flexible probability distributions in state-of-the-art
157

DL models 159
Multinomial distribution as a flexible distribution 160
Making sense of discretized logistic mixture 162
6.2 Case study: Bavarian roadkills 165
6.3 Go with the flow: Introduction to normalizing
flows (NFs) 166
The principle idea of NFs 168 The change of variable technique
■

for probabilities 170 Fitting an NF to data 175 Going

■ ■

deeper by chaining flows 177 Transformation between higher

■

dimensional spaces* 181 Using networks to control flows 183

■

Fun with flows: Sampling faces 188

x CONTENTS

PART 3 BAYESIAN APPROACHES FOR PROBABILISTIC

DL MODELS ....................................................195

7 Bayesian learning
7.1
197
What’s wrong with non-Bayesian DL:
The elephant in the room 198
7.2 The first encounter with a Bayesian approach 201
Bayesian model: The hacker’s way 202 ■
What did we
just do? 206
7.3 The Bayesian approach for probabilistic models 207
Training and prediction with a Bayesian model 208 A coin toss ■

as a Hello World example for Bayesian models 213 Revisiting ■

the Bayesian linear regression model 224

8 Bayesian neural networks

8.1
229
Bayesian neural networks (BNNs) 230
8.2 Variational inference (VI) as an approximative
Bayes approach 232
Looking under the hood of VI* 233 ■ Applying VI to the toy
problem* 238
8.3 Variational inference with TensorFlow Probability 243
8.4 MC dropout as an approximate Bayes approach 245
Classical dropout used during training 246 ■
MC dropout used
during train and test times 249
8.5 Case studies 252
Regression case study on extrapolation 252 ■ Classification case
study with novel classes 256

Glossary of terms and abbreviations 264

index 269
 preface
Thank you for buying our book. We hope that it provides you with a look under the
hood of deep learning (DL) and gives you some inspirations on how to use probabilis-
tic DL methods for your work.
All three of us, the authors, have a background in statistics. We started our journey
in DL together in 2014. We got so excited about it that DL is still in the center of our
professional lives. DL has a broad range of applications, but we are especially fasci-
nated by the power of combining DL models with probabilistic approaches as used in
statistics. In our experience, a deep understanding of the potential of probabilistic DL
requires both insight into the underlying methods and practical experience. There-
fore, we tried to find a good balance of both ingredients in this book.
In this book, we aimed to give some clear ideas and examples of applications
before discussing the methods involved. You also have the chance to make practical
use of all discussed methods by working with the accompanying Jupyter notebooks.
We hope you learn as much by reading this book as we learned while writing it. Have
fun and stay curious!

xi
 acknowledgments
We want to thank all the people who helped us in writing this book. A special thanks
go out to our development editor, Marina Michaels, who managed to teach a bunch of
Swiss and Germans how to write sentences shorter than a few hundred words. Without
her, you would have no fun deciphering the text. Also, many thanks to our copyeditor,
Frances Buran, who spotted uncountable errors and inconsistencies in the text (and
also in the formulas, kudos!). We also got much support on the technical side from Al
Krinkler and Hefin Rhys to make the text and code in the notebooks more consistent
and easier to understand. Also, thank you to our project editor, Deirdre Hiam; our
proofreader, Keri Hales; and our review editor, Aleksandar Dragosavljević. We would
also like to thank the reviewers, which at various stages of the book helped with their
very valuable feedback: Bartek Krzyszycha, Brynjar Smári Bjarnason, David Jacobs,
Diego Casella, Francisco José Lacueva Pérez, Gary Bake, Guillaume Alleon, Howard
Bandy, Jon Machtynger, Kim Falk Jorgensen, Kumar Kandasami, Raphael Yan, Richard
Vaughan, Richard Ward, and Zalán Somogyváry.
Finally, we would also like to thank Richard Sheppard for the many excellent
graphics and drawings making the book less dry and friendlier.
I, Oliver, would like to thank my partner Lena Obendiek for her patience as I
worked on the book for many long hours. I also thank my friends from the “Tatort”
viewing club for providing food and company each Sunday at 8:15 pm and for keep-
ing me from going crazy while writing this book.
I, Beate, want to thank my friends, not so much for helping me to write the book,
but for sharing with me a good time beyond the computer screen—first of all my

xii
 ACKNOWLEDGMENTS xiii

partner Michael, but also the infamous Limmat BBQ group and my friends and family
outside of Zurich who still spend leisure time with me despite the Rösti-Graben, the
country border to the big canton, or even the big pond in between.
I, Elvis, want to thank everyone who supported me during the exciting time of writ-
ing this book, not only professionally, but also privately during a good glass of wine or
a game of football.
We, the Tensor Chiefs, are happy that we made it together to the end of this book.
We look forward to new scientific journeys, but also to less stressful times where we not
only meet for work, but also for fun.
 about this book
In this book, we hope to bring the probabilistic principles underpinning deep learn-
ing (DL) to a broader audience. In the end (almost), all neural networks (NNs) in DL
are probabilistic models.
There are two powerful probabilistic principles: maximum likelihood and Bayes.
Maximum likelihood (fondly referred to as MaxLike) governs all traditional DL.
Understanding networks as probabilistic models trained with the maximum likeli-
hood principle helps you to boost the performance of your networks (as Google did
when going from WaveNet to WaveNet++) or to generate astounding applications
(like OpenAI did with Glow, a net that generates realistic looking faces). Bayesian
methods come into play in situations where networks need to say, “I’m not sure.”
(Strangely, traditional NNs cannot do this.) The subtitle for the book, “with Python,
Keras, and TensorFlow Probability,” reflects the fact that you really should get your
hands dirty and do some coding.

Who should read this book

This book is written for people who like to understand the underlying probabilistic
principles of DL. Ideally, you should have some experience with DL or machine learn-
ing (ML) and should not be too afraid of a bit of math and Python code. We did not
spare the math and always included examples in code. We believe math goes better
with code.

xiv
 ABOUT THIS BOOK xv

How this book is organized: A roadmap

The book has three parts that cover eight chapters. Part 1 explains traditional deep
learning (DL) architectures and how the training of neural networks (NNs) is done
technically.
■ Chapter 1—Sets the stage and introduces you to probabilistic DL.
■ Chapter 2—Talks about network architectures. We cover fully connected neural
networks (fcNNs), which are kind of all-purpose networks, and convolutional
neural networks (CNNs), which are ideal for images.
■ Chapter 3—Shows you how NNs manage to fit millions of parameters. We keep
it easy and show gradient descent and backpropagation on the simplest network
one can think of—linear regression.
Part 2 focuses on using NNs as probabilistic models. In contrast to part 3, we discuss
maximum likelihood approaches. These are behind all traditional DL.
■ Chapter 4—Explores maximum likelihood (MaxLike), the underlying princi-
ple of ML and DL. We start by applying this principle to classification and
(simple regression problems).
■ Chapter 5—Introduces TensorFlow Probability (TFP), a framework to build
deep probabilistic models. We use it for not-so-simple regression problems like
count data.
■ Chapter 6—Begins with more complex regression models. At the end, we
explain how you can use probabilistic models to master complex distributions
like describing images of human faces.
Part 3 introduces Bayesian NNs. Bayesian NNs allow you to handle uncertainty.
■ Chapter 7—Motivates the need for Bayesian DL and explains its principles. We
again look at the simple example of linear regression to explain the Bayesian
principle.
■ Chapter 8—Shows you how to build Bayesian NNs. Here we cover two approaches
called MC (Monte Carlo) dropout and variational inference.
If you already have experience with DL, you can skip the first part. Also, the second
part of chapter 6 (starting with section 6.3) describes normalizing flows. You do not
need to know these to understand the material in part 3. Section 6.3.5 is a bit heavy on
math, so if this is not your cup of tea, you can skip it. The same holds true for sec-
tions 8.2.1 and 8.2.2.

About the code

This book contains many examples of source code both in numbered listings and in
line with normal text. In both cases, source code is formatted in a fixed-width font,
like this to separate it from ordinary text.
The code samples are taken from Jupyter notebooks. These notebooks include
additional explanations and most include little exercises you should do for a better
xvi ABOUT THIS BOOK

understanding of the concepts introduced in this book. You can find all the code in
this directory in GitHub: https://github.com/tensorchiefs/dl_book/. A good place to
start is in the directory https://tensorchiefs.github.io/dl_book/, where you’ll find
links to the notebooks. The notebooks are numbered according to the chapters. So,
for example, nb_ch08_02 is the second notebook in chapter 8.
All the examples in this book, except nb_06_05, are tested with the TensorFlow v2.1
and TensorFlow Probability (TFP) v0.8. The notebooks nb_ch03_03 and nb_ch03_04,
describing the computation graphs, are easier to understand in TensorFlow v1. For
these notebooks, we also include both versions of TensorFlow. The nb_06_05 note-
book only works with TensorFlow v1 because we need weights that are only provided
in that version of TensorFlow.
You can execute the notebooks in Google’s Colab or locally. Colab is great; you can
simply click on a link and then play with the code in the cloud. No installation—you
just need a browser. We definitely suggest that you go this way.
TensorFlow is still fast-evolving, and we cannot guarantee the code will run in sev-
eral years’ time. We, therefore, provide a Docker container (https://github.com oduerr/
dl_book_docker/) that you can use to execute all notebooks except nb_06_05 and the
TensorFlow 1.0 versions of nb_ch03_03 and nb_ch03_04. This Docker container is
the way to go if you want to use the notebooks locally.

liveBook discussion forum

Purchase of Probabilistic Deep Learning includes free access to a private web forum run
by Manning Publications where you can make comments about the book, ask techni-
cal questions, and receive help from the authors and from other users. To access the
forum, go to https://livebook.manning.com/book/probabilistic-deep-learning-with-
python/welcome/v-6/. You can also learn more about Manning’s forums and the
rules of conduct at https://livebook.manning.com/#!/discussion.
Manning’s commitment to our readers is to provide a venue where a meaningful
dialogue between individual readers and between readers and the authors can take
place. It is not a commitment to any specific amount of participation on the part of
the authors, whose contribution to the forum remains voluntary (and unpaid). We
suggest you try asking the authors some challenging questions lest their interest stray!
The forum and the archives of previous discussions will be accessible from the pub-
lisher’s website as long as the book is in print.
 about the authors
Oliver Dürr is professor for data science at the University of Applied Sciences in Kon-
stanz, Germany. Beate Sick holds a chair for applied statistics at ZHAW, and works as a
researcher and lecturer at the University of Zurich, and as a lecturer at ETH Zurich.
Elvis Murina is a research scientist, responsible for the extensive exercises that accom-
pany this book.
Dürr and Sick are both experts in machine learning and statistics. They have
supervised numerous bachelor’s, master’s, and PhD theses on the topic of deep learn-
ing, and planned and conducted several postgraduate- and master’s-level deep learning
courses. All three authors have worked with deep learning methods since 2013, and
have extensive experience in both teaching the topic and developing probabilistic deep
learning models.

xvii
 about the cover illustration
The figure on the cover of Probabilistic Deep Learning is captioned “Danseuse de l’Isle
O-tahiti,” or A dancer from the island of Tahiti. The illustration is taken from a collec-
tion of dress costumes from various countries by Jacques Grasset de Saint-Sauveur
(1757–1810), titled Costumes de Différents Pays, published in France in 1788. Each illustra-
tion is finely drawn and colored by hand. The rich variety of Grasset de Saint-Sauveur’s
collection reminds us vividly of how culturally apart the world’s towns and regions
were just 200 years ago. Isolated from each other, people spoke different dialects and
languages. In the streets or in the countryside, it was easy to identify where they lived
and what their trade or station in life was just by their dress.
The way we dress has changed since then and the diversity by region, so rich at the
time, has faded away. It is now hard to tell apart the inhabitants of different conti-
nents, let alone different towns, regions, or countries. Perhaps we have traded cultural
diversity for a more varied personal life—certainly for a more varied and fast-paced
technological life.
At a time when it is hard to tell one computer book from another, Manning cele-
brates the inventiveness and initiative of the computer business with book covers
based on the rich diversity of regional life of two centuries ago, brought back to life by
Grasset de Saint-Sauveur’s pictures.

xviii
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
 Part 1

Basics of deep learning

P art 1 of this book gives you a first high-level understanding of what probabi-
listic deep learning (DL) is about and which types of tasks you can tackle with it.
You’ll learn about different neural network architectures for regression (that
you can use to predict a number), and about classification (that you can use to
predict a class). You’ll get practical experiences in setting up DL models, learn
how to tune these, and learn how to control the training procedure. If you don’t
already have substantial experience with DL, you should work through part 1 in
full before moving on to the probabilistic DL models in part 2.
 Introduction to
probabilistic deep learning

Let’s use my
probabilistic travel
You’ll get $500 time gadget!
tip if I arrive at
MoMA within
25 minutes!

This chapter covers

 What is a probabilistic model?
 What is deep learning and when do you use it?
 Comparing traditional machine learning and deep
learning approaches for image classification
 The underlying principles of both curve fitting
and neural networks
 Comparing non-probabilistic and probabilistic
models
 What probabilistic deep learning is and why
it’s useful

3
4 CHAPTER 1 Introduction to probabilistic deep learning

Deep learning (DL) is one of the hottest topics in data science and artificial intelli-
gence today. DL has only been feasible since 2012 with the widespread usage of GPUs,
but you’re probably already dealing with DL technologies in various areas of your
daily life. When you vocally communicate with a digital assistant, when you translate
text from one language into another using the free DeepL translator service (DeepL
is a company producing translation engines based on DL), or when you use a search
engine such as Google, DL is doing its magic behind the scenes. Many state-of-the-art
DL applications such as text-to-speech translations boost their performance using
probabilistic DL models. Further, safety critical applications like self-driving cars use
Bayesian variants of probabilistic DL.
In this chapter, you will get a first high-level introduction to DL and its probabilis-
tic variants. We use simple examples to discuss the differences between non-probabilistic
and probabilistic models and then highlight some advantages of probabilistic DL mod-
els. We also give you a first impression of what you gain when working with Bayesian
variants of probabilistic DL models. In the remaining chapters of the book, you will
learn how to implement DL models and how to tweak them to get their more power-
ful probabilistic variants. You will also learn about the underlying principles that
enable you to build your own models and to understand advanced modern models so
that you can adapt them for your own purposes.

1.1 A first look at probabilistic models

Let’s first get an idea of what a probabilistic model can look like and how you can use
it. We use an example from daily life to discuss the difference between a non-probabilis-
tic model and a probabilistic model. We then use the same example to highlight some
advantages of a probabilistic model.
In our cars, most of us use a satellite navigational system (satnav—a.k.a. GPS) that
tells us how to get from A to B. For each suggested route, the satnav also predicts the
needed travel time. Such a predicted travel time can be understood as a best guess.
You know you’ll sometimes need more time and sometimes less time when taking the
same route from A to B. But a standard satnav is non-probabilistic: it predicts only a
single value for the travel time and does not tell you a possible range of values. For an
example, look at the left panel in figure 1.1, where you see two routes going from
Croxton, New York, to the Museum of Modern Art (MoMA), also in New York, with a
predicted travel time that is the satnav’s best guess based on previous data and the cur-
rent road conditions.
Let’s imagine a fancier satnav that uses a probabilistic model. It not only gives you
a best guess for the travel time, but also captures the uncertainty of that travel time.
The probabilistic prediction of the travel time for a given route is provided as a distri-
bution. For example, look at the right panel of figure 1.1. You see two Gaussian bell
curves describing the predicted travel-time distributions for the two routes.
How can you benefit from knowing these distributions of the predicted travel
time? Imagine you are a New York cab driver. At Croxton, an art dealer boards your
 A first look at probabilistic models 5

taxi. She wants to participate in a great art auction that starts in 25 minutes and offers
you a generous tip ($500) if she arrives there on time. That’s quite an incentive!
Your satnav tool proposes two routes (see the left panel of figure 1.1). As a first
impulse, you would probably choose the upper route because, for this route, it esti-
mates a travel time of 19 minutes, which is shorter than the 22 minutes for the other
route. But, fortunately, you always have the newest gadgets, and your satnav uses a
probabilistic model that not only outputs the mean travel time but also a whole distri-
bution of travel times. Even better, you know how to make use of the outputted distri-
bution for the travel times.
You realize that in your current situation, the mean travel time is not very inter-
esting. What really matters to you is the following question: With which route do you

Let’s use my
probabilistic travel
You’ll get $500 time gadget!
tip if I arrive at
MoMA within
25 minutes!

Chance to get tip: 69%

0.03
Density
0.02

19 Min.
7.9 Miles
0.01

0 10 20 30 40
Travel time

Chance to get tip: 93%

0.02

22 Min.
8.2 Miles
Density
0.01
0.00

0 10 20 30 40
Travel time

Figure 1.1 Travel time prediction of the satnav. On the left side of the map, you see a deterministic version—
just a single number is reported. On the right side, you see the probability distributions for the travel time of the
two routes.
Random documents with unrelated
content Scribd suggests to you:
 ‘Nossir.’
‘Drunk as a boiled owl,’ grunted Red Eller. ‘He don’t know what
it’s all about. Let’s have another drink.’
‘I’m not drunk,’ declared Napoleon. ‘I know a will when I shee
one. Gimme shome of yore tannin’ fluid.’
‘What’s the idea of the clothes?’ queried Spike.
‘Duded up f’r a trip to Cañonville.’
‘Napoleon,’ grinned Spike, ‘have you got a girl?’
‘Nossir, I ain’t got no girl; I’m goin’ on ’ficial business to the county
sheat. These are m’ ’ficial clothes. Here’s m’ regards, gents.’
Napoleon drank a full glass of liquor, groped his way to a chair,
where he flopped down heavily. His derby rolled off across the floor,
and Red Eller kicked it the length of the room. But Napoleon was not
too drunk to witness this bit of horseplay, and his hand groped
drunkenly for the butt of his six-shooter. But after several ineffectual
efforts to draw the gun, he made a gesture of despair, slumped down
in the chair and began snoring.
‘If he’d been sober, he’d have killed you, Red,’ declared Spike.
‘If he’d been sober, I wouldn’t have kicked the hat.’
‘Hell!’ snorted Ed Jones. ‘If he had been sober, he wouldn’t have
worn such a damn lookin’ hat.’
‘Who cares what he would have done?’ growled Dave. ‘What I
want to know is, what are we goin’ to do?’
‘Search me,’ said Spike. ‘I know damn well I’m not goin’ back
there ag’in to-night.’
‘Goin’ to craw-fish on this job?’
‘Not craw-fish, Dave. Old man Lane won’t be there; so what could
we gain by goin’ back?’
‘I reckon that’s true.’
Dave explained to Fairweather what had taken place at the Lane
ranch, but the gambler had no suggestions to offer.
‘I’ll ride down in the mornin’ and collect the guns,’ offered Spike.
‘I’m not scared. They said we could have ’em in daylight.’
 Red Eller and Ed Jones decided that they wanted to play a little
poker, and Dave Morgan wanted to go home; so Dave went away
alone. Others drifted in and the games filled up, while Napoleon
Bonaparte Briggs slept off his jag, and awoke with a stiff neck and
the disposition of a grizzly.
He found the brim of his derby hat, which Red Eller had kicked
loose from its crown, and it pained him greatly. He accepted a drink,
went out to his horse, which he mounted and headed for Cañonville.
‘If that horse ever bucks, that collar will slice old Briggs’s ears off,’
declared Spike Cahill. ‘Funny old coot. Him and Pete Morgan was
pretty close friends, even if they did cuss each other out at least
once a day.’
‘What do you suppose he meant—about that will?’ asked Jack
Fairweather.
‘Liquor talkin’. Old Briggs would rather argue than eat. The minute
somebody says “there is,” old Briggs is sure to say “there ain’t.” But I
sure don’t sabe that boiled shirt and collar and the hard hat.’

It was three days after the voluntary surrender of Paul Lane and
his son, when Hashknife, Rex, and Lem Sheeley rode to Mesa City
from the Lane ranch. Spike Cahill had come to the ranch the day
after their attempt to capture Paul Lane and recovered their guns. If
Spike bore any malice toward Hashknife, Sleepy, or Rex he failed to
show it, but at that time he did not know that Paul Lane had
surrendered to the sheriff.
Both men had sworn that they were innocent of the charge, and
they both denied shooting Noah Evans, who was slowly recovering.
Long Lane swore he had not seen Ben Leach after he left the Oasis
saloon, and that he did not take Ben’s horse and gun.
Of course, no one believed them, and every one knew that they
had surrendered to the law rather than take a chance of being
lynched. Their guilt was so firmly fixed in the eyes of the cattlemen
that any twelve men in the county would have convicted them
without leaving the jury-box.
 Sleepy urged Hashknife to forget the case. As far as he could see
there was nothing to keep them in the Black Horse range any longer,
and Sleepy was anxious to get settled in a job for the winter.
But Hashknife was not satisfied. The shooting of Noah Evans was
one thing unexplained. After talking with Paul Lane, he was satisfied
that neither the old man nor his son was bitter enough against the
6X6 to bushwhack one of that outfit, especially when the light was so
bad that they could not identify their target.
And there was Rex Morgan, whom Sleepy had dubbed ‘The
Orejano.’ Who in Mesa City had sent money to his mother?
wondered Hashknife. And that was the reason why Hashknife, Rex,
and the sheriff had ridden to Mesa City. Hashknife had talked it over
with Lem Sheeley, and they decided to seek information at the Mesa
City Bank.
Jim Harker, the Mesa City banker, a small, wiry man, with heavy
glasses, welcomed them cordially. He had known the sheriff for
years.
‘Well, what’s on your mind, Lem?’ he asked, after Lem had
introduced the others.
‘Peter Morgan banked with you, didn’t he, Jim?’
The banker smiled slowly. ‘What banking he did—yes.’
‘What do yuh mean by that, Jim?’
‘He didn’t do much banking, Lem. Peter Morgan was rather a
queer person, and preferred having his money nearer than a bank
vault.’
‘You mean, he kept it at the ranch?’
‘I suspect he did, Lem.’
‘Well, here’s somethin’ we want to find out, Jim.’ And Lem
explained about Rex’s mother receiving the seventy-five-dollar
check. The banker listened closely, and when Lem finished he shook
his head thoughtfully.
‘Was that the only check from here that you have seen?’ he
asked Rex.
 ‘That was the only one, Mr. Harker. But I feel sure that my mother
received money from some one.’
‘The name of Morgan kinda had us guessin’,’ said Lem.
The banker smiled slowly, thoughtfully.
‘I don’t suppose I’d be violating any confidence, now that Peter
Morgan is dead,’ he said, ‘but the fact of the matter is this—Peter
Morgan could not write.’
‘Couldn’t write?’ pondered Lem.
‘He had no education whatever. In fact, when he wished to draw a
check, I signed his name for him. So that answers your question
regarding that particular check.’
‘Could it have been Dave Morgan?’ asked Hashknife.
The banker shook his head quickly. ‘No. Dave Morgan closed his
account with us several months ago. I think I was the only one in this
country who knew that Peter Morgan could not write. He was very
sensitive about it. I don’t believe Dave Morgan knew it. When there
were any papers to be signed, Peter always brought them to me.’
‘You never heard him mention a will, did yuh?’ asked Hashknife.
‘No, I never did. I’m sure I should have known about it if there had
been one. I understand you have Paul Lane in jail for murdering
Peter.’
‘Yeah, he’s down there,’ sighed Lem.
‘Any question about his guilt, Lem?’
‘I hope so, Jim. I dunno what defense the old man will put up. If
he wasn’t a nester, he might get off. Yuh see, he warned Peter to
keep away from his place. There ain’t no direct evidence that
Morgan was killed on the Lane ranch, but the jury will probably think
he was.’
‘How is Noah Evans getting along?’
‘Kinda slow. The doctor seems to think he’s out of danger. I
reckon he is; he cusses all the time.’
They thanked the banker for his information. Lem had other
business to transact; so they left him in Mesa City and rode back to
the ranch. Hashknife was disappointed. He had expected some
information from the banker which might be of some value to him,
but, as far as he could see, they were up against a blank wall.
Rex had nothing to offer. He couldn’t remember what the
signature on that check looked like.
‘I don’t know what to do,’ he told Hashknife.
‘I can’t stay here all my life. I haven’t any money, and no place to
go. Rather a puzzling situation, isn’t it?’
‘Well, we’re in the same boat,’ smiled Hashknife. ‘But our case is
a little different, except that we can’t all pull out and leave that girl
here alone. The Lanes haven’t any money, either. It looks to me as
though we’ve all got to stay here and see what works out. Me and
Sleepy have enough for a grubstake for all of us, I reckon.’
‘Well, I suppose we’ll have to do something like that.’
‘Sure. But the limb of the law waves so damn slow in a country
like this that we’ll wear the seats out of our pants waitin’ for them to
try the Lane family for murder. I reckon we’ll just stick around and
see what happens.’
They talked it over with Nan and Sleepy at the ranch. Nan wanted
to go to Cañonville and try for a job.
‘I might get work in a restaurant,’ she said. ‘That would relieve
you boys of my presence. I didn’t realize the situation until now.’
‘You are not going to work in any restaurant,’ declared Rex
warmly. ‘We can get along out here. I’ve still got my five dollars.’
‘You bloated financier!’ exploded Sleepy. ‘If you knew anythin’
about poker, I’d take that five away from yuh.’
‘I don’t know anything about poker,’ said Rex quickly, ‘but if you
want the five, I’ll give it to you, Sleepy.’
‘Thank yuh,’ grunted Sleepy, rather taken back by Rex’s
generosity. ‘I reckon yore all right, kid; we’ll get along.’
‘Can yuh imagine that?’ he asked Hashknife a little later on.
‘Rex is all right, Sleepy.’
‘Shore, he’s all right. Pretty heavy on education, but he’ll get that
knocked out of him in a short time. Do yuh know, I’ve got a hunch
that Nan thinks quite a lot of him.’
 ‘She’s sorry for him.’
‘Yeah, and he’s sorry for her. He looks at her like a dyin’ calf in
the spring thaw.’
They took care of their horses and wandered back to the house,
where they found Nan and Rex on the porch, talking confidentially.
Nan seemed very determined about something, and Rex seemed
troubled. Hashknife sprawled on the steps and rolled a cigarette.
‘I want to tell you something, Hashknife,’ said Nan. ‘Rex don’t
think I should, but——’
‘I wouldn’t,’ said Rex firmly.
‘But I think you boys understand,’ said Nan. ‘Oh, it won’t hurt
anything, Rex—not now. We haven’t told anybody, except my father.’
‘Go ahead,’ urged Hashknife. ‘I suppose it’s about findin’ Peter
Morgan’s body, ain’t it? And sendin’ it home on the horse?’
Nan gasped, staring at Hashknife. ‘What—why, how did you
know?’
‘Guessed it, Nan. Lem found you and Rex in the corral. Rex had
fainted. And then Lem found Peter Morgan’s six-gun in the corral. It
wasn’t more than an hour or so later that the body of Morgan came
to the 6X6.’
‘And you just guessed it?’ asked Nan wonderingly.
‘Somethin’ like that, Nan. I figured that you and Rex had a secret
between yuh. Would yuh mind takin’ me down to the corral and
showin’ me just how the body lay, and all that?’
‘And—and you don’t blame us for what we done?’ asked Nan.
‘Certainly not; I’d have done the same. C’mon.’
They all went down to the corral, where Nan explained all about
the position of the body, and how they had secured the horse from
the willows across the stream and had managed to rope the body to
the saddle.
Hashknife listened closely, questioning both of them as to small
details, and even examined the dust closely, where Peter Morgan’s
body had lain. Nan pointed out the place where the horse had been
tied, and Rex took Hashknife over to the spot where he secured the
horse.
The ground was fairly soft along the creek, and Hashknife was
able to distinguish the tracks of the shod horse.
‘Mr. Morgan must have left his horse here while he went over to
the corner of the stable,’ said Rex.
Hashknife grunted, as he studied the tracks closely. From where
they stood, the horse would have been invisible to any one at the
ranch-house. The presumption would have been that Morgan rode
up to the willows from that side of the creek, not taking any chances
of being seen; but the tracks showed that the horse had crossed the
stream twice; one set of tracks, of course, being made when Rex
took the animal over to the corral. It proved that the rider had come
in past the corral, crossed the creek, and tied the horse over there.
They came back to the bank of the little creek, where Hashknife
stopped again to examine the tracks. The stream was about four feet
wide and two feet deep at this point. Rex sprang across and went
back to the corral fence, while Hashknife squatted on his heels on
the creek bank.
Suddenly he got to his feet and looked down the stream, where
the water swung around an undercut bank, practically undermining a
heavy growth of willows. Something had attracted his attention, and
he shoved down through the brush to this spot, where he sprawled
along the bank, reaching down in the water.
After some little effort he was able to draw out the object, which
he lugged back to a clear space. It was a Navajo rug, about four feet
wide and five feet long, which had been rolled tightly and tied at both
ends with whang-leather strings.
Hashknife cut the strings and unrolled the rug. It was rather
difficult to tell how long the rug had been in the water. It was rather
discolored, but the pattern was clear enough. The two ends of the
rug were of red and gray design, while the center was dead black,
with a jagged strip of white, representing the Navajo idea of lightning.
Hashknife carried the rug over to the corral, where he spread it
out on the ground. It was a very distinctive pattern, and Nan was
sure she had never seen it before. Just why it was in the creek, none
of them were able to say. It was not a rug that any one would
discard. He hung it over the top-pole of the corral to dry out, and left
it there, dripping down across the poles.
‘That must have been a beautiful rug,’ sighed Nan. ‘I have always
admired Navajo rugs, but we have always been too poor to buy one.’
‘You can have that one,’ smiled Hashknife. ‘Probably take a lot of
washin’ to clean it up. Lot of that silt has soaked up in it, and it’ll take
time to get it out. Might be better to let it dry, and then beat it out.’
‘What would they do to us if they knew what we had done?’ asked
Rex anxiously.
‘I dunno,’ smiled Hashknife. ‘Better not tell anybody else. It would
be a point for the prosecution, yuh know. It would prove just where
Peter Morgan was killed. It’s too bad yuh didn’t think to get that gun.’
‘We were too excited to think of anything except to get the body
away from here,’ said Nan.
‘I’ll betcha. That was shore some job for you two. Now, we’ll just
forget all that.’
It was about an hour later when Lem Sheeley rode in at the
ranch. Nan was busy in the kitchen, but the three men were on the
porch to meet him.
‘I’ve got a little information for yuh,’ said the sheriff, declining to
dismount. ‘After you boys left town, Jim Harker called me back to the
bank. Yuh see, bankers are kinda close when it comes to talkin’
about things, and he didn’t know yuh very well.[’]
‘Here’s what he told me, boys. Napoleon Bonaparte Briggs is the
one who has been sendin’ money to a Mary Morgan. He has been
sendin’ it for years. Jim said he never asked Nap about it, ’cause he
figured it wasn’t his business.[’]
‘He said he wondered where Briggs got the money, until one day
Peter Morgan told him that Briggs had an interest in the 6X6. Now,
mebby yuh can find out from Briggs what it’s all about, Hashknife.
Harker don’t know a thing about it, except that Briggs kept a balance
in the bank, and mostly every month he sent a check away. Harker
says he don’t know any Mary Morgan.’
 ‘Old man Briggs is the cook at the 6X6, ain’t he, Lem?’ asked
Hashknife.
‘Yeah, a queer old pelican. Been with the 6X6 since these hills
were holes in the ground. Don’t start any argument, ’cause it won’t
get yuh anywhere with him. I wish yuh luck in findin’ out anythin’.
Just thought mebby you’d like to know; so I dropped in. Got to get
back before dark.’
Hashknife thanked him for the information, and Lem rode on
toward Cañonville.
‘It shore is worse tangled than ever now,’ said Hashknife. ‘I
reckon the name Morgan is just a coincidence in this case, Rex. But
just where does Briggs come in on it?’
‘Oh, I suppose we’ll never find out,’ sighed Rex. ‘But after all,
what difference does it make? It can’t affect my future in any way.
Still, I’d like to know. Don’t you see the position I am in? Suppose’—
Rex hesitated for a moment—‘suppose I wanted to marry a girl, and
she asked me about my father?’
‘Tell her he died before you was born,’ advised Hashknife.
‘But that would be a lie.’
‘How do you know?’
‘But I couldn’t prove it, Hashknife.’
‘Any girl who likes you well enough to marry yuh won’t make yuh
prove when yore father died, kid.’
‘But I don’t even know I had a father.’
‘Well, yuh won’t have to prove that. Just forget that yuh went
through life kinda one-sided on parents. And don’t argue with me. I
want to set down and think about Napoleon Bonaparte Briggs and a
dirty Navajo rug.’
Sleepy went into the house, where he flopped on the old couch,
burying his nose in an old magazine, while Rex sat down on a corner
of the porch, watching the changing lights on the hills as the sun
sank lower in the west.
It was as though a painter, unsatisfied with an effect, would swiftly
blot out a streak of gold and draw in a full brush of violet; only to
change it to a deep mauve and then to an opaque cobalt; striking
new highlights with glowing gold.
Farther to the north a great flock of birds, like a lot of black sheets
of paper caught in a whirlwind, spiraled up from among the hills,
always traveling in circles. Rex watched them, fascinated. They did
not seem to flap their wings, but mounted higher and higher. Some
of them circled back to earth, but seemed to come back, flapping
their wings, as though in haste to gain altitude.
‘What kind of birds are those, Hashknife?’ asked Rex.
‘Buzzards,’ he said indifferently.
‘I tried to count them, but they weave back and forth so swiftly,
and each one looks like the other.’
Hashknife relaxed and reached for his cigarette papers.
‘Scavengers, Rex; a big bird who smells death, they say. But I
don’t believe it, because I’ve fooled ’em. I’ve stretched out on the
desert, played dead, and had them down so close I pulled feathers
out of their tails.’
‘Is there something dead over there, Hashknife?’
‘Undoubtedly. They’ve been having a feast, and are pulling out
before dark. Mebby a coyote or two came along and started an
argument.’
‘Dead cow, do you suppose?’
Hashknife squinted quizzically at the gyrating flock, slowly
mounting higher. They were not splitting up, as a flock usually does,
when the feast is over; but rather they were acting as though
something had interrupted them. Hashknife grinned and turned to
Rex.
‘Let’s take a rifle and go over there, kid. It’s in a little swale off the
road, and we might knock over a coyote.’
Rex was willing. Hashknife called to Sleepy, asking him to go
along.
‘Goin’ to ride over?’ asked Sleepy.
‘It’s only a little ways,’ replied Hashknife.
 ‘Count me out. I wouldn’t walk a mile for all the coyotes in
Arizona.’
Hashknife took Paul Lane’s thirty-thirty, and they walked up the
road, while the buzzards still circled. It was a little over a mile to
where they left the road, and about a quarter of a mile from where
Ben Leach had been killed.
From the road they went cautiously through the brushy swale,
circling the thickets of mesquite. Suddenly a coyote went streaking
across the swale, almost invisible in the waning light. Hashknife
stepped back, swinging up the Winchester, and as the animal started
up the slope of the hill on the opposite side of the swale, the rifle
cracked sharply and the coyote gave a convulsive sideways leap,
landed in a Spanish dagger, from whence it went yipping along
through the brush, telling the world in coyote language what it
thought of a man who would drive a thirty-thirty bullet in front of the
nose of any well-meaning coyote.
‘Led him too much,’ laughed Hashknife. ‘Didja see him set down
in that dagger? Talk about anythin’ bein’ full of pins and needles! I
had a hunch that some coyotes had chased them buzzards from
their supper.’
They circled another clump of mesquite and found what had
attracted the scavengers. It was what was left of a blue-roan horse,
which was still wearing a saddle and bridle. The buzzards and
coyotes had made a sorry mess of it, but the saddle and bridle were
still intact.
With his pocket-knife, Hashknife cut the latigo, and drew the
saddle away from the carcass. It was a good grade of stock saddle,
with stamped seat and fenders. The skin of the animal had been
literally torn to shreds, obliterating the brand, but leaving enough to
identify its color.
Hashknife examined the head of the animal for possible bullet
holes, but found none. Upon closer examination, however, he found
that the horse’s shoulder had been broken. The bridle reins were
tangled about the other leg, drawing the head of the animal sharply
downward.
 No doubt the coyotes had pulled the body about to some extent,
but Hashknife was able to read the signs fairly plain.
‘I reckon this was Ben Leach’s horse,’ he told Rex. ‘It busted its
shoulder in some way, leavin’ it to hobble on three legs until the reins
got tangled in the other front leg and threw it. Mebby the fall broke its
neck, or mebby it just couldn’t get up, and the coyotes finished it.’
‘Does it mean anything?’ asked Rex.
‘Well, it means that Walter Lane didn’t steal the horse, which is
one point in his favor.’
Hashknife hung the saddle in a mesquite thicket, and they went
back home, leaving the way clear for the coyotes to continue their
interrupted meal. The buzzards had disappeared by this time.
‘If it hadn’t been for those buzzards, we should never have found
that horse,’ said Rex.
‘That’s true,’ thoughtfully. ‘Sometimes it’s a good plan to foller the
buzzards, kid. Yuh never can tell what yuh might find.’
 CHAPTER XII: AN EMPTY SAFE
‘What right have you got to open the safe?’ demanded Dell
Bowen. ‘This ranch ain’t never been turned over to yuh, Dave.’
Dave Morgan, standing in the center of the 6X6 ranch-house
living-room, smiled sarcastically at Dell Bowen. With Morgan was Ed
Jones, his right-hand man. Spike Cahill and Bert Roddy were
standing beside a small, old-fashioned iron safe against the west
wall of the room.
On the table beside Morgan was a collection of papers, some
money; the miscellaneous stuff which had been taken from Peter
Morgan’s pockets. Dave Morgan held a key in his hand, which he
had taken from the table.
‘It will be, as far as that’s concerned,’ said Morgan.
‘Then yuh better wait until it is, Dave.’
‘Yea-a-ah? Well, I’m just a little curious to know what’s in that
safe, Dell. All this talk about wills has kinda made me wonder if there
is such a thing. You boys can check up everythin’ in there, and I
won’t take anythin’.’
‘Well, that might be all right,’ dubiously. ‘As long as we put
everythin’ back, Dave. I don’t want to get in bad with the law, you
know.’
They went over to the safe and watched Dave Morgan insert the
key. The safe opened easily. For a moment there was no sound, and
then Spike Cahill whistled softly.
The safe was empty! Not even a scrap of paper. Dave got to his
feet, looking from face to face.
‘Looks pretty clean, don’t it?’ he asked softly.
‘It does,’ admitted Dell Bowen.
Dave walked back to the table and dropped the key. He did not
bother to lock it again.
 ‘I don’t quite sabe that empty safe,’ said Bowen. ‘The boss must
’a’ cleaned it out before he got killed.’
‘Yuh think he did, eh?’ sneered Morgan.
‘Well, it shore looks that away, Dave.’
‘Yea-a-ah? Huh! I don’t sabe why that key wasn’t turned over to
the sheriff. Leavin’ it layin’ around——’
‘Wait a minute,’ cautioned Bowen. ‘You ain’t aimin’ to put the
dead wood on any of us, are yuh?’
‘I’m not aimin’ at anybody,’ angrily. ‘But it don’t look right for that
safe to be empty, does it?’
‘Aw, what the hell!’ snorted Spike. ‘It’s empty, and that’s all there
is to it.’
‘Mebby not all,’ retorted Dave. ‘Pete must have had some money.
There’s damn little in the bank; I found that out.’
‘Well, what are yuh goin’ to do about it?’ demanded Dell. ‘It shore
looks to me as though yuh was a little previous, Dave.’
‘Yuh think so, do yuh, Bowen?’ Dave hooked his thumbs over his
cartridge belt and glared at the red-faced cowboy who had acted as
Peter Morgan’s foreman. ‘Lemme tell you somethin’. This 6X6
belongs to me. It’s only a matter of a few days until I move in here
and take charge.’
‘Suits me,’ grunted Bowen.
Dave hunched his shoulders and scanned the faces of the three
6X6 cowboys.
‘And I’ll prob’ly hire my own crew,’ he added meaningly.
‘And far as I’m concerned, you can hire ’em right now, Morgan;
I’m through.’
‘Same here,’ nodded Spike, and Bert nodded in agreement.
‘Yuh don’t need to go off half-cocked,’ said Dave quickly.
‘As far as I’m concerned, yuh can go to hell,’ said Bowen coldly.
‘And another thing, Morgan; don’t ever make the crack that any of us
unlocked that safe. If yuh do, we’ll shore take yuh apart to see what
makes yuh tick.’
 ‘I don’t reckon I need to make any cracks, Bowen. And as far as
takin’ me apart is concerned, I’ll be right there, showin’ yuh what
makes me tick.’
The three cowboys headed for the door.
‘Yuh might take Napoleon Bonaparte Briggs along with yuh,’ said
Dave. ‘I reckon I can find a cook.’
‘He’s gone already,’ growled Spike. ‘Pulled out several days ago.
The night the tenderfoot made us back-water with an empty shotgun,
old man Briggs headed for Cañonville.’
‘Well, if yuh see him, yuh can tell him he don’t need to come
back.’
‘If he knows you’re here, it’s a cinch he won’t want to.’
It did not take the three boys long to pack their belongings in their
war-bags. Dave Morgan and Ed Jones sat on the spacious porch of
the ranch-house and watched them saddle their horses and leave
the ranch.
Dave Morgan smiled around at his possessions. The 6X6 was the
biggest ranch in the country. Even the stable was a better building
than the ranch-house at the Flying M.
‘Got kinda salty, didn’t they?’ grinned Dave.
‘They allus have been,’ said Ed slowly. ‘I was wonderin’ about old
man Briggs, Dave. He pulled out several days ago, and he had
plenty chance to open that safe.’
Dave smiled thoughtfully. ‘I’ll betcha you’re right, Ed. That old
pelican cleaned out that safe and skipped the country. I’ll get in touch
with the sheriff as soon as I can, and we’ll see if we can’t find Briggs.
He’s the man.’
‘Another thing,’ said Jones. ‘You’ll have to hire some punchers,
Dave. We can’t run both places. What yuh ort to do is to hire a crew
to run the Flyin’ M, and let us come here.’
‘Might be a good idea. I dunno where I’d find any men right now.’
‘How about them two jiggers at the Lane place? Oh, I know yuh
don’t like ’em, but they look like cowmen, Dave.’
 ‘I know,’ nodded Dave. ‘I kinda hate to go ahead and hire a crew
until this thing is all settled up, but I reckon I might as well. To-
morrow I’ll go down to Cañonville and have a talk with a lawyer. He
can fix it all up for me, and in the meantime, I’ll ride down to the Lane
place and have a talk with them two fellers. I don’t quite sabe what
they’re doin’ around here, and it might be a good scheme to have
’em where we know what they’re doin’.’
‘You don’t think they’re in here to pull anythin’ crooked, do yuh,
Dave?’ asked Ed.
‘Quien sabe? They’re shore backin’ the Lane family, and I might
be doin’ the law a favor by hirin’ ’em away from the Lane ranch. You
can stay here, Ed. Mebby yuh better ride back to the ranch and get
Red to come over with yuh. Move yore stuff over. I’ll stay at the Flyin’
M with Cal until I get things fixed up. And I’ll see if I can rustle a cook
for here.’
That same morning Hashknife had been doing considerable
thinking about the dead horse they had found; so he decided to ride
to Cañonville and have a talk with the sheriff. Nan wanted to send
some clean clothes to the jail; so Sleepy decided to go along. With
both of the Lanes in jail, there was nothing for Nan to be worried
about, as far as the 6X6 was concerned. Anyway, Rex would stay.
After they went away he investigated the Navajo rug, and found it
practically dry. He carried it up beside the house and proceeded to
hammer the dirt out of it. Armed with a section of broomhandle he
beat industriously, and was so engaged when Dave Morgan rode up.
Rex was a bit apprehensive when he saw who the rider was, but
Dave’s grin was reassuring.
‘Cleanin’ house?’ he asked, glancing at the rug.
‘Not exactly,’ said Rex, wiping the perspiration off his nose. ‘Mr.
Hartley discovered this rug in the creek yesterday. I doubt that I shall
ever be able to beat the dirt out of it.’
‘Found it in the creek, eh?’
‘Yes. Queer, isn’t it? Looks like a good rug. It has a very
distinctive pattern, don’t you think?’
‘Yea-a-ah, it has.’
 Nan came to the door and saw Dave Morgan. He smiled at her
and lifted his hat.
‘How do yuh do, Miss Lane?’
Nan nodded coldly.
‘Aw, let’s be friends,’ he laughed. ‘I don’t blame yuh for the way
yuh feel. No? Well, I’m sorry. Where’s Hartley and his pardner?’
‘Oh, they went to Cañonville to-day,’ said Rex quickly. ‘We found
the horse and saddle which belonged to Ben Leach last night, and I
think Mr. Hartley wished to talk to the sheriff about it.’
‘Thasso? Where did yuh find it?’
Rex pointed vaguely toward the hills. ‘Out there. It had a broken
shoulder, and the coyotes had eaten it nearly all up.’
‘Yea-a-ah?’ Dave rubbed his chin thoughtfully. ‘Well, that’s funny.’
He turned to Nan. ‘Mebby yore brother didn’t take Ben’s horse,
after all. That’ll be good evidence in his favor.’
‘I hope something will be in his favor,’ wearily.
‘You just leave it to Hashknife Hartley,’ said Rex. ‘He knows
things. Sleepy told me that Hashknife saw things that other men
overlooked.’
‘Is that so? I wanted to have a talk with ’em to-day. I’ve taken over
the 6X6, and I need some help. In fact, I’d like to get them to run the
Flyin’ M for me, because I want to move my crew over to the 6X6.[’]
‘I might be able to use you, too,’ he said to Rex.
‘You can learn.’
‘Certainly, I can learn,’ said Rex, smiling.
Dave looked at Nan thoughtfully.
‘You’ve been cookin’ for these three men, ain’t yuh?’
Nan nodded. ‘After a fashion.’
‘Well, now here’s what I was thinkin’ about. This place don’t need
much attention. Why don’t yuh all move over to the Flyin’ M, Miss
Lane? You can do the cookin’. Take what little stock you’ve got over
there. It’s better than this.[’]
 ‘Under the circumstances, there ain’t a thing yuh can do here.
That trial won’t come off for another month, at least. Why not try a
whirl at the Flyin’ M? At least you’ll get paid for yore work.’
‘Why, I don’t know—’ faltered Nan. ‘I—I don’t know what to say.’
‘Well, think it over. When Hartley comes back, talk it over with
him. If yuh want the jobs, you’ll probably find me at the Flyin’ M. I’m
goin’ to Cañonville to-morrow to see a lawyer, but I’ll be back late in
the afternoon. Don’t hold it against me for the things my cousin done
to yuh. I’m owner of the 6X6 now, and I don’t mind a nester.’
‘It is nice of you to make us that offer,’ said Nan.
‘Nothin’ of the kind; I need yuh. Let me know what yuh decide to
do. Well, I’ll see yuh later.’
Dave Morgan tipped his sombrero and rode away, while Rex
essayed an awkward dance on the half-cleaned rug. He was
enthused over a chance to get a job. But Nan did not share his
enthusiasm.
‘Why, Nan, he’s all right,’ said Rex. ‘He’s so different from what he
was the last time he was here.’
‘Very much different,’ said Nan seriously. ‘I can’t trust men who
suffer such great changes. Hang that dirty rug on the fence and get
me some wood. I’m trying to bake an apple pie for supper, and I
need a hot oven.’
Rex shouldered the rug and turned toward the corral, but swung
around facing Nan.
‘If I had a lucrative position, would you marry me?’
Nan flushed slightly. ‘That question is a whole lot like saying, “If
we had some ham we could have some ham and eggs, if we had
some eggs.”’
‘I see,’ blankly. ‘Well, I shall get you the wood, Nan. I’m afraid I’ll
never get this rug cleaned. It seems to be badly stained.’
‘Well, hang it on the fence.’
Dave Morgan decided to go to Cañonville that day instead of
waiting until the next, and he was within a couple of miles of the town
when he met the sheriff, Hashknife, and Sleepy. The sheriff had
considered the dead horse worthy of further investigation, and they
were on their way to the Lane ranch.
Dave greeted them in a friendly manner, and told them he had
been at the Lane ranch.
‘Didja see old Briggs in Cañonville?’ he asked Lem.
‘He ain’t there, Dave. What do yuh want Briggs for?’
Dave told them of the empty safe at the 6X6. He did not directly
accuse Briggs of theft, but he intimated that the guilt lay between
Briggs and the other three cowboys, with most of the evidence
against Briggs, because he had seemingly left the country.
‘How do yuh know there was anythin’ in the safe before Peter
Morgan was killed?’ asked Hashknife.
‘I don’t know that there was, Hartley; but there should have been.
Pete wasn’t very strong for banks, and it don’t seem reasonable that
the safe would be empty, does it?’
‘Who had charge of the key?’ asked Lem.
‘Anybody at the ranch. They had the keys, along with the other
stuff they took from Pete’s pockets, in a drawer of the table. Any of
them could have swiped the key and looted the safe, Lem.’
Lem considered the matter gravely.
‘It’s kinda hard to accuse anybody of theft when yuh don’t know
anythin’ has been stolen, Dave. There’s no record of what was in the
safe. Old Briggs has had several days to make his getaway, and
nobody knows which way he went.’
‘Yeah, that’s true. I suppose we might as well forget that safe.’
‘Best thing to do, Dave. Have you taken over the 6X6?’
‘Just about. I’m goin’ to see a lawyer now and have him fix it up.’
He turned to Hashknife. ‘I need two good men, Hartley. The crew
at the 6X6 have quit, and I’m kinda short-handed. If you’ll take the
job, I’ll put yuh in charge of the Flyin’ M. I had a talk with Miss Lane
and the young feller awhile ago, and I made her a proposition to
cook for yuh. She’s cookin’ for yuh now; so she might as well make
some money out of it. I can use the kid, too. He ain’t worth much, but
he’ll learn. What do yuh say?’
 ‘I dunno,’ smiled Hashknife. ‘Sounds all right. I’ll talk it over with
Miss Lane and see what she thinks. We can’t step out and leave her
alone, yuh see.’
‘No, that wouldn’t be right. You talk it over with her and let me
know. I want to move my outfit over to the 6X6 as soon as I can. As
far as Lane’s stock is concerned, he’s only got a few head, and yuh
can herd them over to my place.’
‘Well, that’s nice of yuh, Morgan.’
‘It’ll help me out a whole lot.’
Hashknife promised to talk it over with Nan, and they rode on.
Sleepy was rather jubilant over the chance to go to work, especially
if Hashknife was to have charge of the ranch. But Hashknife had little
to say about it.
‘Don’t yuh think he’s a little previous about takin’ over the 6X6?’
asked Hashknife.
‘Looks thataway,’ replied Lem. ‘Still, he might as well, I suppose.
Now, that he owns the Oasis, he’ll probably drink himself to death.
Dave’s all right, though. He’s all excited over that 6X6.’
‘Probably was an awful blow to find the safe empty,’ grinned
Sleepy.
‘Sure. Just between us, I’ll bet old Napoleon Bonaparte Briggs did
clean out that safe. He just about emptied it and headed out of the
country.’
‘What about his share of the 6X6?’ queried Hashknife. ‘The
banker at Mesa City mentioned it to you, didn’t he?’
‘Mebby Briggs’s share was in the safe,’ chuckled Lem. ‘Briggs
hated Dave Morgan, and he knew Dave would inherit the 6X6; so he
just took his share out of the safe and busted up the partnership.
Anyway, it’s too bad they didn’t turn that key over to me.’
They left the main road and soon came to the spot where Ben
Leach had been killed. Hashknife had never had much of that
incident explained to him, but Lem pointed out where Ben had fallen,
according to what the others had told him, because they had moved
the body before he had a chance to view it.