Handbook of Functional MRI Data Analysis 1st

Edition Russell A. Poldrack pdf download

https://ebookname.com/product/handbook-of-functional-mri-data-
analysis-1st-edition-russell-a-poldrack/

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

Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Handbook of Small Animal MRI 1st Edition Elliott

https://ebookname.com/product/handbook-of-small-animal-mri-1st-
edition-elliott/

Theoretical Foundations of Functional Data Analysis

with an Introduction to Linear Operators 1st Edition
Tailen Hsing

https://ebookname.com/product/theoretical-foundations-of-
functional-data-analysis-with-an-introduction-to-linear-
operators-1st-edition-tailen-hsing/

Purely functional data structures 1st Edition Okasaki

https://ebookname.com/product/purely-functional-data-
structures-1st-edition-okasaki/

Regional Anaesthesia Stimulation and Ultrasound

Techniques 1st Edition Paul Warman

https://ebookname.com/product/regional-anaesthesia-stimulation-
and-ultrasound-techniques-1st-edition-paul-warman/
The Politics of War Trauma The Aftermath of World War
II in Eleven European Countries 1st Edition Jolande
Withuis

https://ebookname.com/product/the-politics-of-war-trauma-the-
aftermath-of-world-war-ii-in-eleven-european-countries-1st-
edition-jolande-withuis/

Sex Allocation Stuart West

https://ebookname.com/product/sex-allocation-stuart-west/

Manganese Catalysis in Organic Synthesis 1st Edition

Jean-Baptiste Sortais

https://ebookname.com/product/manganese-catalysis-in-organic-
synthesis-1st-edition-jean-baptiste-sortais/

In the Beginning Was the Apeiron Infinity in Greek

Philosophy Adam Drozdek

https://ebookname.com/product/in-the-beginning-was-the-apeiron-
infinity-in-greek-philosophy-adam-drozdek/

Environmental Policy New Directions for the Twenty

First Century 11th Edition Norman J. Vig (Editor)

https://ebookname.com/product/environmental-policy-new-
directions-for-the-twenty-first-century-11th-edition-norman-j-
vig-editor/
Video Coding for Wireless Communication Systems Signal
Processing Series 1st Edition King N. Ngan

https://ebookname.com/product/video-coding-for-wireless-
communication-systems-signal-processing-series-1st-edition-king-
n-ngan/
Handbook of Functional MRI Data Analysis

Functional magnetic resonance imaging (fMRI) has become the most popular method for imaging
brain function. Handbook of Functional MRI Data Analysis provides a comprehensive and practical
introduction to the methods used for fMRI data analysis. Using minimal jargon, this book explains
the concepts behind processing fMRI data, focusing on the techniques that are most commonly
used in the field. This book provides background about the methods employed by common data
analysis packages including FSL, SPM, and AFNI. Some of the newest cutting-edge techniques,
including pattern classification analysis, connectivity modeling, and resting state network analysis,
are also discussed.
Readers of this book, whether newcomers to the field or experienced researchers, will obtain
a deep and effective knowledge of how to employ fMRI analysis to ask scientific questions and
become more sophisticated users of fMRI analysis software.

Dr. Russell A. Poldrack is the director of the Imaging Research Center and professor of
Psychology and Neurobiology at the University of Texas at Austin. He has published more than
100 articles in the field of cognitive neuroscience, in journals including Science, Nature, Neuron,
Nature Neuroscience, and PNAS. He is well known for his writings on how neuroimaging can be
used to make inferences about psychological function, as well as for his research using fMRI and
other imaging techniques to understand the brain systems that support learning and memory,
decision making, and executive function.

Dr. Jeanette A. Mumford is a research assistant professor in the Department of Psychology

at the University of Texas at Austin. Trained in biostatistics, her research has focused on the
development and characterization of new methods for statistical modeling and analysis of fMRI
data. Her work has examined the impact of different group modeling strategies and developed
new tools for modeling network structure in resting-state fMRI data. She is the developer of the
fmriPower software package, which provides power analysis tools for fMRI data.

Dr. Thomas E. Nichols is the head of Neuroimaging Statistics at the University of Warwick,
United Kingdom. He has been working in functional neuroimaging since 1992, when he joined
the University of Pittsburgh’s PET facility as programmer and statistician. He is known for his
work on inference in brain imaging, using both parametric and nonparametric methods, and he
is an active contributor to the FSL and SPM software packages. In 2009 he received the Wiley
Young Investigator Award from the Organization for Human Brain Mapping in recognition for his
contributions to statistical modeling and inference of neuroimaging data.
Handbook of Functional
MRI Data Analysis

Russell A. Poldrack
University of Texas at Austin, Imaging Research Center

Jeanette A. Mumford
University of Texas at Austin

Thomas E. Nichols
University of Warwick
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town,
Singapore, São Paulo, Delhi, Tokyo, Mexico City

Cambridge University Press

32 Avenue of the Americas, New York, NY 10013-2473, USA

www.cambridge.org
Information on this title: www.cambridge.org/9780521517669

© Russell A. Poldrack, Jeanette A. Mumford, and Thomas E. Nichols, 2011

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 2011

Printed in China by Everbest

A catalog record for this publication is available from the British Library.

Library of Congress Cataloging in Publication data

Poldrack, Russell A.
Handbook of functional MRI data analysis / Russell A. Poldrack,
Jeanette A. Mumford, Thomas E. Nichols.
p. ; cm.
Includes bibliographical references and index.
ISBN 978-0-521-51766-9 (hardback)
1. Brain mapping – Statistical methods. 2. Brain – Imaging – Statistical methods.
3. Magnetic resonance imaging. I. Mumford, Jeanette A., 1975–
II. Nichols, Thomas E. III. Title.
[DNLM: 1. Magnetic Resonance Imaging. 2. Brain Mapping. 3. Data Interpretation,
Statistical. 4. Image Processing, Computer-Assisted – methods. WN 185]
RC386.6.B7P65 2011


616.8 047548–dc22 2011010349

ISBN 978-0-521-51766-9 Hardback

Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or
third-party Internet Web sites referred to in this publication and does not guarantee that any content on such
Web sites is, or will remain, accurate or appropriate.
 Contents

Preface page ix

1 Introduction 1
1.1 A brief overview of fMRI 1
1.2 The emergence of cognitive neuroscience 3
1.3 A brief history of fMRI analysis 4
1.4 Major components of fMRI analysis 7
1.5 Software packages for fMRI analysis 7
1.6 Choosing a software package 10
1.7 Overview of processing streams 10
1.8 Prerequisites for fMRI analysis 10

2 Image processing basics 13

2.1 What is an image? 13
2.2 Coordinate systems 15
2.3 Spatial transformations 17
2.4 Filtering and Fourier analysis 31

3 Preprocessing fMRI data 34

3.1 Introduction 34
3.2 An overview of fMRI preprocessing 34
3.3 Quality control techniques 34
3.4 Distortion correction 38
3.5 Slice timing correction 41
3.6 Motion correction 43
3.7 Spatial smoothing 50

4 Spatial normalization 53
4.1 Introduction 53
4.2 Anatomical variability 53
v
vi

4.3 Coordinate spaces for neuroimaging 54

4.4 Atlases and templates 55
4.5 Preprocessing of anatomical images 56
4.6 Processing streams for fMRI normalization 58
4.7 Spatial normalization methods 60
4.8 Surface-based methods 62
4.9 Choosing a spatial normalization method 63
4.10 Quality control for spatial normalization 65
4.11 Troubleshooting normalization problems 66
4.12 Normalizing data from special populations 66

5 Statistical modeling: Single subject analysis 70

5.1 The BOLD signal 70
5.2 The BOLD noise 86
5.3 Study design and modeling strategies 92

6 Statistical modeling: Group analysis 100

6.1 The mixed effects model 100
6.2 Mean centering continuous covariates 105

7 Statistical inference on images 110

7.1 Basics of statistical inference 110
7.2 Features of interest in images 112
7.3 The multiple testing problem and solutions 116
7.4 Combining inferences: masking and conjunctions 123
7.5 Use of region of interest masks 126
7.6 Computing statistical power 126

8 Modeling brain connectivity 130

8.1 Introduction 130
8.2 Functional connectivity 131
8.3 Effective connectivity 144
8.4 Network analysis and graph theory 155

9 Multivoxel pattern analysis and machine learning 160

9.1 Introduction to pattern classification 160
9.2 Applying classifiers to fMRI data 163
9.3 Data extraction 163
9.4 Feature selection 164
9.5 Training and testing the classifier 165
9.6 Characterizing the classifier 171

10 Visualizing, localizing, and reporting fMRI data 173

10.1 Visualizing activation data 173
10.2 Localizing activation 176
vii

10.3 Localizing and reporting activation 179

10.4 Region of interest analysis 183
Appendix A Review of the General Linear Model 191
A.1 Estimating GLM parameters 191
A.2 Hypothesis testing 194
A.3 Correlation and heterogeneous variances 195
A.4 Why “general” linear model? 197
Appendix B Data organization and management 201
B.1 Computing for fMRI analysis 201
B.2 Data organization 202
B.3 Project management 204
B.4 Scripting for data analysis 205
Appendix C Image formats 208
C.1 Data storage 208
C.2 File formats 209
Bibliography 211
Index 225
 Preface

Functional magnetic resonance imaging (fMRI) has, in less than two decades,
become the most commonly used method for the study of human brain function.
FMRI is a technique that uses magnetic resonance imaging to measure brain activity
by measuring changes in the local oxygenation of blood, which in turn reflects the
amount of local brain activity. The analysis of fMRI data is exceedingly complex,
requiring the use of sophisticated techniques from signal and image processing and
statistics in order to go from the raw data to the finished product, which is generally
a statistical map showing which brain regions responded to some particular manip-
ulation of mental or perceptual functions. There are now several software packages
available for the processing and analysis of fMRI data, several of which are freely
available.
The purpose of this book is to provide researchers with a sophisticated under-
standing of all of the techniques necessary for processing and analysis of fMRI data.
The content is organized roughly in line with the standard flow of data processing
operations, or processing stream, used in fMRI data analysis. After starting with a
general introduction to fMRI, the chapters walk through all the steps that one takes
in analyzing an fMRI dataset. We begin with an overview of basic image processing
methods, providing an introduction to the kinds of data that are used in fMRI and
how they can be transformed and filtered. We then discuss the many steps that are
used for preprocessing fMRI data, including quality control, correction for vari-
ous kinds of artifacts, and spatial smoothing, followed by a description of methods
for spatial normalization, which is the warping of data into a common anatomical
space. The next three chapters then discuss the heart of fMRI data analysis, which is
statistical modeling and inference. We separately discuss modeling data from fMRI
timeseries within an individual and modeling group data across individuals, fol-
lowed by an outline of methods for statistical inference that focuses on the severe
multiple test problem that is inherent in fMRI data. Two additional chapters focus
on methods for analyzing data that go beyond a single voxel, involving either the

ix
x Preface

modeling of connectivity between regions or the use of machine learning techniques

to model multivariate patterns in the data. The final chapter discusses approaches
for the visualization of the complex data that come out of fMRI analysis. The appen-
dices provide background about the general linear model, a practical guide to the
organization of fMRI data, and an introduction to imaging data file formats.
The intended audience for this book is individuals who want to understand fMRI
analysis at a deep conceptual level, rather than simply knowing which buttons to
push on the software package. This may include graduate students and advanced
undergraduate students, medical school students, and researchers in a broad range
of fields including psychology, neuroscience, radiology, neurology, statistics, and
bioinformatics. The book could be used in a number of types of courses, including
graduate and advanced undergraduate courses on neuroimaging as well as more
focused courses on fMRI data analysis.
We have attempted to explain the concepts in this book with a minimal amount
of mathematical notation. Some of the chapters include mathematical detail about
particular techniques, but this can generally be skipped without harm, though inter-
ested readers will find that understanding the mathematics can provide additional
insight. The reader is assumed to have a basic knowledge of statistics and linear
algebra, but we also provide background for the reader in these topics, particularly
with regard to the general linear model.
We believe that the only way to really learn about fMRI analysis is to do it. To that
end, we have provided the example datasets used in the book along with example
analysis scripts on the book’s Web site: http://www.fmri-data-analysis.org/.
Although our examples focus primarily on the FSL and SPM software packages,
we welcome developers and users of other packages to submit example scripts that
demonstrate how to analyze the data using those other packages. Another great way
to learn about fMRI analysis is to simulate data and test out different techniques. To
assist the reader in this exercise, we also provide on the Web site examples of code
that was used to create a number of the figures in the book. These examples include
MATLAB, R, and Python code, highlighting the many different ways in which one
can work with fMRI data.
The following people provided helpful comments on various chapters in the book,
for which we are very grateful: Akram Bakkour, Michael Chee, Joe Devlin, Marta
Garrido, Clark Glymour, Yaroslav Halchenko, Mark Jenkinson, Agatha Lenartowicz,
Randy McIntosh, Rajeev Raizada, Antonio Rangel, David Schnyer, and Klaas Enno
Stephan. We would also like to thank Lauren Cowles at Cambridge University Press
for her guidance and patience throughout the process of creating this book.
 1

Introduction

The goal of this book is to provide the reader with a solid background in the tech-
niques used for processing and analysis of functional magnetic resonance imaging
(fMRI) data.

1.1 A brief overview of fMRI

Since its development in the early 1990s, fMRI has taken the scientific world by
storm. This growth is easy to see from the plot of the number of papers that mention
the technique in the PubMed database of biomedical literature, shown in Figure 1.1.
Back in 1996 it was possible to sit down and read the entirety of the fMRI literature
in a week, whereas now it is barely feasible to read all of the fMRI papers that were
published in the previous week! The reason for this explosion in interest is that fMRI
provides an unprecedented ability to safely and noninvasively image brain activity
with very good spatial resolution and relatively good temporal resolution compared
to previous methods such as positron emission tomography (PET).

1.1.1 Blood flow and neuronal activity

The most common method of fMRI takes advantage of the fact that when neu-
rons in the brain become active, the amount of blood flowing through that area
is increased. This phenomenon has been known for more than 100 years, though
the mechanisms that cause it remain only partly understood. What is particularly
interesting is that the amount of blood that is sent to the area is more than is needed
to replenish the oxygen that is used by the activity of the cells. Thus, the activity-
related increase in blood flow caused by neuronal activity leads to a relative surplus
in local blood oxygen. The signal measured in fMRI depends on this change in
oxygenation and is referred to as the blood oxygenation level dependent, or bold,
signal.
1
2 Introduction

3000

Number of publications 2500

2000

1500

1000

500

0
1992 1994 1996 1998 2000 2002 2004 2006 2008
Year of publication

Figure 1.1. A plot of the number of citations in the PubMed database matching the query [“fMRI” OR
“functional MRI” OR “functional magnetic resonance imaging”] for every year since 1992.

Figure 1.2 shows an example of what is known as the hemodynamic response, which
is the increase in blood flow that follows a brief period of neuronal activity. There
are two facts about the hemodynamic response that underlie the basic features of
bold fMRI and determine how the data must be analyzed. First, the hemodynamic
response is slow; whereas neuronal activity may only last milliseconds, the increase in
blood flow that follows this activity takes about 5 seconds to reach its maximum. This
peak is followed by a long undershoot that does not fully return to baseline for at least
15–20 seconds. Second, the hemodynamic response can, to a first approximation, be
treated as a linear time-invariant system (Cohen, 1997; Boynton et al., 1996; Dale,
1999). This topic will be discussed in much greater detail in Chapter 5, but in essence
the idea is that the response to a long train of neuronal activity can be determined
by adding together shifted versions of the response to a shorter train of activity.
This linearity makes it possible to create a straightforward statistical model that
describes the timecourse of hemodynamic signals that would be expected given
some particular timecourse of neuronal activity, using the mathematical operation
of convolution.

1.1.2 Magnetic resonance imaging

The incredible capabilities of magnetic resonance imaging (MRI) can hardly be
overstated. In less than 10 minutes, it is possible to obtain images of the human
3 1.2 The emergence of cognitive neuroscience

0.8

0.6
% change in MRI signal

0.4

0.2

−0.2

−0.4
0 2 4 6 8 10 12 14 16
Peristimulus time (seconds)

Figure 1.2. An example of the hemodynamic responses evoked in area V1 by a contrast-reversing checker-
board displayed for 500 ms. The four different lines are data from four different individuals,
showing how variable these responses can be across people. The MRI signal was measured
every 250 ms, which accounts for the noisiness of the plots. (Data courtesy of Stephen Engel,
University of Minnesota)

brain that rival the quality of a postmortem examination, in a completely safe and
noninvasive way. Before the development of MRI, imaging primarily relied upon the
use of ionizing radiation (as used in X-rays, computed tomography, and positron
emission tomography). In addition to the safety concerns about radiation, none
of these techniques could provide the flexibility to image the broad range of tissue
characteristics that can be measured with MRI. Thus, the establishment of MRI as a
standard medical imaging tool in the 1980s led to a revolution in the ability to see
inside the human body.

1.2 The emergence of cognitive neuroscience

Our fascination with how the brain and mind are related is about as old as humanity
itself. Until the development of neuroimaging methods, the only way to understand
how mental function is organized in the brain was to examine the brains of indi-
viduals who had suffered damage due to stroke, infection, or injury. It was through
these kinds of studies that many early discoveries were made about the localization
of mental functions in the brain (though many of these have come into question
4 Introduction

subsequently). However, progress was limited by the many difficulties that arise in
studying brain-damaged patients (Shallice, 1988).
In order to better understand how mental functions relate to brain processes in
the normal state, researchers needed a way to image brain function while individu-
als performed mental tasks designed to manipulate specific mental processes. In the
1980s several groups of researchers (principally at Washington University in St. Louis
and the Karolinska Institute in Sweden) began to use positron emission tomography
(PET) to ask these questions. PET measures the breakdown of radioactive mate-
rials within the body. By using radioactive tracers that are attached to biologically
important molecules (such as water or glucose), it can measure aspects of brain
function such as blood flow or glucose metabolism. PET showed that it was possible
to localize mental functions in the brain, providing the first glimpses into the neural
organization of cognition in normal individuals (e.g., Posner et al., 1988). However,
the use of PET was limited due to safety concerns about radiation exposure, and due
to the scarce availability of PET systems.
fMRI provided exactly the tool that cognitive neuroscience was looking for. First,
it was safe, which meant that it could be used in a broad range of individuals, who
could be scanned repeatedly many times if necessary. It could also be used with
children, who could not take part in PET studies unless the scan was medically
necessary. Second, by the 1990s MRI systems had proliferated, such that nearly every
medical center had at least one scanner and often several. Because fMRI could be
performed on many standard MRI scanners (and today on nearly all of them), it was
accessible to many more researchers than PET had been. Finally, fMRI had some
important technical benefits over PET. In particular, its spatial resolution (i.e., its
ability to resolve small structures) was vastly better than PET. In addition, whereas
PET required scans lasting at least a minute, with fMRI it was possible to examine
events happening much more quickly. Cognitive neuroscientists around the world
quickly jumped on the bandwagon, and thus the growth spurt of fMRI began.

1.3 A brief history of fMRI analysis

When the first fMRI researchers collected their data in the early 1990s, they also
had to create the tools to analyze the data, as there was no “off-the-shelf ” software
for analysis of fMRI data. The first experimental designs and analytic approaches
were inspired by analysis of blood flow data using PET. In PET blood flow studies,
acquisition of each image takes at least one minute, and a single task is repeated
for the entire acquisition. The individual images are then compared using simple
statistical procedures such as a t-test between task and resting images. Inspired by this
approach, early studies created activation maps by simply subtracting the average
activation during one task from activation during another. For example, in the study
by Kwong et al. (1992), blocks of visual stimulation were alternated with blocks of
no stimulation. As shown in Figure 1.3, the changes in signal in the visual cortex
5 1.3 A brief history of fMRI analysis

Photic Stimulation -- IR Images

2860 off on off on

Signal Intensity
2800

2740

2680
0 70 140 210 280
Seconds

Photic Stimulation -- GE Images

6050 off on off on

Signal Intensity
5900

5750

5650
0 60 120 180 240
Seconds

Figure 1.3. Early fMRI images from Kwong et al. (1992). The left panel shows a set of images starting
with the baseline image (top left), and followed by subtraction images taken at different
points during either visual stimulation or rest. The right panel shows the timecourse of a
region of interest in visual cortex, showing signal increases that occur during periods of visual
stimulation.

were evident even from inspection of single subtraction images. In order to obtain
statistical evidence for this effect, the images acquired during the stimulation blocks
were compared to the images from the no-stimulation blocks using a simple paired
t-test. This approach provided an easy way to find activation, but its limitations
quickly became evident. First, it required long blocks of stimulation (similar to PET
scans) in order to allow the signal to reach a steady state. Although feasible, this
approach in essence wasted the increased temporal resolution available from fMRI
data. Second, the simple t-test approach did not take into account the complex
temporal structure of fMRI data, which violated the assumptions of the statistics.
Researchers soon realized that the greater temporal resolution of fMRI relative to
PET permitted the use of event-related (ER) designs, where the individual impact of
relatively brief individual stimuli could be assessed. The first such studies used trials
that were spaced very widely in time (in order to allow the hemodynamic response to
return to baseline) and averaged the responses across a time window centered around
each trial (Buckner et al., 1996). However, the limitations of such slow event-related
designs were quickly evident; in particular, it required a great amount of scan time to
collect relatively few trials. The modeling of trials that occurred more rapidly in time
required a more fundamental understanding of the bold hemodynamic response
(HRF). A set of foundational studies (Boynton et al., 1996; Vazquez & Noll, 1998;
Dale & Buckner, 1997) established the range of event-related fMRI designs for which
6 Introduction

the bold response behaved as a linear time invariant system, which was roughly for
events separated by at least 2 seconds. The linearity of the bold is a crucial result,
dramatically simplifying the analysis by allowing the use of the General Linear Model
and also allowing the study of the statistical efficiency of various fMRI designs. For
example, using linearity Dale (1999) and Josephs & Henson (1999) demonstrated
that block designs were optimally sensitive to differences between conditions, but
careful arrangement of the events could provide the best possible ER design.
The noise in bold data also was a challenge, particularly with regard to the extreme
low frequency variation referred to as “drift.” Early work systematically examined the
sources and nature of this noise and characterized it as a combination of physiological
effects and scanner instabilities (Smith et al., 1999; Zarahn et al., 1997; Aguirre et al.,
1997), though the sources of drift remain somewhat poorly understood. The drift
was modeled by a combination of filters or nuisance regressors, or using temporal
autocorrelation models (Woolrich et al., 2001). Similar to PET, global variation in
the bold signal was observed that was unrelated to the task, and there were debates
as to whether global fMRI signal intensity should be regressed out, scaled-away, or
ignored (Aguirre et al., 1997).
In PET, little distinction was made between intrasubject and group analyses, and
the repeated measures correlation that arises from multiple (at most 12) scans from
a subject was ignored. With fMRI, there are hundreds of scans for each individual.
An early approach was to simply concatenate the time series for all individuals
in a study and perform the analysis across all timepoints, ignoring the fact that
these are repeated measures obtained across different individuals. This produced
“fixed effects” inferences in which a single subject could drive significant results
in a group analysis. The SPM group (Holmes & Friston, 1999) proposed a simple
approach to “mixed effects” modeling, whose inferences would generalize to the
sampled population. Their approach involved obtaining a separate effect estimate
per subject at each voxel and then combining these at a second level to test for effects
across subjects. Though still widely in use today, this approach did not account for
differences in intrasubject variability. An improved approach was proposed by the
FMRIB Software Library (FSL) group (Woolrich et al., 2004b; Beckmann & Smith,
2004) that used both the individual subject effect images and the corresponding
standard error images. Although the latter approach provides greater sensitivity
when there are dramatic differences in variability between subjects, recent work has
shown that these approaches do not differ much in typical single-group analyses
(Mumford & Nichols, 2009).
Since 2000, a new approach to fMRI analysis has become increasingly common,
which attempts to analyze the information present in patterns of activity rather than
the response at individual voxels. Known variously as multi-voxel pattern analysis
(MVPA), pattern information analysis, or machine learning, these methods attempt
to determine the degree to which different conditions (such as different stimulus
classes) can be distinguished on the basis of fMRI activation patterns, and also
7 1.5 Software packages for fMRI analysis

to understand what kind of information is present in those patterns. A particular

innovation of this set of methods is that they focus on making predictions about new
data, rather than simply describing the patterns that exist in a particular data set.

1.4 Major components of fMRI analysis

The analysis of fMRI data is made complex by a number of factors. First, the data
are liable to a number of artifacts, such as those caused by head movement. Second,
there are a number of sources of variability in the data, including variability between
individuals and variability across time within individuals. Third, the dimensionality
of the data is very large, which causes a number of challenges in comparison to
the small datasets that many scientists are accustomed to working with. The major
components of fMRI analysis are meant to deal with each of these problems. They
include
• Quality control: Ensuring that the data are not corrupted by artifacts.
• Distortion correction: The correction of spatial distortions that often occur in
fMRI images.
• Motion correction: The realignment of scans across time to correct for head
motion.
• Slice timing correction: The correction of differences in timing across different
slices in the image.
• Spatial normalization: The alignment of data from different individuals into
a common spatial framework so that their data can be combined for a group
analysis.
• Spatial smoothing: The intentional blurring of the data in order to reduce noise.
• Temporal filtering: The filtering of the data in time to remove low-frequency
noise.
• Statistical modeling: The fitting of a statistical model to the data in order to
estimate the response to a task or stimulus.
• Statistical inference: The estimation of statistical significance of the results,
correcting for the large number of statistical tests performed across the brain.
• Visualization: Visualization of the results and estimation of effect sizes.
The goal of this book is to outline the procedures involved in each of these steps.

1.5 Software packages for fMRI analysis

In the early days of fMRI, nearly every lab had its own home-grown software package
for data analysis, and there was little consistency between the procedures across
different labs. As fMRI matured, several of these in-house software packages began
to be distributed to other laboatories, and over time several of them came to be
distributed as full-fledged analysis suites, able to perform all aspects of analysis of
an fMRI study.
8 Introduction

Table 1.1. An overview of major fMRI software packages

Package Developer Platformsa Licensing

SPM University College MATLAB Open-source

London
FSL Oxford UNIX Open source
University
AFNI NIMH UNIX Open source
Brain Brain Innovation Mac OS X, Commercial
Voyager Windows, Linux (closed-source)

a Those platform listed as UNIX are available for Linux, Mac OS X, and other UNIX flavors.

Today, there are a number of comprehensive software packages for fMRI data
analysis, each of which has a loyal following. (See Table 1.1) The Web sites for all of
these packages are linked from the book Web site.

1.5.1 SPM
SPM (which stands for Statistical Parametric Mapping) was the first widely used and
openly distributed software package for fMRI analysis. Developed by Karl Friston
and colleagues in the lab then known as the Functional Imaging Lab (or FIL) at
University College London, it started in the early 1990s as a program for analysis
of PET data and was then adapted in the mid-1990s for analysis of fMRI data.
It remains the most popular software package for fMRI analysis. SPM is built in
MATLAB, which makes it accessible on a very broad range of computer platforms.
In addition, MATLAB code is relatively readable, which makes it easy to look at
the code and see exactly what is being done by the programs. Even if one does
not use SPM as a primary analysis package, many of the MATLAB functions in
the SPM package are useful for processing data, reading and writing data files, and
other functions. SPM is also extensible through its toolbox functionality, and a large
number of extensions are available via the SPM Web site. One unique feature of
SPM is its connectivity modeling tools, including psychophysiological interaction
(Section 8.2.4) and dynamic causal modeling (Section 8.3.4). The visualization tools
available with SPM are relatively limited, and many users take advantage of other
packages for visualization.

1.5.2 FSL
FSL (which stands for FMRIB Software Library) was created by Stephen Smith
and colleagues at Oxford University, and first released in 2000. FSL has gained
substantial popularity in recent years, due to its implementation of a number of
9 1.5 Software packages for fMRI analysis

cutting-edge techniques. First, FSL has been at the forefront of statistical modeling
for fMRI data, developing and implementing a number of novel modeling, estima-
tion, and inference techniques that are implemented in their FEAT, FLAME, and
RANDOMISE modules. Second, FSL includes a robust toolbox for independent
components analysis (ICA; see Section 8.2.5.2), which has become very popular
both for artifact detection and for modeling of resting-state fMRI data. Third, FSL
includes a sophisticated set of tools for analysis of diffusion tensor imaging data,
which is used to analyze the structure of white matter. FSL includes an increasingly
powerful visualization tool called FSLView, which includes the ability to overlay a
number of probabilistic atlases and to view time series as a movie. Another major
advantage of FSL is its integration with grid computing, which allows for the use of
computing clusters to greatly speed the analysis of very large datasets.

1.5.3 AFNI
AFNI (which stands for Analysis of Functional NeuroImages) was created by Robert
Cox and his colleagues, first at the Medical College of Wisconsin and then at the
National Institutes of Mental Health. AFNI was developed during the very early days
of fMRI and has retained a loyal following. Its primary strength is in its very power-
ful and flexible visualization abilities, including the ability to integrate visualization
of volumes and cortical surfaces using the SUMA toolbox. AFNI’s statistical model-
ing and inference tools have historically been less sophisticated than those available
in SPM and FSL. However, recent work has integrated AFNI with the R statisti-
cal package, which allows use of more sophisticated modeling techniques available
within R.

1.5.4 Other important software packages

BrainVoyager. Brain Voyager, produced by Rainer Goebel and colleagues at Brain
Innovation, is the major commercial software package for fMRI analysis. It is
available for all major computing platforms and is particularly known for its ease
of use and refined user interface.
FreeSurfer. FreeSurfer is a package for anatomical MRI analysis developed by Bruce
Fischl and colleagues at the Massachusetts General Hospital. Even though it is
not an fMRI analysis package per se, it has become increasingly useful for fMRI
analysis because it provides the means to automatically generate both cortical
surface models and anatomical parcellations with a minimum of human input.
These models can then be used to align data across subjects using surface-based
approaches, which may in some cases be more accurate than the more standard
volume-based methods for intersubject alignment (see Chapter 4). It is possi-
ble to import statistical results obtained using FSL or SPM and project them
onto the reconstructed cortical surface, allowing surface-based group statistical
analysis.
10 Introduction

1.6 Choosing a software package

Given the variety of software packages available for fMRI analysis, how can one
choose among them? One way is to listen to the authors of this book, who have each
used a number of packages and eventually have chosen FSL as their primary analysis
package, although we each use other packages regularly as well. However, there are
other reasons that one might want to choose one package over the others. First, what
package do other experienced researchers at your institution use? Although mailing
lists can be helpful, there is no substitute for local expertise when one is learning a
new analysis package. Second, what particular aspects of analysis are most important
to you? For example, if you are intent on using dynamic causal modeling, then SPM
is the logical choice. If you are interested in using ICA, then FSL is a more appropriate
choice. Finally, it depends upon your computing platform. If you are a dedicated
Microsoft Windows user, then SPM is a good choice (though it is always possible
to install Linux on the same machine, which opens up many more possibilities). If
you have access to a large cluster, then you should consider FSL, given its built-in
support for grid computing.
It is certainly possible to mix and match analysis tools for different portions of
the processing stream. This has been made increasingly easy by the broad adoption
of the NIfTI file format by most of the major software packages (see Appendix C
for more on this). However, in general it makes sense to stick largely with a single
package, if only because it reduces the amount of emails one has to read from the
different software mailing lists!

1.7 Overview of processing streams

We refer to the sequence of operations performed in course of fMRI analysis as a
processing stream. Figure 1.4 provides a flowchart depicting some common process-
ing streams. The canonical processing streams differ somewhat between different
software packages; for example, in SPM spatial normalization is usually applied
prior to statistical analysis, whereas in FSL it is applied to the results from the
statistical analysis. However, the major pieces are the same across most packages.

1.8 Prerequisites for fMRI analysis

Research into the development of expertise suggests that it takes about ten years to
become expert in any field (Ericsson et al., 1993), and fMRI analysis is no different,
particularly because it requires a very broad range of knowledge and skills. However,
the new researcher has to start somewhere. Here, we outline the basic areas of
knowledge that we think are essential to becoming an expert at fMRI analysis, roughly
in order of importance.
11 1.8 Prerequisites for fMRI analysis

Image Distortion Motion Slice Timing

Reconstruction Correction Correction Correction

Spatial Statistical Spatial

FSL/AFNI Smoothing
Normalization Analysis

Statistical Spatial Spatial

SPM
Analysis Smoothing Normalization

Figure 1.4. A depiction of common processing streams for fMRI data analysis.

1. Probability and statistics. There is probably no more important foundation for

fMRI analysis than a solid background in basic probability and statistics. With-
out this, nearly all of the concepts that are central to fMRI analysis will be
foreign.
2. Computer programming. It is our opinion that one simply cannot become an
effective user of fMRI analysis without strong computer programming skills.
There are many languages that are useful for fMRI analysis, including MATLAB,
Python, and UNIX shell scripting. The particular language is less important than
an underlying understanding of the methods of programming, and this is a place
where practice really does make perfect, particularly with regard to debugging
programs when things go wrong.
3. Linear algebra. The importance of linear algebra extends across many different
aspects of fMRI analysis, from statistics (where the general linear model is most
profitably defined in terms of linear algebra) to image processing (where many
operations on images are performed using linear algebra). A deep understanding
of fMRI analysis requires basic knowledge of linear algebra.
4. Magnetic resonance imaging. One can certainly analyze fMRI data without know-
ing the details of MRI acquisition, but a full understanding of fMRI data requires
that one understand where the data come from and what they are actually mea-
suring. This is particularly true when it comes to understanding the various ways
in which MRI artifacts may affect data analysis.
5. Neurophysiology and biophysics. fMRI signals are interesting because they are
an indirect measure of the activity of individual neurons. Understanding how
12 Introduction

neurons code information, and how these signals are reflected in blood flow, is
crucial to interpreting the results that are obtained from fMRI analysis.
6. Signal and image processing. A basic understanding of signal and image processing
methods is important for many of the techniques discussed in this book. In
particular, an understanding of Fourier analysis (Section 2.4) is very useful for
nearly every aspect of fMRI analysis.
 2

Image processing basics

Many of the operations that are performed on fMRI data involve transforming
images. In this chapter, we provide an overview of the basic image processing
operations that are important for many different aspects of fMRI data analysis.

2.1 What is an image?

At its most basic, a digital image is a matrix of numbers that correspond to spatial
locations. When we view an image, we do so by representing the numbers in the
image in terms of gray values (as is common for anatomical MRI images such as in
Figure 2.1) or color values (as is common for statistical parametric maps). We gen-
erally refer to each element in the image as a “voxel,” which is the three-dimensional
analog to a pixel. When we “process” an image, we are generally performing some
kind of mathematical operation on the matrix. For example, an operation that
makes the image brighter (i.e., whiter) corresponds to increasing the values in the
matrix.
In a computer, images are represented as binary data, which means that the
representation takes the form of ones and zeros, rather than being represented in a
more familiar form such as numbers in plain text or in a spreadsheet. Larger numbers
are represented by combining these ones and zeros; a more detailed description of
this process is presented in Box 2.1.

Numeric formats. The most important implication of numeric representation is

that information can be lost if the representation is not appropriate. For example,
imagine that we take a raw MRI image that has integer values that range between
1,000 and 10,000, and we divide each voxel value by 100, resulting in a new image
with values ranging between 10 and 100. If the resulting image is stored using
floating point values, then all of the original information will be retained; that is, if
there were 9,000 unique values in the original image, there will also be 9,000 unique

13
14 Image processing basics

288 27 38 364 621

264 21 97 500 640

271 22 133 543 647

312 28 113 521 649

390 53 58 424 635

Figure 2.1. An image as a graphical representation of a matrix. The gray scale values in the image at left
correspond to numbers, which is shown for a set of particular voxels in the closeup section
on the right.

values in the new image (3,280 becomes 32.8, and so on). If the resulting image
is stored instead as integers (like the original image), then information is lost: The
9,000 unique values between 1,000 and 10,000 will be replaced by only 90 unique
values in the new image, which means that information has been lost when values
are rounded to the nearest integer. The tradeoff for using floating point numbers is
that they will result in larger image files (see Box 2.1).

Metadata. Beyond the values of each voxel, it is also critical to store other informa-
tion about the image, known generally as metadata. These data are generally stored
in a header, which can either be a separate file or a part of the image file. There are
a number of different types of formats that store this information, such as Analyze,
NIfTI, and DICOM. The details of these file formats are important but tangential to
our main discussion; the interested reader is directed to Appendix C.

Storing time series data. Whereas structural MRI images generally comprise a
single three-dimensional image, fMRI data are represented as a time series of three-
dimensional images. For example, we might collect an image every 2 seconds for
a total of 6 minutes, resulting in a time series of 180 three-dimensional images.
Some file formats allow representation of four-dimensional datasets, in which case
this entire time series could be saved in a single data file, with time as the fourth
dimension. Other formats require that the time series be stored as a series of separate
three-dimensional datafiles.
 Another Random Document on
Scribd Without Any Related Topics
 The manifold representations of the Palestinian 'good goddess' extend
over a lengthy period, and vary in taste and nuance from the crudest of
specimens to veritable artistic products of the Seleucid age. They indicate
that the fundamental religious conceptions agreed with those of Western
Asia as a whole, and it may be assumed that the conclusions which can be
drawn from the figurines and plaques of this deity would apply, mutatis
mutandis, to others.

Among other objects which hardly belong to public cult, but were
probably for household or private use, may be noticed the small idols; e.g.
one from Megiddo in the clumsy 'snow-man' technique, another from
Jericho with the head of a bull. Numerous small phalli have also been
unearthed. Some are roughly carved in human shape, others approximate
the form of a fish. They do not necessarily belong to the cult of any male
deity, but the true significance of these and other small emblems is often
uncertain. As with the many small models of the heads of bull, cow, or
serpent, or the two small conical stones from the temple at Serabit, each
with a groove along the base, it is often difficult to distinguish the fetishes
and symbols, which involve ideas of some relationship with a supernatural
being, from the charms, amulets, and talismans, wherein other religious
ideas are involved. The possibility that some of the objects are really toys
cannot be excluded.

CHAPTER IV

SACRED RITES AND PRACTICES

General Inferences.—That the old places of cult had their duly

ordained officials may be taken for granted; even the smallest of them, like
those of to-day, must have had appointed attendants. The Amarna letters
mention the wealthy temple of Byblos with the handmaidens of the goddess
of the city, and in Merneptah's reign we hear of a man of Gaza who is
described as a servant of Baal. We may be sure, also, that the rites and
festivals were similar to those usually prevalent among agricultural peoples.
The nature-worship of the age can be realised from a survey of the old cults
of Western Asia, and from the denunciations of the Old Testament, which
prove the persistence of older licentious rites. Popular religion often
continues to tolerate practices which social life condemns, and the fertility
of crops, cattle, and of man himself, was co-ordinated by an uncontrollable
use of analogy in which the example was set by the 'sacred' men and
women of the sanctuaries (kādesh; Deut. xxiii. 17, R.V. marg.). Sympathetic
magic—the imitation of the cause to produce a desired effect—underlay a
variety of rites among a people whose life depended upon the gifts of the
soil, whose religion was a way of life. Here, however, we are restricted
chiefly to some miscellaneous evidence which the excavations suggest.

The Disposal of the Dead.—Incineration or cremation had been

originally practised by a people physically distinct from that among whom
inhumation prevailed. The latter innovation has been ascribed to the
invading Semites. Subsequently, in Carthage, cremation is found to re-enter,
presumably through foreign influence; but the two practices co-exist, even
in the same family, and it is probable that there, at all events, cremation was
only followed in special circumstances. A large burial-cave at Gezer with a
thick layer of burnt ash proves the lengthy duration of the earlier custom.
The same cave was afterwards utilised by those who inhumed their dead,
and thenceforth there is little evolution in the history of early Palestinian
burial. No particular orientation predominates; the dead are placed upon a
layer of stones, or within cists, or in pits in the floor of the caverns. Both the
contracted or squatting and the outstretched attitude occur. From the story
of Sinuhe (p. 9), it would seem that burial in a sheep-skin was also
customary. The needs of the dead are supplied by vessels of food, which
occasionally show traces of burning; drink was more important, and the
large jars sometimes contain small cups for the convenience of the thirsty
soul. In the case of a jug with two mammillary projections one is reminded
of a type usually associated at Carthage with the burial of infants. A variety
of miscellaneous objects provided for other needs: weapons, jewels, ostrich
eggs, seals, scarabs, amulets, small figures in human or animal form, etc.
Especially characteristic of the later tombs are the abundant deposits of
lamps.

The abode of the dead being one of the centres of the religion of the
living, the tomb always possesses sanctity. The internal arrangements, with
platforms or hewn benches, will often suggest some burial-ritual. The cup-
marks, which frequently appear near or even in the tomb itself, like those
still to be seen upon Palestinian dolmens, could serve for sacrifices or
libations, or to collect the refreshing rain for the soul of the deceased. Or,
again, later usage will suggest that they were planted with flowers which,
like the 'Gardens of Adonis,' symbolised the mysteries of death and revival.
Often, the dead are buried beneath the streets (if the narrow windings
deserve that name), or within the houses, under circumstances which
preclude the foundation-sacrifices to be noticed presently. This feature is
scarcely accidental; it is well known elsewhere, and was probably intended
to keep the spirit of the dead near its former abode, over which it could
continue to exercise a benevolent influence.

Jar Burial.—It is sometimes difficult to distinguish between an ordinary

burial and some sacrificial ceremony. The burial of new-born or very young
infants in jars, in or near some sacred locality (p. 16 sq.), points very
strongly to the sacrifice of the first-born to which the Old Testament bears
witness (Micah vi. 7). But where the circumstances make this view less
probable, the special treatment of those who died in early infancy needs
consideration. In inhumation and the return of the dead to the ground we are
in the midst of ideas associated with 'mother-earth,' the begetter of all
things. The burial in a contracted or squatting position might naturally
represent the usual crouched posture of the individual as he sat in life-time
among his fellows; it might also point to a belief in the re-birth of the soul
of the dead. The jar-burials, where the infant is inserted head downwards,
are more suggestive of the latter, and evidence from Africa and Asia shows
that provision is sometimes made for the re-birth of still-born or very young
babes on the conviction that at some future occasion they will enter again
into a mother's womb. The numerous emblems of nature-worship and the
mother-goddess, especially at Gezer, raise the presumption that the deities
of the place were powers of fertility and generation; and, just as the shrines
of saints to-day are visited by would-be mothers who hope for offspring, it
is not improbable that in olden times those who had been prematurely cut
off from the living were interred in sacred sites venerated by the women.
This view, which has been proposed by Dr. J. G. Frazer, will not apply of
course to those jar-burials where human-sacrifice is clearly recognisable.[1]

[1] Adonis, Attis, Osiris, pp. 77 sq., 82 sq..

Human Sacrifice.—A gruesome discovery was made in a cistern at

Gezer where, together with a number of adult skeletons, lay the upper half
of a young girl about sixteen years of age. Near the mouth were the
decapitated heads of two girls. In another case at Gezer (described as a
'foundation deposit') the upper half of the skeleton of a youth had been
placed with two adults. Perhaps we should here include the cases where
only a few bones of the deceased were preserved, e.g. in one tomb the skull
and certain other bones were missing. Vessels, also, were found containing
only one or two human bones: the patella of an adult, the calvaria of a skull;
but in the majority of instances they belonged to infants. Partial burial of
this character has been explained on the theory of cannibalism; this
practice, often based on the idea of absorbing the attributes of the deceased,
has left scattered traces among the Semites. But the dismemberment of the
dead (known at Susa, Egypt, and common to many savage races) admits of
other explanations, whether, for example, we observe the use of bones as
amulets (p. 51 sq.), or recall the story of the severed Osiris. In the latter,
however, it may be suspected that a sacrifice for magical purposes underlies
an aetiological legend.[2] The bank of skulls south of the monoliths of
Gezer (p. 16) may perhaps recall the mound (or pillar?) of heads which
certain Assyrian kings erected in front of the cities they conquered (e.g.
Ashur-nasir-pal I.). Such a deed, like their holocaust of children after a
victory, was no unmeaning ferocity; religion entered profoundly into
ancient life, and every war was a 'holy war.' The horrid rites in honour of
the gods who fought for their followers are to be traced in Egypt, Assyria,
and the Old Testament, and even as late as 307 B.C. the Carthaginians after
their defeat of Agathocles slew the choicest prisoners 'before the altar in
front of the holy tent.'

[2] J. G. Frazer, Adonis, etc., pp. 273 sq., 321, and especially 331 sqq.
Here one may perhaps refer to the tradition that the prophet Isaiah was
sawn in half, hidden as he was in a tree (comp. also Ep. Hebrews, xi.
37).

The widespread custom of Foundation Sacrifice survives in Palestine

when popular opinion requires that blood shall be shed at the inauguration
of every important building, at the breaking-up of unoccupied land, or at the
opening of a new well. Thus, a sheep was sacrificed at the building of a
jetty for the landing of the German Emperor at Haifa in 1898. The rite is a
propitiation to the numen of the place.

Mohammed in his day tried to prohibit such sacrifices to the jinn, but the
inveterate sentiment is summed up in the words of a modern native: 'every
house must have its death, either man, woman, child, or animal.' The
animal-victim is recognised as a substitute, and vulgar superstition still
associates with the foundation of buildings some vague danger to human
life—if not its loss. Traditions of human sacrifice are recorded by mediæval
and older writers, and excavation has disclosed authentic examples. At
Gezer the skeleton of an adult female had been placed under the corner of a
house, and the bones of infants were often found in or under the walls of
houses down to the later Israelite period. At Megiddo, a young girl of about
fifteen was laid across a foundation-stone, and a victim at the foot of a
tower in Taanach was a child scarcely in its teens. A jar with the remains of
a new-born infant rested upon a platform in the Gezer crematorium, and the
evidence allowed the inference that it was a dedicatory sacrifice when the
cave was taken over and used for inhumation. Infants buried in jars were
found, together with bowls and lamps, under the foundations in Gezer as
late as the latter part of the Israelite monarchy, although a modification had
already been introduced in the simple deposits of lamps and bowls, usually
at the corners of houses or chambers or under the jambs of doors. If the
bowls represent the sacrificial offerings, the significance of the lamps is
uncertain. The victim in the rite had not been burned, but probably buried
alive, and it may be conjectured that the identification of life and light
(familiar from the Old Testament) underlies the symbolical lamp. The
modern Palestinian custom of hanging lights in shrines, etc., in cases of
sickness possibly involves the same association of ideas. On the other hand,
the lamps found in tombs naturally recall the widespread custom of lighting
the soul on its dark journey, or of kindling a lamp in the home to enable it to
retrace its steps on the anniversary. These purely burial lamps are very well
known (e.g. in Carthage), and they survive in Palestine to the Christian age,
when they are inscribed with such distinctive mottoes as 'Christ is my light,'
or 'the light of Christ shines for all.'

The Importance of Sacrifice makes itself felt at every sacred site from
the enormous quantities of burnt ash before the caves of Serabit to the
similar accumulations upon the summit of Mount Hermon. The worshipper
believes that the rite brings him into contact with the powers who are to be
nourished, invoked, or recompensed. Its prevalence vividly indicates man's
dependence upon them throughout the seasons of the year and on the great
occasions of life: birth, circumcision (already practised in our period),
marriage and death. Underlying the sacrifice is the profound significance of
blood. It is the seat of existence; it has potent virtues whether for protection,
expiation, or purification; and the utmost care is taken to dispose of it
according to established usage. The fat, too, has no less its living qualities,
and since the oldest unguents were animal fats—modern usage is often
content with butter—it is probable that anointing originally had a deeper
meaning than would at first appear. Wanton bloodshed called for
vengeance, and when a Babylonian king demanded that Ikhnaton should
slay the Canaanites who had killed his merchants, and thus 'bring back their
blood' and prevent retaliation, the inveterate blood-revenge of primitive
social life finds an early illustration. But as a sacrifice, the slaughter of
human victims, though perhaps not regular, was at least not exceptional,
and the frightful bloodshed which the Old Testament attests emphasises the
difficulties which confronted those teachers of Israel who would
disassociate their national God from an inveterate practice (Ezek. xvi. 20
sq., xx. 31).

For a striking illustration of the diffusion and persistence of human

sacrifice we may refer to Carthage where the distress caused by Agathocles
in 310 B.C. was attributed to the wrath of the god to whom the rich had
been offering purchased children instead of their own. But there is a general
tendency in religion to soften crude rites, save when a particularly
efficacious offering is felt necessary in the midst of some grave crisis, and
of the changes in that background of cult which has survived throughout the
history of Palestine, the substitution of the animal for the human victim is
the most significant. Yet, as we have seen, the idea of human sacrifice has
not entirely disappeared (p. 40). The animal is still recognised as a 'ransom,'
and in the present rite of that name loss of human life is averted by the
sacrifice of some animal, and it is explained that the sacrifice will combat
and overcome the cause of the impending danger. It would be only logical,
therefore, to proceed on the assumption that the greater the danger the more
powerful and efficacious must be the sacrifice. Current beliefs thus afford
suggestive hints for earlier usage, and when we learn that to-day a natural
death finds consolation in the thought that it may have been the ransom for
another, we meet with an idea that could be put into practice: it is no great
step to the ceremonies (observed in Africa) which give effect to the
conviction that a man's life may be prolonged or his old age recuperated by
the actual sacrifice of another human being. It is essentially the same idea
when Egyptian kings, like Amenhotep II. and Ramses II. slew the prisoners
of war that they themselves or their name 'might live for ever.' (On the
name, see below, p. 60.)

Sacrificial rites were never irrational, however difficult it may be to

perceive their object, and from a survey of comparative custom one can
sometimes picture the scenes by which they were accompanied. It is only
by such means that one can conjecturally explain the discovery near the
temple-area at Gezer of animal bones, sliced, hacked, and broken into
fragments, with no signs of having been cooked. One is tempted to refer to
a rite practised by the Arabs of the Sinaitic desert towards the close of the
fourth century A.D. The old ascete, Nilus, describes a solemn procession of
chanting worshippers who move around an altar of rude stones upon which
is bound a camel. The beast is stabbed, and the leader drinks of the gushing
blood. At once the assembly hack the victim to pieces, devouring it raw
until the whole is consumed—the entire ceremony begins with the rise of
the morning star (in whose honour it was performed) and ends with the
rising sun. Was some rite of this kind practised in Palestine? It must be a
matter for conjecture; the least that can be said is that the scene is not too
barbaric for our land and period.

Broken Offerings, e.g. figurines, models, and other articles, when found
deposited in tombs, have been explained in the light of comparative custom
as destroyed or 'killed' to the end that their 'soul' may accompany that of the
deceased. But other ideas are evidently involved when the area of the
sanctuary at Serabit proved to be covered with a mass of pottery, plaques,
bracelets, wands, sistra, etc., so fragmentary that no single specimen could
be pieced together. At Gezer, also, although the plaques of the goddess were
fairly tough, all had been broken, and apparently with intention. We may
compare the modern custom of breaking pottery in fulfilment of a vow, an
interesting illustration of which was furnished by the late Professor Curtiss
from Bludan on the road from Zebedany to Damascus. At a spot, familiarly
known as the 'mother of pieces,' is a rock-platform with cave, shrine, sacred
grove and hereditary ministers. Hither come the women to break a jar when
they have gained their one wish, and it is singular to observe that the
traditions which are attached to the custom include the belief that a girl, the
patroness of the shrine, lies buried there. The likeness to the suggested rites
at Gezer will be noticed (p. 37). But the stories do not elucidate the peculiar
treatment of the offerings, and the usage finds its most probable explanation
in the persuasion that things once dedicated or put to a sacred use are 'holy,'
and cannot be used for ordinary purposes. We touch upon a fundamental
institution embodying a series of apparently paradoxical ideas—the
universal 'tabu.'

'Holy' and 'Unclean.'—The terms Holy or Sacred (comp. the Latin

sacer) are not to be understood in the ethical or moral sense. A holy thing is
one which has been set aside, dedicated, or restricted; it is charged with
supernatural influence which is contagious; everything that comes in
contact with it also becomes holy. In some cases it is provided that this
inconvenient sanctity may be purged; in others, the thing has to be
destroyed. When the Talmud says that a Canonical Book of the Old
Testament 'defiles' the hand, it means that the very sanctity of the book
demands that the hand should be ceremonially purified or cleansed before
touching anything else. 'Holy and unclean things,' to quote Robertson
Smith, 'have this in common, that in both cases certain restrictions lie on
men's use of and contact with them, and that the breach of these restrictions
involves supernatural dangers. The difference between the two appears, not
in their relation to man's ordinary life, but in their relation to the gods. Holy
things are not free to man, because they pertain to the gods; uncleanness is
shunned, according to the view taken in the higher Semitic religions,
because it is hateful to the god, and therefore not to be tolerated.'

Sacred Animals, in the light of the above, are those associated with cults
which might be regarded as illegitimate. An example is afforded by the pig
which enters into the rites and myths of Adonis, Attis, Ninib, and Osiris. In
a cavern south of the monoliths of Gezer a number of pig-bones lay
underneath a shaft which led to the cup-marked surface above (p. 16); the
circumstances recall the Thesmophoria, the caves and vaults in the Greek
area connected with Demeter and Proserpine, and the use of the pig in
mystic rites of chthonic and agricultural deities. In Palestine and Syria the
animal was used in certain exceptional sacrifices which were recognised as
idolatrous (Isaiah lxv. 4; lxvi. 17), and it was an open question whether it
was really polluted or holy. If, as the excavations suggest, the sacrifice of
the swine dates from the earliest inhabitants of Gezer, with whom it was
also a domestic animal, it is interesting to observe the persistence of its
character as a proper sacrificial animal from pre-Semitic times by the side
of the apparently contradictory belief that it was also unclean.

The camel bones at Tell es-Sāfy, also, are of interest since Robertson
Smith has shown that the animal (which became 'unclean' to the Israelites),
though used by the Arabs for food and sacrifice, was associated with ideas
of sanctity, and its flesh was forbidden to converts to Christianity. The
model of a bronze cobra found in a temple-enclosure (p. 15) might be
conjecturally explained, but it will suffice to remember that serpents were
and still are connected with spirits both benevolent and malevolent. The
recurrence of models of the animal-world, the numerous representations
upon seals of deer, gazelles, etc. (animals connected with Astarte), or the
predilection for the lion upon objects discovered at Megiddo need not have
any specific meaning for the religious ideas. On the other hand, the animal-
like attributes which appear upon some plaques of the mother-goddess are
scarcely meaningless. There is no ground for the assumption that Palestine
was without the animal-deities and the deities with special sacred animals,
which have left their traces in the surrounding lands, and it would be
misleading to suppose that the myths and legends which have grown up
around these features account for their origin. The conviction that man was
made in the likeness of the gods (who are therefore anthropomorphic)
implies certain conceptions of their nature, the development of which
belongs to the history of religion, and in turning next to the spirit-world of
Ancient Palestine it is necessary that we should be prepared to appreciate a
mental outlook profoundly different from our own.

CHAPTER V

THE WORLD OF SPIRITS

Awe.—A fundamental sense of awe was felt in the presence of anything

unusual or contrary to experience, and man's instinctive philosophy shaped
his ideas from the suggestions of daily life, accounting for all cases of
causation by assimilating them to the intentional acts of voluntary agents
like himself. There was no doubt of the existence and influence of
surrounding unseen powers; they must be cajoled, appeased, bribed and
rewarded. Some were inevitably malevolent; with others man could enter
into relations which were mutually beneficial. Even at the present day there
is no clear distinction between what we should call the natural and the
supernatural; a demon or a saint can appear in human or animal form; and
the marvel or miracle is that which happens to lie outside the intellectual
horizon of the individual. The modern phenomena can be traced back
through early sources and appear now in grosser and now in more elevated
forms; even the presence of any advanced material culture, or of more
spiritual conceptions of the Godhead does not annihilate that lower
supernaturalism which flourishes uncontrolled among more rudimentary
races. It would be unreasonable to suppose that the religion of our period
was more free from imprecision than that of more progressive peoples: the
whole routine of life brought the individual into constant contact with
unseen agencies, and the world of spirits involved a medley of beliefs, more
embarrassing to the modern inquirer who seeks to systematise them, than to
the Oriental mind which has always been able and willing to accept the
incredible and the contradictory.

Man's relations with the spirits whom he shuns or seeks are illustrated in
magical practices; e.g. incantation, symbolic magic (p. 34). Charms, on the
other hand, possess a magical virtue which is effective without interference
on the part of the possessor. Many little objects of this character have been
unearthed: pendants of red coral (still a prophylactic against the evil eye),
beads (still supposed to possess curative properties), small articles cut out
of bone (especially the heads of human femora, sawn off and perforated).
Here may be included the occasional jewels (e.g. a silver pendant crescent)
—amulets and ornaments were closely associated, and the latter continue to
convey ideas which could be regarded as idolatrous (compare Gen. xxxv.
4). The representations of Egyptian gods and the 'Horus-eyes' should also
be mentioned here. 'Eyes' are still on sale in the East, they are expected to
be on the watch for evil influences. But the anxiety to avert evil and to
procure favour need not involve an intelligent interest in the means
employed, and some of the objects (when not originally possessed by
Egyptian settlers) may have as much bearing upon the question of Egyptian
influence upon the religion of Palestine as the use of foreign (Phoenician?)
formulæ in Egyptian magical texts.
 Oracles are obtained at those places where supernatural beings have
manifested themselves, or from their symbols or their human
representatives. In the stone enclosures at Serabit Professor Petrie would
recognise the sacred places visited by those who worked the mines and
hoped for useful dreams. The value attached to visions of the night needs no
telling, and when the Egyptian king Merneptah saw in his sleep the god
Ptah offering him the sword of victory, or when the god Ashur directed the
Lydian Gyges to 'lay hold of the feet' of Ashurbanipal (i.e. place himself
under his protection), we perceive among relatively advanced societies
important factors in the growth of all religions. Divine advice and help
could be granted by the statues of the gods: a cuneiform tablet from
Taanach refers to an omen given by the finger of the goddess Ashirat, and
the writer asks for the sign and its interpretation. As in the 'nodding' of the
gods in Egyptian records the modus operandi must not be too closely
examined. Some of the old caverns of Palestine were certainly used for
magical or religious purposes, and when we find them connected by small
and curved passages, it is not improbable that they were the scenes of
oracles, theophanies, and the like (p. 15 sq.). As Mr. Macalister has
observed, apropos of such caverns in the lowlands of Judah and at Gezer,
mysterious responses and wonders could be easily contrived, and would be
as convincing to the ignorant as the Miracle of the Holy Fire is to the
modern Russian pilgrim in the Church of the Holy Sepulchre at Jerusalem.
The association of caves and other hidden resorts with the worship of
deities and oracles is well-known in other fields (e.g. Greece). Susa also had
a god of oracles who dwelt in secret retreats, and other deities whose
remote haunts were burned by Ashurbanipal when he carried them off.

The representatives of the supernatural powers include prophets,

priests, and even kings; they are also the possessors of supernatural
qualities, the one involved the other. Between the modern Palestinian
majnūn ('possessed by the jinn') and fakir, and the prophet of old—
contemptuously called 'madman'—the difference is one of degree. The
frenzied utterer is capable of incalculable good or harm, and often enjoys a
respect out of all keeping with his merits. His very sanctity places him in a
class by himself, and he is allowed a licence which would not be tolerated
in others. An early example of inspiration appears in the story of Wenamon
of Egypt who visited Zakarbaal of Byblos, probably in the reign of Ramses
XII. (about 1100 B.C.). Although the envoy had with him the statue of the
great Egyptian god Amon, for nineteen days he received scant courtesy and
was unable to obtain the desired interview. At length, as the king was
sacrificing to his gods, one of his noble youths was seized with ecstasy
which lasted the whole night, and in this state he demanded that 'the
messenger of Amon' be summoned (for the sequel see below, p. 74 sq.).
Prophecy, as Dr. Frazer has shown, by means of numerous examples, is 'a
phenomenon of almost world-wide occurrence,' and it is important to
remember that the relations between man and the spirit-world are not to be
estimated in the light of modern preconceptions. There were orthodox and
unorthodox relations, legitimate and illegitimate communion, true and false
representatives of the supernatural powers; distinctions were maintained
although the evidence is often insufficient for us to appreciate older
standpoints. Broadly speaking, it may be affirmed that the test lay in the
communal aspect of religion (whether of clan, tribe, or people) which was
opposed to practices which were private or independent of the official cult.

The dead, in their turn, depart into the mysterious unseen which looms
so largely in the thoughts of the living, and burial and mourning rites are
shaped by many different principles depending upon theories of the nature
of spirits, affection for the dead, the safety of his soul, fear of malignant
influences, etc. But the interpretation of the religious rites which attended
every crisis in life becomes unusually difficult when the community suffer a
loss, and perhaps no other study stands so much in need of careful
'comparative' treatment. Unfortunately Palestine has furnished no funerary
texts, and little direct evidence; the dead 'go to their fate,' the king of
Mitanni fasts on the day he hears of the death of Amenhotep III., and
Zakarbaal of Byblos offers to show Wenamon the tomb where the members
of a former embassy sleep (lit. lie, or pass the night). A people accustomed
to the annual death and revival of nature might easily formulate theories of
the survival of the dead, and care is accordingly taken to provide for the
needs of the deceased (p. 35). But the same thoughts are not necessarily
symbolised by the same rites. Thus, cremation, the earlier custom, may
have been intended to sever the soul from the body, to destroy the haunting
spirit, or to prevent contamination and contagion. However, the subsequent
use of the Gezer crematorium by those who practised inhumation involved
a continuity of thought, albeit with some adaptation and adjustment, since
identical conceptions of death and the dead scarcely encircled the two
distinct customs. This is instructive for the growth of complex ideas, and
the subsequent prohibition in Palestine of certain mourning rites may find a
probable explanation in their association with cults which were regarded as
illegitimate.

The attitude of the living towards the dead raises the problem of
ancestor-worship and the relation between deified ancestors and gods. In
the absence of contemporary evidence from Ancient Palestine, we may
notice the inscription of King Panammu of North Syria (eighth century),
where he acknowledges his indebtedness to his gods, especially Hadad, to
whose honour he erects a colossal statue of the deity. The text invokes the
god's blessing upon the successor to the throne, provided that the latter
when he sacrifices makes mention of Panammu's soul with Hadad or prays
that Panammu's soul may eat and drink with the god. Should these duties be
neglected, Hadad is besought not to accept the sacrifices, to refuse his
requests; and sleeplessness and other troubles are called down upon the
unfilial descendant. It appears from this, therefore, that while the dead relies
upon the attentions of the living, and it was necessary that his name should
be kept fresh; the dead could only exert an indirect influence, and the soul
or vital principle, apart from the body, could be regarded as potent only
through its companionship with the deity. This may be supplemented from
Egypt in the account of the relations between Ramses II. and his dead
father, Sety I. The latter is reminded of the benefits which his son had
conferred upon him, his statue, and his ka or vital force. These he may still
continue to enjoy, and, since he now has the companionship of the gods,
Ramses beseeches him to influence them to grant him a long reign. The
deceased king acknowledges the bread and water which had been regularly
offered to him; and relates that he has become a god more beautiful than
before; he now mingles with the great gods, and he declares that he has
successfully interceded on his son's behalf.
 The dead relied upon his descendants and upon the benevolence of
future generations, and Egyptian kings (at least) hoped to partake of the
food offered to the recognised deities. Religious and other works were
undertaken that the 'name' might 'live.' Promises and threats were freely
made to ensure due attention, and were usually respected by the living; but
the frequent acts of desecration would indicate that fear of the dead was not
necessarily a predominating or lasting feeling, at all events outside a man's
own family. The above-mentioned Panammu and Ramses are somewhat
exceptional cases since individuals, distinguished by rank, sanctity, or even
more ordinary qualifications, readily acquire distinguished positions in
after-life. Moreover, Ramses, at all events, was already a god, in his life-
time, in accordance with Egyptian belief, and all those who had had the
advantage of being representatives of the supernatural powers scarcely lost
this relative superiority. The protection afforded by famous tombs and the
virtues of the dust taken from such sacred spots are recognised to the
present day. The venerated shrines regularly found their justification in the
traditions which encircled the illustrious occupant: to violate them was not
merely an insult, it struck a blow at one of the centres of cult and prosperity.
Unfortunately for the problem, by the side of the tendency to elevate an
illustrious ancestor must be placed the very human and inveterate weakness
of tracing for oneself a noble ancestry. Like the claim of the modern
Palestinian peasant to be descended from the alleged occupant of the local
shrine which he venerates, every apparent case of ancestor-worship stands
in need of a critical examination. As in most problems of religion,
ambiguity of terminology (viz. 'worship') is responsible for much
confusion. It must be admitted that there would be a natural inclination for
every individual to regard his dead ancestor in the spirit-world as more
powerful and influential than himself. If this were so even when there were
recognised gods, it is obvious that allowance must be made for the crucial
stages, before the deities gained that recognition, and after they had lost it.

Space prevents any adequate reference to the part which animism has
held in the history of Palestinian religion; without a recognition of this
fundamental factor in all religions much of our evidence would be
unintelligible.[1] When we take the ideas which are associated with the
name, we find that it has magical powers, its use enlists or confers
protection or possession; it is the nature or essence of the thing which bears
it—indeed, almost identical with it (comp. Is. xxx. 27). Hence the meaning
of names is always instructive. The supposition that the child who bears an
animal-name will acquire something of the quality of the animal in question
(whatever be the original motive) preserves more than metaphor, and
indicates a stage when man saw little difference between animals and
himself. Even at the present day it is still believed that the soul of an
ancestor can reappear in an animal (comp. p. 50). In like manner, the
personal names of our period which denote kinship with a deity point to a
belief in a physical relationship as natural as the conviction of the modern
native when he refers to Allah in terms which imply that man is in every
detail the literal image of the Almighty. A difference between human and
superhuman is scarcely recognised at the present day. The women of the
land continue to visit the holy sites to obtain offspring, and it is freely
acknowledged that welis and spirits of the dead can be physical fathers.
This absence of any clear dividing-line between natural and supernatural is
inveterate. The Egyptian Pharaoh of old was both a god and the son of a
god, and a record is preserved of the visit of the god Amon to queen
Ahmose in the form of her husband. The halo of divinity was perhaps not so
distinct as in earlier times, but in their king the people still saw the earthly
likeness of the deity.

[1] It must suffice to refer to works dealing with primitive religion, see
E. Clodd, Animism, the Seed of Religion (London, 1905), A. C. Haddon,
Magic and Fetishism (1906), in this series.

The Divinity of kings was a fundamental belief which reveals itself in a

variety of forms through Western Asia and Egypt. The inscriptions of
Gudea, the code of Khammurabi, the Assyrian records and the praises of the
Pharaohs reflect conceptions which are materialised now in the insignia of
the kings, and now in their costume and toilet. In a Babylonian myth the
royal ornaments lay before the supreme god awaiting the monarch; in Egypt
the king is the god's ka, his first-born; chosen, created and crowned by the
divine father. The kings stood in the closest relationship to the gods; they
were not only the heads of the state, they were also (in early Assyria) priest-
kings, and in Egypt theoretically all offerings for the living and the dead
were made by the Pharaoh. All this was neither mere empty formality nor
an isolated eccentricity. It is quite in accordance with the powers commonly
ascribed to divine representatives, that the control of the rain and storm is
held to depend upon the influence of Ramses II. with the weather-god. It is
equally intelligible (from anthropological evidence) when the same king
caused the gods to take up their abode in the images which had been
prepared for them!

Khammurabi could declare that he carried in his bosom the people of

Sumer and Akkad, and the Pharaoh could call himself the husband of
Egypt, while Egypt was 'the only daughter of Re (the sun-god) whose son
sits upon the throne.' Not only was he the incarnation and the son of the
deity (or of all the recognised deities), but he was the cause of the land's
fruitfulness, prosperity, and protection. The Pharaoh, 'the god of all people'
(as he is once called), received the adoration of his subjects, and one could
sometimes believe that he was more essentially a deity than the gods
themselves, were it not that the subordinate gods always maintained their
hold upon the people locally. With all allowance for the difference between
conventional and practical religion, the fundamental relations between land,
people, ruler and the deity persisted in many related though varying forms,
which are extremely interesting in any consideration of the social changes
at the rise of a monarchy and after its downfall.

This digression is necessary, because, although the practical working of

such beliefs as these may perplex us, the fact remains that they were shared
in Palestine. The petty rulers in the Amarna letters thoroughly recognise the
divine nature of the king who was a god and had the god for his father (see
p. 78 sq.). Later, when Palestine had its own king, the 'Lord's anointed' was
almost as the deity himself (Ex. xxii. 28, cp. 2 Sam. xiv. 17); king and cult
were one (Hos. iii. 4), and the king's death could be regarded as the
extinction of the nation's lamp (2 Sam. xxi. 17). Not to mention other
details, the Messianic ideals of the divinely-begotten son and of the ruler
whose origin was of aforetime preserve the inveterate belief in the divine
ancestry of rulers, an honour which in other lands continued to be conferred
upon rather than claimed by them.
 Recognised gods.—It is very important to find that the representatives
or possessors of divine powers are the worshippers of their deity in life and
his inferiors in death. The recognised gods have their definite circles of
clients, and if their human representatives are subsequently worshipped or
even deified, this is a not unnatural development, especially as the official
deities are apt to be at the mercy of political and religious changes. The
older gods can be degraded and sink to the rank of demons (from newer
stand-points), but the petty deities and the lower supernatural beings are as
little influenced by external vicissitudes as the lower ranks of humanity
with whom they always stand in closer relationship. Their persistence in
popular belief is as typical as the descent of the more august beings,
although even the latter are understood to retain an influence which those of
more recent introduction have not yet acquired or are unable to exert. While
the general fundamental conceptions remain virtually unchanged, they are
shaped by the social and political institutions, for religious and political life
formed part of the same social organism.

CHAPTER VI

THE GODS

Their vicissitudes.—The deities were not originally personifications of

any one power of nature; like the secular heads of small local groups they
were the supreme patrons of their little circle. They were usually nameless,
but were known by an epithet, or were styled 'god' (el) or 'lord, owner'
(baal), with the corresponding feminine form. Each might be distinguished
by the name of its locality. The 'god' of Sidon was otherwise the 'Baal' of
Sidon, the 'goddess' of Byblos was known as the 'Baalath' of the city; the
Baal of Tyre was called Melkart, i.e. simply 'king of the city'; the proper-
name of the Baal of Harran was Sin (the moon-god); the Baal of Heaven,
according to Philo of Byblos, was the Sun. When Baal and El were used as
generic terms, their application was perfectly intelligible locally; and when
they occur in forty or more place-names, and numerous old personal names
in Palestine, it is unnecessary to suppose that they represent two distinct and
definite deities. From the old Palestinian names we learn that the deity is
high, great and good; he opens, builds, heals, sows, gathers; he remembers,
hastens, helps, protects, blesses, etc.[1] Such conceptions would be
generally true of all; the power of each was not unlimited, but it extended to
all that man usually desired.

[1] Apart from names whose meaning is uncertain (e.g. Jacob-el, God
supplants?), the list could be easily enlarged; a number of names of
western (as opposed to the usual Babylonian) type can be gleaned from
the records of the First Babylonian Dynasty.

From the general resemblance subsisting between the distinct local gods
it was possible to regard them as so many forms of a single god; and when
groups combined and individual gods were fused, multiplicity of types
ensued. The status of a local tutelary was affected when commercial
intercourse widened the horizon of both the traveller and the native; and in
the growth of political power and the rise of a kingship the conceptions
entertained of the deity's attributes and powers were elevated. Through the
extension of authority the way lay open to groups of gods who could not be
fused, and equally to the superiority of one national patron deity over the
rest. Political or other changes led to the promotion of this or the other god,
and prominent or specialised deities in superseding others acquired fresh
attributes, though local divergencies were again necessarily retained. This
does not complete the vicissitudes of the gods or the intricacies caused by
assimilation or identification. A popular epithet or appellative could appear
by the side of the proper deity as a new creation, or the deity was sub-
divided on cosmical and astral theories. The female deity (whose name may
be without the usual distinguishing mark of gender) could even change her
sex; the specific name could also become employed as a common term for
any deity, and the plural 'gods' could be applied to a single being as a
collective representation of the characteristics it embodied.
 Amid the intricate careers of the great names, the local deities
obstinately survived in popular religious life. They have found their parallel
in the welis or patrons, saints and holy sheikhs of the modern shrines (see
pp. 21 sqq.). The modern analogy is instructive in many points of detail,
particularly when we observe the vicissitudes which the occupants of the
shrines have experienced. It is natural to ask for the ancient counterparts of
the Allah, the supreme god in the official religion, who, as we have said, is
vague and remote in the practical religious life of the peasant of to-day. A
series of well-defined historical events made him pre-eminent over all other
gods and goddesses and established Mohammedanism; internal and external
causes shaped the varying conceptions of his nature, and gave birth to
numerous sects. All the Oriental religions have this twofold aspect: the
historical circumstances which affected the vicissitudes of the deities, and
the more subtle factors which have influenced forms of belief. But we have
no direct information upon the rise of the general conditions in Palestine
during our period, and such problems as the origin of the term El 'God'
(common to all the Semitic peoples) belong to the pre-historic ages.

Their representative character.—When the gods reign like feudal

princes over their principalities their sphere is limited and other districts or
kingdoms belong to other gods. Residence in an alien land brought one
under the influence of alien gods, whose reality was not denied, though
their power could be variously estimated. At Serabit, for example, the
Egyptians had combined the worship of their god Sopdu with that of the
local 'lady of turquoise,' whom they identified with their Hathor, and the
caves and temples of both stood side by side. The Egyptian inscriptions
there refer to 'the gods and goddesses of this land,' and an officer, probably
of the Twelfth Dynasty, encourages a cult which was largely utilitarian:
'Offer ye to the mistress of heaven, appease ye Hathor; if ye do it, it will be
profitable to you; if ye increase to her, it shall be well among you.' Some
centuries later, we read that Ramses III., desirous of the precious treasures,
sent clothes and rich presents to his 'mother' Hathor, 'lady of the turquoise.'

The relationship between countries and their respective national gods

(cp. Judges xi. 24) is frequently illustrated. When Tushratta, king of
Mitanni, writes to Amenhotep III., he ascribes a victory to the weather-god
Teshub (if that was the native name), and trusts that his lord Teshub will
never permit him to be angry with his 'brother' the king of Egypt. Similarly,
he prays that the sun-god (Shamash) and the goddess Ishtar may go before
his daughter and make her in accord with the king's heart.[2] On the other
hand, it is 'the gods,' or Teshub and Amon (of Egypt),[3] who will make the
present alliance a lasting one, and his gods and those of his 'brother,' or
Ishtar, 'lady of ladies,' and Amon who will guard the damsel on the journey
and give her favour with the king.

[2] In the names in chapters vi. and vii., the more familiar Astarte is
employed for Ashtart (Old Testament, Ashtoreth). Where cuneiform
evidence is used the Babylonian form (e.g. Shamash, Ishtar) is usually
retained.

[3] Amon, the predominant god of Egypt, owed his rise from an obscure
local deity of Thebes to the political growth of the city. He was then
assimilated to Re (the solar-orb) of Heliopolis.

Towards the close of the reign of Amenhotep III. Tushratta despatched to

Egypt, Ishtar of Nineveh 'lady of lands, lady of heaven,' in pursuance to her
oracle 'to the land that I love I will go.' She was doubtless sent to exercise
her powers in Egypt, and Tushratta expresses the hope that the king may
revere her tenfold more than on the occasion of a previous visit. He also
invokes a hundred thousand years and great joy for his 'brother' and
himself. There is a parallel to this in the late popular story where Ramses II.
sent one of the images of Khonsu (moon-god and god of healing) to cure a
Hittite princess, the sister of his queen, of an evil spirit. The god
accompanied by a priest was received with all reverence, the demon was
expelled and allowed to depart in peace to the place he desired, and a great
feast was celebrated. Indeed, the Hittite chief kept the useful god with him
for nearly four years, when, frightened by a vision of the god flying
upwards towards Egypt, he restored it to its rightful soil. The very human
limitations of the deities render it necessary that some representation or
emblem should be employed when their help is required.
 In political Treaties and Covenants the representative gods of the
respective countries are invoked as witnesses, and their curses are expected
to fall upon the defaulter. It was generally felt that curses as well as
blessings had a very real potency, and the thrilling denunciations at the end
of Khammurabi's Code of Laws and contemporary examples from Egypt
threaten desolation, hunger, thirst, flaming fire, and the avenging pursuit of
the gods. Political treaties are instructive for the light they throw upon the
ruling powers. In Esarhaddon's treaty with Baal, king of Tyre (677-6 B.C.),
the gods of the latter are Baal-shamen (Baal of heaven), two other specified
Baals, Melkart of Tyre, Eshmun and the goddess Astarte. Later, in
Hannibal's covenant with Philip of Macedon, the Carthaginian gods are
enumerated in two triads, then follow the gods who took part in war, and
finally, sun, moon, earth, rivers, harbours (?) and streams. But the most
illuminating example is the Egyptian version of the treaty (about 1290),
between Ramses II. and the Kheta (Hittites), the two great rival influences
over the intervening lands. Here the representative heads of Egypt and the
Kheta are respectively the sun-god Re and Sutekh (i.e. Set, the Egyptian
equivalent of a weather- or storm-god whose native name can only be
conjectured). Formerly, we learn, '[the] god prevented hostilities' between
the two lands by treaty, and this new pact is made for 're-establishing the
relations which Re made and Sutekh made for the land of Egypt with the
land of Kheta' to prevent future warfare. The thousand gods male and
female both of the Kheta and of Egypt are called to witness. Those of the
former are particularly interesting, they comprise the sun-god lord of
heaven, the sun-god of the city of Ernen (also called 'lord of every land'),
Sutekh lord of heaven, Sutekh of Kheta, Sutekh of the city of Ernen, and the
Sutekh of various specified cities, Antheret (probably Astarte) of the land of
Kheta, nine gods and goddesses of certain named cities. Next come 'the
queen of heaven; gods, lords of swearing; the mistress of the soil, the
mistress of swearing, Teshker, the mistress of the mountains, and the rivers
of the land of Kheta,' and, finally, the gods of a North Syrian ally of Kheta.
On the Egyptian side are Amon, the sun-god, Sutekh (here an Egyptian
deity, see p. 83), the male and female gods of the mountains and the rivers
of Egypt, of the heavens, the soil, the great sea, the wind and the storms.
The treaty also bore a representation of the king of the Kheta and his queen
embraced respectively by Sutekh the ruler of the heavens, and a goddess
whose name is lost. To the gods of Palestine there is no reference; Palestine
did not enjoy political independence.

The Influence of Egypt.—Our latest source is the Egyptian account of

the visit of Wenamon to Byblos to procure cedar-wood from Lebanon for
the sacred-barge of Amon-Re, King of Gods (about 1100). The human
messenger took with him the divine messenger in the shape of a statue of
'Amon-of-the-Way,' reputed to confer life and health; a sacred image upon
which no common eye might gaze. When at length Zakarbaal granted an
interview (see p. 54), he was inclined to ignore the political supremacy of
Egypt, although he appears to allow that Amon had civilised Egypt and
thence all lands, and that artisanship and teaching had come from Egypt to
his place of abode. Wenamon, for his part, showed that former kings not
only sold cedars to Egypt, but spent their lives sacrificing to Amon. Even
the evidence of 'the journal of his fathers' did not remove the king's
reluctance. But the envoy urged the claim of Amon to be lord and possessor
of the sea and of Lebanon, and solemnly warned Zakarbaal: 'wish not for
thyself a thing belonging to Amon-Re, yea the lion loves his own.'
Ultimately the king sent the wood, and he commemorated his obedience to
Amon-Re by an inscription which was likely to be profoundly beneficial.
For, as the envoy observed, should Byblos be visited by Egyptians who
were able to read the stele with his name (the all-essential adjunct), he
would 'receive water in the West (the world of the dead where the sun-god
descended nightly) like the gods who are here' (presumably at Byblos).

Although the narrative is written from an Egyptian standpoint, the

conviction which is ascribed to Zakarbaal finds a parallel in the familiar
story of the journey of Osiris, the founder of Egyptian civilisation, from the
Delta to Byblos. Even before the Hyksos period Egyptian women named
themselves after the Baalath of Byblos whom they identified with Hathor
and evidently regarded as an appropriate patroness.[4] The connection
between Egypt and the port of Lebanon may have been exceptionally close,
but there were Egyptian settlements at Gezer, Megiddo, and the north at an
equally early age. Under the conquerors of the Eighteenth Dynasty, the
daughters of the small tributaries were taken into the royal harem, and the
sons were removed as hostages and safely guarded in Egypt. Some of the
latter settled down, others were appointed in due course to the thrones of
their fathers, after having received the necessary anointing-oil from the
great king. One of the latter recalls in the Amarna letters how he had served
the king in Egypt and had stood at the royal gate, and from the grave-stone
of a Palestinian soldier at El-Amarna we may see how settlement upon
Egyptian soil had led to the acceptance of Egyptian ideas of the other
world.

[4] A. Erman, Zeitschrift fur Aegyptische Sprache, xlii. p. 109.

Meanwhile Palestine and Syria were under the direction of Egyptian

authorities, to whose presence the Amarna letters frequently allude, and one
of the writers quaintly likens the solicitude of a certain official on his behalf
to that of a mother or a father. Where there were Egyptians or where princes
had been in Egypt, some trace of the national religion may be expected, and
it is probable that every military garrison possessed some kind of sanctuary.
Moreover, Thutmose III. had dedicated three cities in the Lebanon district
to Amon; later, Egyptian gods 'dwelt' in the north at Tunip. A stele, a few
miles south of Tell 'Ashtarah (cp. the name Ishtar) in Bashan represents
Sety I. offering a libation to Amon, and the pure Egyptian workmanship
points to a strong foreign influence in the locality. Ramses II. set up a statue
of his majesty in Tunip, and a city in South Lebanon was called after his
name. Still descending, we read that cities were set apart for Amon-Re in
the reign of Ramses III., and this king built in Canaan 'a mysterious house
like the horizon of heavens which is in the sky' (i.e. the abode of the sun-
god), with a great statue of 'Amon-of-Ramses-ruler-of-Heliopolis,' to which
the natives brought tribute, 'for it was divine.'

Elsewhere, Ramses III. asserts that he built strongholds in Asia in honour

of Amon, taxing them year by year to bring their offerings to the ka of the
'lord of gods.' Accordingly, down to the first half of the twelfth century the
cult of Amon followed the extension of Egyptian supremacy, and although
the subsequent political history is obscure, the story of Wenamon would
indicate that some sixty or seventy years later the prestige of the god's name
was not entirely lost. Wenamon's claim corresponds to the explicit
recognition (in the Amarna letters) that the land belonged to Egypt's gods; it
was the natural corollary of political extension. Like Zakarbaal and his
ancestors, all the tributary princes were expected to acknowledge the
suzerainty of Egypt's king and his deity. To refrain from sacrificing to the
conqueror's gods was one of the signs of open revolt, as we know from
Assyria and Babylonia. The king identified himself with the sun, like the
contemporary Hittite king Subbiluliuina and other monarchs, from
Khammurabi 'the Sun of Babylonia' who 'caused light to go forth over the
lands of Sumer and Akkad' to the Assyrian Shalmaneser II. Although the
result is confusing, the subordinate chiefs of Palestine and Syria were
accustomed to the thought. They address the king as their gods, their Sun,
the son of the Sun whom the Sun loves, the Sun in heaven, the Sun of the
lands, or the everlasting Sun. This deified Sun or Shamash (to retain the
Babylonian form) answers to the Egyptian Re or Amon. So Abimilki
(Abimelech) of Tyre writes, 'My lord is the Sun which goes up over the
lands daily according to the decision of the Sun (Shamash) his gracious
father.' And again, 'I have said to the Sun, the father of the king, my lord,
"when shall I see the face of the king, my lord?"' Another writer ascribes his
victory to the king's gods and Sun which went before his face. The chief of
Megiddo, in a letter interesting for its glosses in the native language,
announces his intentions should the king's gods assist him, and other writers
invoke the god or gods of the king and acknowledge the might of Shamash.
Nevertheless, the identification of the Egyptian and the Asiatic sun-god
would not, and probably did not, prevent them from being regarded as two
deities, and a private tablet at Taanach not only recognises the god Amon
and the weather-god Addu, but even appears to add Shamash. It is natural to
suppose from the identification of the king, the sun, and the national sun-
god Amon (or Shamash) that many apparently ordinary rites had a deeper
significance, whether it was the anointing of a vassal or the fasting for a
dead monarch (p. 56). The custom of offering sacrifices on behalf of kings
is well attested, and it is possible that the position of divine kings throws
light upon the fact that the king of Cyprus has to explain his failure to send
a representative to Egypt when Amenhotep was celebrating a sacrificial
feast.
 The Treatment of Alien Gods depends largely upon political relations
(cp. pp. 69 sqq.). New settlers might add the established deities of the soil
to their own. A conqueror might recognise the deities of the district to
which he laid claim. The gods of a defeated land were not invariably
deposed, although the Assyrian kings would sometimes destroy them or
present them to their own deities. Mesha king of Moab (about 850 B.C.)
records that he brought before his god certain captured objects of cult, and
it is possible that the pillar at Gezer which is not of local origin had a
history of this kind (p. 14). The Philistines were dismayed at the 'mighty
gods' which the Hebrews, in accordance with a familiar custom, took with
them into battle, and, on another occasion, their own gods, left behind in
their flight, were carried away by David (1 Sam. iv. 8, 2 Sam. v. 21). The
mere capture of the gods was sometimes enough to lead to overtures for
peace. But an Assyrian king would even repair the dilapidated captive
deities, and having inscribed upon them the 'might' of his god and the
'writing of his name' would restore them to a trusted vassal. In Palestine the
petty rulers enjoyed considerable freedom provided they paid their tribute,
and supported their suzerain. We do not learn that Egypt sought to
amalgamate subdued peoples and make of them 'one folk' (lit. mouth), as
was claimed by Tiglath-Pileser I. and other Assyrian kings. Nor do we find
that the Egyptian king sent skilled emissaries to teach (as Sargon II. says)
'the fear of God and the king,' although, if the reference be merely to the
promulgation of the official cult, this was probably the chief results also of
Egypt's supremacy.

On the other hand, a Syrian prince who had recaptured his Sun-god from
the Hittites besought Amenhotep III. (whom he addresses as 'Son of
Shamash') to put his name upon it as his fathers had done in the past. The
text is somewhat obscure, but the recognition of the Asiatic Shamash is
clear, and intelligible on the identification of Shamash and Amon-Re. So,
also, when the king of Byblos asserts that 'the gods, Shamash, and the
Baalath' of the city had brought about the king's accession, we have to
remember that the goddess had long before been identified with the
Egyptian Hathor. At a later date, a stele found north of Tell 'Ashtarah
depicts Ramses II. paying homage to a deity whose crown, horn, and
Semitic title prove him (or her) to be a native deity whom the king
evidently respected.[5] Respect for alien gods ceases when they are found
to be powerless; but Egypt was constantly troubled by her warlike Asiatics,
and so far from their gods being ignored or rejected, they entered Egypt and
found an extremely hospitable reception (see Chapter vii). Asiatic
conquerors in Egypt appear to have been less tolerant. The Hyksos ruled 'in
ignorance of Re,' and their god (Sutekh) was planted in the land; and, later,
during the brief period of anarchy when a Palestinian or Syrian chief held
Egypt until his overthrow by Setnakht, the upstarts 'made the gods like men
and no offerings were presented in the temples.' We may assume then that
the religion of our land remained practically unchanged during Egyptian
supremacy except in so far as this involved the official recognition of the
Egyptian national god and his representative upon the throne.

[5] Zeitschrift des Deutschen Palästina-Vereins, xiv. p. 142, xv. p. 205.

The stele, known as the 'stone of Job,' has entered into the worship of a
Moslem place of prayer, and is appropriately connected with a story of
the patriarch, many traditions of whom are current in this part of Hauran.

CHAPTER VII

THE PANTHEON

Until the necessary evidence comes to light it is scarcely possible to do

more than collect a few notes upon some of the gods and goddesses of our
period. The most important sources are from Babylonia, Assyria, and
Egypt; but some additional information can be gleaned from Palestinian
names, allowance being made for the fact that a personal name
compounded with that of a deity is not enough to prove that the bearer was
its worshipper.
 Asiatic Deities in Egypt date from before the age of the Hyksos
invasion, as can be gathered from the history of the mixed cult at Serabit
and from the introduction of Baalath of Byblos (p. 75). Apophis, a Hyksos
king, has left an altar dedicated to his 'father Sutekh,' who had set all lands
under his feet, and after the expulsion of the Hyksos, this foreign deity,
Egyptianised as Set (or Sutekh), became firmly established. Both SUTEKH
and BAAL were regarded as essentially gods of battle, and the latter often
occurs in descriptions of the prowess of the Pharaohs of the Nineteenth and
Twentieth Dynasties. Thus, the king is like Baal in the lands, mighty in
strength, far-reaching in courage, strong-horned; he is like Sutekh great in
might. He is the equal of Baal, 'his real son for ever,' and he is as Baal in his
hour (i.e. of manifestation). When he appears upon the battlefield like Baal,
his flame consumes the foe, and Amon-Re announces to Ramses III., 'I
overthrow for thee every land, when they see thy majesty in strength, like
my son, Baal in his wrath.' Baal is in his limbs; his roaring is like Baal in
heaven, and his enemies fall down in fear of him like Baal. Baal was
virtually identical with Sutekh who is represented as a foreign god and is
sometimes horned (e.g. at Serabit). A curious scarab shows a winged Sutekh
with horned cap and long streamer standing upon a lion.

Another foreigner is RESHEPH, lord of heaven, lord of eternity, or

governor of the gods; he is the warrior, the god of fire and lightning
(subsequently identified with Apollo). Valiant Egyptian officers are likened
to him. He appears on the Egyptian monuments with Semitic profile, and
conical hat (or otherwise a fillet) from which projects the head of a gazelle;
he holds a lance and shield in the left hand, and in his right a club.
According to a magical text his consort was '-t-m, a deity who seems to be
combined with Shamash in an old North Palestinian place-name, and may
recur in the familiar Obed (servant of) -Edom. In Egypt Resheph also
formed a triad with Min (the old harvest-deity and god of reproduction) and
the goddess KADESH ('holy'). The last, whose name suggests the sacred
licentious rites of Asiatic cults (p. 33 sq.), is called lady of heaven, mistress
of the gods, the eye of Re, etc. She was assimilated to Hathor, and stands
nude upon a lion with lotus flowers in her right hand and a serpent in her
left; her head framed with heavy tresses of hair is sometimes surmounted by
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!

ebookname.com