Full download test bank at ebook textbookfull.

com

Understanding Multivariate Research

a Primer For Beginning Social
Scientists First Edition. Edition

CLICK LINK TO DOWLOAD

https://textbookfull.com/product/understandin
g-multivariate-research-a-primer-for-
beginning-social-scientists-first-edition-
edition-berry/

textbookfull
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Applied Univariate Bivariate and Multivariate

Statistics Understanding Statistics for Social and
Natural Scientists With Applications in SPSS and R 2nd
Edition Daniel J Denis
https://textbookfull.com/product/applied-univariate-bivariate-
and-multivariate-statistics-understanding-statistics-for-social-
and-natural-scientists-with-applications-in-spss-and-r-2nd-
edition-daniel-j-denis/

Multivariate Statistical Methods: A Primer, Fourth

Edition Bryan F.J. Manly

https://textbookfull.com/product/multivariate-statistical-
methods-a-primer-fourth-edition-bryan-f-j-manly/

A Primer of Permutation Statistical Methods Kenneth J.

Berry

https://textbookfull.com/product/a-primer-of-permutation-
statistical-methods-kenneth-j-berry/

Differential Equations A Primer for Scientists and

Engineers 2nd Edition Christian Constanda

https://textbookfull.com/product/differential-equations-a-primer-
for-scientists-and-engineers-2nd-edition-christian-constanda/
Practical Bayesian Inference A Primer for Physical
Scientists 1st Edition Coryn A. L. Bailer Jones

https://textbookfull.com/product/practical-bayesian-inference-a-
primer-for-physical-scientists-1st-edition-coryn-a-l-bailer-
jones/

Marine Carbon Biogeochemistry A Primer for Earth System

Scientists Jack J. Middelburg

https://textbookfull.com/product/marine-carbon-biogeochemistry-a-
primer-for-earth-system-scientists-jack-j-middelburg/

Beginning Julia Programming: For Engineers and

Scientists 1st Edition Sandeep Nagar

https://textbookfull.com/product/beginning-julia-programming-for-
engineers-and-scientists-1st-edition-sandeep-nagar/

Practical Social Engineering: A Primer for the Ethical

Hacker 1st Edition Joe Gray

https://textbookfull.com/product/practical-social-engineering-a-
primer-for-the-ethical-hacker-1st-edition-joe-gray/

SAS for R users : a book for budding data scientists

First Edition Ohri

https://textbookfull.com/product/sas-for-r-users-a-book-for-
budding-data-scientists-first-edition-ohri/
 Understanding

BERRY/SANDERS
Although nearly all major social science departments offer gradu-
ate students training in multivariate quantitative methods, the
typical sequencing of topics often delays training in regression
analysis and other multivariate techniques until a student’s second
Multivariate Research
year. William D. Berry and Mitchell S. Sanders’s Understanding
Multivariate Research fills this gap with a concise introduction to

Understanding Multivariate Research

regression analysis and other multivariate techniques. Their book
is designed to give new graduate students a grasp of multivariate
analysis sufficient to understand the essential elements of research
strategy and the major findings of research they must read, which
rely on such analysis, prior to their formal training in quantitative
methods. Readers are assumed to have no background in descrip-
tive or inferential statistics, making the book highly accessible to
students with no prior graduate course work.

William D. Berry is professor of political science at Florida State Univer-

sity. He is a contributor to the American Political Science Review, the
American Journal of Political Science, and the Journal of Politics and has
served on the editorial boards of the latter two. He has published several
books on research methods, and he has received the Policy Studies
Organization’s Harold Laswell Award for outstanding career contribu-
tions to the study of the policymaking process. Berry has also served as
president of the State Politics and Policy section of the American Political
Science Association.
Mitchell S. Sanders is assistant professor of political science at Florida
State University. His research can be found in the American Journal of
Political Science, Political Research Quarterly, Public Choice, and
Political Analysis.

WESTVIEW PRESS ISBN-13: 978-0-8133-9971-3

William D. Berry &

ISBN-10: 0-8133-9971-8
5500 Central Avenue
Boulder, Colorado 80301-2877
12 Hid’s Copse Road
Cumnor Hill • Oxford OX2 9JJ Mitchell S. Sanders
www.westviewpress.com
 Understanding
Multivariate Research
Understanding
Multivariate
Research
A PrilTIer for
Beginning Social Scientists

William D. Berry
Florida State University

Mitchell S. Sanders
Florida State University

t';e~rm
'~""-.-"""/ A Member of the Perseus Books Group
All rights reserved. Printed in the United States of America. No part of this pub-
lication may be reproduced or transmitted in any form or by any means, elec-
tronic or mechanical, including photocopy, recording, or any information stor-
age and retrieval system, without permission in writing from the publisher.

Copyright © 2000 by Westview Press, A Member of the Perseus Books Group

Published in 2000 in the United States of America by Westview Press, 5500 Cen-
tral Avenue, Boulder, Colorado 80301-2877, and in the United Kingdom by
Westview Press, 12 Hid's Copse Road, Cumnor Hill, Oxford OX2 9JJ

Find us on the World Wide Web at www.westviewpress.com

Library of Congress Cataloging-in-Publication Data

Berry, William Dale.
Understanding multivariate research: a primer for beginning social
scientists / William D. Berry, Mitchell S. Sanders.
p. cm.
Includes bibliographical references and index.
ISBN-10: 0-8133-9971-8 ISBN-13: 978-0-8133-9971-3
1. Social sciences-Research-Methodology. 2. Multivariate analysis.
3. Regression analysis. I. Sanders, Mitchell S. II. Title. III. Series.

H62 .B454 2000

300'.7'2-dc21
99-057762

The paper used in this publication meets the requirements of the American
National Standard for Permanence of Paper for Printed Library Materials
Z39.48-1984.
 Contents

List of Tables and Figures Vll

Preface for Teachers and Students ix
Acknowledgments X1ll

1 Introduction 1
The Concept of Causation, 1
Experimen tal Research, 2
The Logic Underlying Regression Analysis,S
Some Necessary Math Background, 7

2 The Bivariate Regression Model 15

The Equation, 15
The Intercept, 17
The Slope Coefficient, 18
The Error or Disturbance Term, 19
Sorne Necessary Assumptions, 20
Estimating Coefficients with Data from a San1ple, 24

3 The Multivariate Regression Model 29

The Value of Multivariate Analysis, 29
Interpreting the Coefficients of
a Multivariate Regression Model, 31
Dichotomous and Categorical Independent Variables, 33
The Assun1ptions of Multivariate Regression, 37
Choosing the Independent Variables for
a Regression Model, 38

v
vi Contents

4 Evaluating Regression Results 41

Standardized Coefficients, 41
Strong Relationships Among the Independent Variables:
The Problem of Multicollinearity, 43
Measuring the Fit of a Regression Model, 44
Statistical Significance, 45
Cross-Sectional vs. Time-Series Data, 49

5 Some Illustrations of Multiple Regression 51

Lobbying in Congress, 51
Population Dynan1ics and Economic Development, 57

6 Advanced Topics 63
Interaction vs. Nonlinearity, 63
Interactive Models, 64
Nonlinear Models, 68
Dichotomous Dependent Variables:
Probit and Logit, 72
Multi-equation Models: Simultaneous
Equation Models and Recursive Causal Models, 76

7 Conclusion 79

Glossary 81
References 83
Index 85
 Tables and Figures

Tables
1.1 SynonYlTIS for independent and dependent variable 2
1.2 Observations for food intake and body weight 8

3.1 Regression coefficients for multivariate

model explaining body weight 33
3.2 Some assumptions of multivariate regression 38

5.1 Wright's model of lobbying 55

5.2 Crenshaw, AlTIeen, and Christenson's lTIodel
of GDP growth 61

Figures
1.1 Data on food intake and body weight
for four individuals 8
1.2 Two variables that are completely unrelated 9
1.3 Two scatterplots with correlation
coefficients of +0.75 11
1.4 Two lines 13

2.1 Graphical representation of

bivariate regression equation 2.4 17
2.2 Which is the best-fitting line? 25
2.3 Vertical distances between points and two lines 26
2.4 Population regression line and estimated line
based on a salTIple 27

vii
viii Tables and Figures

6.1 AI-Sharideh and Goe's interactive model 67

6.2 A quadratic equation 69
6.3 Raj's nonlinear model 71
6.4 Estimated effects in Johnson and
Scheuble's logit model 75
 Preface for
Teachers and Students

If social science departments wanted to structure their graduate

programs to allow students to make the best use of their training
in research methodology, they would probably devote their stu-
dents' first semester of course work to methods training, using ap-
plications from the literature as illustrations but delaying substan-
tive courses until students had completed a department's core
methods courses. Of course, very few programs are structured in
this fashion. Indeed, such a course schedule would likely leave stu-
dents frustrated with having to delay beginning the study of hu-
man affairs until the second sen1ester.
In most graduate programs the core methods sequence is
stretched over two or three semesters. A typical three-semester
version includes philosophy of science and research design in the
first semester, introductory statistics in the second, and multivari-
ate analysis-emphasizing regression-in the third. So in practice,
exposure to regression analysis and other multivariate techniques
often does not come until well into a student's second or third se-
mester in graduate school. However, before this exposure, stu-
dents are taking substantive courses and reading literature that re-
lies on regression and other, sin1ilar techniques. In effect, we have
been teaching students to evaluate quantitative social science re-
search much like parents who teach their young children to swim
by throwing them into the n1iddle of a pool without any prior in-
struction. Our students may learn to «swin1" -to survive-but
they certainly won't have any fun doing so.
As teachers, we ought to do better. Yet virtually all textbooks
covering multivariate methods-even those intended for intro-

ix
x Preface for Teachers and Students

ductory courses-present the techniques assuming students (1)

have learned the basic principles of probability theory and statisti-
cal inference (so that they can make sense of standard errors,
t-tests for coefficient estimates, and confidence intervals), and (2)
are familiar with the concept of a distribution and the use of uni-
variate descriptive statistics for 11leasuring central tendency and
dispersion of distributions.
It is entirely reasonable that presentations of techniques such as
regression analysis assume such a background in statistics, since if
one is going to learn both introductory inferential statistics and
regression, it is more efficient to start with statistics. But students
can learn a great deal about regression and other multivariate
11lethods with virtually no background in statistics; they can ac-
quire a conceptual understanding of some of the key assumptions
of these techniques and an ability to interpret the meaning of the
coefficients estimated. Thus, we can give students an entry-level
background in multivariate analysis very early in their graduate
training, thereby allowing them to understand the essential ele-
ments of research that relies on such analysis before they under-
take more thorough training in multivariate analysis later in their
careers.
This book offers a conceptual introduction to regression analy-
sis and related techniques that can be assigned to graduate stu-
dents at the beginning of their first semester, or even the SUlll11ler
before starting school. We assume that students come to the book
with a "clean slate"-that they have no knowledge of descriptive
or inferential statistics or of social science research design. We
present all topics without relying on the 11lathematicallanguage of
probability theory and statistical inference. In fact, the math is
limited to some simple algebra, which we review early on. Fur-
thermore, we believe that the material presented can be learned by
a conscientious student, through 11lultiple readings, with little or
no time devoted to attending accompanying class lectures. The
book is short, so that even three or four close readings do not re-
quire an excessive investlllent of tinle.
The text is divided into two major parts. The first covers basic
topics in regression analysis and is restricted to linear additive
models. The second extends the regression model in several ways:
Preface for Teachers and Students xi

to nonlinear and nonadditive models, to "causal models" contain-

ing more than a single equation, and to probit and logit models
with dichotomous dependent variables. Unless students are going
to have frequent exposure to research articles relying on these ad-
vanced techniques, the second part of the book can be treated as
reference material-to be read a couple of times, and then re-
viewed in greater detail when faced with a study relying on one of
these techniques.
The book is also appropriate for some undergraduate students.
Since it has no statistics prerequisite, its appropriateness for un-
dergraduates is determined by whether they will be expected to
read original research relying on multivariate analysis. If the stu-
dents are deemed capable of reading articles reporting on quanti-
tative research, clearly they should be able to understand this in-
troduction to quantitative methods. We can envision this book
being assigned as a text in an undergraduate methods course in a
department in which students in junior- and senior-level courses
are asked to read quantitative research. However, we can also see it
being assigned by instructors in departments without an under-
graduate research methods requirement, as preliminary reading in
a course in which students will be reading original quantitative re-
search. If the articles to be read rely only on linear, additive regres-
sion models, students might be assigned just the first part of this
book, through Chapter 5.
It would be terrific if we could publish a different version of this
book for each social science discipline, so that each student could
see exan1ples drawn exclusively from the discipline he or she has
chosen to study. Since this is not feasible, we wrote a single version
that relies on illustrations drawn from research across a variety of
disciplines. To make them accessible to students with varied back-
grounds, we chose examples that we believe could be understood
without any background in the parent discipline. Indeed, for one re-
curring illustration, we turn away from the social sciences to con-
sider the factors determining a person's weight, since this is a topic
about which all readers should have a certain an10unt of intuition.
We close this preface with some encouragement and advice for
students. Assuming the material in this book is new to you, do not
expect to understand-or remen1ber-all of it after just one read-
xii Preface for Teachers and Students

ing. There is much to learn and synthesize. However, with each

additional reading, more and more aspects of the material should
beconle clear. Thus, we hope you have the patience to undertake
several readings even if your principal reaction to your first expo-
sure is frustration. We are confident that your persistence will be
rewarded!
At various points in the text, we include brief exercises that you
may use to test your knowledge of the topics discussed. These ex-
ercises are enclosed in brackets (i.e., [... ]) and prefaced, for easy
identification, with a pair of exclamation points (!!). The text im-
mediately following an exercise gives the correct answer. However,
the design of the text allows these bracketed exercises to be ig-
nored without any loss of continuity of our presentation. Indeed,
we suggest skipping the exercises when reading the book for the
first tinle, but pausing to work through them on additional read-
ings. If you can complete these exercises successfully, this is a
strong indication that you are absorbing the material presented.

Willianz D. Berry
Mitchell S. Sanders
 Acknowledgll1ents

This book began as a paper written for graduate students in polit-

ical science at Florida State University in 1994. Since then it has
gone through nunlerous revisions before assunling its current
form. We wish to thank the colleagues who provided helpful COln-
ments on various previous versions (even though several of these
folks read the book so long ago that they will probably will not re-
member having done so!): Charles Barrilleaux, Daniel Berry,
Frances Berry, Belinda Davis, Aubrey Jewett, Gary King, Tatiana
Kostadinova, Doug Lemke, Andrew Long, Andrew Martin, Glenn
Mitchell, Evan Ringquist, and Kevin Wang. The students in two
Florida State graduate courses on research methods also reviewed
drafts of the book: Berry's spring 1994 seminar, and Will Moore's
fall 1999 seminar. We are very grateful to both classes. Because our
writing stretched over such a long period, and the years have taken
a toll on our memories, we have undoubtedly failed to thank SOlne
who reviewed early drafts. We apologize to any who have been in-
advertently omitted.

W.D.B.
M.S.S.

xzu
 Understanding
Multivariate Research
 1
Introduction

The Concept of Causation

Much social science research is designed to test hypotheses (or
propositions) about causation. Such hypotheses take the form of an
assertion that if something (e.g., some event) occurs, then something
else will happen as a result. Among nations, we might assert that
population growth causes (or influences) economic growth. Among
individuals, we nlight believe that body weight is influenced by food
consumption. In a causal hypothesis, the phenomenon that is ex-
plained is called the dependent variable. It is called a variable be-
cause we are conceiving of something that can "vary" across a set of
cases (e.g., persons or nations); it is called dependent because of the
assertion of causation: its value is hypothesized to be dependent on
the value of some other variable. In our examples the dependent
variables are a nation's economic growth and an individual's weight.
The other variable in the hypothesis-the one that is expected to in-
fluence the dependent variable-is called the independent (or ex-
plana tory) variable. Population growth is thought to be an indepen-
dent variable affecting a nation's economic growth; a person's food
consumption is conceived as an independent variable influencing
his or her weight. There are numerous synonynls for the terms in-
dependent and dependent variable in the social sciences. Table 1.1
lists the most common ternlS.
Let us take a closer look at what it means to claim that one vari-
able influences another. The most conlmon conception of causa-
2 Introduction

TABLE 1.1 Synonyms for Independent and Dependent Variable

Independent Variable Dependent Variable

Explanatory variable Explained variable
Exogenous variable Endogenous variable
Predictor variable Response variable
Target variable

SOURCE: Modified from Maddala (1992,61).

tion focuses on the responsiveness of one variable to a change in

the value of the other. When we claim that food intake influences
body weight, we are implicitly arguing that if we were able to in-
crease a person's food consumption while holding everything else
constant, the individual's weight will change. The clause "while
holding everything else constant" is important, because if other
variables change at the same time as food consumption, the indi-
vidual's weight change could be a response to a change in one or
more other factors rather than the increase in food consumption.
More generally, when we claim that some variable, X, influences
another variable, Y, we mean that if all other variables could be held
constant, then a change in the value of X would result in a change in
the value of Y. We can also develop a measure of the magnitude (or
strength) of the impact of X on Y by focusing on the size of the
change in the value of Y occurring in response to some fixed in-
crease in X. If a given increase in X leads to a 10- unit decrease in Y
in one environment, but to a 5-unit decrease in another context, the
former impact can be deemed twice as strong as the latter. (Several
expressions are used interchangeably by social scientists to convey
an assertion of causation; "X causes Y," "X influences y;" "X affects
Y," and «X has an impact on Y" are synonymous. The custom of us-
ing the symbol Y to denote a dependent variable and X to indicate
an independent variable is deeply ingrained in the social science lit-
erature, and we shall follow this custom throughout the book.)

Experimen tal Research

Suppose we wish to test the hypothesis that an independent vari-
able X influences a dependent variable Y using empirical analysis.
Introduction 3

(Empirical analysis refers to analysis based on observation.) The

ideal way to do so would be to conduct an experiment. Your famil-
iarity with experinlents probably dates back to your first science
class in elementary school. However, it is important to refresh our
memories on the specific features of an experiment. To illustrate,
say we design an experilnent to test the clailn that a fictitious new
drug-a pill called Mirapill-helps to prevent children from get-
ting the fictitious disease turkey pox. The population in ques-
tion-that is, the cases to which the hypothesis is meant to ap-
ply-is children who have not had turkey pox. The independent
variable is whether or not a child is given Mirapill, and the depen-
dent variable is the probability that the child will get the disease.
In an experiment designed to test whether Mirapill reduces the
probability of getting turkey pox, we would begin by taking a ran-
donl sanlple-perhaps 1,000 subjects-from the population of
children who have never had turkey pox. (For the sample to be ran-
dom, every member of the population must have the same chance
of being included in the sample.) These LOOO children then would
be randomly assigned to two groups. One group of 500 would be
called the experimental group, and the other, the control group.
Randolnness-both in the selection of subjects froln the popu-
lation and in the assignment of subjects to the experimental and
control groups-is critical to the validity of an experiment. Statis-
ticians have discovered that if a sample is selected randomly and is
large enough (LOOO is certainly sufficient), it is likely to be repre-
sentative, in every respect, of the larger population from which it
is drawn. I This means that we can learn almost as much by ob-
serving the sample as by observing the full population, yet the for-
mer is generally far less expensive and time consuming. In an ex-

I A large sample is necessary for this claim to hold) because the smaller a sam-

ple, the more likely it is to be unrepresentative of the population.1b get a sense of

why this is true, imagine flipping a fair coin four times. Although getting half
heads and half tails is the most likely outcome, you probably would not be
amazed to get three heads and one tails (i.e.) three-quarters heads). But if you
were to flip the coin 100 times, you likely would be quite surprised if you got
three-fourths heads (i.e., a 75-25 outcome). This is because a small sample of
coin flips is more likely to be unrepresentative of the population (in which heads
and tails occur equally often) than a large sample.
4 Introduction

periment, we observe the random sample and, on the basis of

what we learn, draw an inference about whether the hypothesis is
likely to be true in the overall population. Similarly, randonl as-
signnwnt of the children in the sample to the two groups makes it
very likely that the groups will be nearly equivalent in every way.
For example, the two groups of children should be almost equally
likely to be genetically predisposed to get turkey pox. The two
groups also should be nearly equally likely to be exposed to turkey
pox through contact with other children.
In the next step of the experiment, the children in the experi-
mental group would be given MirapilL whereas those in the con-
trol group would receive a placebo. (A good placebo would be a
pill that looks exactly like Mirapill but contains no nledicine.) Af-
ter the pills are administered, both groups would be observed for a
period, and we would determine how nlany children in each
group contracted turkey pox. If fewer children in the experimental
group than in the control group got the illness, this would be evi-
dence supporting the hypothesis that Mirapill helps to prevent
turkey pox. Furthermore, the difference between the two groups
in the number of children contracting the disease would serve as a
measure of the strength of Mirapill's impact as a preventive. If
many fewer children in the experimental group got sick, this
would suggest that Mirapill has a strong effect. If only slightly
fewer experimental group children canle down with turkey pox,
this would mean that the effect is probably weak.
Suppose we conduct this experiment and find that the inci-
dence of turkey pox is substantially lower in the experinlental
group. Why would this be convincing evidence that Mirapill pre-
vents turkey pox? To see why, recall what we mean when we say
that X causes Y: if all other variables were held constant, then a
change in the value of X would lead to a change in the value of Y.
Our experiment gives us just the information we need to assess a
claim of causation. We find out what happens to the dependent
variable (the probability of getting turkey pox) when we change
the value of the independent variable (receiving or not receiving
Mirapill) when all other variables are held constant. (Saying that
the experimental and control groups are nearly equivalent in
every way-as a consequence of random assignment of children
Introduction 5

to the two groups-is the same as saying that all variables are held
nearly constant from one group to the other.) In other words, ran-
dom assignment of children to the control and experimental
groups eliminates all explanations other than Mirapill for the dif-
ference in disease incidence between the two groups. For instance,
a difference between the two groups in genetic susceptibility to
turkey pox is unlikely to be responsible for the difference in dis-
ease incidence, because randomness of assignment makes it likely
that varying degrees of genetic susceptibility are distributed
evenly between the two groups, and thus likely that the two
groups are similarly predisposed to getting turkey pox. Also, the
fact that both groups were given some pill-either Mirapill or a
placebo-allows us to reject the mere taking of sonle pill as a pos-
sible cause of the lower incidence of turkey pox in the experimen-
tal group.

The Logic Underlying Regression Analysis

Consider the hypothesis that food intake influences body weight.
In principle, we could test this hypothesis experimentally, by ran-
domly selecting subjects from the population of adults and then
randomly assigning different levels of food intake to these sub-
jects. But in practice, such a strategy would not be feasible, since it
would require those being studied to tolerate an enormous degree
of intrusion into their daily lives by permitting us to control how
much they eat. An alternative (and more realistic) plan would be
to allow the subjects to eat what they want, but to measure their
food consumption and, in one way or another, compare the body
weights of "small eaters" to those of "big eaters." If we find that
those who eat a lot tend to be heavier than those who consume
less, we would claim support for the hypothesis.
For nlany social science hypotheses, it is also infeasible to nla-
nipulate the independent variable experimentally. For example, if
we were studying lobbying (i.e., attempts by individuals and
groups to influence governnlent officials) and we wanted to test
the hypothesis that nlembers of Congress in leadership positions
are lobbied more heavily than members who are not in leadership
positions, we would not be able to intervene and inlpose our own
6 Introduction

leadership choices on Congress. Instead, we can observe Congress

as it exists and determine whether there are different patterns of
lobbying activity for leaders than for nonleaders. Similarly, to test
the hypothesis that population growth affects economic develop-
ment, it obviously would be impossible to run an experiment in
which we randonlly assign different levels of population growth to
nations. However, we can observe the nations of the world and de-
termine whether those countries whose populations are growing
fastest are also those whose economies are expanding most
quickly. Regression analysis is a nonexperimental technique for ex-
tracting this sort of information from a sample of data. In this
chapter and the next, we exallline the simplest form of regression,
bivariate (or two-variable) regression, which involves a single in-
dependent variable hypothesized to influence a dependent vari-
able. In subsequent chapters we consider multivariate (or multiple)
regression, which involves two or more independent variables pre-
sumed to influence the same dependent variable.
Although regression analysis often is more feasible than experi-
mental research, it cannot provide as convincing evidence of causa-
tion as an experiment. As we have seen, in an experilllent, randolll
assignment of the value of the independent variable to subjects en-
ables the researcher to assess the response of the dependent variable
to a change in the independent variable when all other variables that
influence the dependent variable are held constant. In nonexperilllen-
tal research relying on regression, we forgo random assignment of
the value of the independent variable and instead accept the values
that the cases being analyzed happen to have. The result is that when
we use regression, we estilllate the change in the dependent variable
associated with a given change in an independent variable when the
other independent variables in the regression analysis are held con-
stant. This falls well short of the experimental design's ability to hold
constant all variables that influence the dependent variable.
Consequently, researchers should use regression analysis to test a
hypothesis that X influences Y only when they have a theory that ex-
plains why it lllakes sense to expect this causal relationship. Note
that in experimental research, such a theory is not essentiaL For ex-
ample, sometimes scientists confirm that a new drug is effective us-
ing experilllental research, even when they do not know why the
Introduction 7

drug works. But conducting nonexperimental research without a

theory can lead to highly deceptive conclusions. For instance, if we
were to examine fires, we would probably find that fire damage was
most severe when a large number of fire trucks were on the scene-
since both fire damage and number of fire trucks should rise with
the size of the fire (Weisberg, Krosnick, and Bowen 1989). If so, bi-
variate regression analysis with fire damage as the dependent vari-
able and the number of fire trucks present as the independent vari-
able would generate apparent support for the (preposterous)
hypothesis that fire trucks cause fire damage. What should prevent
us from arriving at this erroneous conclusion is the absence of a
plausible theory to suggest why fire trucks should cause damage-
indeed, without such a theory, we should be unwilling to conduct
regression analysis to test this hypothesis in the first place.

Some Necessary Math Background

Representing Data in a Graph
Suppose we have measured the values (or scores) of a set of cases
(e.g., persons, organizations, or nations) on two variables, X and
Y. Returning to one of our earlier examples, let us say that we ob-
serve the food intake and the body weight of four people. We mea-
sure weight in pounds and denote this variable by the label
WEIGHT. Our measure of food intake is average daily food con-
sumption in calories during the year prior to the observation of
weight, and we denote this variable by the label FOOD.2 Our ob-
servations are presented in Table 1.2.
We can portray these data in a graph by designating food intake
as the horizontal axis (or what mathematicians call the x-axis) and
body weight as the vertical axis (the y-axis) and plotting four
points on the graph, one for each person. (The nearly uniform
custom in the social sciences is to use the horizontal axis for the
independent variable and the vertical axis for the dependent vari-
able.) The position of each point is determined by the values of
FOOD and WEIGHT for the associated person. For example, in

2This example is a modified version of an illustration presented in Berry 1993.

8 Introduction

TABLE 1.2 Observations for rood Intake and Body \Neight

X FOOD Y WEIGHT

Bob 3,000 190

Carol 1)100 120
Ted 2,300 155
Alice 1,400 140

•
Bob

Ted
150 Alice •
•
- - - - - - - - - - - - .. Carol

~ 100
:c:

~II
~ 50

o~~~~~~~~~~~~~~~~~~~~~~~

o 1)00 2)000 3)000

X =: FOOD (food intake)

FIGURE 1.1 Data on food intake and body weight for four individuals

the graph in Figure 1.1, the lower left point representing Carol is
positioned 1,100 units out on the horizontal axis (denoting
FOOD) and 120 units up on the vertical axis (denoting
WEIGHT).
This type of graph is called a scatterplot (or son1etimes scatter
diagram or scattergram), because it plots the scattered X and Y val-
ues. The great advantage of the graphical presentation of these
data over the tabular format is that the graph allows us more easily
to observe the relationship between food intake and body weight.
Introduction 9

•
•
• •
•
•
• •
•
y

• •

• • •
•
I

• •• • I
I
•
Cases with very Cases with very
low scores on X high scores on X
x
FICURE 1.2 Two variables that are completely unrelated

This is particularly true when there are a large number of cases)

which makes it difficult or impossible to observe patterns within
data presented in columns. For the four individuals whose food
intake and body weight are plotted in Figure 1.1, it appears that
high values on WEIGHT tend to be associated with high values on
FOOD. In contrast, the graph in Figure 1.2 shows a situation in
which there is no relationship between two variables for a group
of twenty cases. We say that two variables, X and Y, are completely
unrelated (or that there is no relationship between them) ifknowl-
edge of the value of X for a case would provide no help at all in
predicting the value of Y for that case. For the data plotted in Fig-
ure 1.2, if we are told that a case has a very high score on X, it is of
no help to us in predicting the case's score on Y) since cases having
high values on X have scores on Y spread all the way fron1 very low
to very high. The same is true for cases having a low value on X. In
contrast, for the data in Figure 1.1, knowledge that a case has a
high value on X would n1ake a prediction that the case also has a
10 Introduction

high score on Y much more reasonable than a prediction of a low

score on Y.
Social scientists use the term correlation as one way to de-
scribe the strength of the relationship between two variables.
You may have heard this term used in everyday discourse to ex-
press the closeness of a relationship. The formal nleasure of cor-
relation is called the correlation coefficient, and it is nearly always
denoted by a lowercase r. The correlation coefficient ranges
from -1 to + 1. A correlation coefficient of zero means that there
is no relationship between X and Y; the relationship in Figure
1.2, for example, is one for which r = O. As r increases or de-
creases from zero, the relationship becomes stronger. A positive
correlation coefficient-a coefficient greater than zero-implies
a relationship in which, as the value of X increases over cases,
the value of Y tends to increase too (as in Figure 1.1). In con-
trast, a negative correlation coefficient-less than zero-means
that the value of Y tends to decrease as the value of X increases.
The extrenle values for r of + 1 and -1 indicate a perfect linear re-
lationship, that is, one in which all points in the scatterplot fall
exactly on a line.
Although the correlation coefficient provides useful informa-
tion about how well we can predict values of Y if we know X, it
does not tell us anything about how responsive Y is to a change in
X-that is, how much Y changes for a given change in X. There-
fore, the correlation coefficient does not help us to assess the
strength of X's impact on Y. For example, consider the two scatter-
plots in Figure 1.3. The relationships depicted are equal in
strength in ternlS of the correlation coefficient; in both r +0.75.
Yet, in a different sense, we would consider the relationship shown
in scatterplot A to be stronger than that shown in scatterplot B.
Note how the scattering of dots rises more rapidly from left to
right in A than in B. Thus, as the value of X increases, the value of
Y increases by a larger amount in scatterplot A than in scatterplot
B. This makes it clear that Y is nl0re responsive to a change in X in
A than in B. Regression analysis nleasures the strength of the rela-
tionship between X and Y in this sense of responsiveness. For this
reason, regression analysis is more useful for studying causation
than is calculating a correlation coefficient.
 Scatterplot A Scatterplot B

~
25

. • •~
25
.
20
... 20

... . ..., ... . .•.. •.• ..I ••

.. . ... · . . .. .
y 15 ~ # • -I Y 15

.,. -. .:...·
l~ r o

00

.
-I

-1
10

5 . ... .
. .,........ ~..". . ,.... .-
.. .. :.,...'......
~ . .:.,,...... #....
0' o
.. .. ..
-
-
L· •
,
.. ..
0 0.2 0.4 0.6 O.S 1.0 0 0.2 0.4 0.6 O.S 1.0
X X

FIGURE 1.3 Two scatterplots with correlation coefficients of +0.75

......
......
12 Introduction

Lines and Their Equations

Scatterplots-such as those graphed in Figures 1.1, 1.2, and 1.3-
show the values of specific cases on the variables X and Y. We can
also graph lines expressing the relationship between X and Y, such
as the lines in Figure 1.4. Social scientists use the term intercept to
denote the value of Y when X equals zero, or equivalently, the
point at which the line intersects the vertical axis. (Mathemati-
cians would call the intercept the y-intercept. But social scientists
refer to it simply as the intercept, since they seldom have any rea-
son to calculate the x-intercept.) [!! Try to determine the intercept
of each line in Figure 1.4.] Thus, the intercept of the dotted line in
Figure 1.4 is 1, and that of the solid line is 3.
The slope of a line is defined as the change in Y associated with a
one-unit increase in X. Because a line is straight, the slope of a line
is the same regardless of the level of X at which it is calculated. [!!
Determine the slope of each line in Figure 1.4.] In the case of the
dotted line in Figure 1.4, an increase of one unit in X is associated
with an increase of two units in Y, so the slope is 2.In contrast, for
the solid line, a one-unit increase in X yields a decrease of one-
quarter unit in Y (i.e., a change of nlinus one-fourth in Y), so the
slope of the line is -0.25. (Another way to say this is that an in-
crease of four units in X is associated with a decrease of one unit
in Y.) Paralleling the distinction between positive and negative
correlations, lines that slope upward from left to right (like the
dotted line in Figure 1.4), and thus have a positive slope, are de-
scribed as reflecting positive relationships, whereas lines that slope
downward (like the solid line in Figure 1.4) are said to indicate
negative relationships.
The mathematical equation for a line takes a simple form:

This equation may be familiar to you from high school algebra

courses, where you probably learned that the equation for a line
takes the form

Y 11lX + b,
Another random document with
no related content on Scribd:
 Part III

SIEGFRIED

Motif of Mime’s Meditation

[Listen] [MusicXML]

PRELUDE

When Sieglinde ran into the woods with the pieces of the broken
sword, Nothung, she took shelter in a cave where a wicked old dwarf
lived alone. There a little boy was born. But Sieglinde had never
thoroughly recovered from the shock of her husband’s death. The
way through the woods had been difficult, and she had endured
great hardships; so one day she called the Dwarf to her and gave
him the broken sword, telling him to keep it for her son until he grew
old enough to have a weapon of his own, and she told the Dwarf that
she was Sieglinde, and that her husband had been Siegmund, the
Volsung, and she finally said that she wanted the child to be named
Siegfried; then she sank back and died. And so Siegfried, who was a
very little baby then, never, really, saw either his father or mother.
The only father he knew, as he grew older, was the Dwarf, who was
none other than Mime, Alberich’s half-brother. And he could not help
knowing that Mime was wicked and sly, though the Dwarf pretended
to love his foster-son, and tried to arouse some love in return.
Now, perhaps, you wonder, if Mime was so wicked, why he took care
of the boy. I will tell you.
Mime, like every one else, wanted the Rhinegold, and could not get
it, for Fafner, the Dragon, guarded it by night and day at Hate Hole.
And being as sly and evil-minded as the rest of the Nibelungs, he
had concocted a plot by which he thought he could obtain it. He
hoped Siegfried, when he grew older, would slay Fafner with the
sword Nothung, and win the Rhinegold. You see he hoped to
accomplish Fafner’s death through Siegfried, just as Wotan had
once tried to do through Siegmund. Only, after Siegfried had attained
the Gold, Mime hoped to be able to poison him and steal from him
the treasure.
But, to accomplish this, the broken sword must be mended, and this
Mime could not do. Its splintered edges baffled even him—clever
smith as he was. So he set to work forging other swords, and trying
to fashion a blade keen enough to satisfy the boy-Volsung, and also
to kill the Dragon at Hate Hole. But every weapon he made Siegfried
broke into pieces, and demanded a stronger and still stronger sword,
until Mime was in despair.
It angered him terribly, too, that Siegfried, more by instinct than
anything else, knew how wicked his heart was, and how full of bad,
cruel thoughts. The little, dark Nibelung could not understand how
the boy, beautiful as the sun, golden-haired and keen-eyed, strong of
limb and true of heart, loved to roam in the wide forests all the day,
merrily blowing his silver horn and making friends with the woodland
creatures, only returning to Mime’s cave at night. He could not
realize the pleasure that the soft forest voices gave to the youth just
growing into manhood; how he loved the wolves and bears better
than the cringing, evil-eyed, horrible little Dwarf in the cave at home
—the only home he knew.
As for Siegfried, the only thing he wondered at was that he ever went
back to the cave at all. Why did he not roam away forever into the
forest, search out that far, strange place called the world, that really
seemed as if it must be a different universe from the one in which he
lived? He could not tell. He only knew that a strange, irresistible
something seemed to draw him back to Mime’s side every night—a
something he could not explain or even understand. Meanwhile time
passed.

Motif of Forest Life, sometimes called Motif of Love Life

[Listen] [MusicXML]
 Motif of the Forging of Nothung
“Nothung! Nothung! Notable Sword!”
[Listen] [MusicXML]

CHAPTER I

SIEGFRIED AND MIME

The cave was a dark one, but it was not altogether a bad place in
which to live. It was as lofty as a stately cathedral, and the Dwarf’s
forge, built on one side, lent a fitful red light and a little warmth to the
dim, cold atmosphere. Skins of animals gave it a semblance of
comfort; and, indeed, to a wild creature like Siegfried, it would have
been a most desirable home had it not been for the continual
presence of Mime. On the day on which I will open my story, Mime
was sitting on a low stool trying to fashion a sword which would not
break in the hands of the impetuous young Volsung, who, at that
particular moment, was, as usual, out in the woods with his friends,
the wild beasts. As he hammered, Mime grumbled crossly because
he had to work forever with swords that seemed of no use to the
crazy boy, who insisted on smashing them all, and racing off to the
woods, merely demanding as he went a better and a stronger
weapon.
“There is a blade that he could not break,” muttered the Nibelung, as
he worked. “Nothung he would find firm in his hands, but I cannot
weld the splinters. Ah! if I could, I should be well repaid.” He paused,
and then went on, mysteriously murmuring to himself:
“Fafner, the great, wicked worm! Well guards he the Rhinegold. Only
Siegfried can overthrow him. This can only be done by Nothung, I
feel sure. And, alas! I cannot shape Nothung, the sword.”
He began to hammer once more, grumbling continually because
Siegfried insisted that he should make swords, and snarling with
rage because every weapon he forged fell to pieces in the boy’s
strong hands.
Suddenly, from without, came a clear, merry voice, shouting a blithe
“Hoyho!” and the next moment in came Siegfried himself, leading a
great bear, which he had harnessed with a bit of rope.
“Ask the foolish smith if he has finished the sword, Bruin!” he cried to
the bear, and, holding back the great creature firmly, he pretended to
chase Mime, who, springing behind the anvil, cried, savagely:
“Take him away! I don’t want the bear! I have done my best with your
sword.”
“Good!” laughed the boy. “Good-bye, Bruin; run away,” and he freed
the great creature, sending him lumbering off into the woods again.
Then, turning to the trembling Nibelung, he again asked for the
sword, and Mime handed it to him. The young Volsung took it into his
hands quickly, scorn on his handsome face and anger in his eyes.
He was dressed in a wild forest costume of wolf-skins, and his yellow
hair curled over his shoulders. He, indeed, made a great contrast to
Mime, and one could not wonder that they did not get on well
together.
“What a toy!” he cried out. “Do you call this a sword?” and, striking it
on the anvil, he broke the blade into a hundred slivers, and then
burst into a rage with the smith, who had pretended to give him a
sword fit for battle, and had shaped him so foolish a switch, as he
called it. And finally, thoroughly out of breath, he flung himself upon
the stone couch at one side, and not all Mime’s coaxings could
appease his anger. He finally confessed that he did not know why he
ever returned to the cave, because, he said frankly, he could not
help detesting the Dwarf, and was much happier when away from
him. And then he broke into a passionate description of the wood-life
he loved so well; the mating of the birds in the spring-time, and the
way they loved and helped each other; the care that the mother deer
lavished upon her little ones; the tenderness among all the forest
creatures that seemed so beautiful and mysterious to him.
“I learned watching them,” said Siegfried, almost sorrowfully, “what
love must be. Mime, where is she whom I may call mother?”
“Nonsense!” said Mime, and tried to draw Siegfried’s mind away
from the dangerous topic; for he had never told him anything about
his parents, always calling him his own son. And he feared the boy’s
anger if he should ever know that he had been deceived.
But, thoroughly aroused, the young Volsung fiercely demanded the
names of his father and mother, declaring that he was far too unlike
Mime to be his son. At last the Nibelung confessed the truth, and told
him the story of his mother’s death, and of how she had left her child
in his care. And, when the boy asked for proof, he slowly crept away,
to return with the broken sword Nothung, the mending of which was
so hard a riddle even to his sly brain.
Wildly excited, Siegfried commanded him to work at it anew and do
his best to weld the pieces; and, with a shout of delight and hope, he
went merrily away into the woods, leaving Mime in saddest, deepest
perplexity.
Despairing, he murmured at the hopelessness of the task, which his
rather unruly young charge had set him, and was sitting, a picture of
discouragement and misery, when from the dark woods came a
stranger clad as a wanderer, and bearing a great spear. He
advanced to the door of the cave and asked in slow, grave tones for
rest and shelter. Mime was at first frightened, then angry, and finally
refused to harbor the strange guest, until the Wanderer made the
following proposal: Mime was to ask him three questions, and if they
were not correctly answered the host should have the privilege of
cutting off his guest’s head. To this Mime consented, and, after a
little thought, thus chose his first question:
“Tell me what is the race down in the earth’s depths?”
And the Wanderer made answer: “In the earth’s depths dwell the
Nibelungs. Nibelheim is their land. Once they were ruled by Black
Alberich, who owned a magic Ring by which he possessed untold
wealth. What is the next question?”
Again Mime pondered.
“Now, Wanderer, since you know so much of the earth’s depths,” he
said, “tell me what is the race that dwells upon its surface?”
“The giants dwell upon its surface. Two of them, Fasolt and Fafner,
fought for Black Alberich’s hoard. Fafner guards it now as a dragon.
Put your third question!”
“What race dwells in the sky above?” demanded Mime.
And the Wanderer answered, majestically:
“The gods dwell above in Walhalla. Their King is Wotan, who owns a
spear made of the World-Ash. With that spear he rules the world.”
And, as he spoke, Wotan, the Wanderer, struck the earth with the
haft he held, and a peal of thunder crashed suddenly out upon the
silence.
As Mime cowered, terror-stricken, recognizing his guest, the
Wanderer again spoke.
He said it was only fair that he should have the same right he had
given to Mime, and declared that he should ask three questions with
the privilege of cutting off the Nibelung’s head if they were not
answered aright.
“Tell me, O Dwarf,” he began, “what was that race which Wotan
loved, and yet treated harshly?”
“The Volsungs,” answered Mime, partially recovering from his terror.
“Siegmund and Sieglinde were descended from the race. Siegfried is
their son—the strongest Volsung who ever lived.”
“Well answered!” said the Wanderer. “Now listen and reply! A sly
Nibelung watches Siegfried, knowing that he is fated to kill Fafner,
the Dragon. What sword must he use to kill him?”
“Nothung!” cried Mime, eagerly. “Nothung is the name of the sword.
Siegmund once drew it from a great tree. It was broken by the spear
of Wotan. Now a clever smith”—and he rubbed his hands gleefully
—“understands all this, and he hoards well the splinters, knowing
that with these alone can Siegfried kill the Dragon.”
The Wanderer burst out into laughter.
“But who will mend the sword?” he asked.
Mime sprang to his feet in despair, filled with terror and rage; for that
was the one question he could not answer—that was his riddle, his
everlasting mystery.
Quietly Wotan rose from the hearth where he had been sitting.
“I gave you three chances to ask me the question which I have now
asked you. Foolishly, you let them all slip by. Listen while I answer it!
Only he who has never felt fear can forge Nothung anew.”
He strode to the door of the cave, and there paused, looking back.
“Guard well your head, O Dwarf! I leave it to him who knows not
fear.”
Smiling quietly, the Wanderer disappeared in the wood’s depths, and
thunder and lightning followed him as he went.
Mime was left—puzzled, despairing, terror-stricken. His vivid
imagination began to conjure up before him visions of Fafner, the
Dragon, and he had fallen behind the anvil, so great was his fear,
when Siegfried came hastily in, asking once more for the sword.
Mime, creeping out from behind the anvil, could not at once collect
his scattered wits, and merely muttered:
“Only he who has never felt fear can forge Nothung anew. My wits
are too wise for that job.”
Finally, as Siegfried demanded why he had not worked at the sword,
he said, slowly:
“I was fearing for your sake.”
“Fearing!” said Siegfried. “What do you mean by fearing?”
Mime described the tremblings, shudderings, and quakings aroused
by fear, and Siegfried remarked, as he finished:
“All that must seem very queer. I rather think I should like to feel all
that—but how shall I learn?”
Mime, delighted, told him of Fafner, and said that the Dragon would
teach him, or any one else, the art of fearing, and ended by
promising to lead him to Hate Hole the next day.
“Does the world lie that way?” asked the boy.
“To Hate Hole it is close at hand,” responded the wicked little
Nibelung, beginning to feel rapture glow in his heart.
But, when Siegfried again demanded the sword, the smith fell once
more into despair, wailing that he could not shape it, that only one
who knew not fear could forge it anew.
Straight to the hearth sprang the strong young Volsung with the
splinters of Nothung.
“My father’s blade will I forge!” he cried; and he began to move about
merrily, brightening the fire and hunting for the file with which to work
on the broken blade.
Mime watched him with wondering eyes. So swiftly and well did he
work that even the clever smith could not understand. And, as he
dragged at the rope of the bellows and blew up the fire in the forge,
this is the song that Siegfried sang:

“Nothung, Nothung, notable sword!

Who did thy bright steel shiver?
To shreds I have shattered the noble blade,
 In the pot I shall melt each sliver.

“Oho, oho, aha, aha, oho!

Bellows blow,
Brighten the glow!

“Far in the woodlands wild and fair,

’Mid the thickets, a tree felled I;
I have burned the brown ash into coal,
On the hearth I have piled it high.

“Oho, oho, aha, aha, oho!

Bellows blow,
Brighten the glow!

“The coal from the tree how bravely it flames!

The fire how fierce to see!
It sends its wild sparks scattering far,
And the steel shreds it smelts for me.

“Oho, oho, aha, aha, oho!

Bellows blow,
Brighten the glow!”
 SIEGFRIED AT THE FORGE

Meanwhile Mime was busy about something, too. He was preparing

a poison for Siegfried, which he did not intend to give him until after
he had slain the Dragon. Round, round the cave capered the Dwarf,
filled with delight at the pleasant prospect he saw before him.
At last the sword was finished, and Siegfried fitted it into its handle. It
was mended anew.
Waving it aloft, he broke into a new verse of his song:

“Nothung, Nothung, new and young!

I have given thee life and might!
Dead and desolate hast thou lain,
Now leapest thou fearless and bright.
Show now thy sheen to the cowards all,
Shatter deceit, and on falsehood fall.”

He sprang to the anvil and swung the blade high in the air.
“See, Mime, so serves Siegfried’s sword!” he shouted, exultantly.
Down came the flashing steel, and the anvil was shattered in pieces.
Mime sank to the ground in terror, but, holding his father’s sword
above his head, and filled with absolute joy and triumph, stood
young Siegfried—he who had never felt fear, and who had forged
Nothung anew.
 Siegfried Motif
[Listen] [MusicXML]

Motif of the Niebelungs’ Hate

[Listen] [MusicXML]
 Siegfried’s Horn-call
[Listen] [MusicXML]

CHAPTER II

HATE HOLE
To Hate Hole, in the dark time before dawn, came the Wanderer,
and found Alberich waiting and watching near the entrance. The
Dwarf was fearfully enraged at the sight of the old god, whom he
hated with all the strength of his wicked Nibelung soul. He burst into
a torrent of abuse and anger as Wotan drew near, speaking of the
broken promise of the giants and the deceit by which the Gold had
been obtained from the Nibelungs, and again threatening the
downfall of the gods when the Ring should come back to his hands.
The Wanderer answered quietly that a hero was even then drawing
near through the woods—a hero fated to kill Fafner and obtain the
Gold; and, with hidden sarcasm, he bade the Dwarf attempt to use
the youth for his own ends.
The King God believed in the workings of Fate. The Norns wove
continually, and all that they wove came to pass. No one could
change the histories wound into their golden cord, until the Dusk of
the Gods had come, when they also would, in the Last Twilight, be
gone forever. So, feeling as he did, it mattered very little whom he
aided, whom he harmed. He even went so far as to arouse Fafner
for Alberich, and ask him to give the Dwarf the Ring. The old Dragon
snarled and yawned and went to sleep again. The Wanderer turned
to the Nibelung, with a great laugh.
“Listen!” he said. “Remember, O Alberich, what I say. All things work
in nature’s course. You can alter nothing.”
And, so saying, he vanished in the dark woods, and a faint, pale
flicker of lightning shot through the forest as he went. Alberich crept
hastily into a crevice in the rocks on one side, and the dawn broke
just as two figures came into the little green glade by Hate Hole.
The figures were those of Siegfried and Mime; for the Nibelung, true
to his word, had led the boy to the place where he was to learn to
fear.
“If you do not hastily discover fear here, my dear boy, you never will
anywhere,” said the Dwarf, with a chuckle. And he described at great
length the means which Fafner would use to teach the art, saying
that the Dragon’s breath was fire, and his twisting tail strong enough
to crush any hero. But Siegfried merely laughed, and said that he
would find the great worm’s heart and strike Nothung into that; and
then he bade Mime be gone. The Nibelung crept away out of sight
among the trees, and as he went he muttered, in an exasperated
undertone:
“Fafner and Siegfried! Siegfried and Fafner! Oh, that each might kill
the other!”
The boy, left alone, sat under a linden-tree, looking up through the
branches. At first Mime’s figure pervaded his brain, and he could not
help remembering the horrible little creature. But, after a while,
thoughts of his mother crept in—very vague and formless thoughts—
for this forest youth had never in his life seen a woman. Leaning
back, he gave himself up to the enchantment of the summer-day,
dreaming boyish dreams, and listening to the forest voices all around
him.
Have you ever sat in a great, green wood and watched the soft
flickering shadows from the little leaves overhead dance back and
forth on the moss? Have you heard the great surge of music made of
a thousand tiny sounds, the hum of little, unseen insects, the ripple
of far-away brooks, the faint sigh of the wind in the tall reeds, the
rustling of the trees, the melodies that seemed made by the touch of
some master-hand on a great harp? That was what Siegfried saw
and heard that summer day when he lay under the linden-tree and
dreamed day-dreams.
After a while a little bird began to sing in the tree above him, and
after listening for a moment, and wondering whether it brought him a
message from his mother, he resolved to try to imitate it,
remembering that Mime had once said that some people were able
to talk with the birds. So he fashioned a flute out of a reed and tried
to play upon it the melody that the bird sang. Finally, however, he
gave it up in despair, and instead, as he began to feel lonely, he blew
a loud blast on his horn—to bring him a friend, he said to himself.
 THE DEATH OF THE DRAGON

And what sort of a friend do you think it brought him?

Well, it waked Fafner, the monster worm; and he dragged his huge
scaly body to the door of the cave and peered out, and you may
fancy like what sort of a friend he looked.
Siegfried burst out into laughter when he saw him.
“At last!” he cried, merrily. “My call has brought me something truly
lovely!”
“What is that?” growled Fafner, glaring at him as though he were a
small insect of some sort.
“Hey! You can talk, can you?” cried Siegfried. “Being so wise, you
should be able to teach me how to fear. I have come for that.”
Fafner laughed, and showed his teeth, bidding the boy come and be
eaten.
“I come, growler!” said the young Volsung; and, drawing his sword,
he sprang boldly at the great, hideous creature at the cave’s
opening. Fafner reared to receive him, and the combat began. It was
fierce, but not very long, for the boy was strong and Nothung was
sharp, and soon Alberich’s spell had again worked its misery; and,
indeed, it could be said of the dying Dragon that his death was sad—
his life had been a failure.
Before he died he told Siegfried to beware of Mime, and then spoke
slowly and sadly of the race of giants that had come to an end.
“Siegfried,” he began once more—but he never finished, poor old
Dragon; for, just at the word, he rolled over and died. And that was
the end of the race of giants.
Stooping down, the young warrior drew his sword from out the
Dragon’s heart. In so doing, a drop of blood fell on his hand. It
burned like the cruellest fire. He raised it quickly to his mouth to
relieve the smarting; and, as the blood touched his lips, a strange
thing happened—he could understand the language of birds. Yes, as
the same little singer that he had heard before began to twitter, he
could understand what it was saying to him.
“Hey! Siegfried will have now the Nibelung’s hoard! He will find the
hoard in the hole. The Tarnhelm would aid him through wonderful
deeds; but the Ring would give him might over the world.”
With a laugh and a word of thanks to the little singer, the boy
stepped into the cave to look for the treasure. At the same minute
Mime crept near from behind the clump of bushes. Alberich sprang
out from his rocky crevice, and the two little Nibelungs met, snarling,
capering, and making faces with rage.
Each claimed the Ring, and called the other names, and each
proved himself a marvel in wickedness and greed, and they were
nearing a point when blows were not far off when the hero himself
stepped out from the cave with the Tarnhelm thrust into his belt, and
the Rhinegold Ring upon his finger. The dwarfs hastened out of
sight.
The heaped-up hoard of the Nibelungs, Siegfried had left, for he
knew little of its use, and he cared nothing for wealth. Indeed, both
Helm and Ring he had taken only because the bird had so advised
him. He could not fancy what good either of them would do him.
“Hey! Siegfried has now the Helm and the Ring!” sang the wood-bird
in the tree. “Trust not in Mime! The Dragon’s blood will tell Siegfried
what the treacherous Dwarf really means.”
At this point, Mime himself appeared, smiling and bowing, and
holding in his hands a horn of wine for Siegfried. He said that it
would refresh the boy after his labors, but we know that it was
poisoned. Thanks to the Dragon’s blood, Siegfried knew it too, and
read all the cruel thoughts that were passing through Mime’s brain,
and, in a burst of anger, he finally raised his sword and killed the
treacherous Dwarf with one blow.
So that was the end of the Nibelung Mime, the cleverest smith, they
say, that ever lived in the world—even though he could not fashion
Nothung, the sword. From a black crevice in the rocks came
Alberich’s laugh, loud and mocking—the echo of his own wicked
thoughts.
Siegfried turned away wearily, and, seating himself under the linden,
listened for the bird’s song again. As it did not come at once, he
looked up into the branches and spoke:
“You seem very happy, flying among your brothers and sisters,
birdie. But I am all alone. I have no brothers nor sisters, and my
father and mother are both dead. Tell me where I may find a loving
friend. I have called one so often, but none ever comes.” He sighed.
“Sing now, sing,” he begged; and again the bird’s twitter sounded
from among the leaves above him.
“Hey! Siegfried has slain now the wicked Dwarf. I know where he’ll
find a glorious bride. On a rock she sleeps amid fire. If he passed
through the blaze and awakened her, Brünnhilde would then be his.”
Wild with excitement and joy, Siegfried sprang to his feet and asked
if he would really be able to do this.
“Brünnhilde is won only by him who knows not fear,” said the wood-
bird, and flew off before him, guiding him through the woods.
In a transport of joy Siegfried followed, and, shouting with delight, he
began his journey to the far-away rock in its circle of flame, where
the Walküre, in her long penance of sleep, waited for the hero brave
enough to pass through the fire and awaken her.