Intermediate Statistics A Modern Approach 3rd

Edition James P Stevens - PDF Download (2025)

https://ebookultra.com/download/intermediate-statistics-a-modern-
approach-3rd-edition-james-p-stevens/

Visit ebookultra.com today to download the complete set of

ebooks or textbooks
We have selected some products that you may be interested in
Click the link to download now or visit ebookultra.com
for more options!.

The Modern Presidency Sixth Edition James P. Pfiffner

https://ebookultra.com/download/the-modern-presidency-sixth-edition-
james-p-pfiffner/

Modern Rhetorical Criticism 3rd Edition Roderick P Hart

https://ebookultra.com/download/modern-rhetorical-criticism-3rd-
edition-roderick-p-hart/

Respiratory Disease A Case Study Approach to Patient Care

3rd Edition James Dexter

https://ebookultra.com/download/respiratory-disease-a-case-study-
approach-to-patient-care-3rd-edition-james-dexter/

Project Planning Scheduling Control 3rd Edition James P.

Lewis

https://ebookultra.com/download/project-planning-scheduling-
control-3rd-edition-james-p-lewis/
Elementary and intermediate algebra a combined approach
6th Edition Jerome E. Kaufmann

https://ebookultra.com/download/elementary-and-intermediate-algebra-a-
combined-approach-6th-edition-jerome-e-kaufmann/

Computer Vision A Modern Approach David A. Forsyth

https://ebookultra.com/download/computer-vision-a-modern-approach-
david-a-forsyth/

An Integrated Approach to Intermediate Japanese Workbook

Akira Miura

https://ebookultra.com/download/an-integrated-approach-to-
intermediate-japanese-workbook-akira-miura/

Intermediate Algebra with Applications 3rd Edition Wesner

https://ebookultra.com/download/intermediate-algebra-with-
applications-3rd-edition-wesner/

SPSS for intermediate statistics use and interpretation

2nd ed Edition Nancy Leech

https://ebookultra.com/download/spss-for-intermediate-statistics-use-
and-interpretation-2nd-ed-edition-nancy-leech/
Intermediate Statistics A Modern Approach 3rd Edition
James P Stevens Digital Instant Download
Author(s): James P Stevens
ISBN(s): 9780805854664, 0805854665
Edition: 3
File Details: PDF, 5.12 MB
Year: 2007
Language: english
James P. Stevens

Lawrence Erlbaum Associates

New York London

Cover design by Kathryn Houghtaling.

Lawrence Erlbaum Associates Lawrence Erlbaum Associates

Taylor & Francis Group Taylor & Francis Group
270 Madison Avenue 2 Park Square
New York, NY 10016 Milton Park, Abingdon
Oxon OX14 4RN

© 2007 by Taylor & Francis Group, LLC

Lawrence Erlbaum Associates is an imprint of Taylor & Francis Group, an Informa business

Printed in the United States of America on acid-free paper

10 9 8 7 6 5 4 3 2 1

International Standard Book Number-13: 978-0-8058-5466-4 (Softcover) 978-0-8058-5465-7 (Hardcover)

No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic,
mechanical, or other means, now known or hereafter invented, including photocopying, microfilming,
and recording, or in any information storage or retrieval system, without written permission from the
publishers.

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

Preface xi

Chapter 1
INTRODUCTION 1

1.1 Focus and Overview of Topics

1.2 Some Basic Descriptive Statistics
1.3 Summation Notation
1.4 t Test for Independent Samples
1.5 t Test for Dependent Samples
1.6 Outliers
1.7 SPSS and SAS Statistical Packages
1.8 SPSS for Windows—Release 12.0
1.9 Data Files
1.10 Data Entry
1.11 Editing a Dataset
1.12 Splitting and Merging Files
1.13 Two Ways of Running Analyses on SPSS
1.14 SPSS Output Navigator
1.15 SAS and SPSS Output for Correlations, Descriptives, and t Tests
1.16 Data Sets on Compact Disk
Appendix Obtaining the Mean and Variance on the Tl–30Xa Calculator

Chapter 2
ONE WAY ANALYSIS OF VARIANCE 45

2.1 Introduction
2.2 Rationale for ANOVA
2.3 Numerical Example
2.4 Expected Mean Squares
2.5 MSw and MSb as Variances

v
vi CONTENTS

2.6 A Linear Model for the Data

2.7 Assumptions in ANOVA
2.8 The Independence Assumption
2.9 ANOVA on SPSS and SAS
2.10 Post Hoc Procedures
2.11 Tukey Procedure
2.12 The Scheffé Procedure
2.13 Heterogeneous Variances and Unequal Group Sizes
2.14 Measures of Association (Variance Accounted For)
2.15 Planned Comparisons
2.16 Test Statistic for Planned Comparisons
2.17 Planned Comparisons on SPSS and SAS
2.18 The Effect of an Outlier on an ANOVA
2.19 Multivariate Analysis of Variance
2.20 Summary
Appendix

Chapter 3
POWER ANALYSIS 105

3.1 Introduction
3.2 t Test for Independent Samples
3.3 A Priori and Post Hoc Estimation of Power
3.4 Estimation of Power for One Way Analysis of Variance
3.5 A Priori Estimation of Subjects Needed for a Given Power
3.6 Ways of Improving Power
3.7 Power Estimation on SPSS MANOVA
3.8 Summary

Chapter 4
FACTORIAL ANALYSIS OF VARIANCE 123

4.1 Introduction
4.2 Numerical Calculations for Two Way ANOVA
4.3 Balanced and Unbalanced Designs
4.4 Higher Order Designs
4.5 A Comprehensive Computer Example Using Real Data
4.6 Power Analysis
4.7 Fixed and Random Factors
4.8 Summary
Appendix Doing a Balanced Two Way ANOVA With a Calculator
 CONTENTS vii

Chapter 5
REPEATED MEASURES ANALYSIS 181

5.1 Introduction
5.2 Advantages and Disadvantages of Repeated Measures Designs
5.3 Single Group Repeated Measures
5.4 Completely Randomized Design
5.5 Univariate Repeated Measures Analysis
5.6 Assumptions in Repeated Measures Analysis
5.7 Should We Use the Univariate or Multivariate Approach?
5.8 Computer Analysis on SAS and SPSS for Example
5.9 Post Hoc Procedures in Repeated Measures Analysis
5.10 One Between and One Within Factor—A Trend Analysis
5.11 Post Hoc Procedures for the One Between and One Within Design
5.12 One Between and Two Within Factors
5.13 Totally Within Designs
5.14 Planned Comparisons in Repeated Measures Designs
5.15 Summary

Chapter 6
SIMPLE AND MULTIPLE REGRESSION 219

6.1 Simple Regression

6.2 Assumptions for the Errors
6.3 Influential Data Points
6.4 Multiple Regression
6.5 Breakdown of Sum of Squares in Regression and F Test for Multiple Correlation
6.6 Relationship of Simple Correlations to Multiple Correlation
6.7 Multicollinearity
6.8 Model Selection
6.9 Two Computer Examples
6.10 Checking Assumptions for the Regression Model
6.11 Model Validation
6.12 Importance of the Order of Predictors in Regression Analysis
6.13 Other Important Issues
6.14 Outliers and Influential Data Points
6.15 Further Discussion of the Two Computer Examples
6.16 Sample Size Determination for a Reliable Prediction Equation
6.17 ANOVA as a Special Case of Regression Analysis
6.18 Summary of Important Points
Appendix The PRESS Statistic
viii CONTENTS

Chapter 7
ANALYSIS OF COVARIANCE 285

7.1 Introduction
7.2 Purposes of Covariance
7.3 Adjustment of Posttest Means
7.4 Reduction of Error Variance
7.5 Choice of Covariates
7.6 Numerical Example
7.7 Assumptions in Analysis of Covariance
7.8 Use of ANCOVA with Intact Groups
7.9 Computer Example for ANCOVA
7.10 Alternative Analyses
7.11 An Alternative to the Johnson–Neyman Technique
7.12 Use of Several Covariates
7.13 Computer Example with Two Covariates
7.14 Summary

Chapter 8
HIERARCHICAL LINEAR MODELING 321

8.1 Introduction
8.2 Problems Using Single-Level Analyses of Multilevel Data
8.3 Formulation of the Multilevel Model
8.4 Two-Level Model—General Formulation
8.5 HLM6 Software
8.6 Two Level Example—Student and Classroom Data
8.7 HLM Software Output
8.8 Adding Level One Predictors to the HLM
8.9 Addition of a Level Two Predictor to a Two Level HLM
8.10 Evaluating the Efficacy of a Treatment
8.11 Final Comments on Hlm

Appendix A
DATA SETS 365

A.1 Clinical Data

A.2 Alcoholics Data
A.3 Sesame Street Data
A.4 Headache Data
A.5 Cartoon Data
 CONTENTS ix

A.6 Attitude Data

A.7 National Academy of Sciences Data
A.8 Agresti Home Sales Data

Appendix B
STATISTICAL TABLES 399

Table B.1 Critical Values for F

Table B.2 Percentile Points of Studentized Range Statistic
Table B.3 Critical Values for Dunnett’s Test
Table B.4 Critical Values for F (max) Statistic
Table B.5 Critical Values for Bryant-Paulson Procedure

Appendix C
POWER TABLES 413

Table C.1 Power of F Test at α = .05, u = 1

Table C.2 Power of F Test at α = .05, u = 2
Table C.3 Power of F Test at α = .05, u = 3
Table C.4 Power of F Test at α = .05, u = 4
Table C.5 Power of F Test at α = .10, u = 1
Table C.6 Power of F Test at α = .10, u = 2
Table C.7 Power of F Test at α = .10, u = 3
Table C.8 Power of F Test at α = .10, u = 4

References 423

Answers to Selected Exercises 431

Author Index 453

Subject Index 457

 Preface
This book is written for behavioral and social science students at the advanced un-
dergraduate or beginning graduate level. The text emphasizes conceptual under-
standing, the effective use of statistical software to run the analyses, and the correct
interpretation of results. Two statistical software packages, SAS and SPSS, are an
integral part of each chapter. An annotated printout is given from at least one of the
programs for each analysis. The annotations highlight what the numbers mean and
how to interpret the results. The explanation appears on the printout or on the same
page to enhance learning efficiency. The assumptions underlying each analysis are
given special attention, and the reader is shown how to test the critical assump-
tion(s) using SAS and SPSS. Power analysis is an integral part of the book. There
are no computational formulas in this text. I took the position that they were not
needed many years ago, and it is even truer today.
The instructional mix of strategies that is employed to illustrate each statistical
technique consists of two parts (a) First, I use definitional formulas on small data
sets to convey conceptual insight into what is being measured, and (b) Then, I pro-
ceed directly to the packages to efficiently process data. I feel very strongly about
using these strategies.
The most significant change in this edition is the addition of a chapter on hierar-
chical linear modeling using HLM6. This material is important because correlated
observations occur frequently in social science research and just a SMALL amount
of dependence causes the type I error rate to be several times greater than one
wishes! Since HLM involves a series of regressions, this new chapter is placed af-
ter the material on regression. The distinction between fixed and random factors is
important, and so it is emphasized. The chapter on HLM was written by Dr.
Natasha Beretvas of the University of Texas at Austin. I thank her very much for
her contribution.
The third edition features newer versions of SPSS (Release 12.0) and SAS (Re-
lease 8.0). Much of the material on importing data into SAS or SPSS that previ-
ously appeared in chapter 1 was deleted. Importing data into these two programs is
now much easier so this material was no longer necessary.
The exercises involve a mixture of numerical, conceptual and computer related
problems. I have de-emphasized purely numerical exercises, for I agree entirely
with Cobb (l987, p. 323) that, “computing rules are just the skin of our subject; it is
focus that reveals the skeleton of fundamental concepts and connections that hold

xi
xii PREFACE

the body of knowledge together.” Regarding exercises, it is important to note that

there are 3 new exercises for each chapter. Answers are provided for half of the ex-
ercises and an Instructor’s Solutions CD is available to adopters. A computer ex-
ample of real data integrates many of the concepts. A CD containing all of the
book’s data sets is included in the back of the book.
The reader should have a background of a one quarter course in statistics that
covered at least the t tests for independent and dependent samples.
I am very grateful to the reviewers of this text: Dale Berger of Claremont Grad-
uate University, Michael Milburn of University of Massachusetts, Mary Lou
Kerwin of Rowan University, Gordon Brooks of Ohio University, and Roderick
Gillis of the University of Miami. I am also indebted to some individuals at my
publisher. Larry Erlbaum continues to be very supportive. Debra Riegert was in-
strumental in motivating me to write this third edition.

Jim Stevens
 1
Introduction

CONTENTS
1.1 Focus and Overview of Topics
1.2 Some Basic Descriptive Statistics
1.3 Summation Notation
1.4 t Test for Independent Samples
1.5 t Test for Dependent Samples
1.6 Outliers
1.7 SPSS and SAS Statistical Packages
1.8 SPSS for Windows—Release 12.0
1.9 Data Files
1.10 Data Entry
1.11 Editing a Dataset
1.12 Splitting and Merging Files
1.13 Two Ways of Running Analyses on SPSS
1.14 SPSS Output Navigator
1.15 SAS and SPSS Output for Correlations, Descriptives, and t Tests
1.16 Data Sets on Compact Disk
Appendix Obtaining the Mean and Variance on the Tl–30Xa Calculator

1.1 FOCUS AND OVERVIEW OF TOPICS

This book has been written for applied social science researchers at the advanced
undergraduate or beginning graduate level. It is assumed that you have had a one
quarter course in beginning statistics that covered measures of central tendency,
measures of variability, standard scores (z, T, stanines, etc.), correlation, and infer-
ential statistics, including at least the t tests for independent and dependent sam-
ples. In the next four sections of this chapter, we review briefly some descriptive
statistics, summation notation, and testing for a “significant” difference. These

1
2 CHAPTER 1

sections are not intended to thoroughly teach this material again, but to refresh
your memory.
The emphasis in the book is on conceptual understanding of the statistical tech-
niques, learning how to effectively use statistical software to run the analyses, and
learning how to interpret the computer printout that results from such runs. The
two major statistical packages, SAS (Statistical Analysis System) and SPSS (Sta-
tistical Package for the Social Sciences), are an integral part of this book. Details
on SAS and SPSS are given in Section 1.7. I have attempted to make the text as
practical as possible. To accent the practical emphasis, nine real data sets have
been provided in Appendix A in the back of the book. For convenience, these data
sets are also available on a CD. Some of the exercises in the chapters involve run-
ning these data sets, or a part of a real data set. Singer and Willett (1988) have pro-
vided an excellent annotated bibliography, indicating where numerous other real
data sets may be found.
The instructional mix of strategies adopted to illustrate each statistical tech-
nique involves two parts:

1. First, we illustrate each technique using definitional formulas on small data

sets. These formulas are useful in yielding conceptual insight into what is being
measured or quantified. As a simple example, the definitional formula for sample
variance is

s2 = [( x1 - x )2 + ( x2 - x )2 +L + ( xn - x )2 ] /(n -1)

This formula shows very clearly that variance is measuring how much the
scores for the subjects scatter or disperse about the mean.
2. Then we move directly to the computer, that is, to the statistical packages,
to show how to efficiently process data. And more importantly, how to interpret the
printout from the packages. In practice, analyses will very likely be run on one or
more of these packages, and thus it is important to become familiar with them.

Now we give an overview of the topics in the book. The reader may recall
that the t test for independent samples is appropriate for comparing two groups
to determine whether they differ on the average on a dependent variable. But
what if we wish to compare more than two groups simultaneously on a depend-
ent variable? For example, we wish to compare the effect of four counseling
methods on attitude toward education. Then a statistical procedure called analy-
sis of variance is needed. This technique is covered in Chapter 2. Suppose that
for this example there was reason to believe that the sex of the subjects might
moderate the effect of the counseling methods, and we wanted to check this pos-
sibility. This would lead us to a more complicated analysis of variance design,
since we are examining the effect of two independent variables (sex and counsel-
 INTRODUCTION 3

ing method) on attitude toward education. It is an example of a factorial design.

These designs are covered in Chapter 4.
Chapter 3 deals with power analysis. The power of a statistical test is the proba-
bility of rejecting the null hypothesis when it is false. Although it may seem obvi-
ous that we would want to achieve this, many researchers in the literature have
failed to do so, as Cohen (1969) and others have pointed out. The reason is that
power is generally inadequate with small group sizes (especially with 20 or less
subjects per group), and in some areas of research such sample sizes are quite com-
mon for pragmatic or other reasons. Chapter 3 provides a detailed and practical ap-
proach to estimating the power of completed studies and also for estimating sam-
ple size required for adequate power in an upcoming study.
In Chapter 5 we treat the class of situations in which the same subjects are mea-
sured more than twice on a dependent variable. For example, suppose a dietitian
wishes to assess the immediate and long term effects of a behavior modification
approach on weight loss for a group of overweight men. She measures the weight
loss immediately following treatment and then 7 additional times (in three month
intervals) over a two year period. The appropriate statistical analysis here is a dif-
ferent type of analysis of variance from that in Chapter 2, called repeated measures
analysis. The simplest case of a repeated measures design measures the subjects
just twice, e.g., pretest—treatment—posttest. The investigator is interested in test-
ing for a significant gain or change on the dependent variable, and the appropriate
test is the t test for correlated (dependent) samples that you studied in beginning
statistics.
Chapter 6, which is a new addition to my intermediate text, deals with multiple
regression. Much of the material is taken from my multivariate text (Stevens,
1996). Multiple regression is a much used and abused technique. One of the prob-
lems is that many researchers use multiple regression without validating their re-
sults on an independent sample of data. I have made validating the model a major
theme in this chapter.
Analysis of covariance is now found in Chapter 7. This technique combines
analysis of variance and regression analysis. Because of this, and because of the
suggestions of two reviewers of this edition, I have covariance after regression and
ANOVA. A covariate is a variable that is significantly correlated with the depend-
ent variable. Analysis of covariance can be quite helpful in randomized studies,
that is, studies where the subjects have been randomly assigned to the treatments,
in increasing the sensitivity (power) of an experiment.
Analysis of variance procedures and multiple regression are used very often in
the literature. Thus it is important to learn this material in order to be able to intelli-
gently and critically read the literature.
4 CHAPTER 1

1.2 SOME BASIC DESCRIPTIVE STATISTICS

The measure of central tendency that is used most frequently is the mean or aver-
age for a set of scores. It is defined as

x = ( x1 + x2 + L + xn ) / n

where n is the number of subjects and x1, is the score for subject 1 on variable x,
x2 is the score for subject 2, etc. The mean is an example of a summary statistic—it
summarizes an important or salient feature of a set of data. For example, if you are
told that the average weight for a pro football lineman is 280 pounds, or that the av-
erage income of people living in a certain community is $80,000, each of these
numbers packs a message. The average weight of 280 pounds indicates that the
weights of most linemen tend to cluster around that value, and the income of
$80,000 means that the incomes of most people in that community cluster around
$80,000. These statements are accurate provided that there are no extreme values
or outliers (see Section 1.6).
Although the mean is useful in characterizing one important feature of a set of
data, it can be misleading just by itself. To see this consider the following scores
for three groups of 10 children each on a 20 item pretest in mathematics:

Group 1 Group 2 Group 3

10 10 10
13 11 18
7 11 2
12 10 13
13 12 17
11 11 3
8 11 8
14 10 15
9 12 19
12 11 4

x1 = 10.9 x2 = 10.9 x3 = 10.9

On the average there is no difference between these three groups of children.

However, there is a major difference among the groups in terms of variability of the
scores about the mean. One can see intuitively that there is the least variability for
group 2 (since the scores cluster very tightly about the mean of 10.9), while vari-
ability is greatest for group 3. This differential amount of variability would have
definite instructional implications, if you had to teach one of these three groups
mathematics. Other things being equal, group 2 would be easier to teach since they
are all at about the same level of ability.
 INTRODUCTION 5

To quantify the amount of variability in a set of scores we use the sample vari-
ance s2, the definitional formula of which is
( x1 - x )2 + ( x2 - x )2 + L + ( xn - x )2
s2 =
n -1
Notice that variance simply measures how much the scores vary about the
mean. Now we find the variances for the three groups of children. Although the
emphasis in this book is on using the computer for doing statistical analysis, there
is a wide array of very inexpensive calculators that are conveniently used for calcu-
lating the mean and variance for a set of data. In Appendix 1 at the end of this chap-
ter we give the details for the TI–30Xa for the children in group 1. The variances
for the three groups are: s12 = 5.43, s22 = .54, and s32 = 41.43.
Summary statistics like the mean and variance are especially useful in compar-
ing different data sets (groups of subjects) on the same variable. Consider the fol-
lowing two sets of scores, which represent the age of 25 automobile salesmen in
the United States and 25 automobile salesmen in Western Europe:

United States Western Europe

23 63 25 22 32 43 26 30 27 40
56 30 34 56 30 35 48 36 47 41
25 48 44 27 26 34 45 30 38 33
38 26 30 39 30 35 44 24 33 40
36 32 36 38 33 31 23 29 37 28

It is far from obvious by just looking at these sets how the ages for the two
groups differ, if at all. Computation of the mean and variance for the groups yields:
U.S. (x = 35.16, s2 = 117.22) and Western Europe (x = 35.08, s2 = 51.16). These
statistics indicate that the average age is about the same for the groups and that the
variability in age for the U.S. salesmen is over twice that for the Western Europe-
ans.

1.3 SUMMATION NOTATION

The reader probably was exposed to the summation operator in an introductory sta-
tistics course. Nevertheless, a brief review of some basic properties of Σ (sigma)
will be helpful. The symbol Σ means “take the sum of.” Suppose we had measured
50 subjects on anxiety. The sum of their scores is

x1 + x2 + x3 … + x50

This sum can be expressed concisely using Σ as follows:

6 CHAPTER 1

50
å xi
i =1

The first term (x1) is obtained by setting i = 1, the second term (x2) by setting i =
2, on down to the last term (x50) for i = 50. The quantity i is called the index of sum-
mation; it is what we are summing on. Let us consider a few more examples to il-
lustrate. Suppose we have measured 75 subjects on a variable y and wish to
represen the sum of those scores using Σ. Then it would look like this:
75
y1 + y2 + L + y75 = å yi
i=1

Or if we had 100 subjects measured on variable z and wish the sum of the scores
for subjects 3 through 100, then we have

100
z3 + z4 + L + z100 = å zi
i =3

If the limits are understood, then they are dropped, and we would just write Σ zi.
Note that the mean for a set of n scores can be written using Σ:

x = ( x1 + x2 +L + xn ) / n = å xi / n

Often we may wish to concisely represent a sum of squares of some type. Sup-
pose we have n subject scores (x1, x2, …xn) and wish to denote the sum of the
squared scores. This is

x12 + x22 + L + xn2 = å xi2

The sample variance for a set of scores involves a sum of squares (squared devi-
ations):
s2 = [( x1 - x )2 + ( x2 - x )2 +L + ( xn - x )2 ] /(n -1)
n
= å ( xi - x )2 /(n -1)
i =1

or as s2 = S( x - x )2 /(n -1) if the limits are understood.

Example
Evaluate Σ xi,2, where x1 = 10, x2 = 8, x3 = 13, and x4 = 5.
 INTRODUCTION 7

å xi2 = x12 + x22 + x32 + x42 = 102 + 82 + 132 + 52 = 358

The following four properties of the summation operator are useful to know:

å ( x + y) = å x + å y
1. summation sum
of sum of summations

å ( x - y) = å x-å y
2. summation of difference
difference in summations

3. å cxi = cå xi (a constant c can be moved across the summation)

To show that property 3 holds note that

å cxi = cx1 + cx2 + L + cxn = c( x1 + x2 + L + xn ) = cå xi

4. å c = nc (summing over n subjects)
The constant c mentioned in properties 3 and 4 can appear in many different
subtle ways. To illustrate that and also to show how to apply several of the above
properties, we will prove that the mean of a set of z scores is 0.
Denote the z scores by z1, z2, . . ., zn. Then by definition of a mean we have

z = å zi / n

To show that z = 0 it suffices to show that Σz = 0.

By definition zi = ( x1 - x ) / s . Therefore by substitution:

å zi = å ( xi - x ) / s
Note that 1/s is a constant here; that is, it does not depend on i (index of summa-
tion). By property 3 we can move it across the summation and write

å zi = (1/ s)å ( xi - x )
Now by property 2 we can further rewrite this as

å zi = (1/ s)[å xi - å x ]
Other documents randomly have
different content
 LEHMBRUCK
Kneeling Woman

the human form, but the human form today is never so symmetrical, so
perfect as in classic sculpture, and one suspects the Greeks themselves
idealized their young men and maidens.
Long before Matisse, Rodin started the “return to nature.” His “Age of
Bronze,” 1877, was so literal a transcript it was denounced as a cast from
life; sculptors and critics refused to believe human fingers could model so
perfect an impression. His “Saint John,” “Eve,” “Bourgeois of Calais,” “Le
Penseur,” “La Belle Heaulmière,” to mention only a few, were all created in a
spirit diametrically opposed to the classic—yet Rodin is a most intelligent
lover of the classic.
Per contra, most of Rodin’s marbles are a fine mixture of the classic and
purely modern—of the classic and the romantic.
The point here is that in some of his bronzes he exhibits as clear and
merciless an observation of nature as Matisse or any other modern. It may be
said once for all that in the number and variety of things he does, in the
manner in which he links past and present, Rodin stands quite alone among
sculptors. If he has little sympathy with the extreme sculpture of the hour it is
because life is short and in his life time he has covered so vast a territory,
responded to so many impulses, ancient and modern, he is not unnaturally
reluctant to embark upon new experiments or interest himself vitally in what
others are doing.

The best American sculpture, even more than American painting, is

solidly virile-impressionistic, notably the work of such men as Barnard and
Borghlum. Davidson has one foot firmly planted within the confines of Post-
Impressionism, but he has by no means cut loose from the past. His
“Decorative Panel” in the Exhibition was purely post-impressionistic, a work
of the imagination, while his figures were virile-impressionistic.

It is only by comparing the work of these new men with that of St.
Gaudens, French, MacMonies—to mention no others—that one begins to
rightly understand what is meant by the “reaction to nature.”
 There is plenty of pure observation and plenty of fine imagination in the
work of those three men, but there is also much of the purely classical, and
not one of them showed or shows any desire to break with tradition, while the
very essence of the modern movement is a disregard, conscious or
unconscious, for tradition; in many of the new men there is a violent revolt
against the domination of the past.

It is when we come to the work of Brancusi and Archipanko that we find

the most startling examples of the reaction along purely creative lines.
Nature is purposely left far behind, as far behind as in Cubist pictures, and
for very much the same reasons.
Of Brancusi something has been said already.
Of all the sculpture in the International Exhibition the two pieces that
excited the most ridicule were Brancusi’s egg-shaped portrait of Mlle.
Pogany and “Family Life” by Archipanko.
Both are creative works, products of the imagination, but in their
inspiration they are fundamentally different.

In his symmetrical oval head with the spiral masses where the neck would
be, it is apparent the sculptor’s interest is in the play of line and relation of
masses, no profound human problem troubled him. That there is a relation
between the strange shape of the head and his theories of life and art no
BOCCIONI
Spiral Expansion of
Muscles in Action
 MATISSE
Portrait Heads

serious observer of his other work could doubt, but his unusual technic over-
shadows other interest.

In his “Family Life,” the group of man, woman, child, Archipanko

deliberately subordinated all thought of beauty of form to an attempt to
realize in stone the relation in life that is at the very basis of human and
social existence.
Spiritual, emotional, and mathematical intellectuality, too, is behind the family group of
Archipanko. This group, in plaster, might have been made of dough. It represents a
featureless, large, strong male—one gets the impression of strength from humps and lumps
—an impression of a female, less vivid, and the vague knowledge that a child is mixed up in
the general embrace. The faces are rather blocky, the whole group with arms intertwined—
arms that end suddenly, no hands, might be the sketch of a sculpture to be. But when one
gets an insight it is intensely more interesting. It is, eventually, clear that in portraying his
idea of family love the sculptor has built his figures with pyramidal strength; they are grafted
together with love and geometric design, their limbs are bracings, ties of strength, they
represent, not individuals, but the structure itself of family life. Not family life as one sees it,
but the unseen, the deep emotional unseen, and in making his group when the sculptor found
himself verging upon the seen—that is, when he no longer felt the unseen—he stopped.
Therefore the hands were not essential. And this expression is made in the simplest way.
Some will hoot at it, but others will feel the respect that is due one who simplifies and
expresses the deep things of life. You may say that such is literature in marble—well, it is the
modernest sculpture.[66]

The group is so angular, so Cubist, so ugly according to accepted notions,

that few look long enough to see what the sculptor means; yet strange as the
group was it undeniably gave a powerful impression of the binding, the
blending character of the family tie, a much more powerful impression than
groups in conventional academic pose could give.
In considering the extreme modern movement in sculpture it must not be
forgotten that groups and figures just as strange have been done in the past—
that even queerer and more grotesque things have been used to adorn
churches and altars.
True, those sculptures and carvings are naive and primitive, but may not
the naive and primitive be closer to life and to life’s great truths than the
sophisticated and classical?
That is the question.
The answer of the moderns is that the swing of the pendulum in art is
from the naive and primitive through the more and more conventional to the
fixed and lifeless mold of the classic and academic, then back again to the
naive, traversing the romantic, in its course, both ways.
MATISSE
Back of Woman
ERBSLOH
Young Woman
 XIV

IN CONCLUSION

T O gather the loose ends of the argument in one skein.

Impressionism was the natural, the inevitable reaction from the romantic
and story-telling art of the forties, fifties, and sixties—a return to nature from
the studio, to works of the observation from works of the imagination.
Impressionism developed along three diverging lines:
A. Superficial Impressionism—Monet.
B. Realistic Impressionism—Manet.
C. Substantial Impressionism—Cézanne.

A. Superficial—the painting of light effects, the impressionism of Monet,

culminated in the extreme refinements of the pointillists, the Neo-
Impressionists, Seurat and Signac.
In superficial Impressionism the last word seems to have been said for the
time being. Any number of delightful pictures—light effects—are being
painted, and will continue to be painted, but the early enthusiasm has largely
subsided.
Superficial Impressionism leads naturally to the painting of pure color
effects—color music, orphism, compositional painting. After the last word in
the observation of light effects Post-Impressionistic attempts to create pure
color effects, irrespective of natural—that is a logical reaction.
B. Realistic Impressionism penetrates a little deeper. While Monet and his
followers, Signac and Seurat, dealt more and more with the play of light on
the surface of things, Manet and his followers painted closer to the heart of
things.
While Monet was content to paint a hay stack twenty times in as many
different lights, Manet preferred a touch of life and character in his pictures.
While he was first and last a painter, he was not so absorbed in securing
purely technical effects as to be wholly blind to the human element, hence his
wonderful portraits, his bullfights, his glimpses of city life—pictures big in
more senses than one.
Still he and his followers were primarily interested in the aspect of things,
the characteristics as distinguished from the fundamental character of
things. He penetrated far deeper than Monet, so much deeper the two had
little in common, but he did not get so close to the heart that he forgot the
skin; he was always a painter of appearances, but in a big as distinguished
from a superficial way.
The realistic Impressionism of Manet has by no means run its course.
Some of the finest painting in the world has been done and is being done
along this line. It is the line of Franz Hals and Velasquez; it is the line of men
so different as Whistler and Sargent in their best portraits.
The natural reaction from perfection in this line is higher accentuation of
characteristics—in the extreme caricature.
That is, given the last word in the painting of character by great men in a
solid way, the logical attempts of new men or lesser men will be the
indication of character in a lighter and more superficial way. The penetrating
observation of the older men gives way to the keen and playful fancies of the
younger. The same sitter yields with the former a powerful portrait, with the
latter a fascinating picture which may be quite as revealing both as a likeness
and as a characterization.
C. Substantial Impressionism is not so easy to define and differentiate. It
is far from superficial but has much in common with realistic.
It is easiest to simply say it is the Impressionism of Cézanne and those
who have read what has already been said about Cézanne will understand.
Cézanne was not content to paint either the surface or the characteristics
of things or people; he sought to go deeper, to get at the very substance and
to place on canvas their elemental qualities.
As a natural result the longer he painted the less interesting his pictures
became superficially, but the greater their interest fundamentally.
While Monet became more and more a popular painter, a painter for the
dealer and the buyer, Cézanne became more and more a painter’s painter,
doing things that only the technically skilled could rightly appreciate.
Interested solely in the profoundest problems of his art and painting only
for those who had a very great knowledge of art, he attracted comparatively
few followers; the path he followed promised little in the way of immediate
fame and rewards.
Still during his last years he had his ardent admirers and after his death his
simple, strong constructive, elemental pictures began to be widely
appreciated.
They make no pretense to the superficial charm of color or composition
that attracts the average observer, but they fascinate every man who studies
things long enough to even partially understand what the artist was so
earnestly trying to do.
Substantial or Cézanne Impressionism led naturally to the Virile-
Impressionism of today, a way of seeing and painting things that is a
compound of the Impressionism of Manet with that of Cézanne.
There is a great and glorious future for Virile-Impressionism. Some of the
greatest portraits and pictures in the world will be painted with the
penetrating vision of a Cézanne, modified by the clear, cool observation of a
Manet.
The logical reaction from carrying observation of nature to the extent
Cézanne carried it is painting of the substance of things creatively,
theoretically, as in Cubism.
Cézanne carried the use of planes imitatively so far that it was but a step to
their use arbitrarily and scientifically.
Substantial Impressionism leads naturally to substantial Post-
Impressionism; or in other words, the substance of things painted
impressionistically (more or less imitatively) leads logically to the painting of
the substance of things creatively = Post-Impressionistically.
KROLL
Still Life
 APPENDIX I

EXHIBITIONS AT 291 FIFTH AVENUE

D URING a number of years prior to 1913 Mr. Alfred Stieglitz gave

exhibitions of extreme modern work in his Small Photo-Secession
Gallery, 291 Fifth Avenue, New York, and the International was the
outcome, the logical culmination of these earlier efforts.
Mr. Stieglitz prepared the following chronological narrative:
In the end of November, 1906, “291” (“Photo-Secession Gallery,” “Little
Gallery,” etc., etc.) was opened with an exhibition of pictorial photography.
The exhibition represented the best work of Steichen, Frank Eugene,
Kasebier, Clarence White, Stieglitz, Coburn, Brigman, Herbert G. French,
and about thirty others, all Americans.
This exhibition was followed up by a series of exhibitions—usually one-
man—of the picked work which had been done in pictorial photography the
world over.
In 1907 the first exhibition not devoted to photography was that of Miss
Pamela Coleman Smith. This exhibition created a sensation. At the time it
aroused the ire of most of the New York critics.
Following this there were shown Willie Geiger’s (Munich) best etchings
and Ex Libris. This was the first show of his in America.
But the real beginning, I suppose, of the so-called Modern work shown at
“291” was the exhibition of about sixty of Rodin’s choicest drawings. These
were selected by Rodin and Steichen for the special exhibition. The
exhibition aroused intense indignation in New York amongst the critics and
amongst most painters (men like Chase, Alexander, and others of this type
feeling that such things were not meant for the public).
April, 1908, Matisse was introduced to the American public for the first
time. This exhibition of Matisse’s represented the complete evolution of
Matisse from his academic period up to date. It included etchings, drawings,
water colors, lithographs, and oil paintings.
January, 1909, the work of Marius De Zayas was introduced for the first
time.
 March, 1909, John Marin and Alfred Maurer (the “new” Maurer) were
introduced. The work of these Americans seemed to upset the equilibrium of
the academicians even more than the “jokes” of Rodin and Matisse.
May, 1909, Marsden Hartley was introduced to the public for the first
time.
December, 1909, Toulouse Lautrec Exhibition. A very choice collection of
his lithographs. First Lautrec Exhibition in America.
February, 1910, second Marin Exhibition.
March, 1910, exhibition of the work of “Younger American Painters”:
Arthur G. Dove, Arthur B. Carles, L. Fellows, Marsden Hartley, Putnam
Brindley, John Marin, Alfred Maurer, Steichen, Max Weber. This was the
first collective exhibition of Modern work by Americans.
April, 1910, second Rodin Exhibition. The very latest drawings of Rodin
were shown, together with eleven of his earliest ones. At the same time the
best small bronze of the “Penseur” (loaned by Mrs. John W. Simpson) was
exhibited.
November, 1910, Exhibition of lithographs by Cézanne, Renoir, Manet,
and Toulouse Lautrec. Together with these, drawings and paintings by Henri
Rousseau, just deceased. This exhibition introduced Rousseau for the first
time to America, as well as it introduced Cézanne.
January, 1911, Exhibition by Max Weber, American.
February, 1911, Marin Exhibition (third).
March, 1911, a series of Cézanne water colors. The first one-man show of
Cézanne’s in America. These water colors were most carefully selected and
really represent a side of Cézanne which is underestimated by all those, even
Cézanne lovers, who do not fully understand Cézanne’s importance.
April, 1911, Picasso. Drawings, lithographs, water colors, etc. A series of
eighty showing the complete evolution of Picasso. The first introduction of
Picasso to America and the first exhibition anywhere of Picasso held in this
sense.
February, 1912, second Hartley exhibition.
February, 1912, first Arthur G. Dove exhibition.
March, 1912, sculptures and latest drawings by Matisse. First introduction
to America of Matisse, the sculptor.
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!

ebookultra.com