A Guide to Doing Statistics in Second

Language Research Using SPSS and R

(Second Language Acquisition
Research Series) – Ebook PDF Version
Visit to download the full and correct content document:
https://ebookmass.com/product/a-guide-to-doing-statistics-in-second-language-resear
ch-using-spss-and-r-second-language-acquisition-research-series-ebook-pdf-version/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Second Language Research: Methodology and Design

https://ebookmass.com/product/second-language-research-
methodology-and-design/

Research Methods, Statistics, and Applications Second

Edition – Ebook PDF Version

https://ebookmass.com/product/research-methods-statistics-and-
applications-second-edition-ebook-pdf-version/

Introduction to 80×86 Assembly Language and Computer

Architecture – Ebook PDF Version

https://ebookmass.com/product/introduction-to-8086-assembly-
language-and-computer-architecture-ebook-pdf-version/

A Simple Guide to IBM SPSS Statistics – version 23.0 –

Ebook PDF Version

https://ebookmass.com/product/a-simple-guide-to-ibm-spss-
statistics-version-23-0-ebook-pdf-version/
Reading Statistics and Research – Ebook PDF Version

https://ebookmass.com/product/reading-statistics-and-research-
ebook-pdf-version/

Statistics Using IBM SPSS: An Integrative Approach –

Ebook PDF Version

https://ebookmass.com/product/statistics-using-ibm-spss-an-
integrative-approach-ebook-pdf-version/

Short Guide to Action Research, A – Ebook PDF Version

https://ebookmass.com/product/short-guide-to-action-research-a-
ebook-pdf-version/

Doing Qualitative Research in Language Education 1st

ed. Edition Seyyed-Abdolhamid Mirhosseini

https://ebookmass.com/product/doing-qualitative-research-in-
language-education-1st-ed-edition-seyyed-abdolhamid-mirhosseini/

Talking About Second Language Acquisition Karim Sadeghi

https://ebookmass.com/product/talking-about-second-language-
acquisition-karim-sadeghi/
Downloaded by [University of California, San Diego] at 02:46 19 March 2017
 CONTENTS
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

Preface xv
Acknowledgments xix

PART I
Statistical Ideas 1

1 Getting Started with the Software and Using the Computer

for Experimental Details 3
1.1 Getting Started with SPSS 3
1.1.1 Opening a Data File 5
1.1.2 Entering Your Own Data 5
1.1.3 Application Activity for Getting Started with SPSS 10
1.1.4 Importing Data into SPSS 11
1.1.5 Saving Your Work in SPSS 11
1.1.6 Application Activities for Importing and Saving Files 13
1.2 Getting Started with R 13
1.2.1 Downloading and Installing R 13
1.2.2 Customizing R in Windows 15
1.2.3 Loading Packages and R Commander 17
1.2.4 A List of All the R Packages Used in this Book 19
1.3 Working with Data in R and R Commander 19
1.3.1 Entering Your Own Data 20
1.3.2 Importing Files into R through R Commander 22
1.3.3 Viewing Entered Data 25
1.3.4 Saving Data and Reading It Back In 25
1.3.5 Saving Graphics Files 27
1.3.6 Closing R and R Commander 27
1.3.7 Application Activities Practicing Entering Data into R 28
1.4 Understanding the R Environment 28
1.4.1 Using R as a Calculator 28
1.4.2 Using R as a Calculator Practice Activities 30
1.4.3 Objects in R 31
1.4.4 Creating Objects in R Practice Activities 32
1.4.5 Types of Data in R 33
1.4.6 Types of Data Practice Activities 35
 viii Contents

1.4.7 Functions in R 35
1.4.8 Functions in R Practice Activities 36
1.4.9 The R Workspace 37
1.4.10 Specifying Variables within a Data Set, and Attaching
and Detaching Data Sets 37
1.5 Missing Data 38
1.5.1 Missing Data and Multiple Imputation in SPSS 38
1.5.2 Missing Data Application Activity in SPSS 43
1.5.3 Missing Data and Multiple Imputation in R 43
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

1.5.4 Missing Data Application Activity in R 46

1.6 Getting Help 46
1.6.1 Getting Help with SPSS 46
1.6.2 Getting Help with R 46
1.7 Summary 49

2 Some Preliminaries to Understanding Statistics 50

2.1 Variables 51
2.1.1 Levels of Measurement of Variables 51
2.1.2 Application Activity: Practice in Identifying
Levels of Measurement 53
2.1.3 Dependent and Independent Variables 54
2.1.4 Application Activities: Practice in Identifying Variables 56
2.1.5 Summary of Variables 57
2.1.6 Fixed versus Random Effects (Advanced Topic) 57
2.2 Understanding Hidden Assumptions about How Statistical Testing Works 58
2.2.1 Hypothesis Testing 59
2.2.2 Application Activities: Creating Null Hypotheses 60
2.2.3 Who Gets Tested? Populations versus Samples and Inferential
Statistics 60
2.2.4 What Does a P-Value Mean? 62
2.2.5 Effect Sizes 65
2.2.6 Understanding Statistical Reporting 65
2.2.7 Application Activities: Understanding Statistical Reporting 69
2.2.8 The Inner Workings of Statistical Testing 70
2.2.9 Application Activity: The Inner Workings of Statistical Testing 73
2.2.10 Summary of Hidden Assumptions 73
2.3 Parametric and Non-Parametric Statistics 73
2.3.1 Why Robust Statistics? 74
2.4 Summary 76

3 Describing Data Numerically and Graphically and

Assessing Assumptions for Parametric Tests 77
3.1 Numerical Summaries of Data 77
3.1.1 The Mean, Median and Mode 78
3.1.2 Standard Deviation, Variance and Standard Error 80
3.1.3 Confidence Intervals 85
3.1.4 The Number of Observations and Other Numerical Summaries
You Might Want to Report 89
 Contents ix

3.1.5 Reporting Numerical Summaries 89

3.1.6 Data for this Chapter 90
3.2 Using SPSS to Get Numerical Summaries 91
3.2.1 Obtaining Numerical Summaries with SPSS
and Splitting Groups 91
3.2.2 Application Activities for Numerical Summaries in SPSS 94
3.3 Using R to get Numerical Summaries 95
3.3.1 Basic Descriptive Statistics in R 95
3.3.2 Application Activities for Numerical Summaries in R 99
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

3.4 Satisfying Assumptions for Parametric Tests 99

3.5 Graphic Summaries of Data: Examining the Shape of Distributions for Normality 100
3.5.1 Histograms 101
3.5.2 Skewness and Kurtosis 104
3.5.3 Stem and Leaf Plots 106
3.5.4 Quantile-Quantile Plots 107
3.6 Obtaining Exploratory Visual Summaries in SPSS 107
3.6.1 Application Activities: Looking at Normality Assumptions 112
3.7 Obtaining Exploratory Visual Summaries in R 112
3.7.1 Creating Histograms with R 113
3.7.2 Creating Stem and Leaf Plots with R 115
3.7.3 Creating Q-Q Plots with R 117
3.7.4 Testing for Normality with R 119
3.7.5 Application Activities: Looking at Normality Assumptions with R 120
3.8 Examining the Shape of Distributions: The Assumption of Homogeneity 121
3.8.1 Checking Homogeneity of Variance (with SPSS or R) 123
3.9 Dealing with Departures from Expectations 124
3.9.1 Outliers 124
3.9.2 Transforming Data 124
3.10 Summary 126

4 Changing the Way We Do Statistics: The New Statistics 128

4.1 Introduction to Confidence Intervals 130
4.1.1 Application Activity for ESCI and Confidence Intervals 131
4.1.2 Interpreting Confidence Intervals 133
4.1.3 Application Activities with Confidence Intervals 137
4.1.4 Confidence Intervals and the Imprecision of P-Values 139
4.1.5 Application Activities with Confidence Intervals and Precision 141
4.2 Introduction to Effect Sizes 141
4.2.1 Understanding Effect Size Measures 143
4.2.2 Interpreting Effect Sizes 144
4.2.3 Calculating Effect Sizes Summary 146
4.2.4 Effect Size Confidence Intervals 149
4.3 Some Explanations of the “Old” Statistics 151
4.3.1 Null Hypothesis Significance Tests 151
4.3.2 One-Tailed versus Two-Tailed Tests of Hypotheses 154
4.3.3 Outcomes of Null Hypothesis Significance Testing 156
4.3.4 Power Analysis 157
4.3.5 Calculating Effect Sizes for Power Analysis 158
 x Contents

4.3.6 Examples of Power Analyses 158

4.3.7 Application Activities with Power Calculation 162
4.4 Precision instead of Power 163
4.4.1 Application Activities with Precision Calculation 165
4.5 Summary 166
4.5.1 Power through Replication and Belief in the “Law of Small Numbers” 166

PART II
Statistical Tests 169
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

5 Choosing a Statistical Test 171

5.1 Statistical Tests that are Covered in this Book 171
5.2 A Brief Overview of Correlation 172
5.2.1 Correlation: A Test of Relationships 172
5.3 A Brief Overview of Partial Correlation 174
5.3.1 Partial Correlation: A Test of Relationships 174
5.4 A Brief Overview of Multiple Regression 174
5.4.1 Multiple Regression: A Test of Relationships 174
5.5 A Brief Overview of the Chi-Square Test of Independence 176
5.5.1 Chi-Square: A Test of Relationships 176
5.6 A Brief Overview of T-Tests 177
5.6.1 T-Test: A Test of Group Differences 177
5.6.2 A Brief Overview of the Independent Samples T-Test 178
5.6.3 A Brief Overview of the Paired Samples T-Test 179
5.7 A Brief Overview of One-Way Analysis of Variance 180
5.7.1 One-Way Analysis of Variance: A Test of Group Differences 180
5.8 A Brief Overview of Factorial Analysis of Variance 181
5.8.1 Factorial Analysis of Variance: A Test of Group Differences 181
5.9 A Brief Overview of Analysis of Covariance 183
5.9.1 Analysis of Covariance: A Test of Group Differences 183
5.10 A Brief Overview of Repeated-Measures Analysis of Variance 184
5.10.1 Repeated-Measures Analysis of Variance: A Test of Group Differences 184
5.11 Summary 185
5.12 Application Activities for Choosing a Statistical Test 185

6 Finding Relationships Using Correlation: Age of Learning 188

6.1 Visual Inspection: Scatterplots 190
6.1.1 The Topic of Chapter 6 190
6.2 Creating Scatterplots in SPSS 190
6.2.1 Adding a Regression or Loess Line 192
6.2.2 Viewing Simple Scatterplot Data by Categories 195
6.3 Creating Scatterplots in R 196
6.3.1 Modifying a Scatterplot in R Console 197
6.3.2 Viewing Simple Scatterplot Data by Categories 200
6.3.3 Application Activities with Scatterplots 202
6.3.4 Multiple Scatterplots 203
6.3.5 Creating Multiple Scatterplots with SPSS 203
 Contents xi

6.3.6 Creating Multiple Scatterplots with R 204

6.3.7 Interpreting Multiple Scatterplots 205
6.4 Assumptions of Parametric Statistics for Correlation 206
6.4.1 Effect Size for Correlation 208
6.4.2 Confidence Intervals for Correlation 210
6.5 Calculating Correlation Coefficients and Confidence Intervals 210
6.5.1 Calculating Correlation Coefficients and Confidence Intervals in SPSS 211
6.5.2 Calculating Correlation Coefficients and Confidence Intervals in R 213
6.5.3 Robust Correlations 218
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

6.5.4 Application Activities for Correlation 221

6.5.5 Reporting a Correlation 222
6.6 Summary 222

7 Looking for Groups of Explanatory Variables through Multiple

Regression: Predicting Important Factors in First Grade Reading 224
7.1 Understanding Regression Design 225
7.1.1 Standard Multiple Regression 227
7.1.2 Sequential (Hierarchical) Regression 228
7.1.3 Data Used in this Chapter 228
7.2 Visualizing Multiple Relationships 229
7.2.1 Graphs in R for Understanding Complex Relationships:
Conditioning Plots 230
7.2.2 Graphs in R for Understanding Complex Relationships:
3-D Graphs 234
7.2.3 Graphs in R for Understanding Complex Relationships:
Tree Models 235
7.2.4 Application Activities in R with Graphs for Understanding
Complex Relationships 237
7.3 Assumptions of Multiple Regression 238
7.3.1 Assumptions about Sample Size 238
7.4 Performing a Multiple Regression 240
7.4.1 Starting the Multiple Regression in SPSS 241
7.4.2 Regression Output in SPSS 242
7.4.3 Examining Regression Assumptions Using SPSS 249
7.4.4 Robust Regression with SPSS 250
7.4.5 Linear Regression in R: Doing the Same Type of Regression as
in SPSS 252
7.4.6 Examining Regression Assumptions in R 259
7.4.7 Robust Linear Regression in R 263
7.4.8 Reporting the Results of a Regression Analysis 264
7.4.9 Application Activities: Multiple Regression 266
7.5 Summary 268

8 Looking for Differences between Two Means with T-Tests: Think-Aloud

Methodology and Teaching Sarcasm 269
8.1 Types of T-Tests 269
8.1.1 Application Activity: Choosing a T-Test 271
 xii Contents

8.2 Data Summaries and Numerical Inspection 272

8.2.1 Visual Inspection: Box Plots 273
8.2.2 Box Plots for One Variable Separated by Groups in SPSS 275
8.2.3 Box Plots for One Variable Separated by Groups in R 276
8.2.4 Box Plots for More than One Variable Plotted on the Same
Graph in SPSS 280
8.2.5 Box Plots for More than One Variable Plotted on the Same
Graph in R 281
8.2.6 Box Plots for More than One Variable Separated by Groups
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

in SPSS and R 283

8.2.7 Application Activities with Box Plots 285
8.3 Assumptions of T-Tests 286
8.3.1 Adjustments for Multiple T-Tests (Bonferroni Adjustment,
False Discovery Rate) 287
8.3.2 Data Formatting for Tests of Group Differences (the “Wide
Form” and “Long Form”) 287
8.4 The Independent Samples T-Test 289
8.4.1 Performing an Independent Samples T-Test in SPSS 290
8.4.2 Performing an Independent Samples T-Test in R 293
8.4.3 Performing a Bootstrapped Independent Samples T-Test in R 295
8.4.4 Performing a Bootstrapped, 20% Trimmed Means, Independent
Samples T-Test in R 296
8.4.5 Effect Sizes for Independent Samples T-Tests 298
8.4.6 Reporting the Results of an Independent Samples T-Test 299
8.4.7 Application Activities for the Independent Samples T-Test 300
8.5 The Paired Samples T-Test 301
8.5.1 Performing a Paired Samples T-Test in SPSS 301
8.5.2 One-Sided versus Two-Sided Confidence Intervals 304
8.5.3 Performing a Paired Samples T-Test in R 305
8.5.4 Performing a Robust Paired Samples T-Test in R 306
8.5.5 Effect Sizes for Paired Samples T-Tests 308
8.5.6 Application Activities with Paired Samples T-Tests 308
8.5.7 Reporting the Results of a Paired Samples T-Test 309
8.8 Summary of T-Tests 310

9 Looking for Group Differences with a One-Way Analysis of Variance:

Effects of Planning Time 311
9.1 Understanding the Analysis Of Variance Design 313
9.2 The Topic of Chapter 9 315
9.2.1 Numerical and Visual Inspection of the Data in this Chapter 316
9.3 Assumptions for an Analysis of Variance 318
9.4 One-Way Analysis of Variance 318
9.4.1 Omnibus Tests with Post-Hoc Tests or Planned Comparisons 318
9.4.2 Testing for Group Equivalence before an Experimental Procedure 319
9.4.3 Performing an Omnibus One-Way Analysis of Variance Test in SPSS
with Subsequent Post-Hoc Tests 321
9.4.4 Performing an Omnibus One-Way Analysis of Variance in R
with Subsequent Post-Hoc Tests 326
 Contents xiii

9.4.5 Performing a Bootstrapped One-Way Analysis of Variance in R 331

9.4.6 Conducting a One-Way Analysis of Variance Using Planned Comparisons 333
9.4.7 Conducting Planned Comparisons in SPSS 334
9.4.8 Conducting Planned Comparisons in R 336
9.4.9 Effect Sizes in One-Way Analysis of Variance 338
9.4.10 Application Activities with One-Way Analysis of Variance 341
9.4.11 Reporting the Results of a One-Way Analysis of Variance 343
9.5 Summary of One-Way Analysis of Variance 344
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

10 Looking for Group Differences with Factorial Analysis of Variance

When there is More than One Independent Variable: Learning with Music 345
10.1 Analysis of Variance Design 347
10.1.1 Analysis of Variance Design: Interaction 347
10.1.2 Application Activity in Understanding Interaction 348
10.1.3 Analysis of Variance Design of the Obarow Study 352
10.1.4 Analysis of Variance Design: Variable or Level? 352
10.1.5 Application Activity: Identifying Independent Variables and Levels 353
10.2 Numerical and Visual Inspection 355
10.2.1 Creating a Combination Box Plot and Means Plot in R 358
10.3 Assumptions of a Factorial Analysis of Variance 361
10.4 Getting Ready to Perform a Factorial Analysis of Variance 362
10.4.1 Making Sure Your Data is in the Correct Format for a Factorial
Analysis of Variance 362
10.4.2 Rearranging Data for a Factorial Analysis of Variance Using SPSS 364
10.4.3 Rearranging Data for a Factorial Analysis of Variance Using R 366
10.4.4 Excursus on Type II vs. Type III Sums of Squares (Advanced Topic) 368
10.5 Factorial Analysis of Variance: Extending Analyses to More than
One Independent Variable 369
10.5.1 Performing a Three-Way Factorial Analysis of Variance with SPSS 369
10.5.2 Performing a Three-Way Factorial Analysis of Variance Using R 382
10.5.3 A Confidence Interval Approach to Factorial ANOVA
(Advanced Topic) 385
10.5.4 Planned Comparisons in a Factorial Analysis of Variance 394
10.5.5 Performing Planned Comparisons in a Factorial Analysis of
Variance for SPSS and R 395
10.5.6 Effect Sizes for Factorial Analysis of Variance 396
10.5.7 Application Activities with Factorial Analysis of Variance 397
10.5.8 Reporting the Results of a Factorial Analysis of Variance 397
10.6 Summary 399

11 Looking for Group Differences When the Same People are Tested
More than Once: Repeated-Measures Analysis of Variance with Wug
Tests and Instruction on French Gender 401
11.1 Understanding Repeated-Measures Analysis of Variance Designs 403
11.1.1 Repeated-Measures Analysis of Variance Design of the
Murphy (2004) Study 405
11.1.2 Repeated-Measures Analysis of Variance Design of the Lyster
(2004) Study 406
 xiv Contents

11.1.3 Application Activity: Identifying Between-Groups and

Within-Groups Variables to Decide between Repeated-Measures
and Factorial Analysis of Variance Designs 406
11.2 Arranging the Data for a Repeated-Measures Analysis of Variance 410
11.2.1 Arranging the Data for Repeated-Measures Analysis of Variance in SPSS 410
11.2.2 Changing from Wide Form to Long Form in SPSS 411
11.2.3 Arranging the Data for a Repeated-Measures Analysis of Variance in R 413
11.2.4 Application Activities for Changing Data from the Wide to the Long Form
(Necessary for Use with the R Program Only) 416
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

11.3 Visualizing Repeated-Measures Data 417

11.3.1 Exploring the Murphy (2004) and Lyster (2004) Data with the
Combination Interaction Plot and Box Plot 417
11.3.2 Parallel Coordinate Plots 419
11.3.3 Creating a Parallel Coordinate Plot in SPSS 420
11.3.4 Creating a Parallel Coordinate Plot in R 422
11.3.5 Application Activities with Parallel Coordinate Plots 424
11.4 Repeated-Measures Analysis of Variance Assumptions 424
11.4.1 Exploring Model Assumptions 424
11.5 Performing a Repeated-Measures Analysis of Variance with the
Least-Squares Approach 426
11.5.1 Least-Squares Repeated-Measures Analysis of Variance in SPSS 427
11.5.2 Repeated-Measures Analysis of Variance Output 428
11.5.3 Least-Squares Repeated-Measures Analysis of Variance in R 435
11.5.4 Application Activities with Least-Squares, Repeated-Measures
Analysis of Variance 440
11.6 Furthering a Repeated-Measures Analysis by Exploring Simple Interaction
Effects and Simple Main Effects 441
11.6.1 Exploring Simple Interaction Effects and Simple Main Effects in the
Murphy (2004) Data (SPSS and R) 441
11.6.2 Reporting the Results of a Repeated-Measures Analysis of Variance 447
11.6.3 Application Activities with Further Exploration of Repeated-Measures
Analysis of Variance Using Simple Interaction Effects and
Simple Main Effects 449
11.7 Summary 450

Appendix A: Doing Things in R 453

Glossary 473
Bibliography 486
Author index 497
R commands 500
Subject index 502
 PREFACE
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

I wrote this book for myself. I remember being a graduate student, working on my dissertation and
struggling to make sense of the statistics I needed to analyze my dissertation. I wanted a book to help
me understand SPSS that would use examples from my field, but none existed. This book was born of
that longing. This book is written to help those new to the field of statistics feel they have some idea
of what they are doing when they analyze their own data, and how to do it using a statistical program.
In the first edition of the book only SPSS was included. In this edition, the statistical programs
SPSS and R are illustrated side by side so that those who want to use SPSS can continue to do so,
those who know SPSS but want to learn R can do so more easily, and those who want to start learning
statistics with R can also do so. I originally wrote the book that way, and I am pleased to have that
format now, although it makes the book quite a bit larger. Including R increased the size enough that I
needed to cut down the finished product in my revised edition. I have thus cut out most of the sections
that were formerly labeled “Advanced Topic,” as well as the three chapters on chi-square, analysis of
covariance (ANCOVA) and non-parametric statistics. In my experience, chi-square and ANCOVA
are not used as frequently in second language acquisition (SLA) as the other statistical tests that
remain in the second edition. Nevertheless, they are good and useful chapters, so if you need informa-
tion on them, please look at them online! I was especially pleased with the many interesting graphics
that are available for the categorical variables that are found in the Chi-Square chapter, so if you need
any of those, take a look online. I cut out the Non-Parametric Statistics (i.e. the classical types of non-
parametric statistics, such as the Kruskal–Wallis or Mann–Whitney U tests). I judged that robust sta-
tistics, which are mostly non-parametric, would be better substitutes than these tests, but the chapter
is still available online, if readers would prefer these traditional tests. There are also alternative ways
of conducting tests (usually in R), such as the sections cut from Chapter 11 on RM-ANOVA, which
tell how to use mixed-effect models instead of the least-squares approach.
I have included a graphical user interface to R called R Commander. I think this program makes
it easier to work into R, but there are many places in the book where you will need to go to the R
Console (the command-line interface) to do things in the book, so I have tried to ease the reader into
the R code by showing the code from any R Commander menu choices, and explaining the parts of
these. There are quite a number of advantages, as I see it, to learning R instead of SPSS if you are just
learning about statistics:

x It is free.
x It is supported by the statistical community and, as such, continues to be updated with new
packages that can do different things.
x R has more sophisticated analyses and is extremely strong in the area of graphics.
x Using command syntax for statistics helps users understand better what they are doing in the
statistical analysis.
 xvi Preface

Some users may be concerned that because R is free, that means it is of low quality. Muenchen (2008)
noted that, at the time, comparisons of R with SPSS found that the quality of R was just as high as
SPSS, and that the R base packages were reviewed and revised just as carefully as SPSS. (This is not
necessarily true for all the R packages, however, and Muenchen noted that you can find reviews of
individual packages at http://crantastic.org/.)
Muenchen also noted that R could do everything that SPSS could do and more, and if you bought
a commercial copy of SPSS with all of the functionality of R, it would cost tens of thousands of dol-
lars. In Japan, where I am currently working at a university, access to SPSS is not automatic and
requires purchase of an individual site license. Perhaps your institution does not mind spending over
USD100,000 for licenses for multiple users. However, as more institutions seek to pare down expenses,
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

learning R (which can do everything that SPSS can do) makes more and more sense.
Now that I can use R, I have included information about how to conduct various kinds of robust
statistics in the book, although some of the sections on robust analyses in R have had to be moved
online. For more information on how to do robust analyses in different ways, see those online sections.
Currently, SPSS can perform bootstrapping, and I have shown how to do that in this book, but there
are several other robust tests that can only be performed in R. It is true that SPSS has a way to run
R, but one must use scripts to do so, and I have not felt moved to try to figure this system out, feeling
that if one is going to learn scripts, one might just as well start using R. In addition, SPSS cannot run
the current version of R and is instead using R from a number of versions back. (R releases come out
twice a year.)
I am a strong advocate for using robust statistics, which basically rely on the techniques of para-
metric statistics but use computer-intensive techniques that eliminate the requirement that data be
normally distributed (an assumption that probably is often not true in our field; see Larson-Hall &
Herrington, 2009).
Almost all of the data sets analyzed in the book and the application activities are real data, gath-
ered from recently published articles or theses. I am deeply grateful to these authors for allowing me
to use their data, and I feel strongly that those who publish work based on a statistical analysis of the
data should make that data available to others. The statistical analysis one does can affect the types of
hypotheses and theories that go forward in our field, but we can all recognize that most of us are not
statistical experts and we may make some mistakes in our analyses. Providing the raw data serves as a
way to check that analyses are done properly, and, if provided at the submission stage, errors can be
caught early. I do want to note, however, that authors who have made their data sets available for this
book have done so in order for readers to follow along with the analyses in the book and do application
activities. If you wanted to do any other kind of analysis on your own that would be published using
these data, you should contact the authors and ask for permission to either coauthor or use their data
(depending on your purposes).
This book was updated with the most recent version of IBM SPSS Statistics available at the time
of writing, and I was working on a PC. As of this writing, the most recent version of SPSS was 22.0
(2013). SPSS 22 is quite different in some ways from previous versions, and for those who may be
working with an older version I have tried to provide some guidance about possible differences. I have
used small capitals to distinguish commands in SPSS and R and make them stand out on the page.
For example, if you need to first open the File menu and then choose the Print option, I would write
FILE > PRINT.
I also used the most recent version of R, which was version 3.1.3 (released 9 March 2015) and R
Commander version 2.1–7 (released 19 February 2015). I used R on both a PC and a Mac. Because
these programs are updated frequently, there may be some times especially when certain packages do
not work. I urge you at those times to google the package and see what might have replaced it.
I have written this book mainly as a way for those totally new to statistics to understand some of the
most basic statistical tests that are widely used in the field of SLA and applied linguistics. I suggest that
the best way to read most chapters is to open the data sets that are included with the book and work
 Preface xvii

along with me in SPSS or R. You can learn a lot just by recreating what you see me do in the text. The
application activities will then help you move on to try the analysis by yourself, but I have included
detailed answers to all activities so you can check what you are doing. My ultimate goal, of course, is
that eventually you will be able to apply the techniques to your own data. I assume that readers of this
book will be familiar with how to use Microsoft Windows or Mac OS X and pull-down menus, and how
to open and close files. These data files will be available on the website that accompanies this book.
SPSS files have the .sav extension, while most R files are .csv, which are comma-delimited text files.
Some sections in the chapters are labeled “Advanced Topic.” These sections can be skipped by
novices who are already feeling overwhelmed by the regular statistics, but they provide additional
information or justifications for choices I have made to the more statistically knowledgeable readers.
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

For the novice to statistics, the book is meant to be read in chronological order for both parts of the
book. Part I introduces the fundamental concepts in statistics that are necessary to beginning a statisti-
cal analysis. Part II then provides information in individual chapters about basic statistical procedures
that are commonly used in second language research. In some cases these chapters also build upon one
another, although users who are familiar with statistics and just looking up information on one specific
technique can skip directly to the relevant chapter.
Part I consists of four chapters. In Chapter 1, I start things off by telling you how to set yourself
up with SPSS or R and how to start entering data and manipulating it. One important point here is
how to deal with missing data. It seems most researchers believe they need to throw out incomplete
data but I find many statisticians enthusiastic about the use of missing-data imputation, and there are
sophisticated statistical methods to do so. If your data are missing at random, this is a better choice
than eliminating partial data entries that researchers have worked hard to gather, so if you do not
know about imputation, be sure to take a look at Chapter 1. Throughout the book important terms
are highlighted in bold, and summaries of their meanings are found in the glossary. In Chapter 2, I
discuss some essential ideas that are necessary to understanding further chapters, such as identifying
whether variables are independent or dependent and what null hypothesis testing is. Hidden statistical
assumptions authors have are revealed in this chapter and can help those new to statistics to better
understand the statistical jargon that they will often see. In Chapter 3, I show how to produce numeri-
cal summaries of data, such as means, medians and standard deviations, and how to check for measures
that might indicate problems with standard statistical assumptions, such as a normal distribution and
equal variances for all groups. Graphic ways of evaluating the same types of information have been
intentionally placed in the same chapter, for I feel that graphic summaries are just as informative as,
if not more so, numerical summaries, and I argue strongly for more use of graphic summaries in our
research. Chapter 4 is the result of my first reading many papers that argue that we are often on the
wrong track with our statistical analyses when we rely on a single dichotomous measure such as a p
value (less or greater than 0.05) to tell us whether our results are important, and then reading Geoff
Cumming’s recent (2012) book called Understanding the New Statistics. Cumming is persuasive and
clear in his argument that we need to bring better ways of measuring and hypothesizing about our
data into use, and this includes mostly confidence intervals and effect sizes. In the first edition, I did
urge readers to use confidence intervals, but I did not understand then as well as I do now, thanks to
Cumming, how to do so in a way that actually used the confidence intervals differently from p values.
I hope the information in this chapter conveys the beauty of confidence intervals, the problems with p
values and the excitement I feel about the “new statistics” to you as well as Cumming did to me (and
I do urge you to read Cumming for more information!).
In Chapter 5, I lay out the different statistical tests that will be covered in Part II. I provide infor-
mation about how to choose a test, depending on the questions and the types of data you have. I
illustrate how these statistical tests are used in actual studies in second language research. I hope this
will help you understand the differences between the tests, although you should know that figuring out
statistics does take some time and you should not expect to be an expert after reading Chapter 5! But
maybe you will have a small inkling of how the various tests are used. In the remaining six chapters of
 xviii Preface

Part II, I go over various types of statistical analyses that are common in the field of second language
research. In Chapter 6, I cover correlation; in Chapter 7, I introduce the reader to multiple regression;
in Chapter 8, I look at t-tests; in Chapter 9, I look at the one-way analysis of variance (ANOVA); and
in Chapter 10, I extend ANOVA to the case where there is more than one independent variable (fac-
torial ANOVA). In Chapter 11, I extend ANOVA even further to the case where there are repeated
measures. Three more chapters are available online: Chi-Square, ANCOVA (an ANOVA analysis
with covariates) and Non-Parametric Tests. In general, the chapters go from conceptually easier tests
to more difficult ones, although this is not true for regression analysis.
Writing this book has given me a great appreciation for the vast world of statistics, which I do
not pretend to be a master of. Like our own field of second language research, the field of statistics is
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

always evolving and progressing, and there are controversies throughout it as well. If I have advocated
an approach that is not embraced by the statistical community wholeheartedly, I have tried to provide
reasons for my choice and additional reading that can be done on the topic. Although I have tried
my hardest to make this book accessible and helpful for readers, I would appreciate being informed of
any typos or errors in this book or answers to application activities, or any areas where my explana-
tions have not been clear enough. Because of the watchful eye of many readers, I am convinced that
this edition would be error-free if I were simply updating the SPSS version. But because I have added
in a whole new statistical program, there are bound to be places where I have again made mistakes. I
understand that this is frustrating, especially when you are trying to learn a new topic, and I apologize
in advance. I cannot give you a free copy of the book if you catch things, but I hope you will feel the
satisfaction of helping others if you do find something and pass it along. Please check the website for
errata as I will post these as readers send them to me.
 ACKNOWLEDGMENTS
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

I would like to thank the many individuals who have helped me to write this book. First is Richard
Herrington, who guided and helped me discover many treasures of statistics, including R. He
provided many articles, many answers and just fun philosophizing along the way. Susan Gass and
Alison Mackey were instrumental in helping me get this book accepted into the Second Language
Acquisition Research Series and providing encouragement. Elysse Preposi and Leah Babb-Rosenfeld,
editors at Routledge, were very helpful in the laborious process of getting the book into its best form
possible and I appreciate their warm help. Many thanks to the students in my Research Methods
courses who provided feedback on the book in its various stages. I also thank the reviewers of the book
for their careful work in reading the manuscript and providing useful feedback. A number of readers
of the first edition helpfully sent me corrections, the most thorough and enthusiastic being Gabriele
Pallotti from Italy, whose notes of a careful reading of the entire book were quite helpful. Other readers
who have written more recently and whose names are thus not lost to the mists of time include Yves
Bestgen, Tuan Phu, Matthias Raess and Lawrence Cheung. I sincerely appreciate their help, and the
only reason I am sorry to be expanding this book and including R is that there will most likely be a
number more mistakes that will occur. I welcome input on this and apologize in advance. I again want
to profusely thank the authors who have given permission for the reproduction of their data in the
examples presented in the book. They are cited individually in their respective chapters, but I feel this
book is much more relevant and interesting to readers because of the data that they provided me with.
Last but not least is my family who did not see much of me in the final stages of finishing this revision,
and were very patient in giving me the time to get it finished. A special thank you to my son, Lachlan,
who graphed out the data for the small multiples figure in the online chapter on ANCOVA.
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

This page intentionally left blank

Downloaded by [University of California, San Diego] at 02:46 19 March 2017

Part I
Statistical Ideas
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

This page intentionally left blank

 Chapter 1

Getting Started with the Software

and Using the Computer for
Experimental Details
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

In my view R is such a versatile tool for scientific computing that anyone contemplating a career in
science, and who expects to [do] their own computations that have a substantial data analysis component,
should learn R.
John Maindonald (R Help thread, 5 January 2006)

In this chapter I will provide instructions for getting started with both SPSS and R. SPSS—now
owned by the IBM corporation and called IBM SPSS Statistics—is probably the most commonly
used statistical program in the field of social sciences for which it was originally developed, and is
used by many researchers working in the field of second language research. It uses a graphical user
interface (GUI) with drop-down menus that should be familiar to those who use programs such as
Microsoft Word or Excel. I will also explain how to get started with R, which is the programming
language and software environment most statisticians use and has grown greatly in popularity in the
last decade, including with linguists. (See an R-lang mailing list created by Roger Levy in 2007, which
also contains an archive of all prior messages.) By itself R does not have an intuitive GUI, but in this
book I will show you how to use a GUI called R Commander that will help simplify the use of R for
beginners.
In this chapter I will first explain how to physically get started by opening the software and either
importing or entering your own data into a spreadsheet. (Data are the information you have.) Another
important part of getting started is being able to save data. Once you have some data to work with,
this chapter then goes on to explain ways you might like to manipulate your data, such as combining
columns or calculating percentages, filtering out some cases of participants that you do not want to
include in the analysis, or creating new groups from existing columns of data. I will also give you some
help as to what to do if you have missing data, and finally, some resources for getting more help with
these programs. The SPSS instructions throughout have been tested only with a PC (not Mac) with
SPSS. The R instructions have been tested on both a PC and a Mac and any differences between these
platforms will be noted in the text or on the website.

1.1 Getting Started with SPSS

First, open SPSS by either clicking on the SPSS icon or going through the Programs
menu from Windows. When you open SPSS you will see an “SPSS Data Editor” as in Figure 1.1,
which looks like a spreadsheet for entering data. In SPSS version 10.0 and later, before you get to the
Data Editor you will have a pop-up window which tries to help you get started. Usually you will
choose either to type in new data or to open up an existing data set, like the ones provided for this
book. Data files saved using SPSS have the extension .sav.
 4 Statistical Ideas
Downloaded by [University of California, San Diego] at 02:46 19 March 2017

Figure 1.1 What you’ll see when starting up SPSS.

Note that there are two tabs in the Data Editor at the bottom of the spreadsheet. One is the “Data
View,” which is the default view. The other is the “Variable View.” If you click on this tab, you will
see a different spreadsheet that is used to enter specific information about your variables, such as their
names, type and number of decimal points you want to see, among others.
SPSS has two different types of windows. One is the Data Editor, which looks like a spreadsheet
and is where you enter your data. The other type of window is where any output appears. This window
is called the “Statistics Viewer.” In SPSS, any calls to perform data analysis, such as creating a graph,
doing a statistical test or creating a table result in objects automatically appearing in the Statistics
Viewer (see Figure 1.2). If you save data from the Statistics Viewer, these files have a .spv extension.

Figure 1.2 The Statistics Viewer window.

 Getting Started with SPSS 5

(Versions of SPSS older than 16.0 created .spo files as the output files, and these cannot be opened in
SPSS 16 unless you install the SPSS 15.0 Smart Viewer, available on the installation CD, according
to the SPSS help files.)
Section 1.1 deals with entering and importing data into SPSS as well as saving files. If you want
information about and practice with manipulating variables in SPSS, such as moving columns or
rows or simply deleting them, combining two or more variables into one variable, recoding groups or
excluding cases from your data, see the online section titled “Manipulating variables in SPSS.”

1.1.1 Opening a Data File

Downloaded by [University of California, San Diego] at 02:46 19 March 2017

Once you have SPSS open, I am sure you will want to get started trying to do things! If you have not
used the initial pop-up menu shown in Figure 1.1 to open a file that has already been created, you can
also do this through the menus using the sequence F i l e > O p e n > D a t a and then navigating to
the .sav file you want. If you do this, you will see that the spreadsheet in the SPSS Data Editor becomes
filled with data, as shown in Figure 1.3.

Figure 1.3 The Data Editor in “Data View” (“Data View” tab is highlighted at bottom) with data in it.

1.1.2 Entering Your Own Data

If you are going to create your own data file and would like to open a new spreadsheet through the
menu system then you would go to File > New > Data, which opens a blank spreadsheet. Since
SPSS version 16.0, whenever you carry out any command through the menu system, the SPSS syntax
for that command is documented in the Statistics Viewer. For example, Figure 1.4 shows the Viewer
after opening up a new file.
There are a few things you need to know about data entry before you start inputting your numbers:

x Rows in SPSS are cases. This means that each participant should occupy one row.
x Columns in SPSS are separate variables, such as ID number, score on a test or category in a certain
group.
x You can name each case (participant) by making your first column the ID number of each
participant.
Another random document with
no related content on Scribd:
 CHAPTER IX
THE SHELL, ITS FORM, COMPOSITION AND GROWTH—DESIGNATION OF
ITS VARIOUS PARTS

The popular names of ‘shells,’ ‘shell-fish,’ and the like, as

commonly applied to the Mollusca, the intrinsic beauty and grace of
the shells themselves, resulting in the passion for their collection,
their durability and ease of preservation, as compared with the non-
testaceous portion,—all these considerations tend to unduly exalt the
value of the shell as part of the organism as a whole, and to obscure
the truth that the shell is by no means the most important of the
organs.
At the same time it must not be forgotten that the old systems of
classification, which were based almost entirely on indications drawn
from the shell alone, have been strangely little disturbed by the new
principles of arrangement, which depend mainly on structural points
in the animal. This fact only tends to emphasise the truth that the
shell and animal are in the closest possible connexion, and that the
shell is a living part of the organism, and is equally sensitive to
external influences.
A striking instance of the comparative valuelessness of the shell
alone as a primary basis of classification is furnished by the large
number of cases in which a limpet-shaped shell is assumed by
genera widely removed from one another in cardinal points of
organisation. This form of shell occurs in the common limpet
(Patellidae), in Ancylus (Limnaeidae), Hemitoma (Fissurellidae),
Cocculina (close to Trochidae), Umbrella and Siphonaria
(Opisthobranchiata), while in many other cases the limpet form is
nearly approached.
Roughly speaking, about three-quarters of the known Mollusca,
recent and fossil, possess a univalve, and about one-fifth a bivalve
shell. In Pholas, and in some species of Thracia, there is a small
accessory hinge plate; in the Polyplacophora, or Chitons, the shell
consists of eight plates (see Fig. 2, p. 8), usually overlapping. A
certain proportion of the Mollusca have no shell at all. In many of
these cases the shell has been present in the larva, but is lost in the
adult.
The shell may be
(1) External, as in the great majority of both univalves and
bivalves.
(2) Partly external, partly internal; e.g. Homalonyx, Hemphillia,
some of the Naticidae, Scutum, Acera, Aplustrum (Figs. 148 and
149).

Fig. 148.—Aplustrum
aplustre L. Mauritius,
showing the partly
internal shell (S); F,
foot; LL, cephalic
lappets; TT, double
set of tentacles. (After
Quoy and Gaimard.)
 Fig. 149.—Sigaretus
laevigatus Lam.,
showing shell partially
immersed in the foot;
F, anterior
prolongation of the
foot. (After Souleyet.)
(3) Internal; e.g. Philine, Gastropteron, Pleurobranchus, Aplysia,
Limax, Arion, Hyalimax, Parmacella, Lamellaria, Cryptochiton, and,
among bivalves, Chlamydoconcha.
(4) Absent; e.g. all Nudibranchiata and Aplacophora, many
Cephalopoda, a few land Mollusca, e.g. all Onchidiidae, Philomycus,
and Vaginula.
The Univalve Shell.—In univalve Mollusca the normal form of the
shell is an elongated cone twisted into a spiral form round an axis,
the spiral ascending to the left. Probably the original form of the shell
was a simple cone, which covered the vital parts like a tent. As these
parts tended to increase in size, their position on the dorsal side of
the animal caused them gradually to fall over, drawing the shell with
them. The result of these two forces combined, the increasing size of
the visceral hump, and its tendency to pull the shell over with it,
probably resulted in the conversion of the conical into the spiral shell,
which gradually came to envelop the whole animal. Where the
visceral hump, instead of increasing in size, became flattened, the
conical shape of the shell may have been modified into a simple
elliptical plate (e.g. Limax), the nucleus representing the apex of the
cone. In extreme cases even this plate dwindles to a few calcareous
granules, or disappears altogether (Arion, Vaginula).
 Varieties of the Spiral.—Almost every conceivable modification
of the spiral occurs, from the type represented by Gena, Haliotis,
Sigaretus, and Lamellaria, in which the spire is practically confined to
the few apical whorls, with the body-whorl inordinately large in
proportion, to a multispiral form like Terebra, with about twenty
whorls, very gradually increasing in size.

Fig. 150.—Examples of shells with A, a flattened

spire (Polygyratia); B, a globose spire (Natica);
C, a greatly produced spire (Terebra).
As a rule, the spire is more or less obliquely coiled round the axis,
each whorl being partially covered, and therefore hidden by, its
immediate successor, while the size of the whorls, and therefore the
diameter of the spire as a whole, increases somewhat rapidly. The
effect of this is to produce the elevated spire, the shell of six to ten
whorls, and the wide aperture, of the normal type of mollusc, the
whelk, snail, periwinkle, etc.
Sometimes, however, the coil of the whorls, instead of being
oblique, tends to become horizontal to the axis, and thus we have
another series of gradations of form, from the excessively produced
spire of Terebra to the flattened disc of Planorbis, Polygyratia,
Euomphalus, and Ammonites. The shell of many species of Conus
practically belongs to the latter type, each whorl folding so closely
over its predecessor that the spiral nature of the shell is not
perceived until it is looked at at right angles to the spire.

Fig. 151.—Examples of
shells with
disconnected whorls;
A, Cyathopoma cornu
Mf., Philippines; B,
Cylindrella hystrix
Wright, Cuba. (Both ×
4.)
 Fig. 152.—Example of a
shell whose apical
whorls alone are
coiled, and the
remainder produced in
a regular curve.
(Cyclosurus Mariei
Morel., Mayotte.)
In some cases the regularly spiral form is kept, but the whorls are
completely disconnected; e.g. some Scalaria, Spirula; among fossil
Cephalopoda, Gyroceras, Crioceras, and Ancyloceras; and, among
recent land Mollusca, Cylindrella hystrix and Cyathopoma cornu (Fig.
151). Sometimes only the last whorl becomes disconnected from the
others, as in Rhiostoma (see Fig. 180, p. 266), Teinostoma, and in
the fossil Ophidioceras and Macroscaphites. Sometimes, again, not
more than one or two whorls at the apex are spirally coiled, and the
rest of the shell is simply produced or coiled in an exceedingly
irregular manner, e.g. Cyclosurus, Lituites, Orygoceras, Siliquaria
(Fig. 153), Vermetus. In Coecum (Fig. 170, p. 260) the spiral part is
entirely lost, and the shell becomes simply a cylinder. In a few cases
the last whorl is coiled irregularly backwards, and is brought up to
the apex, so that the animal in crawling must carry the shell with the
spire downwards, as in Anostoma (Fig. 154), Opisthostoma (Fig.
208, p. 309), Strophostoma, and Hypselostoma (Fig. 202 A, p. 302).
Fig. 153.—Siliquaria anguina Lam.,
showing scalariform coil of upper
whorls and irregular extension of
the lower.

Fig. 154.—Anostoma globulosum Lam.,

Brazil. (After P. Fischer.)
 Fig. 155.—Various forms of the internal plate in
Capulidae: A, Calyptraea (Mitrularia) equestris
Lam., E. Indies; B, Crucibulum scutellatum Gray,
Panama; C, Ergaea plana Ad., and Reeve,
Japan; D, Galerus chinensis L., Britain; E,
Crepipatella dilatata Lam., Callao; F, Trochita
maculata Quoy, N. Zealand; G, Crepidula
fornicata Lam., N. America.
Some genera of the Capulidae, in which the shell is of a broadly
conical form or with scarcely any spire, develop an internal plate or
process which serves the purpose of keeping the animal within the
shell, and does the work of a strong attachment muscle. In Mitrularia
this process takes the form of a raised horse-shoe; in Crucibulum it
is cup-shaped, with the edge free all round; in Galerus, Ergaea,
Crepipatella, and Trochita we get a series of changes, in which the
edge of the cup adheres to the interior of the shell, and then
gradually flattens into a plate. In Crepidula proper this plate becomes
a regular partition, covering a considerable portion of the interior
(Fig. 155 G). Hipponyx secretes a thin calcareous plate on the
ventral surface of the foot, which intervenes like an operculum
between the animal and the substance to which it adheres.
Sinistral, or Left-handed Shells.—The vast majority of univalve
spiral shells are normally dextral, i.e. when held spire uppermost,
with the aperture towards the observer, the aperture is to the right of
the axis of the spire. If we imagine such a shell to be a spiral
staircase, as we ascended it we should always have the axis of the
spire to our left.
 Sinistral or ‘reversed’ forms are not altogether uncommon, and
may be grouped under four classes:—
(1) Cases in which the genus is normally sinistral; (2) cases in
which the genus is normally dextral, but certain species are normally
sinistral; (3) cases in which the shell is indifferently dextral or
sinistral; (4) cases in which both genus and species are normally
dextral, and a sinistral form is an abnormal monstrosity.

Fig. 156.—Fulgur
perversum L., Florida. ½.

Fig. 157.—Illustration of the gradation of forms in Ampullaria between a dextral (A)

and an ultra-dextral species (F).

In all cases of sinistral monstrosity, and all in which a sinistral and

dextral form are interchangeable (sections 3 and 4 above), the
position of the apertures of the internal organs appears to be
relatively affected, i.e. the body is sinistral, as well as the shell. This
has been proved to be the case in all specimens hitherto examined,
and may therefore be assumed for the rest. The same uniformity,
however, does not hold good in all cases for genera and species
normally sinistral (sections 1 and 2). As a rule, the anal and genital
apertures are, in these instances also, to the left, but not always. In
Spirialis, Limacina, Meladomus, and Lanistes the shell is sinistral,
but the animal is dextral. This apparent anomaly has been most
ingeniously explained by Simroth, Von Ihering, and Pelseneer. The
shell, in all these cases, is not really sinistral, but ultra-dextral.
Imagine the whorls of a dextral species capable of being flattened,
as in a Planorbis, and continue the process, still pushing, as it were,
the spire downwards until it occupies the place of the original
umbilicus, becoming turned completely ‘inside out,’ and we have the
whole explanation of these puzzling forms. The animal remains
dextral, the shell has become sinistral. A convincing proof of the truth
of this is furnished by the operculum. It is well known that the twist of
the operculum varies with that of the shell; when the shell is dextral,
the operculum is sinistral, with its nucleus near the columella, and
vice versâ. In these ultra-dextral shells, however, where it is simply
the method of the enrolment of the spire that comes in question, and
not the formation of the whorls themselves, the operculum remains
sinistral on the apparently sinistral shell.
The reverse case to this, when the shell is dextral but the orifices
sinistral, is instanced by the two fresh-water genera Pompholyx
(from N. America), and Choanomphalus (L. Baikal). A similar
transition in the enrolment of the whorls may be confidently assumed
to have taken place, and the shells are styled ultra-sinistral.
Yet another variation remains, in which the embryonic form is
sinistral, but the adult shell dextral, the former remaining across the
nucleus of the spire. This is the case with Odostomia, Eulimella,
Turbonilla, and Mathilda, all belonging to the Prosobranchiata, with
Actaeon, Tornatina, and Actaeonina among the Opisthobranchs, and
Melampus alone among Pulmonates.
Monstrosities of the Shell.—Abnormal growths of the shell
constantly occur, some of them being scarcely noticeable, except by
a practised eye, others of a more serious nature, involving an entire
change in the normal aspect of the creature. Scalariform
monstrosities are occasionally met with, especially in Helix and
Planorbis, when the whorls become unnaturally elevated, and
sometimes quite disjoined from one another; carinated monstrosities
develop a keel on a whorl usually smooth; acuminated monstrosities
have the spire produced to an extreme length (Fig. 158); sinistral
monstrosities (see above) have the spire reversed: dwarfs and
giants, as in our own race, are occasionally noticed among a crowd
of individuals.
More serious forms of monstrosity are those which occur in
individual cases. Mr. S. P. Woodward once observed[332] a specimen
of an adult Helix aspersa with a second, half-grown individual fixed
to its spire, and partly embedded in the suture of the body whorl. The
younger snail had died during its first hibernation, as was shown by
the epiphragm remaining in the aperture, and its neighbour, not
being able to get free of the incubus, partially enveloped it in the
course of its growth. In the British Museum two Littorina littorea have
become entangled in a somewhat similar way (Fig. 160 B), possibly
as a result of embryonic fusion. Double apertures are not
uncommon[333] in the more produced land-shells, such as Cylindrella
and Clausilia (Fig. 160 A). In the Pickering collection was a Helix
hortensis which had crawled into a nutshell when young, and,
growing too large to escape, had to carry about this decidedly extra
shell to the end of its days. A monstrosity of the cornucopia form, in
which the whorls are uncoiled almost throughout, is of exceedingly
rare occurrence (Fig. 161).
 Fig. 158.—Monstrosities of
Neptunea antiqua L., and
Buccinum undatum L., with a
greatly produced spire (from
specimens in the Brit. Mus.).

Fig. 159.—Monstrosities of Littorina

rudis Mat, The Fleet, Weymouth.
(After Sykes.)
Some decades ago ingenious Frenchmen amused themselves by
creating artificial monstrosities. H. aspersa was taken from its shell,
by carefully breaking it away, and then introduced into another shell
of similar size (H. nemoralis, vermiculata, or pisana). At the end of
several days attachment to the columella took place, and then
growth began, the new shell becoming soldered to the old, and the
spiral part of the animal being protected by a thin calcareous
envelope. A growth of from one to two whorls took place under these
conditions. The individuals so treated were always sordid and
lethargic, but they bred, and naturally produced a normal aspersa
offspring.[334] In the British Museum there is a specimen of one of
these artificial unions of a Helix with the shell of a Limnaea stagnalis.

Fig. 160.—Monstrosities with two

apertures: A, Cylindrella
agnesiana C. B. Ad.,
Jamaica; B, Littorina littorea
(from specimens in the
British Museum).

Fig. 161.—Cornucopia-
shaped monstrosity of
Helix aspersa, from
Ilfracombe. (British
Museum.)
Composition of the Shell.—The shell is mainly composed of
pure carbonate of lime, with a very slight proportion of phosphate of
lime, and an organic base allied to chitin, known as conchiolin. The
proportion of carbonate of lime is known to vary from about 99 p.c. in
Strombus to about 89 p.c. in Turritella. Nearly 1 p.c. of phosphate of
lime has been obtained from the shell of Helix nemoralis, and nearly
2 p.c. from that of Ostrea virginica. The conchiolin forms a sort of
membranous framework for the shell; it soon disappears in dead
specimens, leaving the shell much more brittle than it was when
alive. Carbonate of magnesia has also been detected, to the extent
of ·12 p.c. in Telescopium and ·48 p.c. in Neptunea antiqua. A trace
of silica has also occasionally been found.
When the shell exhibits a crystalline formation, the carbonate of
lime may take the form either of calcite or aragonite. The calcite
crystals are rhombohedral, optically uniaxal, and cleave easily, while
the aragonite cleave badly, belong to the rhombic system, and are
harder and denser, and optically biaxal. Both classes of crystal may
occur in the same shell.
Two main views have been held with regard to the formation and
structure of the shell—(1) that of Bowerbank and Carpenter, that the
shell is an organic formation, growing by interstitial deposit, in the
same manner as the teeth and bones of the higher animals; (2) that
of Réaumur, Eisig, and most modern writers, that the shell is of the
nature of an excretion, deposited like a cuticle on the outside of the
skin, being formed simply of a number of calcareous particles held
together by a kind of ‘animal glue.’ Leydig’s view is that the shell of
the Monotocardia is a secretion of the epithelium, but that in the
Pulmonata it originates within the skin itself, and afterwards
becomes free.[335]
According to Carpenter, when a fragment of any recent shell is
decalcified by being placed in dilute acid, a definite animal basis
remains, often so fine as to be no more than a membranous film, but
sometimes consisting of an aggregation of ‘cells’ with perfectly
definite forms. He accordingly divides all shell structure into cellular
and membranous, according to the characteristics of the animal
basis. Cellular structure is comparatively rare; it occurs most notably
in Pinna, where the shell is composed of a vast multitude of tolerably
regular hexagonal prisms (Fig. 162 B). Membranous structure
comprises all forms of shell which do not present a cellular tissue.
Carpenter held that the membrane itself was at one time a
constituent part of the mantle of the mollusc, the carbonate of lime
being secreted in minute ‘cells’ on its surface, and afterwards
spreading over the subjacent membrane through the bursting of the
cells.
The iridescence of nacreous shells is due to a peculiar lineation of
their surface, which can be readily detected by a lens. According to
Brewster, the iridescence is due to the alternation of layers of
granular carbonate of lime and of a very thin organic membrane, the
layers very slightly undulating. Carpenter, on the other hand, holds
that it depends upon the disposition of a single membranous layer in
folds or plaits, which lie more or less obliquely to the general surface,
so that their edges show as lines. The nacreous type of shell occurs
largely among those Mollusca which, from other details in their
organisation, are known to represent very ancient forms (e.g.
Nucula, Avicula, Trigonia, Nautilus). It is also the least permanent,
and thus in some strata we find that only casts of the nacreous shells
remain, while those of different constitution are preserved entire.
Porcellanous shells (of which the great majority of Gasteropoda
are instances) usually consist of three layers, each of which is
composed of a number of adjacent plates, like cards on edge. The
inclination of the plates in the different layers varies, but that of the
plates in the inner and outer layer is frequently the same, thus if the
plates are transverse in the middle stratum, they are longitudinal in
the inner and outer strata, and, if longitudinal in the middle, they are
transverse in the other two. Not uncommonly (Fig. 163 B) other
layers occur. In bivalves the disposition and nature of the layers is
much more varied.
Fig. 162.—A, Section of shell of Unio: a, periostracal layer; b, prismatic layer; c,
nacreous layer. B, Horizontal section of shell of Pinna, showing the hexagonal
prisms.

In Unio the periostracal or uppermost layer is very thin; beneath

this is a prismatic layer of no great depth, while the whole remainder
of the shell is nacreous (Fig. 162 A). Many bivalves show traces of
tubular structure, while in the Veneridae the formation and character
of the layers approaches closely to that of the Gasteropoda. Further
details may be gathered from Carpenter’s researches.[336]
Formation of Shell.[337]—The mantle margin is the principal
agent in the deposition of shell. It is true that if the shell be fractured
at any point, the hole will be repaired, thus showing that every part of
the mantle is furnished with shell-depositing cells, but such new
deposits are devoid of colour and of periostracum, and no
observation seems to have been made with regard to the layers of
which they are composed. As a rule the mantle, except at its margin,
only serves to thicken the innermost layer of shell.
It is probable that the carbonate of lime, of which the shell is
mainly composed, is separated from the blood by the epithelial cells
of the mantle margin, and takes the crystalline or granular form as it
hardens on exposure after deposition. The three layers of a
porcellanous shell are deposited successively, and the extreme edge
of the mouth, when shell is forming, will contain only one layer, the
outermost; a little further in, two layers appear, and further still, three.
The pigment cells which colour the surface are situated at the front
edge of the mantle margin.

Fig. 163.—Sections of shells. A, Conus: a, outer layer; b, middle prismatic layer,

with obliquely intersecting laminae above and below; c, inner layer. B, Oliva:
a, outer layer; b, layer of crossed and curved laminae; c, prismatic layer,
succeeded by layer of laminae at right angles to one another; d, inner layer. C,
Cypraea: a, outer layer; b, middle layer; c, inner layer.
Shelly matter is deposited, and probably secreted, not only by the
mantle, but also in some genera by the foot. This is certainly the
case in Cymbium, Oliva, Ancillaria, Cassis, Distortio, and others, in
several of which the foot is so large that the shell appears to be quite
immersed in it.[338]
The deposition of shell is not continuous. Rest periods occur,
during which the function is dormant; these periods are marked off
on the edge of the shell, and are known as lines of growth. In some
cases (Murex, Triton, Ranella), the rest period is marked by a
decisive thickening of the lip, which persists on the surface of the
shell as what is called a varix (see p. 263).
 Fig. 164.—Murex
tenuispina L.,
Ceylon. × ⅔.

Fig. 165.—Neritina
longispina Récl., Mauritius.
(Operculum removed.)
The various details of sculpture on the exterior surface of the
shell, the striae, ribs, nodules, imbrications, spines, and other forms
of ornamentation are all the product of similar and corresponding
irregularities in the mantle margin, and have all been originally
situated at the edge of the lip. Spines, e.g. those of Murex and
Pteroceras, are first formed as a hollow thorn, cleft down its lower
side, and are afterwards filled in with solid matter as the mantle edge
withdraws. What purpose is served by the extreme elaboration of
these spiny processes in some cases, can hardly be considered as
satisfactorily ascertained. Possibly they are a form of sculptural
development which is, in the main, protective, and secures to its
owners immunity from the attacks of predatory fishes.
‘Attached’ genera (e.g. Chama, Spondylus) when living on smooth
surfaces have a flat shell, but when affixed to coral and other uneven
surfaces they become very irregular in shape. The sculpture of the
base on which they rest is often reproduced in these ‘attached’
shells, not only on the lower, but also on the upper valve, the
growing edge of which rests on the uneven surface of the base.
Oysters attached to the branches of the mangrove frequently display
a central convex rib, modelled on the shape of the branch, from
which the plaits of sculpture radiate, while specimens fixed to the
smooth trunk have no such rib. Crepidula, a genus which is in the
habit of attaching itself to other shells, varies in sculpture according
to that of its host. Sometimes the fact may be detected that a
specimen has lived on a ribbed shell when young, and on a smooth
one when old, or vice versâ. A new genus was actually founded by
Brown for a Capulus which had acquired ribs through adhesion to a
Pecten. A specimen of Hinnites giganteus in the British Museum
must at one period of its growth have adhered to a surface on which
was a Serpula, the impression of which is plainly reproduced on the
upper valve of the Hinnites.[339]
 Fig. 166.—A specimen of Anomia
ephippium L., Weymouth,
taken upon Pecten maximus,
the sculpture of which is
reproduced on the upper
valve of the Anomia, and
even on a young Anomia
attached to the larger
specimen.
Growth of the Shell.—Nothing very definite is known with regard
to the rate of growth of the shell in marine Mollusca. Under
favourable conditions, however, certain species are known to
increase very rapidly, especially if the food supply be abundant, and
if there is no inconvenient crowding of individuals. Petit de la
Saussaye mentions[340] the case of a ship which sailed from
Marseilles for the west coast of Africa, after being fitted with an
entirely new bottom. On arriving at its destination, the vessel spent
68 days in the Gambia River, and took 86 days on its homeward
voyage. On being cleaned immediately on its return to Marseilles, an
Avicula 78 mm. and an Ostrea 95 mm. long (both being species
peculiar to W. Africa) were taken from its keel. These specimens had
therefore attained this growth in at most 154 days, for at the period
of their first attachment they are known to be exceedingly minute. P.
Fischer relates[341] that in 1862 a buoy, newly cleaned and painted,
was placed in the basin at Arcachon. In less than a year after, it was
found to be covered with thousands of very large Mytilus edulis, 100