Download the full version of the textbook now at textbookfull.

com

Quantitative Data Analysis for Language

Assessment Volume II Advanced Methods Vahid
Aryadoust

https://textbookfull.com/product/quantitative-
data-analysis-for-language-assessment-volume-ii-
advanced-methods-vahid-aryadoust/

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

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

Quantitative Data Analysis for Language Assessment Volume

I Fundamental Techniques Vahid Aryadoust (Editor)

https://textbookfull.com/product/quantitative-data-analysis-for-
language-assessment-volume-i-fundamental-techniques-vahid-aryadoust-
editor/
textbookfull.com

Quantitative Methods of Data Analysis for the Physical

Sciences and Engineering 1st Edition Douglas G. Martinson

https://textbookfull.com/product/quantitative-methods-of-data-
analysis-for-the-physical-sciences-and-engineering-1st-edition-
douglas-g-martinson/
textbookfull.com

Quantitative Analysis of Market Data 1st Edition Adam

Grimes

https://textbookfull.com/product/quantitative-analysis-of-market-
data-1st-edition-adam-grimes/

textbookfull.com

Distinction Exclusivity and Whiteness Elite Nigerian

Parents and the International Education Market Pere Ayling

https://textbookfull.com/product/distinction-exclusivity-and-
whiteness-elite-nigerian-parents-and-the-international-education-
market-pere-ayling/
textbookfull.com
The 2008 Global Financial Crisis in Retrospect: Causes of
the Crisis and National Regulatory Responses Robert Z.
Aliber
https://textbookfull.com/product/the-2008-global-financial-crisis-in-
retrospect-causes-of-the-crisis-and-national-regulatory-responses-
robert-z-aliber/
textbookfull.com

Negotiate the Best Lease for Your Business 4th Edition

Janet Portman

https://textbookfull.com/product/negotiate-the-best-lease-for-your-
business-4th-edition-janet-portman/

textbookfull.com

Medical image recognition, segmentation and parsing :

machine learning and multiple object approaches 1st
Edition Zhou
https://textbookfull.com/product/medical-image-recognition-
segmentation-and-parsing-machine-learning-and-multiple-object-
approaches-1st-edition-zhou/
textbookfull.com

Treasury Notes Investing 101 Step by Step Guide and Smart

Investor Starter s Handbook Usiere Uko

https://textbookfull.com/product/treasury-notes-investing-101-step-by-
step-guide-and-smart-investor-starter-s-handbook-usiere-uko/

textbookfull.com

Computed Tomography of the Lung A Pattern Approach 2nd

Edition Johny A. Verschakelen

https://textbookfull.com/product/computed-tomography-of-the-lung-a-
pattern-approach-2nd-edition-johny-a-verschakelen/

textbookfull.com
Handbook of Research on Herbal Liver Protection:
Hepatoprotective Plants 1st Edition T Pullaiah

https://textbookfull.com/product/handbook-of-research-on-herbal-liver-
protection-hepatoprotective-plants-1st-edition-t-pullaiah/

textbookfull.com
Quantitative Data Analysis for
Language Assessment Volume II

Quantitative Data Analysis for Language Assessment Volume II: Advanced

Methods demonstrates advanced quantitative techniques for language assess-
ment. The volume takes an interdisciplinary approach and taps into expertise
from language assessment, data mining, and psychometrics. The techniques cov-
ered include Structural Equation Modeling, Data Mining, Multidimensional
Psychometrics and Multilevel Data Analysis. Volume II is distinct among avail-
able books in language assessment, as it engages the readers in both theory
and application of the methods and introduces relevant techniques for theory
construction and validation. This book is highly recommended to graduate stu-
dents and researchers who are searching for innovative and rigorous approaches
and methods to achieve excellence in their dissertations and research. It is also
a valuable source for academics who teach quantitative approaches in language
assessment and data analysis courses.

Vahid Aryadoust is assistant professor of language assessment literacy at the

National Institute of Education of Nanyang Technological University, Singapore.
He has led a number of language assessment research projects funded by, for
example, the Ministry of Education (Singapore), Michigan Language Assessment
(USA), Pearson Education (UK), and Paragon Testing Enterprises (Canada)
and has published his research in Language Testing, Language Assessment
Quarterly, Assessing Writing, Educational Assessment, Educational Psychology,
and Computer Assisted Language Learning. He has also (co)authored a number
of book chapters and books that have been published by Routledge, Cambridge
University Press, Springer, Cambridge Scholar Publishing, Wiley Blackwell, and
so on. He is a member of the advisory board of multiple international jour-
nals including Language Testing (Sage), Language Assessment Quarterly (Taylor
& Francis), Educational Assessment (Taylor & Francis), Educational Psychology
(Taylor & Francis), and Asia Pacific Journal of Education (Taylor & Francis). In
addition, he has been awarded the Intercontinental Academia Fellowship (2018–
2019), which is an advanced research program launched by the University-Based
Institutes for Advanced Studies. Vahid’s areas of interest include theory-building
and quantitative data analysis in language assessment, neuroimaging in language
comprehension, and eye tracking research.
Michelle Raquel is a senior lecturer at the Centre of Applied English Studies,
University of Hong Kong, where she teaches language testing and assessment
to postgraduate students. She has worked in several tertiary institutions in
Hong Kong as an assessment developer and has either led or been part of a
group that designed and administered large-scale diagnostic and language profi-
ciency assessments such as Hong Kong Institute of Education’s Tertiary English
Language Test (TELT), Hong Kong University of Science and Technology’s
English Language Proficiency Assessment (ELPA), and Diagnostic English
Language Tracking Assessment (DELTA), a government-funded inter-institu-
tional project tasked to develop a computer-based academic English diagnostic
test. She specializes in data analysis, specifically Rasch measurement, and has
published several articles in international journals on this topic as well as on
academic English diagnostic assessment, English as a second language (ESL)
testing of reading and writing, dynamic assessment of second language dramatic
skills, and English for specific purposes (ESP) testing.
Quantitative Data Analysis
for Language Assessment
Volume II
Advanced Methods

Edited by Vahid Aryadoust

and Michelle Raquel
First published 2020
by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
and by Routledge
52 Vanderbilt Avenue, New York, NY 10017
Routledge is an imprint of the Taylor & Francis Group,
an informa business
© 2020 selection and editorial matter, Vahid Aryadoust and Michelle
Raquel; individual chapters, the contributors
The right of Vahid Aryadoust and Michelle Raquel to be identified
as the authors of the editorial material, and of the authors for their
individual chapters, has been asserted in accordance with sections 77
and 78 of the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or
reproduced or utilised in any form or by any electronic, mechanical,
or other means, now known or hereafter invented, including
photocopying and recording, or in any information storage or
retrieval system, without permission in writing 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.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging in Publication Data
A catalog record for this book has been requested

ISBN: 978-1-138-73314-5 (hbk)

ISBN: 978-1-315-18780-8 (ebk)

Typeset in Galliard
by Cenveo® Publisher Services

Visit the eResources: https: ⁄ ⁄ www.routledge.com ⁄ 9781138733145

Table of contents

List of figures vii

List of tables ix
List of volume II contributors xi
Preface xv

Introduction 1
VAHID ARYADOUST AND MICHELLE RAQUEL

SECTION I
Advanced item response theory (IRT) models
in language assessment 13

1 Applying the mixed Rasch model in assessing

reading comprehension 15
PURYA BAGHAEI, CHRISTOPH J. KEMPER, MONIQUE REICHERT,
AND SAMUEL GREIFF

2 Multidimensional Rasch models in first language

listening tests 33
CHRISTIAN SPODEN AND JENS FLEISCHER

3 The log-linear cognitive diagnosis modeling (LCDM)

in second language listening assessment 56
TUĞBA ELIF TOPRAK, VAHID ARYADOUST, AND CHRISTINE GOH

4 Application of a hierarchical diagnostic classification

model in assessing reading comprehension 79
HAMDOLLAH RAVAND
vi Table of contents

SECTION II
Advanced statistical methods in language assessment 99

5 Structural equation modeling to predict performance

in English proficiency tests 101
XUELIAN ZHU, MICHELLE RAQUEL, AND VAHID ARYADOUST

6 Student growth percentiles in the formative assessment

of English language proficiency 127
HUSEIN TAHERBHAI AND DAERYONG SEO

7 Multilevel modeling to examine sources of variability

in second language test scores 150
YO IN’NAMI AND KHALED BARKAOUI

8 Longitudinal multilevel modeling to examine changes

in second language test scores 171
KHALED BARKAOUI AND YO IN’NAMI

SECTION III
Nature-inspired data-mining methods
in language assessment 191

9 Classification and regression trees in predicting

listening item difficulty 193
VAHID ARYADOUST AND CHRISTINE GOH

10 Evolutionary algorithm-based symbolic regression

to determine the relationship of reading and
lexicogrammatical knowledge 215
VAHID ARYADOUST

Index 234
Figures

1.1 Item parameter profiles of latent classes

(Class 1 = 57.1%, Class 2 = 42.9%). 24
2.1 Item characteristic curves of two Rasch model
items with lower (Item 1) and higher (Item 2)
item difficulty, and individuals with lower (θ1) and higher (θ2)
ability (figure adapted from Fleischer, Wirth, & Leutner, 2007). 35
2.2 Between- and within-item dimensionality models
for two measurement dimensions. 37
3.1 ICBC for Item 34. 71
3.2 ICBC for Item 14. 71
3.3 ICBC for Item 17. 72
4.1 Some possible attribute hierarchies.  89
4.2 Hypothesized hierarchies. 90
5.1 A hypothesized SEM model including two latent
variables and multiple observed variables. 103
5.2 A hypothesized model of factors predicting IELTS scores. 113
5.3 DELTA track. 114
5.4 A CFA model with the IELTS latent factor causing
the variance in the observable variables (IELTSL:
IELTS listening score, IELTSR: IELTS reading score,
IELTSW: IELTS writing score, and IELTSS: IELTS
speaking score) (n = 257). 115
5.5 An SEM model with observable DELTA data predicting
the latent variable, IELTS (n = 257). 116
5.6 An SEM model with faculty predicting the latent variable,
IELTS, and the observable DELTA variable (n = 257). 116
6.1 Student’s score estimates at predetermined percentiles
conditioned on a prior score. (This figure is adapted from
Seo, McGrane, & Taherbhai, 2015. While this student is
functioning in the 30th percentile [dashed line], he ⁄she
needs to perform in the 80th percentile to meet
proficiency [dash-dot-dot line]. 132
6.2 An example of smoothing the data set. 136
viii Figures
7.1 a–e Examples of nested data structures. 151
7.2 a–d Fixed and random coefficients with intercepts
and slopes at Level 2. 155
7.3 Schematic graph of the nested structure of the current data. 158
9.1 A hypothetical classification model for decision-making
in a biased language assessment. 195
9.2 Train-test cross-validation (top) versus k-fold
cross-validation (bottom). 197
9.3 CART model for the data (based on Gini method
of tree-growing). 208
10.1 A two-plate model depicting the relationship between
reading and lexicogrammatical knowledge. 229
Tables

1.1 Model fit statistics for the estimated models 23

1.2 Item parameters and fit indices in the two latent classes 23
1.3 Class size and classification accuracy for the latent classes 24
1.4 Mean ability estimates and reliability 25
1.5 Class comparison 26
2.1 Summary of skill sets, languages, and models investigated
in Analysis 1 and Analysis 2 45
2.2 Parameter estimates and global fit statistics for three
Rasch measurement models applied to L1 listening and
reading comprehension test data 46
2.3 Parameter estimates and global fit statistics for three
Rasch measurement models applied to L1 and L2 listening
comprehension test data 47
3.1 Model modifications and fit indices for the MET listening test 66
3.2 The MET listening test Q-matrix and LCDM item
parameter estimates 67
3.3 Probabilities of a correct response for nonmasters
and masters on all the MET items 70
3.4 Tetrachoric correlations among attributes 73
3.5 Attribute mastery classes and their respective attribute
mastery profiles 73
4.1 DCM categorization 82
4.2 DCM studies of language assessment 83
4.3 The final Q-matrix 88
4.4 Absolute fit indices 91
4.5 Relative fit indices 91
4.6 Attribute profile prevalence 92
5.1 Language assessment SEM research reporting guidelines
(based on Ockey and Choi, 2015, and In’nami
and Kiozumi, 2011a) 110
5.2 Fit indices of the hypothesized models and the data 117
5.3 Standardized and unstandardized estimates of
SEM model 3 (see Figure 5.6) 118
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
x Tables
6.1 Summary statistics of five years’ ELP assessment:
Total and modality score 137
6.2 Regression coefficient estimates of SGP model
at the 50th percentile 138
6.3 Examples of students’ predicted scale scores across
the specified percentiles 139
6.4 Comparing each student’s predicted growth
percentiles across total and modality scores 142
7.1 Sample items from the vocabulary size and depth tests 156
7.2 Descriptive statistics 159
7.3 Multilevel model results 161
8.1 Descriptive statistics for number of times test
was taken (N = 1,000) 175
8.2 Descriptive statistics for PTE total scores by test occasion 175
8.3 Models 1–4 for PTE total scores 179
8.4 Models 5–7 for PTE total scores 183
9.1 Demographic information of the listening tests and test takers 203
9.2 Independent variables in the CART model 204
9.3 Distribution of data in test-train cross-validation CART analysis 206
9.4 Classification accuracy, specificity, sensitivity, and ROC 207
9.5 CART-estimated variable importance indices 207
9.6 Sample IF-THEN rules generated by the CART model 209
10.1 Competing models with their R, R2, MSE, and MAE indices 225
10.2 Sensitivity analysis of the variables in the models 227
Volume II contributors

Purya Baghaei is an associate professor in the English Department, Islamic

Azad University, Mashhad Branch, Mashhad, Iran. He holds a PhD in applied
linguistics from Alpen-Adria Universität, Klagenfurt, Austria. He is a scholar
of the Alexander von Humboldt Foundation in Germany and has conducted
post-doctoral research at universities in Vienna, Berlin, Jena, and Bamberg. His
major research interest is in foreign language proficiency testing with a focus on
the applications of item response theory models in test validation and scaling.
He has published numerous articles on language testing and cognitive compo-
nents of second language acquisition in international journals.
Khaled Barkaoui is an associate professor at the Faculty of Education, York
University, Canada. His current research focuses on second-language (L2)
assessment, L2 writing, L2 program evaluation, longitudinal and mixed-­
methods research, and English for academic purposes (EAP). His publications
have appeared in Annual Review of Applied Linguistics, Applied Linguistics,
Assessing Writing, Language Testing, Language Assessment Quarterly, System,
and TESOL Quarterly.
Jens Fleischer is a post-doc researcher at the Department of Instructional Psychol-
ogy of the University of Duisburg-Essen (Germany). His research interests include
the assessment of competencies in secondary and higher education and research
on cognitive and motivational predictors of academic learning. Most recently, he
has worked on the assessment and modeling of cross-curricular problem-solving
competence and was a member of the coordination team of the interdisciplinary
priority research program “Models of Competence.” Currently, he is working in a
research group investigating factors that influence academic learning and academic
success in the entry phase of science and technology study programs.
Christine Goh is professor of linguistics and language education at the National
Institute of Education, Nanyang Technological University, Singapore. Her areas
of interest and expertise are cognitive and metacognitive processes in L2 listen-
ing and speaking, teaching and assessment of L2 listening and speaking, and
teacher cognition. She publishes extensively in these areas in books, book chap-
ters, and journal articles, and her work has been widely cited internationally.
xii Volume II contributors
Samuel Greiff is research group leader, principal investigator, and ATTR ACT-­
fellow at the University of Luxembourg. He holds a PhD in cognitive and exper-
imental psychology from the University of Heidelberg, Germany. He has been
awarded national and international research funds by diverse organizations, is
currently a fellow in the Luxembourg research programme of excellency, and
has published articles in national and international scientific journals and books.
He has an extensive record of conference contributions and invited talks and
serves as editor for several journals, such as the European Journal of Psychological
Assessment and the Journal of Educational Psychology and Thinking Skills and
Creativity.
Yo In’nami is a professor of English at Chuo University, Japan. He is also a
PhD candidate’s adviser and an external PhD examiner at Temple University
Japan Campus. He has taught various undergraduate- and graduate-level courses
in language testing, second language acquisition, and statistics for language
research. He currently is interested in meta-analytic inquiry into the variability
of effects and the longitudinal measurement of change in language proficiency.
His publications have appeared in International Journal of Testing, Language
Assessment Quarterly, Language Learning, Language Testing, System, and
TESOL Quarterly. His website is https: ⁄ ⁄sites.google.com ⁄site ⁄yoinnami
Christoph J. Kemper worked at different universities and research institutes in
Germany and abroad. He currently holds a position as Professor for Differen-
tial Psychology and Assessment at the HSD University of Applied Sciences in
Cologne, Germany. He was recently appointed head of the Center for Psycho-
logical Assessment of HSD University. He teaches psychological assessment, dif-
ferential psychology, and research methods in bachelor’s and master’s programs
(psychology). His research on individual differences and their assessment such as
personality, motives, and anxiety, as well as different aspects of assessment ⁄sur-
vey methodology, is widely published in more than 50 research papers, including
25 peer-reviewed publications. He is also first and coauthor of many assessment
instruments.
Hamdollah Ravand received a PhD in English language teaching from the Uni-
versity of Isfahan, Iran, in 2013. He joined the English Department of Vali-e-
Asr University of Rafsanjan where he is teaching master’s courses on research
methods in Teaching English as a Foreign Language (TEFL), language testing,
statistics and computers, and advanced writing. Hamdollah has been a visiting
researcher to Jena University and the Institute for Educational Quality Improve-
ment (IQB), Germany in 2012 and 2016, respectively. His major research interests
are applications of diagnostic classification models, structural equation models,
item response theory models, and multilevel models to second language data.
Monique Reichert is a psychologist specializing in language assessment, cog-
nitive science issues, and empirical methods. Since 2004, she has worked at the
Luxembourg Centre for Educational Testing (LUCET; formerly: EMACS) of
the University of Luxembourg. As a head of the domain of language test
 Volume II contributors xiii
development, her main areas of work lie in the development and implementa-
tion of quality assurance procedures in language test development. She took
part in different European projects relating to language evaluation, such as in
the EBAFLS, or the CEF-ESTIM project (http: ⁄ ⁄cefestim.ecml.at  ⁄), and she
continuously develops and conducts training in language test development. She
received her PhD for her work on the validity of C-tests.
Daeryong Seo is a senior research scientist at Pearson. He has led various state
assessments and brings international psychometric experience through his work
with the Australian NAPLAN and Global Scale of English. He has published
several studies in international journals and presented numerous psychomet-
ric issues at international conferences, such as for the American Educational
Research Association (AER A). He also served as a program chair of the Rasch
special interest group, AER A. In 2013, he and Dr. Husein Taherbhai received an
outstanding paper award from the California Educational Research Association.
Their paper is titled, “What Makes High School Asian English Learners Tick?”
Christian Spoden is research methods consultant at the German Institute for
Adult Education—Leibniz Centre for Lifelong Learning, Bonn (Germany). His
research focuses on item response theory modeling, the assessment of competen-
cies in secondary and higher education, and large-scale assessment methods for
educational sciences practice. Previously, he was engaged in the implementation
of statewide standardized scholastic achievement tests for German and English
language assessment and mathematical competencies assessment. Dr. Spoden
also worked on the topic of quality of physics instruction and, recently, on psy-
chometric methods for computerized adaptive testing.
Husein Taherbhai is a retired principal research scientist who led large-scale
assessments in the United States, such as in Arizona, Washington, New York,
Maryland, Virginia, and Tennessee, and for the National Physical Therapists’
Association’s licensure examination. Internationally, Dr. Taherbhai led the Edu-
cational Quality and Assessment Office in Ontario, Canada, and worked for the
Central Board of Secondary Education’s Assessment in India. He has published
in various scientific journals and has reviewed and presented at the NCME,
AER A, and Florida State conferences, with papers relating to language learners,
rater effects, and students’ equity and growth in education.
Tuğba Elif Toprak is an assistant professor of applied linguistics  ⁄ ELT at Izmir
Bakircay University, Izmir, Turkey. Her primary research interests are imple-
menting cognitive diagnostic assessment by using contemporary item response
theory models (named cognitive diagnostic psychometric models) and blending
cognition with language assessment. Dr. Toprak has been collaborating with
international researchers on several projects that are largely situated in the fields
of language assessment, psycholinguistics, and the learning sciences. Her future
plan of study includes focusing on intelligent real-time assessment systems by
combining the techniques from several areas such as the learning sciences, cog-
nitive science, and psychometrics.
xiv Volume II contributors
Xuelian Zhu h  olds a master of arts in applied linguistics from Nanyang Tech-
nological University, Singapore, with special focus on language assessment and
psychometric analysis, such as Rasch measurement and structural equation
modeling. She is a lecturer in English-Chinese translation and interpretation
at Sichuan International Studies University in China, where she oversees the
designing of the listening component of the in-house placement test for English
majors. She is also a member of the panel of test designers for a national place-
ment test for the Foreign Language Test and Research Press (FLTRP). Xuelian’s
current research focuses on eye tracking and on the design and validation of
language assessments for academic purposes.
Preface
Vahid Aryadoust and Michelle Raquel

The two-volume book, Quantitative Data Analysis for Language Assessment

(Fundamental Techniques and Advanced Methods), together with the Companion
website, were motivated by the growing need for a comprehensive sourcebook
of quantitative data analysis for the community of language assessment. As the
focus on developing valid and useful assessments continues to intensify in dif-
ferent parts of the world, having a robust and sound knowledge of quantitative
methods has become an increasingly essential requirement. This is particularly
important given that one of the community’s responsibilities is to develop lan-
guage assessments that have evidence of validity, fairness, and reliability. We
believe this would be achieved primarily by leveraging quantitative data analysis
in test development and validation efforts.
It has been the contributors’ intention to write the chapters with an eye
toward what professors, graduate students, and test development companies
would need. The chapters progress gradually from fundamental concepts to
advanced topics, making the volumes suitable reference books for professors
who teach quantitative methods. If the content of the volumes is too heavy for
teaching in one course, we would suggest professors consider using them across
two semesters, or alternatively, choose any chapters that fit the focus and scope
of their courses. For graduate students who have just embarked on their studies
or are writing dissertations or theses, the two volumes would serve as a cogent
and accessible introduction to the methods that are often used in assessment
development and validation research. For organizations in the test development
business, the volumes provide a unique topic coverage and examples of applica-
tions of the methods in small- and large-scale language tests that such organi-
zations often deal with.
We would like to thank all of the authors who contributed their expertise in
language assessment and quantitative methods. This collaboration has allowed us
to emphasize the growing interdisciplinarity in language assessment that draws
knowledge and information from many different fields. We wish to acknowledge
that in addition to editorial reviews, each chapter has been subjected to rigorous
xvi Preface
double-blind peer review. We extend a special note of thanks to a number of
colleagues who helped us during the review process:

Beth Ann O’Brien, National Institute of Education, Singapore

Christian Spoden, The German Institute for Adult Education, Leibniz
Centre for Lifelong Learning, Germany
Guangwei Hu, Hong Kong Polytechnic University, Hong Kong
Hamdollah Ravand, Vali-e-Asr University of Rafsanjan, Iran
Ikkyu Choi, Educational Testing Service, USA
Kirby Grabowski, Teachers College, Columbia University, USA
Mehdi Riazi, Macquarie University, Australia
Moritz Heene, Ludwig-Maximilians-Universität München, Germany
Purya Baghaei, Islamic Azad University of Mashad, Iran
Shane Phillipson, Monash University, Australia
Shangchao Min, Zhejiang University, China
Thomas Eckes, Gesellschaft für Akademische Studienvorbereitung und
Testentwicklung e. V. c  ⁄o TestDaF-Institut Ruhr-Universität Bochum,
Germany
Trevor Bond, James Cook University, Australia
Tuğba Elif Toprak, Izmir Bakircay University, Turkey
Wenshu Luo, National Institute of Education, Singapore
Yan Zi, The Education University of Hong Kong, Hong Kong
Yasuyo Sawaki, Waseda University, Japan
Yo In’nami, Chuo University, Japan
Zhang Jie, Shanghai University of Finance and Economics, China

We hope that the readers will find the volumes useful in their research and
pedagogy.

Vahid Aryadoust and Michelle Raquel

Editors
July 2019
Introduction
Vahid Aryadoust and Michelle Raquel

Quantitative techniques are mainstream components in most of the published

literature in language assessment as they are essential in test development and
validation research (Chapelle, Enright, & Jamieson, 2008). There are three fam-
ilies of quantitative methods adopted in language assessment research: measure-
ment models, statistical methods, and data mining (although admittedly, setting
a definite boundary in this classification of methods would not be feasible).
Borsboom (2005) proposes that measurement models, the first family of quan-
titative methods in language assessment, fall in the paradigm of classical test the-
ory (CTT), Rasch measurement, or item response theory (IRT). The common
feature of the three measurement techniques is that they are intended to predict
outcomes of cognitive, educational, and psychological testing. However, they
do have significant differences in their underlying assumptions and applications.
CTT is founded on true scores that can be estimated by using the error of meas-
urement and observed scores. Internal consistency reliability and generalizability
theory are also formulated based on CTT premises. Rasch measurement and
IRT, on the other hand, are probabilistic models that are used for the measure-
ment of latent variables—attributes that are not directly observed. There are a
number of unidimensional Rasch and IRT models, which assume the attribute
underlying test performance is comprised of only one measurable feature. There
are also multidimensional models, which postulate that latent variables meas-
ured by tests are many and multidivisible. Determining whether a test is unidi-
mensional or multidimensional requires theoretical grounding, the application
of sophisticated quantitative methods, and an evaluation of the test context. For
example, multidimensional tests can be used to provide fine-grained diagnostic
information to stakeholders; thus a multidimensional IRT model can be used
to derive useful diagnostic information from test scores. In these two volumes,
CTT and unidimensional Rasch models are discussed in Volume I and multidi-
mensional techniques are covered in Volume II.
The second group of methods is statistical and consists of the commonly used
methods in language assessment such as t-tests, analysis of variance (ANOVA), anal-
ysis of covariance (ANCOVA), multivariate analysis of covariance (MANCOVA),
regression models, and factor analysis, which are covered in Volume I.
In addition, multilevel modeling and structural equation modeling (SEM)
2 Aryadoust and Raquel
are presented in Volume II. The research questions that these techniques aim
to address range from comparing average performances of test takers to predic-
tion and data reduction. The third group of models falls under the umbrella
of data-mining techniques, which we believe are relatively under-researched
and underutilized techniques in language assessment. Volume II presents two
data-mining methods: classification and regression trees (CART) and evolution-
ary algorithm-based symbolic regression, both of which are used for prediction
and classification. These methods detect the relationship between dependent
and independent variables in the form of mathematical functions and confirm
the relationships across separate data sets. This feature of the two data-mining
techniques, discussed in Volume II, improves the precision and generalizability
of the detected relationships.
We provide an overview of the two volumes in the next sections.

Quantitative Data Analysis for Language

Assessment Volume I: Fundamental Techniques
This volume is comprised of 11 chapters that are contributed by a number of
experts in the field of language assessment and quantitative data analysis tech-
niques. The aim of the volume is to revisit the fundamental quantitative topics
that have been used in the language assessment literature and shed light on their
rationales and assumptions. This is achieved through delineating the technique
covered in each chapter, providing a (brief) review of its application in previous
language assessment research, and giving a theory-driven example of the appli-
cation of the technique. The chapters in Volume I are grouped into three main
sections, which are discussed as follows.

Section I. Test development, reliability, and generalizability

Chapter 1: Item analysis in language assessment (Rita Green)

This chapter deals with a fundamental but, as R ita Green notes, often-delayed
step in language test development. Item analysis is a quantitative method that
allows test developers to examine the quality of test items, i.e., which test
items are working well (constructed to assess the construct they are meant
to assess) and which items should be revised or dropped to improve overall
test reliability. Unfortunately, as the author notes, this step commonly is
done after a test has been administered and not when items have just been
developed. The chapter starts with an explanation of the importance of this
method at the test development stage. Then, several language testing studies,
which have utilized this method to investigate test validity and reliability, to
improve standard setting sessions, and to investigate the impact of test for-
mat and different testing conditions on test taker performance, are reviewed.
The author further emphasizes the need for language testing professionals to
 Introduction 3
learn this method and its link to language assessment research by suggesting
five research questions in item analysis. The use of this method is demon-
strated by an analysis of a multiple-choice grammar and vocabulary test. The
author concludes the chapter by demonstrating how the analysis can answer
the five research questions proposed, as well as offering suggestions on how
to improve the test.

Chapter 2: Univariate generalizability theory in language

assessment (Yasuyo Sawaki and Xiaoming Xi)
In addition to item analysis, investigating reliability and generalizability is a fun-
damental consideration of test development. Chapter 2 presents and extends the
framework to investigate reliability within the paradigm of classical test theory
(CTT). Generalizability theory (G theory) is a powerful method of investigating
the extent to which scores are reliable as it is able to account for different sources
of variability and their interactions in one analysis. The chapter provides an over-
view of the key concepts in this method, outlines the steps in the analyses, and
presents an important caveat in the application of this method, i.e., conceptu-
alization of an appropriate rating design that fits the context. A sample study
demonstrating the use of this method is presented to investigate the dependabil-
ity of ratings given on an English as a foreign language (EFL) summary writing
task. The authors compared the results of two G theory analyses, the rating
method and the block method, to demonstrate to readers the impact of rating
design on the results of the analysis. The chapter concludes with a discussion of
the strengths of the analysis compared to other CTT-based reliability indices,
the value of this method in investigating rater behavior, and suggested references
should readers wish to extend their knowledge of this technique.

Chapter 3: Multivariate generalizability theory in language

assessment (Kirby C. Grabowski and Rongchan Lin)
In performance assessments, multiple factors contribute to generate a test tak-
er’s overall, score such as task type, the rating scale structure, and the rater,
meaning that scores are influenced by multiple sources of variance. Although
univariate G theory analysis is able to determine the reliability of scores, it
is limited in that it does not consider the impact of these sources of variance
simultaneously. Multivariate G theory analysis is a powerful statistical technique
because, in addition to results generated by univariate G theory analysis, it is
able to generate a reliability index accounting for all these factors in one analysis.
The analysis is also able to consider the impact of subscales of a rating scale. The
authors begin the chapter with an overview of the basic concepts of multivariate
G theory. Next, they illustrate an application of this method through an analysis
of a listening-speaking test where they make clear links between research ques-
tions and the results of the analysis. The chapter concludes with caveats in the
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
4 Aryadoust and Raquel
use of this method and suggested references for readers who wish to complement
their MG theory analyses with other methods.

Section II. Unidimensional Rasch measurement

Chapter 4: Applying Rasch measurement in language assessment:

Unidimensionality and local independence (Jason Fan and
Trevor Bond)
This chapter discusses the two fundamental concepts required in the application
of Rasch measurement in language assessment research, i.e., unidimensionality
and local independence. It provides an accessible discussion of these concepts
in the context of language assessment. The authors first explain how the two
concepts should be perceived from a measurement perspective. This is followed
by a brief explanation of the Rasch model, a description of how these two meas-
urement properties are investigated through Rasch residuals, and a review of
Rasch-based studies in language assessment that reports the existence of these
properties to strengthen test validity claims. The authors demonstrate the inves-
tigation of these properties through the analysis of items in a listening test using
the Partial Credit Rasch model. The results of the study revealed that the listen-
ing test is unidimensional and that the principal component analysis of residuals
analysis provides evidence of local independence of items. The chapter concludes
with a discussion of the practical considerations and suggestions on steps to take
should test developers encounter situations where these properties of measure-
ment are violated.

Chapter 5: The Rasch measurement approach to differential item

functioning (DIF) analysis in language assessment research
(Michelle Raquel)
This chapter continues the discussion of test measurement properties. Differential
item functioning (DIF) is the statistical term used to describe items that inad-
vertently have different item estimates for different subgroups because they are
affected by characteristics of the test takers such as gender, age group, or ethnic-
ity. The author first explains the concept of DIF and then provides a brief over-
view of different DIF detection methods used in language assessment research.
A review of DIF studies in language testing follows that includes a summary of
current DIF studies, the DIF method(s) used, and whether or not the studies
investigated the causes of DIF. The chapter then illustrates one of the most com-
monly used DIF detection methods, the Rasch-based DIF analysis method. The
sample study investigates the presence of DIF in a diagnostic English listening
test where students were classified according to the English language curriculum
they have taken, Hong Kong versus Macau. The results of the study revealed that
although there were a significant number of items flagged for DIF, overall test
results did not seem to be affected.
 Introduction 5
Chapter 6: Application of the rating scale model and the partial
credit model in language assessment research (Ikkyu Choi)
This chapter introduces two Rasch models that are used to analyze polyto-
mous data usually generated by performance assessments (speaking or writ-
ing tests) and questionnaires used in language assessment studies. First, Ikkyu
Choi explains the relationship of the rating scale model (RSM) and the partial
credit model (PCM) through a gentle review of their algebraic representations.
This is followed by a discussion of the differences of these models and a review
of studies that have utilized this method. The author notes in his review that
researchers rarely provide a rationale for the choice of model and neither do they
compare models. In the sample study investigating the scale of a motivation
questionnaire, the author provides a thorough and graphic comparison and eval-
uation of the RSM and the PCM models and their impact on the scale structure
of the questionnaire. The chapter concludes by providing justification as to why
the PCM was more appropriate for the context, the limitations of the parameter
estimation method used by the sample study, and a list of suggested topics to
extend the reader’s knowledge of the topic.

Chapter 7: Many-facet Rasch measurement: Implications

for rater-mediated language assessment (Thomas Eckes)
This chapter discusses one of the most popular item-response theory (IRT)-
based methods to analyze rater-mediated assessments. A common problem in
speaking and writing tests is that the marks or grades are dependent on human
raters who most likely have their own conceptions of how to mark despite
their training, and this impacts test reliability. Many-facet Rasch measurement
(MFRM) provides a solution to this problem in that the analysis simultaneously
includes multiple facets such as raters, assessment criteria, test format, or the
time when a test is taken. The author first provides an overview of rater-mediated
assessments and MFRM concepts. The application of this method is illustrated
through an analysis of a writing assessment where the author demonstrates how
to determine rater severity, consistency of ratings, and how to generate test scores
after adjusting for differences in ratings. The chapter concludes with a discussion
on advances in MFRM research and controversial issues related to this method.

Section III. Univariate and multivariate statistical analysis

Chapter 8: Analysis of differences between groups: The t-test and

the analysis of variance (ANOVA) in language assessment
(Tuğba Elif Toprak)
The third section of this volume starts with a discussion of two of the most
fundamental and commonly used statistical techniques for comparing test score
results and determining whether differences between the groups are due to
6 Aryadoust and Raquel
chance. For example, language testers often find themselves trying to com-
pare two or multiple groups of test takers or to compare pre-test and post-
test scores. The chapter starts with an overview of t-tests and the analysis of
variance (ANOVA) and the assumptions that must be met before embarking
on these analyses. The literature review provides summary tables of recent
studies that have employed each method. The application of the t-test is demon-
strated through a sample study that investigated the impact of English songs
on students’ pronunciation development where the author divided the students
into two groups (experimental versus control group) and then compared the
groups’ results on a pronunciation test. The second study utilized ANOVA to
determine if students’ academic reading proficiency differed across college years
(freshman, sophomores, juniors, and seniors) and which group was significantly
different from the others.

Chapter 9: Application of ANCOVA and MANCOVA in

language assessment research (Zhi Li and Michelle Y. Chen)
This chapter extends the discussion of methods used to compare test results.
Instead of using one variable to classify groups that are compared, analysis of
covariance (ANCOVA) and multivariate analysis of covariance (MANCOVA)
consider multiple variables of multiple groups to determine whether or not dif-
ferences in group scores are statistically significant. ANCOVA is used when there
is only one independent variable, while MANCOVA is used when there are two
or more independent variables that are included in the comparison. Both tech-
niques control for the effect of one or more variables that covary with the depend-
ent variables. The chapter begins with a brief discussion of these two methods,
the situations in which they should be used, the assumptions that must be ful-
filled before analysis can begin, and a brief discussion of how results should be
reported. The authors present the results of their meta-analyses of studies that
have utilized these methods and outline the issues related to results reported in
these studies. The application of these methods is demonstrated in the analyses
of the Programme for International Student Assessment (PISA 2009) reading
test results of Canadian children.

Chapter 10: Application of linear regression in language assessment

(Daeryong Seo and Husein Taherbhai)
There are cases when language testers need to determine the impact of one
variable on another variable such as if someone’s first language has an impact on
their learning of a second language. Linear regression is the appropriate statis-
tical technique to use when one aims to determine the extent to which one or
more independent variables linearly impact a dependent variable. This chapter
opens with a brief discussion of the differences between single and multiple
linear regression and a full discussion on the assumptions that must be fulfilled
before commencing analysis. Next, the authors present a brief literature review
 Introduction 7
of factors that affect English language proficiency as these determine what var-
iables should be included in the statistical model. The sample study illustrates
the application of linear regression by predicting students’ results on an English
language arts examination based on their performance in English proficiency
tests of reading, listening, speaking, and writing. The chapter concludes with a
checklist of concepts to consider before doing regression analysis.

Chapter 11: Application of exploratory factor analysis in

language assessment (Limei Zhang and Wenshu Luo)
A standard procedure in test and survey development is to check and see whether
a test or questionnaire measures one underlying construct or dimension. Ideally,
test and questionnaire items are constructed to measure a latent construct
(e.g., 20 items to measure listening comprehension), but each item is designed to
measure different aspects of the construct (e.g., items that measure the ability to
listen for details, ability to listen for main ideas, etc.). Exploratory factor analysis
(EFA) is a statistical technique that examines how items are grouped together
into themes and how they ultimately measure the latent trait. The chapter com-
mences with an overview of EFA, the different methods to extract the themes
(factors) from the data, and an outline of steps in conducting an EFA. This is
followed by a literature review that highlights the different ways the method
has been applied in language testing research, with specific focus on studies
that confirm the factor structure of tests and questionnaires. The sample study
demonstrates how EFA can do this by analyzing the factor structure of the
Reading Test Strategy Use Questionnaire used to determine the types of read-
ing strategies that Chinese students utilize as they complete reading compre-
hension tests.

Quantitative Data Analysis for Language Assessment

Volume II: Advanced Methods
Volume II comprises three major categories of quantitative methods in language
testing research: advanced item response theory (IRT), advanced statistical
methods, and nature-inspired data-mining methods. We provide an overview of
the sections and chapters as follows.

Section I. Advanced item response theory (IRT) models

in language assessment

Chapter 1: Applying the mixed Rasch modeling in assessing reading

comprehension (Purya Baghaei, Christoph J. Kemper, Monique
Reichert, and Samuel Greif)
In this chapter, the authors discuss the application of the mixed Rasch
model (MR M) in assessing reading comprehension. MR M is an advanced
8 Aryadoust and Raquel
psychometric approach for detecting latent class differential item function-
ing (DIF) that conflates the Rasch model and latent class analysis. MR M
relaxes some of the requirements of conventional Rasch measurement while
preserving most of the fundamental features of the method. MR M further
combines the Rasch model with latent class modeling that classifies test tak-
ers into exclusive classes with qualitatively different features. Baghaei et al.
apply the model to a high-stakes reading comprehension test in English as a
foreign language and detect two latent classes of test takers for whom the dif-
ficulty level of the test items differs. They discuss the differentiating feature
of the classes and conclude that MR M can be applied to identify sources of
multidimensionality.

Chapter 2: Multidimensional Rasch models in first language

listening tests (Christian Spoden and Jens Fleischer)
Since the introduction of Rasch measurement to language assessment, a
group of scholars have contended that language is not a unidimensional phe-
nomenon and, accordingly, unidimensional modeling of language assessment
data (e.g., through the unidimensional Rasch model) would conceal the role
of many linguistic features that are integral to language performance. The
multidimensional Rasch model could be viewed as a response to these con-
cerns. In this chapter, the authors provide a didactic presentation of the multi-
dimensional Rasch model and apply it to a listening assessment. They discuss
the advantages of adopting the model in language assessment research, spe-
cifically the improvement in the estimation of reliability as a result of the
incorporation of dimension correlations, and explain how model compari-
son can be carried out, while elaborating on multidimensionality in listening
comprehension assessments. They conclude the chapter with a brief summary
of other multidimensional Rasch models and their value in language assess-
ment research.

Chapter 3: The log-linear cognitive diagnosis modeling (LCDM)

in second language listening assessment (Tuğba Elif Toprak,
Vahid Aryadoust, and Christine Goh)
Another group of multidimensional models, called cognitive diagnostic
models (CDMs), combine psychometrics and psychology. One of the differ-
ences between CDMs and the multidimensional Rasch models is that the for-
mer family estimates subskills mastery of test takers, whereas the latter group
provides a general estimation of ability for each subskill. In this chapter, the
authors introduce the log-linear cognitive diagnosis modeling (LCDM), which
is a flexible CDM technique for modeling assessment data. They apply the
model to a high-stakes norm-referenced listening test (a practice that is known
as retrofitting) to determine whether they can derive diagnostic information
Random documents with unrelated
content Scribd suggests to you:
 In the Florida Everglades

By William A. Stimpson

“G ood-by, fellows; don’t expect me back before supper time.”

Waving his hand to his friends, Alfred Whyte pushed the
bateau into the water, took his seat in the centre, and with a
few strong, even strokes of the paddle sent the frail craft out of sight
around a bend in the stream.
It was on the edge of the Florida Everglades, those low, marshy
tracts of swamp land that cover the whole of the lower end of the
peninsula. Two New York boys, Willard King and Marvin Stebbins,
had homesteaded a claim in the heart of the morass and were
engaged in growing tomatoes for the northern markets. Alfred, a
former schoolmate, was spending a few weeks with them in their
southern home.
The piece of land upon which the two northerners had settled was
about fifty acres in extent. It rose, island-like, from out the midst of
the network of little creeks and streams that crisscrossed in every
direction and made a veritable land-and-water spider’s web of that
part of the State.
The tomato plants were set out in February and now, the first of
April, the tomatoes had begun to turn red and were large enough to
be picked. They had to be handled very carefully, wrapped in tissue
paper, and packed in light wooden crates, so as to permit the
process of ripening to be completed on the trip north. Picking and
packing them was tedious and took considerable time. Both the
young truck farmers had their hands full, and when a flock of wild
ducks flew overhead on their way to the feeding grounds half a mile
further inland, they merely directed a passing glance upward and
then, stifling their sportsmen’s instinct, turned to their work again.
All the morning the wild fowl could be heard thrashing about in the
tall grass at the lagoon, and both King and Stebbins were sorely
tempted several times to slip up stream in the hope of bagging a
couple. But the steamer on which they intended shipping their
produce sailed from Lincoln, fifteen miles east, the next afternoon,
and by working persistently until dark they could hardly get their
crop ready for an early start on the following morning for the river
town.
“If neither of you fellows can spare the time to go duck shooting,
why can’t I paddle up there and try a shot or two?” asked Alfred,
late in the afternoon.
“All the reason in the world, Al,” replied King. “No one except a
native, or a person who has lived here as long as we have, can
traverse this swamp in safety. Why, before you reach the lake where
the ducks are you will pass eight or ten little streams, any one of
which you are just as likely to enter as to keep on up the main
channel. We’re afraid you’ll get lost, Al. Don’t you think so?” he
asked, turning to Stebbins.
“But I’ve been all around there with you fellows,” explained Alfred,
trying in vain to conceal his disappointment. “I’ve been up to the
lake, too, and I know the main stream perfectly well. I’m going to
try it, for I must have some roast duck.”
Both the boys tried to dissuade him from the undertaking, but he
was insistent, and finally they gave a reluctant consent. Realizing
fully his lack of acquaintance with the swamp, Whyte paid particular
attention to his surroundings as he paddled on, fearing that he
might turn into one of those little side streams of which King had
warned him.
Suddenly, ahead of him, he saw the ducks. Paddling noiselessly,
scarcely rippling the water as he passed through, he got within
range of the flock without alarming them. Bang! bang! went both
barrels of his twelve-bore, and at the reports the ducks rose from
the water with a loud whirr. One bird was wounded and lagged
behind the others. It fluttered along a hundred yards or so, then
sank in a clump of marsh grass, took wing again, but went less than
ten yards, when it turned a somersault in the air and dropped.
A few strokes of the paddle carried the bateau close to where the
bird had fallen, but when he reached the spot Whyte found that a
stretch of marsh lay between the edge of the water and his prize. He
tried to reach the duck with the paddle but could not do so. It was a
fine, fat bird, as he could plainly see, but it lay beyond his reach.
“Just my luck,” he muttered, after several unsuccessful attempts to
reach the bird. “I wonder if those hummocks will hold me,” noticing
the tufts of thick, coarse grass that dotted the morass in every
direction.
The hummocks looked firm enough to bear his weight, so pushing
the prow of the boat as far into the edge of the bank as he could, he
stepped out and tried the first one. It was solid and unyielding.
Certain, then, that his plan was a feasible one, he sprang to the next
hummock and on until he had the bird in his hand. In returning, he
rested too much weight upon one of the tufts of thick grass. The
treacherous mud gave way, his foot slipped, and down he went into
the black ooze up to his thighs.
With an exclamation of impatience, he endeavored to withdraw his
feet and legs. They stuck fast. He tried a second time, but the mud
held him as in a vise. Putting forth all his strength and seizing
several blades of the long, coarse grass within his reach, he tried his
best to extricate himself, but to his dismay he found the sticky mud
to be as unyielding as quicksand. What was worse, when he ceased
his efforts he discovered that he had sunk deeper in the mire and
was now embedded nearly up to his breast.
Thoroughly frightened, he remained perfectly passive and began to
think. He realized that he was in a serious predicament, held a
prisoner, as he was, in the black, slimy mud of the swamp, and it
was cold there, too. His gun lay within reach, and, resting the arm
lengthwise, he made another attempt to release himself, but his
efforts were unavailing. The gun sank in the ooze, and in extracting
it he found that his exertions had caused him to sink several inches
deeper. The top of the mud now reached to his armpits.
He glanced at the sun, and, seeing it low in the west, was
comforted. King and Stebbins, becoming alarmed at his non-
appearance, would soon be setting out to look for him, he thought,
if they were not already doing so. His eyes wandered towards the
opposite bank, and he was struck with its unfamiliar appearance.
Instead of the low, flat marsh that lined that side of the stream, as
he well knew, he was looking upon a patch of higher land similar to
the one upon which King and Stebbins had their home. It dawned
upon him then for the first time that he had left the main channel.
As the realization of his true position came home to him, hope died.
Thinking that he was somewhere along the stream, he had felt sure
of rescue, but his discovery altered the situation completely. How far
out of his true course he was he had no way of knowing, and the
thought of the awful days and nights that would pass while he stood
there dying, if the mud did not eventually bury him and make his
death even a more horrible one, was far from pleasant.
Frantically he struggled to free himself, but he was held fast as
though he had been shackled in irons, and his struggles only left him
exhausted. Great beads of perspiration stood out on his brow. His
mouth was dry and parched and his head began to swim. He felt
that he was losing his reason, but he pulled himself together with a
herculean effort. His legs and feet were cold and numb, and the
keen night wind nipped his ears and nose cruelly. The mud under his
arms had begun to freeze, and unless he kept breaking it continually
with his hands, a stiff crust would form at the top.
He racked his brain to devise some plan of escape from his terrible
position, but could think of nothing except to shout. That, he
supposed, would only be a waste of energy, but he must do
something. Gathering himself together, he essayed to call, but his
mouth was so parched that his voice did not penetrate further than
ten yards. He tried again, and this time found himself shouting
louder. Again and again he shouted until his voice echoed and re-
echoed through the everglades.
As the sounds died away his ear caught a faint call that seemed like
an answer to his own. Flushed with hope, he shouted again and
then strained his ears to listen. But silence, broken only by the
twittering of the night birds, reigned about him.
Once more he shouted, and again he thought he heard a reply, or
was it an echo of his own voice? The ordeal was too much for him,
and with a groan his head drooped and he lost consciousness.
With King and Stebbins the time passed until sundown before they
realized how late it was, and then they dropped their work and
looked along the stream in the direction taken by their guest.
“It is nearly seven o’clock, Marvin,” remarked King, consulting his
watch. “Al said he would be back by supper time, and here it is an
hour after. I believe he’s lost.”
“If that’s the case, we must find him before dark, or he’ll have to
stay in the swamp all night,” said Stebbins.
Both young men were hurrying towards the boat landing as they
spoke. “Maybe he’ll row around there a week before he finds his way
out,” declared King.
Stepping into the remaining boat, they both seized a paddle and
sent the light skiff whirling along towards the lake, keeping a sharp
lookout for any signs of the missing boat. “He promised not to go
further than the lake,” said Stebbins, as they reached a point where
the stream began to widen. “Let’s course over some of those creeks
back there,” indicating a part of the swamp in the rear of their island
home.
The boat’s prow was accordingly turned in that direction, and they
had proceeded but a few yards when King’s ear detected a faint call
somewhere in the distance. It was so low and indistinct that he was
unable to tell from what direction it came, but shouted loudly in
answer.
“Did you hear anything?” asked Stebbins, whose hearing was not so
keen.
“I thought I did,” answered King, “and shouted in the hope that it
might be Alfred. He’s certainly out of the channel and is calling us.
Halloo! halloo! we’re coming! Where are you?” he shouted.
The boys rested a moment or two and listened for a reply. None
came. “We don’t know which way to go,” said King. “Let’s go south
on a venture.”
“Call again,” said Stebbins, after they had been paddling for a few
minutes. King did so, and in answer came a faint shout that both
boys heard. “We’re right, keep on straight ahead,” said King,
excitedly. “Where are you?” he called, but they did not receive any
further answer.
They paddled an eighth of a mile along this course, calling
constantly without seeing anything of the person for whom they
were looking. “Strange he doesn’t answer us,” remarked Stebbins,
thoughtfully. “I’m afraid something’s happened to him.”
King said nothing, but kept peering ahead into the gathering gloom.
Darkness had fallen by this time and objects were hardly
distinguishable. Rounding a bend in the stream, they suddenly saw a
boat—the one in which Alfred had rowed away—drawn up on the
bank. With a shout the boys pushed ahead with rapid strokes.
“Alfred, where are you?” they called. As there was no response, they
backed water, and bringing their bateau to a stop, looked with
blanched faces into the empty boat.
“Where can he be?” muttered Stebbins.
“Look there! look there!” exclaimed King, rising in the skiff and
nearly upsetting it.
Stebbins followed the direction indicated, and saw what appeared to
be a man’s head upright on the ground.
“It’s Alfred, and he’s fast in the mud,” exclaimed Stebbins, grasping
the situation. “He’s dead!” he groaned.
Without further words, the boat was driven to the bank, and,
stepping on the very hummocks that had supported Whyte, they
reached his side. “Quick, Stebbins, get your paddle under his left
arm; I will do the same on my side,” said King, and, working
together, they succeeded in raising the apparently lifeless form from
its position. In another moment they had placed the unfortunate
youth in the boat beside them, and while one sent the skiff
skimming towards home, the other rubbed and chafed the cold
hands and feet. At last they were rewarded by seeing the eyes open
and feeling the heart beat faintly.
By the time the party reached the house, Whyte was himself again,
but so weak and sick that he had to be carried from the landing and
put to bed. A doctor was brought from Lincoln the next day and left
some medicine and a few directions, but Alfred’s robust health and
good constitution did more for him than all the pills and powders,
and in a few days he had recovered from all traces of his terrible
experience, except the memory of it. That will stay with him always.
 Audubon at Bird Rock

A n interesting account, showing the numbers in which birds often

live together, is the following, written by Audubon. The great
ornithologist was, at the time of writing, visiting Bird Rock, a
little granite island in the Gulf of St. Lawrence, so named from its
only inhabitants, birds, mostly of a species called Gannet.
“About ten, a speck rose on the horizon, which I was told was the
Rock. We sailed well, the breeze increased fast, and we neared the
object apace. At eleven, I could distinguish its top plainly from the
deck, and thought it covered with snow to the depth of several feet.
This appearance existed on every portion of the flat, projecting
shelves. Godwin (the guide) said, with the coolness of a man who
had visited this rock for successive seasons, that what we saw was
not snow, but Gannets. I rubbed my eyes, took my spy-glass, and in
an instant the strangest picture stood before me. They were birds
we saw—a mass of birds of such size as I never before cast my eyes
on. The whole of my party stood astounded and amazed, and all
came to the conclusion that such a sight was of itself sufficient to
invite anyone to come across the gulf to view it at this season. The
nearer we approached, the greater our surprise at the enormous
number of these birds, all calmly seated on their eggs or newly-
hatched brood, their heads all turned to the windward and toward
us. The air above for a hundred yards, and for the same distance
around the Rock, was filled with Gannets on the wing, which, from
our position, made it appear as if a heavy fall of snow was directly
above us. The whole surface (of the island) is perfectly covered with
nests, placed about two feet apart, in such regular order that you
may look through the lines as you would look through those of a
planted patch of sweet potatoes or cabbages. When one reaches the
top, the birds, alarmed, rise with a noise like thunder, and fly off in
such a hurried, fearful confusion as to throw each other down, often
falling on each other until there is a bank of them many feet high.”
This was in 1833. If Audubon could visit the island now, how he
would find the “snows” melted. There is to-day not a single Gannet
nesting on the top of the rock. On the ledges and in the crannies
about its sides, the birds still dwell in great numbers, even in
thousands, but not in the countless myriads of the past.
 A DAUGHTER OF THE FOREST
By Evelyn Raymond

CHAPTER XVIII
In the Hour of Darkness

SYNOPSIS OF PREVIOUS CHAPTERS.

Brought up in the forests of northern Maine, and seeing

few persons excepting her uncle and Angelique, the
Indian housekeeper, Margot Romeyn knows little of life
beyond the deep hemlocks. Naturally observant, she is
encouraged in her out-of-door studies by her uncle, at one
time a college professor. Through her woodland instincts,
she and her uncle are enabled to save the life of Adrian
Wadislaw, a youth who, lost and almost overcome with
hunger, has been wandering in the neighboring forest. To
Margot the new friend is a welcome addition to her small
circle of acquaintances, and after his rapid recovery she
takes great delight in showing him the many wonders of
the forest about her home. But finally, after many weeks,
the uncle decides, because of reasons which will be
known later, that it would be better for Margot if Adrian
left them. Accordingly, he puts the matter before the
young man, who, although reluctant to leave his new
friends, volunteers to go. Under the guidance of Pierre
Ricord, a young Indian, the lad sets out for the nearest
settlement. After many adventures, including a narrow
 escape from the dangerous rapids, in which the travelers
lost the canoe and nearly all their possessions, the two
reach Donovan’s, their destination. Here they separate,
Adrian going straight to New York and the home which he
left seemingly so long ago. We leave him on the threshold
of his father’s city mansion, wondering what welcome
there will be for the prodigal. Pierre returns to Peace
Island, where, with Margot and her uncle, we again take
up the story.

“N o sign yet?”
“No sign.” Margot’s tone was almost hopeless. Day after day,
many times each day, she had climbed the pine-tree flagstaff and
peered into the distance. Not once had anything been visible, save
that wide stretch of forest and the shining lake.
“Suppose you cross again, to Old Joe’s. He might be back by this
time. I’ll fix you a bite of dinner, and you better, maybe—”
The girl shook her head and clasped her arms about old Angelique’s
neck. Then the long repressed grief burst forth in dry sobs that
shook them both, and pierced the housekeeper’s faithful heart with a
pain beyond endurance.
“Pst! pouf! Hush, sweetheart, hush! ’Tis nought. A few days more,
and the master will be well. A few days more, and Pierre will come.
Ah! but I had my hands about his ears this minute. That would teach
him—yes—to turn his back on duty—him. The ingrate! Well, what
the Lord sends the body must bear, and if the broken glass—”
Margot lifted her head, shook back her hair, and smiled wanly. The
veriest ghost of her old smile it was, yet, even such, a delight to the
other’s eyes.
“Good. That’s right. Rouse up. There’s a wing of a fowl in the
cupboard, left from the master’s broth—”
“Angel, he didn’t touch it, to-day. Not even touch it.”
“’Tis naught. When the fever is on the appetite is gone. Will be all
right once that is over.”
“But, will it be over? Day after day, just the same. Always that
tossing to and fro, the queer, jumbled talk, the growing thinner—all
of the dreadful signs of how he suffers. Angelique, if I could bear it
for him. I am so young and strong and worth nothing to this world,
while he’s so wise and good. Everybody who ever knew him must be
the better for Uncle Hughie, Angelique.”
“’Tis truth. For that, the good God will spare him to us. Of that be
sure.”
“But I pray and pray and pray, and there comes no answer. He is
never any better. You know that. You can’t deny it. Always before,
when I have prayed, the answer has come swift and sure; but now
—”
“Take care, Margot. ’Tis not for us to judge the Lord’s strange ways.
Else were not you and me and the master shut up alone on this
island, with no doctor near, and only our two selves to keep the
dumb things in comfort. Though, as for dumbness, hark yonder
beast!”
“Reynard! Oh! I forgot. I shut him up because he would hang
around the house and watch your poor chickens. If he’d stay in his
own forest, now, I would be so glad. Yet I love him—”
“Aye, and he loves you. Be thankful. Even a beastie’s love is of God’s
sending. Go feed him. Here—the wing you’ll not eat yourself.”
They were dark days now on the once sunny Island of Peace.
That day when Mr. Dutton had said, “Your father is still alive,”
seemed now to Margot, looking back, as one of such experiences as
change a whole life. Up till that morning she had been a thoughtless,
unreflecting child, but the utterance of those fateful words altered
everything.
Amazement, unbelief of what her ears told her, indignation that she
had been so long deceived, as she put it, were swiftly followed by a
dreadful fear. Even while he spoke, the woodlander’s figure swayed
and trembled, the hoe-handle on which he rested wavered and fell,
and he, too, would have fallen had not the girl’s arms caught and
eased his sudden sinking in the furrow he had worked. Her shrill cry
of alarm had reached Angelique, always alert for trouble and then
more than ever, and had brought her swiftly to the field. Between
them they had carried the now unconscious man within and laid him
on his bed. He had never risen from it since; nor, in her heart, did
Angelique believe he ever would, though she so stoutly asserted to
the contrary before Margot.
“We have changed places, Angelique, dear,” the child often said. “It
used to be you who was always croaking and looking for trouble.
Now you see only brightness.”
“Well, good sooth. ’Tis a long lane has no turnin’, and better late nor
never. Sometimes ’tis well to say, ‘Stay, good trouble, lest worser
comes,’ eh? But things’ll mend. They must. Now, run and climb the
tree. It might be this ver’ minute that wretch, Pierre, was on his way
across the lake. Pouf! but he’ll stir his lazy bones, once he touches
this shore! Yes, yes, indeed. Run and hail him, maybe.”
So Margot had gone, again and again, and had returned to sit beside
her uncle’s bed, anxious and watchful.
Often, also, she had paddled across the narrows and made her way
swiftly to a little clearing on her uncle’s land, where, among giant
trees, old Joseph Wills, the Indian guide, and faithful friend of all on
Peace Island, made one of his homes. Once Mr. Dutton had nursed
this red man through a dangerous illness, and had kept him in his
old home for many weeks thereafter. He would have been the very
nurse they now needed, in their turn, could he have been found. But
his cabin was closed, and on its doorway, under the family sign-
picture of a turtle on a rock, he had printed, in dialect, what signified
his departure for a long hunting trip.
Now, as Angelique advised, she resolved to try once more; and,
hurrying to the shore, pushed her canoe into the water and paddled
swiftly away. She had taken the neglected Reynard with her, and
Tom had invited himself to be a party of the trip; and in the odd but
sympathetic companionship Margot’s spirits rose again.
“It must be as Angelique says. The long lane will turn. Why have I
been so easily discouraged? I never saw my precious uncle ill before,
and that is why I have been so frightened. I suppose anybody gets
thin and says things when there is fever. But he’s troubled about
something. He wants to do something that neither of us understand.
Unless—oh! I believe I do understand. My head is clearer out here
on the water, and I know, I know! It is just about the time of year
when he goes away on those long trips of his. And we’ve been so
anxious we never remembered. That’s it. Surely it is. Then, of
course, Joe will be back now or soon. He always stays on the island
when uncle goes, and he’ll remember. Oh! I’m brighter already, and
I guess, I believe, it is as Angelique claims—God won’t take away so
good a man as uncle and leave me alone. Though I am not alone. I
have a father! I have a father somewhere, if I only knew—all in good
time—and I’m growing gladder and gladder every minute.”
She could even sing to the stroke of her paddle, and she skimmed
the water with increasing speed. Whatever the reason for her
growing cheerfulness, whether the reaction of youth or a prescience
of happiness to come, the result was the same; she reached the
further shore flushed and eager-eyed, more like the old Margot than
she had been for many days.
“Oh! he’s there. He is at home. There is smoke coming out of the
chimney. Joseph! Oh, Joseph! Joseph!”
She did not even stop to take care of her canoe, but left it to drift
whither it would. Nothing mattered, Joseph was at home. He had
canoes galore, and he was help indeed.
She was quite right. The old man came to his doorway and waited
her arrival with apparent indifference, though surely no human heart
could have been unmoved by such unfeigned delight. Catching his
unresponsive hands in hers, she cried:
“Come at once, Joseph! At once.”
“Does not the master trust his friend? It is the time to come.
Therefore, I am here.”
“Of course. I just thought about that. But, Joseph, the master is ill.
He knows nothing any more. If he ever needed you, he needs you
doubly now. Come, come at once.”
Then, indeed, though there was little outward expression of it, was
old Joseph moved. He stopped for nothing, but leaving his fire
burning on the hearth and his supper cooking before it, went out
and closed the door. Even Margot’s nimble feet had ado to keep pace
with his long strides, and she had to spring before him to prevent his
pushing off without her.
“No, no. I’m going with you. Here—I’ll tow my own boat, with Tom
and Reynard—don’t you squabble, pets—but I’ll paddle no more
while you’re here to do it for me.”
Joseph did not answer, but he allowed her to seat herself where she
pleased, and with one strong movement sent his big birch a long
distance over the water.
Margot had never made the passage so swiftly, but the motion
suited her exactly; and she leaped ashore almost before it was
reached, to speed up the hill and call out to Angelique wherever she
might be:
“All is well! All will now be well—Joseph has come.”
The Indian reached the house but just behind her and acknowledged
Angelique’s greeting with a sort of grunt; yet he paused not at all to
ask the way or if he might enter the master’s room, passing directly
into it as if by right.
Margot followed him, cautioning, with finger on lip, anxious lest her
patient should be shocked and harmed by the too-sudden
appearance of the visitor.
Then, and only then, when her beloved child was safely out of sight,
did Angelique throw her apron over her head and give her own
despairing tears free vent. She was spent and very weary; but help
had come; and in the revulsion of that relief nature gave way. Her
tears ceased, her breath came heavily, and the poor woman slept,
the first refreshing slumber of an unmeasured time.
When she waked, at length, Joseph was crossing the room. The fire
had died out, twilight was falling, she was conscious of duties left
undone. Yet there was light enough left for her to scan the Indian’s
impassive face with keen intensity; and though he turned neither to
the right nor left, but went out with no word or gesture to satisfy her
craving, she felt that she had had her answer.
“Unless a miracle is wrought, my master is doomed. Oh, the broken
glass—the broken glass!”

CHAPTER XIX
THE LETTER

F rom the moment of his entrance to the sick room, old Joe
assumed all charge of it, and with scant courtesy banished from
it both Angelique and Margot.
“But he is mine, my own precious uncle. Joe has no right to keep me
out!” protested Margot, vehemently.
Angelique was wiser. “In his own way, among his own folks, that
Indian good doctor. Leave him be. Yes. If my master can be save’,
Joe Wills’ll save him. That’s as God plans; but if I hadn’t broke—”
“Angelique! Don’t you ever, ever let me hear that dreadful talk again.
I can’t bear it. I don’t believe it. I won’t hear it. I will not. Do you
suppose that our dear Lord is—will—”
She could not finish her sentence and Angelique was frightened by
the intensity of the girl’s excitement. Was she, too, growing feverish
and ill? But Margot’s outburst had worked off some of her own
uncomprehended terror, and she grew calm again. Though it had not
been put into so many words, she knew both from Angelique’s and
Joseph’s manner that they anticipated but one end to her guardian’s
illness. She had never seen death, except among the birds and
beasts of the forest, and even then it had been horrible to her; and
that this should come into her own happy home was unbearable.
Then she reflected. Hugh Dutton’s example had been her instruction,
and she had never seen him idle. At times when he seemed most so,
sitting among his books, or gazing silently into the fire, his brain had
been active over some problem that perplexed or interested him.
“Never hasting, never wasting” time, nor thought, nor any energy of
life. That was his rule, and she would make it hers.
“I can, at least, make things more comfortable out-of-doors.
Angelique has let even Snowfoot suffer, sometimes, for want of the
grooming and care she’s always had. The poultry, too, and the poor
garden. I’m glad I’m strong enough to rake and hoe, even if I
couldn’t lift Uncle as Joe does.”
Her industry brought its own reward. Things outside the house took
on a more natural aspect. The weeds were cleared away, and both
vegetables and flowers lifted their heads more cheerfully. Snowfoot
showed the benefit of the attention she received, and the forgotten
family in the Hollow chattered and gamboled in delight at the
reappearance among them of their indulgent mistress. Margot
herself grew lighter of heart and more positive that, after all, things
would end well.
“You see, Angelique dismal, we might as well take that broken glass
sign to mean good things as evil; that uncle will soon be up and
around again, Pierre be at home; and the ‘specimen’ from the old
cave prove copper or something just as rich, and—everybody be as
happy as a king.”
Angelique grunted her disbelief, but was thankful for the other’s
lighter mood.
“Well, then, if you’ve so much time and strength to spare, go yonder
and redde up the room that Adrian left so untidy. Where he never
should have been, had I my own way, but one never has that in this
world; hey, no. Indeed, no. Ever’thin’ goes contrary, else I’d have
cleared away all trace long sin’. Yes, indeed, yes.”
“Well, he is gone. There’s no need to abuse him, even if he did not
have the decency to say good-by. Though, I suppose it was my
uncle put a stop to that. What Uncle has to do he does at once.
There’s never any hesitation about Uncle. But I wish—I wish—
Angelique Ricord, do you know something? Do you know all the
history of this family?”
“Why should I not, eh?” demanded the woman, indignantly. “Is it
not my own family, yes? What is Pierre but one son? I love him, oh,
yes! But—”
 “WHERE IS MY FATHER?”

“You adore him, bad and trying as he is. But there is something you
must tell me, if you know it. Maybe you do not. I did not, till that
awful morning when he was taken ill. But that very minute he told
me what I had never dreamed. I was angry; for a moment I almost
hated him because he had deceived me, though afterward I knew
that he had done it for the best and would tell me why when he
could. So I’ve tried to trust him just the same and be patient. But—
he may never be able—and I must know. Angelique, where is my
father?”
The housekeeper was so startled that she dropped the plate she was
wiping and broke it. Yet even at that fresh omen of disaster she
could not remove her gaze from the girl’s face nor banish the dismay
of her own.
“He told—you—that—that—”
“That my father is still alive. He would, I think, have told me more;
all that there may be yet to tell, if he had not so suddenly been
stricken. Where is my father?”
“Oh, child, child! Don’t ask me. It is not for me—”
“If Uncle cannot and you can, and there is no other person,
Angelique—you must!”
“This much, then. It is in a far, far away city, or town, or place, he
lives. I know not, I. This much I know: he is good, a ver’ good man.
And he have enemies. Yes. They have done him much harm. Some
day, in many years, maybe, when you have grown a woman, old like
me, he will come to Peace Island and forget. That is why we wait.
That is why the master goes, once each summer, on the long, long
trip. When Joseph comes, and the bad Pierre to stay. I, too, wait to
see him, though I never have. And when he comes, we must be ver’
tender, me and you, for people who have been done wrong to, they
—they—pouf! ’Twas anger I was that the master could put the evil-
come into that room, yes.”
“Angelique! Is that my father’s room? Is it? Is that why there are the
very best things in it? And that wonderful picture? And the fresh
suits and clothing? Is it?”
Angelique slowly nodded. She had been amazed to find that Margot
knew thus much of a long-withheld history, and saw no harm in
adding these few facts. The real secret, the heart of the matter—
that was not yet. Meanwhile, let the child accustom herself to the
new ideas, and so be prepared for what she must certainly and
further learn, should the master’s illness be a fatal one.
“Oh, then, hear me. That room shall always now be mine to care for.
I haven’t liked the housewifery, not at all. But if I have a father and I
can do things for him—that alters everything. Oh! you can’t mean
that it will be so long before he comes. You must have been jesting.
If he knew Uncle was ill he would come at once, wouldn’t he? He
would, I know.”
Poor Angelique turned her face away to hide its curious expression,
but in her new interest concerning the “friend’s room,” as it had
always been called, Margot did not notice this. She was all
eagerness and loving excitement.
“To think that I have a father who may come, at any minute, for he
might, Angelique, you know that, and not be ready for him. Your
best and newest broom, please, and the softest dusters. That room
shall, indeed, be ‘redded’—though uncle says nobody but a few
people like you ever use that word, nowadays—better than anybody
else could do it. Just hurry, please, I must begin. I must begin right
away.”
She trembled so that she could hardly braid and pin up her long hair
out of the way, and her face had regained more than its old-time
color. She was content to let all that was still a mystery remain for
the present. She had enough to think about and enjoy.
Angelique brought the things that would be needed and, for once,
forebore advice. Let love teach the child—she had nought to say. In
any case, she could not have seen the dust, herself, for her dark
eyes were misty with tears, and her thoughts on matters wholly
foreign to household cares.
Margot opened the windows and began to dust the various articles
which could be set out in the wide passage, and did not come round
to the heavy dresser for some moments. As she did so finally, her
glance flew instantly to a bulky parcel, wrapped in sheets of white
birch bark, and bearing her own name, in Adrian’s handwriting.
“Why, he did remember me, then!” she cried, delightedly, tearing the
package open. “Pictures! the very ones I liked the best. Xanthippé
and Socrates, and oh! that’s Reynard. Reynard, ready to speak. The
splendid, beautiful creature; and the splendid, generous boy, to have
given it. He called it his ‘masterpiece,’ and, indeed, it was by far the
best he ever did here. Harmony Hollow—but that’s not so fine.
However, he meant to make it like, and—why, here’s a note! Why
didn’t I come in here before? Why didn’t I think he would do
something like this? Forgive me, Adrian, wherever you are, for
misjudging you so. I’m sorry Uncle didn’t like you, and sorry—for lots
of things. But I’m glad—glad you weren’t so rude and mean as I
believed. If I ever see you, I’ll tell you so. Now, I’ll put these in my
own room and then get to work again. This room you left so messed
shall be as spotless as a snowflake before I’ve done with it.”
For hours she labored there—brushing, renovating, polishing; and
when all was finished she called Angelique to see and criticise—if
she could. But she could not; and she, too, had something now of
vital importance to impart.
“It is beautiful’ done, yes, yes. I couldn’t do it more clean myself, I,
Angelique, no. But, ma p’tite I hear, hear, and be calm! The master
is himself! The master has awoke, yes, and is askin’ for his child.
True, true. Old Joe, he says, ‘Come! quick, soft, no cry, no laugh,
just listen.’ Yes. Oh, now all will be well!”
Margot almost hushed her very breathing. Her uncle awake, sane,
asking for her. Her face was radiant, flushed, eager, a face to
brighten the gloom of any sick room, however dark.
But this one was not dark. Joe knew his patient’s fancies. He had
forgotten none. One of them was the sunshine and fresh air; and
though in his heart he believed that these two things did a world of
harm, and that the ill-ventilated and ill-lighted cabins of his own
people were more conducive to recovery, he opposed nothing which
the master desired. He had experimented, at first, but finding a
close room aggravated Mr. Dutton’s fever, reasoned that it was too
late to break up the foolish habits of a man’s lifetime; and as the
woodlander had lived in the sunlight, so he would better die in it,
and easier.
If she had been a trained nurse, Margot could not have entered her
uncle’s presence more quietly, though it seemed to her that he must
hear the happy beating of her heart and how her breath came fast
and short. He was almost too weak to speak at all, but there was all
the old love, and more, in his whispered greeting.
“My precious child!”
“Yes, Uncle. And such a happy child because you are better.”
She caught his hand and covered it with kisses, but softly, oh! so
softly, and he smiled the rare, sweet smile that she had feared she’d
never see again. Then he looked past her to Angelique, in the
doorway, and his eyes roved toward his desk in the corner. A little
fanciful desk that held only his most sacred belongings and had been
Margot’s mother’s. It was to be hers, some day, but not till he had
done with it, and she had never cared to own it, since doing so
meant that he could no longer use it. Now she watched him and
Angelique wonderingly.
For the woman knew exactly what was required. Without question or
hesitation, she answered the command of his eyes by crossing to
the desk and opening it with a key she took from her own pocket.
Then she lifted a letter from an inner drawer and gave it into his thin
fingers.
“Well done, good Angelique. Margot—the letter—is yours.”
“Mine? I am to read it? Now? Here?”
“No, no. No, no, indeed! Would you tire the master with the rustlin’
of paper? Take it, else. Not here, where ever’thin’ must be still as
still.”
Mr. Dutton’s eyes closed. Angelique knew that she had spoken for
him, and that the disclosure which that letter would make should be
faced in solitude.
“Is she right, Uncle, dearest? Shall I take it away to read?”
His eyes assented, and the tender, reassuring pressure of his hand.
“Then I’m going to your own mountain top with it. To think of having
a letter from you, right here, at home! Why, I can hardly wait! I’m so
thankful to you for it, and so thankful to God that you are getting
well. That you will be soon; and then—why, then—we’ll go a-
fishing!”
A spasm of pain crossed the sick man’s wasted features, and poor
Angelique fled the place, forgetful of her own caution to “be still as
still,” and with her own dark face convulsed with grief for the grief
which the letter would bring to her idolized Margot.
But the girl had already gone away up the slope, faster and faster.
Surely, a letter from nobody but her uncle, and at such a solemn
time, must concern but one subject—her father. Now she would
know all, and her happiness should have no limit.
But it was nightfall when she, at last, came down from the
mountain, and though there were no signs of tears upon her face,
neither was there any happiness in it.
[TO BE CONTINUED]

The heights by great men reached and kept

Were not attained by sudden flight;
But they, while their companions slept,
Were toiling upward in the night.

—Longfellow.

STATE FLOWERS

T he following are the “State flowers,” as adopted by the several

States. In Maine, Michigan, and Oklahoma Territory the decision
was made by the Legislature, in the other cases by the votes of
the scholars in the public schools.
Alabama, goldenrod; Arkansas, aster; California, California poppy;
Colorado, columbine; Delaware, peach blossom; Idaho, syringa;
Iowa, wild rose; Maine, pine cone and tassel; Michigan, apple
blossom; Minnesota, moccasin flower; Missouri, goldenrod; Montana,
bitterroot; Nebraska, goldenrod; New Jersey (State tree, maple);
New York, rose (State tree, maple); North Dakota, goldenrod;
Oklahoma Territory, mistletoe; Oregon, Oregon grape; Rhode Island,
violet; Vermont, red clover; Washington, rhododendron. In Kansas,
the sunflower is usually known as the State flower.

The largest bell in the world is the great bell at Moscow, at the foot
of the Kremlin. Its circumference is nearly 68 feet, and its height
more than 21 feet. It is 23 inches thick in its stoutest part, and
weighs 433,722 pounds. It has never been hung.
 THE FLOWERLESS PLANTS

By JULIA McNAIR WRIGHT

T he year around and the world around, journey the plant

pilgrims. Among those perennials which are found in all latitudes
and seasons are the lichens and fungi. In September, while we
wait for fruits and seeds to finish ripening, let us make small studies
in these related groups in the vegetable sub-kingdom called the
thallogens.
This sub-kingdom, one of the chief divisions of the vegetable
kingdom, is known as the class thallophytes. It contains the simplest
forms of vegetable life. Its chief groups are the fungi and algæ, the
lichens being related to both, as if algæ and fungi had united in one
plant, dividing and somewhat changing the characteristic of each.
At any period of the year you can find lichens in abundance. They
cover ragged rocks, dress up old roofs, walls, fence rails and dead
stumps, especially delighting in the north side of trees. If we
examine them through a magnifying glass, we shall see that they
are made up of cells, laid side by side like little chains of beads, or of
cells expanded into short tubes or threads lying like heaps of tiny
fagots. Instead of seeds, lichens have a fine dust, called spores,
from which they develop.
Lichens are exceedingly long-lived and excessively slow of growth.
The lily attains its lovely maturity in a few months; the oaks, elms,
pines, become great trees in twenty or thirty years; the humble
lichen often lives forty or fifty years before it is old enough to
complete its growth by producing spores. Botanists say that the life
of a lichen is fitful and strange, and is practically indefinite as to
duration. Lichens simply live on and on.
Some lichens have been known to live nearly fifty years without
seeming to grow; they appear to dry up, and nearly vanish; then,
suddenly, from some cause there is a revival of growth—they expand
again. Small and insignificant as these lichens are, they often outlive
those longest-lived of trees, the cedar of Lebanon and the California
redwood.
The condition of lichen existence is water, for from moisture alone, in
dew or rain, they secure their food. The carbon, oxygen, ammonia,
hydrogen, in air and rain, afford them their nourishment. The lichen
generally refuses to grow in foul air laden with noxious gases. In the
impure air of cities few appear, but they abound in the open country.
They absorb by all the surface, except the base by which they are
fastened to their place of dwelling. They have no roots, and simply
adhere to bare rocks, sapless wood, even to naked glass, from which
they can receive no nutriment whatever.
In comparison with what is known of plants in general, our
knowledge of lichens is yet very limited. They seem to be made
chiefly of a kind of gelatin which exists in lichens only. Humble as
they appear, they have always been of large importance in arts and
manufactures. They produce exquisite dyes—a rich, costly purple, a
valuable scarlet, many shades of brown, and particularly splendid
hues of blue and yellow are obtained from these common little
growths, which in themselves display chiefly shades of black, gray
green, varied with pink, red, and orange cups, balls, and edges.

Fungi
While not so abundant as lichens, the fungi are well known
everywhere. We cannot claim, as for the lichens, that they are
harmless, for many are a virulent poison: others have a disgusting
odor, and nearly all are dangerous in their decay. On the other hand,
many of them are a useful, delicious food, and nearly all are
beautiful when first developed. Their variety, also, is very fascinating.

THE FLOWERLESS PLANTS

In a walk of less than two miles in a wet summer, may be found
twenty different kinds of fungi—some no larger than a pea, some
eight inches in diameter. They may be round, oval, flat, cup-shaped,
horn-shaped, cushion-shaped, saucer-shaped; they are snow-white,
gray, tan, yellow, lavender, orange, dark brown, pink, crimson,
purple, and variously mottled, scaly or smooth as with varnish.
Placed on a large platter among dark green mosses, they will be, for
one day, a magnificent collection.
One large, egg-shaped variety, growing in pairs, is of a purple shade,
very solid, and when broken open seems filled with glittering matter
like iron or steel filings. Another tan-colored, plum-shaped fungus,
firm and smooth, is of a nearly royal purple within.
September is a good month for the study of fungi, especially after
the early fall rains, when the woods and pastures will be found well-
filled, not only with brilliant, useless, or poisonous varieties, but with
delicious edible kinds. Popularly, people call the edible specimens
“mushrooms,” and the rest “toadstools,” the number of poisonous or
of edible instances so named depending rather upon the amount of
knowledge of the collector than upon the real qualities of the fungi,
for many denominate as “toadstools” what others know to be an
excellent food.
Many varieties not usually eaten are wholesome, and many which
human beings reject, other animals thrive upon. One large, brown
“toadstool” of the woods is, at this season of the year, the chief food
of that epicure, the wood-tortoise.
In general a fungus may be defined as a thallophyte without any
chlorophyl or leaf-green in its composition. Among the brilliant colors
displayed by fungi no green or blue can be found.
The most popular and most useful fungus is the table mushroom.
This rarely ever grows in the woods, in shade, on wet lands, or on
decaying stumps. It prefers the open, breezy, well-sunned pastures,
where the grass is kept short by the grazing of sheep or cattle. Early
in the morning or shortly before sunset, the dainty white or cream-
colored buttons, borne on snow-white stalks, push up through the
soil and gradually expand until the discs are flat or slightly convex.
From two to six inches is the diameter, seldom more than three.
Varieties of the pasture mushroom are few and can readily be
learned. The mushroom is composed of stem and cap; the stem is
finger-shaped, with the roundish end in the earth. About half way up
is usually a ring of the covering skin, where, in the button shape, the
veil of the mushroom was attached.
This veil extends over the cap and is left at the edge of a little frill; it
can be easily stripped off. Under the veil the flesh is ivory-white, and
is smooth and firm.
The under side of the cap is laid in plaits, called gills, from their
resemblance to fish gills. They never grow fast to, or down upon,
the stem, usually stopping short off, about one-tenth of an inch from
its juncture with the cap. Mushrooms are cultivated in gardens or
cellars. They grow from spores or little finger-like lengths, called
spawn, which are produced by the spores. Mushrooms turn black or
purplish after the first twenty hours of growth. When the gills have
taken this dark hue, the mushroom is unfit to eat.
Some fungi grow in very wet places; the woods are likely to be full
of them after a few rainy days. They are all short-lived.

INDEPENDENCE

A lthough not new to some of our readers, we think the following

anecdote, illustrating one phase of Benjamin Franklin’s
character, will bear repeating:
Not long after he began editing his newspaper, Franklin’s free
manner of criticism called forth the disapprobation of many of his
patrons. One of them in particular felt so greatly moved as to make
it his duty to tell him so. “The doctor listened with patience to the