ArticlePDF Available

Investigating the Instructional Supportiveness of Leveled Texts

Authors:

Abstract

LEVELED BOOKS originally selected by or produced for use in Reading Recovery or its regular classroom initiative are now also widely used in regular and special classrooms having no affiliation with Reading Recovery. The frequent use of these leveled books in settings other than Reading Recovery raises an important question: Do books leveled for use in Reading Recovery support other reading instructional emphases in addition to the ones that Reading Recovery teachers are trained to provide? The purpose of this study was to examine the curricular dimensions of books leveled for use in Reading Recovery in order to judge how supportive such texts are for early reading instruction emphasizing word recognition or decoding instead of, or in addition to, the three main cueing systems. The study found that Reading Recovery books, as a category of early reading instructional texts, provide only a moderate amount of support for word recognition instruction and almost none for decoding instruction in the use of onsets and rimes. The study also found that books leveled for use in Reading Recovery do not consistently increase in word-level demands as their levels increase.
Reading Research Quarterly
Vol. 40, No. 4
October/November/December 2005
© 2005 International Reading Association
(pp. 410–427)
doi:10.1598/RRQ.40.4.2
F
Investigating the instructional
supportiveness of leveled texts
JAMES W. CUNNINGHAM
Emeritus, University of North Carolina at Chapel Hill, USA
STEPHANIE A. SPADORCIA
Lesley University, Cambridge, Massachusetts, USA
KAREN A. ERICKSON
University of North Carolina at Chapel Hill, USA
DAVID A. KOPPENHAVER
Appalachian State University, Boone, North Carolina, USA
JANET M. STURM
Central Michigan University, Mount Pleasant, USA
DAVID E. YODER
Emeritus, University of North Carolina at Chapel Hill, USA
or almost a century and a half, graded or leveled texts were generally considered an
important, if not essential, component of elementary reading instruction (Hoffman,
Roser, Salas, Patterson, & Pennington, 2001). Samuel Wood wrote the first graded
readers early in the 1800s (Barr, 1989), and, by the turn of the 20th century, basal
reading programs were often titled or described as “progressive..., not as a description
of a ‘new approach’ to teaching reading but as a description of the leveled nature of
the books in the program” (Hoffman, Sailors, & Patterson, 2002, p. 270).
While some may have questioned the need for texts of gradually increasing
difficulty, controversies centered on how, not whether, texts were to be sequenced or
410
411
LEVELED BOOKS originally selected by or produced for use in Reading Recovery or its regular classroom initia-
tive are now also widely used in regular and special classrooms having no affiliation with Reading Recovery. The fre-
quent use of these leveled books in settings other than Reading Recovery raises an important question: Do books
leveled for use in Reading Recovery support other reading instructional emphases in addition to the ones that
Reading Recovery teachers are trained to provide? The purpose of this study was to examine the curricular dimen-
sions of books leveled for use in Reading Recovery in order to judge how supportive such texts are for early read-
ing instruction emphasizing word recognition or decoding instead of, or in addition to, the three main cueing sys-
tems. The study found that Reading Recovery books, as a category of early reading instructional texts, provide
only a moderate amount of support for word recognition instruction and almost none for decoding instruction in
the use of onsets and rimes. The study also found that books leveled for use in Reading Recovery do not consistently
increase in word-level demands as their levels increase.
Investigating
the instructional
supportiveness
of leveled texts
LOS LIBROS nivelados, originalmente seleccionados o producidos para su uso en el programa Reading Recovery
(Recuperación en Lectura) o en aulas regulares que adoptaban el programa, se usan actualmente en aulas regulares
y especiales sin relación alguna con dicho programa. El uso frecuente de estos libros en contextos diferentes del de
Reading Recovery introduce una pregunta importante: ¿los libros nivelados para Reading Recovery son adecuados
en enfoques didácticos diferentes de aquellos para los que fueron capacitados los docentes de Reading Recovery?
El propósito de este estudio es examinar las dimensiones curriculares de los libros nivelados de Reading Recovery
con el fin de evaluar el apoyo que proporcionan esos textos en una didáctica de la lectura inicial que enfatiza el re-
conocimiento de palabras o la decodificación en lugar de, o además de, los tres principales sistemas de pistas. El es-
tudio halló que, como libros de enseñanza de la lectura inicial, los textos de Reading Recovery proporcionan sólo un
apoyo moderado a la enseñanza del reconocimiento de palabras y casi ninguno para enseñar decodificación con el
uso de ataques y rimas. Asimismo el estudio encontró que en los libros de Reading Recovery no se halla un au-
mento de las demandas en el nivel de las palabras consistente con el aumento de los niveles.
Investigando el
apoyo didáctico
brindado por
textos nivelados
EINGESTUFTE LESEBÜCHER, ursprünglich ausgewählt für oder hergestellt zur Verwendung bei der
Leseverbesserung durch Reading Recovery
®
oder in ihrer regulären Klassenzimmerverwendung, werden jetzt auch
weitgehend in regulären und Sonderschulklassen benutzt, die keine Bindung zu Reading Recovery haben. Der häu-
fige Gebrauch dieser eingestuften Lesebücher in andere Bereichen als Reading Recovery wirft eine wichtige Frage
auf: Unterstützen Bücher, die zur Verwendung bei Reading Recovery in der Leseverbesserung benutzt werden, an-
dere leseanleitende Schwerpunkte in Ergänzung zu jenen, die entsprechend ausgebildete Lehrer bereits vermitteln?
Der Zweck dieser Studie ist es, die lehrplanerischen Dimensionen von solchen Büchern zu untersuchen, die zur
Verwendung in Reading Recovery eingestuft sind, um zu beurteilen, wie weit solche Texte den frühzeitigen
Leseunterricht unterstützen, unter Betonung der Worterkennung oder Entzifferung, anstatt von oder in Ergänzung
zu den wesentlichen drei aufgezeigten Systemen. Die Studie ergab, daß, als eine Kategorie frühzeitiger
Leseanweisungstexte, Reading Recovery Lesebücher nur einen gemäßigten Unterstützungwert für Anweisungen zur
Worterkennung liefern und nahezu keinen für Entzifferungsanweisungen in der Verwendung von Einleitungen und
Reimen. Die Studie ergab ebenfalls, dass die für den Reading Recovery Gebrauch eingestuften Bücher nicht in
dem Maße übereinstimmend Wortschatzanforderungen steigern, in dem sich ihre Schulstufen anheben.
Ergründen
anleitender
Unterstützung
durch eingestufte
Texte
ABSTRACTS
412
DES LIVRES standardisés par niveau, choisis ou produits à l’origine pour être utilisés dans le programme Reading
Recovery (programme de rééducation de la lecture) ou à son initiative dans des classes ordinaires, sont maintenant
utilisés également à grande échelle dans des classes ordinaires ou d’enseignement spécialisé non affiliées à Reading
Recovery. L’utilisation fréquente de ces livres standardisés par niveau dans d’autres contextes que Reading Recovery
soulève une question importante : des livres standardisés pour être utilisés en Reading Recovery peuvent-ils avoir une
valeur instructive centre que celle que les enseignants de Reading Recovery sont formés à apporter ? Cette étude a
pour but d’examiner l’intérêt par rapport aux programmes des livres standardisés par niveau pour être utilisés par
Reading Recovery, afin d’évaluer l’intérêt de ces textes dans le cadre d’un enseignement des débuts de la lecture
qui met l’accent sur la reconnaissance des mots ou le décodage au lieu de ou en plus des trois principaux systèmes
d’indices. La recherche a montré que, en tant que textes pour l’enseignement des débuts de la lecture, les livres de
Reading Recovery fournissent un apport modeste à l’enseignement de la reconnaissance des mots et presque aucun
à l’enseignement du décodage en ce qui concerne l’utilisation des attaques et des rimes. La recherche a trouvé
également que les livres standardisés par niveau pour une utilisation dans Reading Recovery n’augmentent pas les
exigences au niveau mot au fur et à mesure que le niveau des textes s’élève.
Examen de la
valeur instructive
de textes
standardisés
par niveau
ABSTRACTS
leveled. In the mid-1980s, however, a sea change oc-
curred in mainstream basal readers marked by the
lack of “any systematic attention to the decoding de-
mands of the texts” (Hoffman et al., 2002, p. 272).
Soon, reading educators from different perspectives
were expressing concern over the near abandonment
of appropriate text difficulty as a principle in elemen-
tary reading instruction. Clay (1991) observed,
There is an exciting enthusiasm among teachers in some
countries today for teaching from story books but this is of-
ten associated with a strong disregard for any gradient of
difficulty in the texts used. Any levelling of books is seen to
be unnecessary and an impediment to learning.... However,
many children learning to read will be confused without as-
sistance from some form of a gradient of difficulty in reading
books. (p. 201)
In a similar manner, Chall, Conard, and Harris-
Sharples (1991) summarized a number of recent arti-
cles and books:
The common theme of these professional articles was that
the difficulty level of [children’s] textbooks was an unim-
portant issue. (p. 27)
It has...become popular to say that [children’s] textbooks
are suitable if they are interesting and contain high-quality
writing, and that it matters little how difficult they are. (p.
107)
During the time that the literature-based and
whole language movements were bringing the princi-
ple of gradually increasing text difficulty into ques-
tion, the Reading Recovery program was
simultaneously gaining a foothold in the United
States. The principle of gradually increasing text dif-
ficulty in small increments has been central to the
Reading Recovery early intervention program from
the beginning (Clay, 1985, 1991; Peterson, 1991;
Pinnell, 1990). Its application of the principle can be
seen as consistent with the long-standing practice of
determining a child’s reading instructional level
(Betts, 1946), although Reading Recovery uses dif-
ferent procedures and criteria when leveling books
and matching them to readers (Peterson).
Today the use of leveled texts in teaching pri-
mary reading appears to have returned to historical
levels, largely as a result of the popularity of the lev-
eled books originally used by or developed for the
Reading Recovery program (Fry, 2002; Hoffman,
2002). Yet the future role of leveled texts in begin-
ning reading instruction remains in doubt because of
longstanding controversies concerning what factors
should be considered when sequencing or leveling
texts for early reading instruction.
Curricular dimensions of
increasing text difficulty
Although the principle of gradually increasing
text difficulty has historically, and again recently,
been widely considered crucial for early reading in-
struction, this consensus does not extend to the
characteristics the sequence of texts should have.
Rather, most advocates of different approaches to
teaching early reading have argued that texts should
gradually increase in difficulty along the
dimension(s) of emphasis in the instructional ap-
proach they promote. Three major dimensions of in-
creasing challenge have been emphasized in texts
selected or written to complement various approach-
es to early reading instruction.
Text with vocabulary control by word
frequency
In the 1920s and 1930s, the “vocabulary con-
trol movement” began in foreign language learning.
While the first words to be learned in a second lan-
guage can be selected on any basis, the vocabulary
control movement was soon dominated by those ar-
guing that more successful language learning is likely
to occur if the vocabulary of instruction generally
consists of the highest frequency words (Schmitt,
2000). In reading education,
One prominent characteristic of the readers of [the
1925–1935] period was the use of standard word lists as a
basis for selecting the vocabulary. Writers of textbooks took
meticulous care to have the vocabulary in their primers and
first readers consist almost wholly of words having the high-
est frequency. (Smith, 1965/2002, p. 203)
By the 1950s, most basal reading series con-
trolled the frequency, number, and repetition of
words in their selections, particularly at the lower
levels (Hoffman et al., 2001; Smith, 1965/2002).
This vocabulary control by word frequency was the
main tool for sequencing the texts and assigning lev-
els to the books in these widely used series (Hoffman
et al., 2002). Vocabulary control based primarily on
word frequency continued to hold sway in the de-
sign of primary materials in mainline basal series
through the mid-1980s (Hiebert, 1999).
In texts with traditional vocabulary control, as
text difficulty increases, the percentage and repetition
of high-frequency words decreases or the list of words
considered familiar due to frequency range increases
in size. Either way, vocabulary control is loosened as
texts become more difficult. The purpose of vocabulary
Instructional supportiveness of leveled texts
413
control by word frequency is to foster the students
word recognition—the ability to access pronuncia-
tions attached in memory to entire printed words.
The instructional emphasis the texts are designed or
selected to support is word study instruction—guid-
ing students’ work with specific high-frequency words
to help them learn the particular pattern of letters that
comprise each one (Cunningham, Koppenhaver,
Erickson, & Spadorcia, 2004).
Text with phonetic control
Decodable text
Phonetic control increases the percentage of
words in a text selected for their word decodability
(Juel & Minden-Cupp, 2000). Word decodability al-
ways includes attention to phonetic regularity—the
predictability of the words’ letter–sound patterns—
and the number of syllables (Juel &
Roper/Schneider, 1985; Martin & Hiebert, 1999;
Menon & Hiebert, 1999). In addition, word decod-
ability may also include consideration of a perspec-
tive attributable to Beck (1981); “Within this
[instructional consistency] perspective the decodabil-
ity of a word is determined by the instruction that
has preceded the appearance of the word in a selec-
tion” (Hoffman et al., 2002, p. 276).
As the relative difficulty of texts with phonetic
control increases, the percentage of decodable words
decreases or the complexity of words considered de-
codable increases. Either way, phonetic control is
loosened as texts become more difficult (Juel &
Roper/Schneider, 1985). The purpose of phonetic
control is to foster students’ decoding—their ability
to use knowledge of letter–sound relationships and
patterns to construct probable pronunciations for
unfamiliar printed words. The instructional empha-
sis the texts are designed or selected to support is
phonics instruction—teaching students general
knowledge about how printed words encode sounds
(Cunningham et al., 2004).
Predictable text
We use the term predictable text to include all
written material selected or constructed to provide
extra cues to readers so they can read it accurately,
even though their word recognition and decoding
abilities are not adequate to identify a satisfactory
percentage of the words presented in isolation. That
is, predictable text is printed language chosen or en-
gineered to amplify the availability of context clues
of various kinds. Some sources of predictability in-
clude the match between illustrations and print, the
familiarity of language patterns and story episodes
(Peterson, 1991), as well as rhyme and repeated
phrases (Hoffman et al., 2002).
As the relative difficulty of a sequence of pre-
dictable texts increases, the availability of extra context
clues decreases. Texts become “longer,” “more com-
plex,” and have “less patterned language and more
varied vocabulary” (Peterson, 1991, p. 123). That is,
predictability declines as texts become more difficult.
The purpose of predictable text is to foster students
coordinated use of the three main cueing systems—
sentence structure (syntax),” “message (semantics),”
and “letters (graphic cues)” (Clay, 1993, pp. 41–42).
The instructional emphasis the texts are designed or
selected to support is coaching instruction—guiding
students during oral reading to use strategies and mul-
tiple sources of information to solve problems as they
arise (Hicks & Villaume, 2000).
Multiple-criteria text
Hiebert (1999) referred to text with a dominant
principle such as decodability or vocabulary control
by word frequency as a “single-criterion text” (p. 563).
She then described how teachers can use texts with
different single criteria to provide a “multiple-criteria
program” (p. 563). She also expressed the hope that
publishers will follow the example of Dr. Seuss and
create “multiple-criteria texts” (p. 563). Presumably,
if such texts were leveled, multiple-criteria texts would
manifest declines in vocabulary control by word fre-
quency, phonetic control, and predictability as their
assigned levels increased.
Rationale for the study
Little is known about the contribution of texts
to early reading instruction. One reason may be an
attitude toward texts of effectiveness by association. For
example, Beck (1997) argued that, because highly
decodable text is so closely associated with synthetic
phonics instruction, research supporting the latter
supports the former. If such an attitude were com-
mon among researchers of early reading, it would
help explain why so much more research exists on
instructional approaches than on instructional texts.
Another reason that so little is known about the
contribution of texts to early reading instruction may
be the lack of an accepted technology for validating
the schemes used to assign relative or absolute difficulty
to the texts. Readability research has proven unable to
account for the small gradations of difficulty that
414
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
415
mark the typical sequencing or leveling of early read-
ing materials (Hoffman et al., 2001; Klare, 1984). As
a consequence, research on texts for early reading in-
struction that depends on the validity of the levels of
those texts is less likely to be done or published.
This study examined the instructional support-
iveness of texts for teaching early reading. We define
instructional supportiveness as the degree to which
texts provide opportunities for children to apply what
they are being taught and gradually increase the chal-
lenge of those applications. Research on instructional
supportiveness proceeds by analyzing curricular di-
mensions of one or more sets of sequenced texts.
Curricular dimensions of any set of sequenced
texts are manifest in two ways that have not been pre-
viously distinguished in the literature. The first one is
the opportunity for readers to apply one or more in-
structional emphases across the texts as a set (curricu-
lar dimensions of the set of texts as a category of
reading instructional texts). The second one is the fac-
tors that become more demanding as the books in the
set increase in difficulty (curricular dimensions of the
set’s leveling). These two ways correspond to the two
aspects of instructional supportiveness: the opportuni-
ties for students to apply what is taught, and the grad-
ual increase in challenge of those applications.
Investigating the instructional supportiveness
of texts avoids the assumption of effectiveness (or in-
effectiveness) by association. That is, it brackets the
question of whether a particular reading instruction-
al approach is effective or best. Instead, it empirically
examines the match between curricular dimensions
of sets of texts and one or more instructional em-
phases. To the extent that “text matters in learning to
read” (Hiebert, 1999, p. 552), any reading instruc-
tional approach can be well or poorly served by the
degree to which the texts that accompany it provide
increasingly challenging opportunities for the chil-
dren to apply what they are being taught in that ap-
proach. As a consequence, investigating the
instructional supportiveness of texts has the potential
to improve the effectiveness of different reading in-
structional approaches without waiting for “the read-
ing wars” to be settled.
Although the lack of a technology for validating
text levels may have hindered research on instruction-
al texts in the past, that obstacle can be avoided by
examining instructional supportiveness. A set of lev-
eled texts is analyzed for evidence that it supports one
or more particular instructional emphases. The as-
signed levels of the texts examined are not assumed to
be valid. If evidence is found that the leveling of the
texts supports a particular instructional emphasis,
that finding provides some validation for the texts
levels, but only in support of that particular instruc-
tional emphasis. The texts still may not be validly lev-
eled for any other instructional approach or with
regard to their difficulty for readers in general on, say,
a cloze task or an oral reading fluency task.
Determining the instructional supportiveness
of texts is prerequisite to determining their instruc-
tional value. Ascertaining the contribution of a type
of text to reading achievement requires intervention
research that varies texts (contrasting different kinds
of texts, or texts versus no texts) without varying the
instruction itself (Allington & Woodside-Jiron,
1998). However, without an analysis of the texts in
the intervention for their supportiveness of the in-
struction provided, the results of the intervention
study are uninterpretable. For example, a significant
effect for texts could occur because they support an
instructional emphasis other than that in the inter-
vention, but that some participants have or had out-
side the intervention. Or the effect for texts may not
be significant because they fail to provide enough
opportunities for participants to apply what they
were being taught in the intervention with gradually
increasing challenge.
In this study, we investigated the instructional
supportiveness of Reading Recovery books for two
reasons. First, research on the instructional support-
iveness of texts, as a new research approach, over-
comes the current obstacles to researching the
contribution of texts to early reading instruction.
The role that instructional texts play is an important
area of reading education where knowledge is both
lacking and needed. Second, this specific study is
needed because books originally leveled for use in
Reading Recovery are now also widely used in regu-
lar and special classrooms having no affiliation with
Reading Recovery. Apparently, these “little books
have filled the gap left when many publishers of
reading instructional materials downplayed the need
for a continuum of difficulty between the mid-1980s
and the mid-1990s.
Currently, many teachers and publishers seem
to have become committed not only to the principle
of gradually increasing text difficulty but also to the
accuracy of Reading Recovery levels themselves for all
reading instructional settings. In many elementary
schools, the Reading Recovery level of a book appears
to determine the perceived difficulty of that book for
all purposes, whether instruction or assessment, and
regardless of instructional setting—whether regular
classroom, Reading Recovery, special education, Title
I reading, or volunteer tutoring (Title I is a U.S. fed-
erally funded program for at-risk students). However,
the frequent use of these leveled books in settings
other than Reading Recovery and its affiliated pro-
grams raises an important question for research: Do
books leveled for use in Reading Recovery support
other reading instructional emphases besides the use
of the three main cueing systems?
Method
The purpose of this study was to examine
books leveled for use in Reading Recovery in order
to judge how supportive such texts are for early read-
ing instruction emphasizing word recognition or de-
coding instead of, or in addition to, the emphasis in
Reading Recovery on the three main cueing systems.
Reading Recovery levels
Books are leveled by the Reading Recovery
Council of North America on the basis of how well
they “support the reader’s present knowledge and, at
the same time, provide some new challenge and op-
portunity for engaging in ‘reading work’” (Peterson,
1991, p. 123). “Present knowledge” includes what
the reader has learned from previous reading instruc-
tion, both in and out of Reading Recovery, as well as
the knowledge of language, literacy, and life that the
child brings to reading instruction (Peterson).
Reading Recovery leveling was the first of sev-
eral schemes that share many of the same conceptual
bases and lead to levels that more or less correspond
with one another. For example, Fountas and Pinnell
(1996) explained that their leveling is very similar to
Reading Recovery’s, but that their levels are less fine-
grained and receive ascending letters rather than
numbers. Three of the most popular schemes that
yield book levels similar to Reading Recovery are
Fountas and Pinnell leveling, Wright Group/
McGraw Hill leveling, and Developmental Reading
Assessment leveling (Wright Group, 2004). We
chose to use Reading Recovery leveling in this study
because of its longevity, wide use, and the large net-
work of users of the book list who have input on the
leveling and releveling of books. We believe the re-
sults of our study would be comparable had we con-
ducted it with books leveled by any of the schemes
conceptually related to Reading Recovery leveling.
Materials
To select the texts for this study, we used the
master list of books leveled by Reading Recovery
(Reading Recovery Council of North America,
1997). The Reading Recovery early intervention
program levels trade books and sets of books pro-
duced for beginning readers by educational publish-
ers into 20 levels (Peterson, 1991; Reading Recovery
Council of North America). Level 1 books are those
that kindergartners can learn to “read” during their
first semester; level 20 books are those considered
readable by average second graders early in the
school year.
We systematically selected four books at each
of the 20 levels for analysis. The books were not ran-
domly selected because we had difficulty locating so
many of the books on the list. Basically, our strategy
was to use public libraries, large bookstores, and
Reading Recovery teachers in our local area until we
had obtained four of the books at each level. We did
not exclude any book on the master list. This process
resulted in a sample of 80 books.
Measures
To our knowledge, this was the first study ever
done on the instructional supportiveness of text. As a
consequence, there were no established measures of
the construct for us to use. Before we could proceed
with our investigation of the instructional support-
iveness of leveled texts, we had to determine mea-
surement principles for the construct. We followed
four measurement principles in developing and se-
lecting the specific measures for this study.
1. Different factors contribute to the instructional supportiveness
of text. In this study, we were interested in print-based
rather than book-design or reader-based factors. For us,
the challenge a text will present to a particular student at a
specific time is dependent upon print-based, book-design,
reader-based, and situational factors. Investigations of the
instructional supportiveness of text can focus on one or
more of these factors. In this study, we restricted our at-
tention to print-based factors because we wanted to inves-
tigate the supportiveness of leveled texts for print-based
reading instruction.
2. Measures of the instructional supportiveness of text should
respect the multilevel nature of texts. The construct of in-
structional supportiveness of text cannot be measured by
examining only one aspect of text. Otherwise the con-
struct would be the instructional supportiveness of items,
whether they be words, sentences, texts, or something else.
While controversies remain, most “theorists distinguish
among component processes of reading at the word, sen-
tence, and text level” (Haberlandt, 1994, p. 2). In keeping
with this consensus, any study of instructional support-
iveness of text should include at least three kinds of mea-
sures corresponding to these three levels of text structure:
word, sentence, and discourse (text). Consistent with this
principle, our main consideration in this study was to
choose measures that focus on each major level of text
structure, not just the word level.
416
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
417
3. Multiple measures of each aspect of instructional supportive-
ness are best. Because there are no established measures of
the construct of the instructional supportiveness of leveled
texts, we concluded that it was important to have several
measures at each of the three major levels of text structure,
rather than trying to choose, in advance of data, the sin-
gle best measure at each one. Any selection of a single
best measure for a level of text structure should be based
on the results of empirical analyses and will be limited to
the set of texts analyzed or the category of texts that set
represents. Because of this principle, we chose several mea-
sures at each level of text structure for this study.
4. Each measure should reflect current best assessment practices
in quantifying the demands at each level of text structure.
Consistent with this principle, each of the measures in this
study was chosen or developed to represent the best mea-
surement technology currently available for capturing the
print-based demands at each of the three major levels of
text structure.
We contend that these four measurement prin-
ciples should guide the selection of measures for any
study of the instructional supportiveness of text, re-
gardless of the specific research question. The re-
search question only influences which word-level,
sentence-level, and discourse-level demands will be
measured using multiple measures that are among
the best currently available.
Measures of word-level demands. This study was
designed to answer the research question: “Do books
leveled for use in Reading Recovery support other
reading instructional emphases besides the use of the
three main cueing systems?” Several approaches were
taken to obtain measures of the word-level demands
of texts that support instruction emphasizing word
recognition or decoding instead of, or in addition to,
an emphasis on the three main cueing systems. First,
five measures were chosen to examine different as-
pects of vocabulary control by word frequency. These
measures, except type-token ratio, were based on statis-
tics from The Educators Word Frequency Guide
(Zeno, Ivens, Millard, & Duvvuri, 1995), a large re-
cent corpus of the words in printed English, and were
computed technologically using their CD-ROM.
Type-token ratio was also generated by computer.
Mean U of the words—The U statistic is the frequency of a
word per million running words of text, weighted by a mea-
sure of how widely distributed that frequency is across texts
from different subject areas (Carroll, 1971). Larger U’s in-
dicate higher frequency with wider distribution. The mean
U of the unique words (types) in a book is a measure of word
frequency that takes every different word into account.
Percentage of high-frequency words (100-word list) —The per-
centage of the running words in a book that are the 100
most common words in printed English.
Percentage of high-frequency words (500-word list) —The per-
centage of the running words in a book consisting of the 500
most common words and their regular morphological vari-
ations.
Type-token ratio—The number of unique words (types) in a
book divided by the number of running words in the book
(tokens). This is a measure of how much word repetition oc-
curs in a book.
U of the word ending the most frequent 75% of the words
This measure was determined by first sorting the unique
words (types) in the book by their U statistics. Then we drew
a line separating the 75% of the words with the highest U’s
from the 25% of the words with the lowest U’s. We then
selected the word in the larger set that had the lowest U (i.e.,
was closest to the line). This is a measure of the minimum
frequency of the 75% of the most frequent unique words in
the book. The U of the least frequent word of the most fre-
quent 75% of the words in a book is a measure of word fre-
quency that is less sensitive to the presence of a few words
with extremely low U’s such as some proper names.
The four measures employing the U statistic
are based on the lognormal model of word-frequency
distribution (Carroll, 1971) as applied in a recent
corpus of words in printed English (Zeno et al.,
1995). The fifth of these measures, type-token ratio,
was recently used by Hiebert (1999) to indicate the
amount of repetition a text provides on the words it
contains.
Second, we developed three measures of onset-
rime decodability. We had no reason to expect that
Reading Recovery books are leveled to support syn-
thetic phonics instruction. However, because pre-
dictable text often includes rhyme (Hoffman et al.,
2002), we did have reason to expect that books lev-
eled for use in Reading Recovery may support in-
struction in the use of onsets and rimes to decode
one-syllable, phonetically regular words.
Percentage of onset-rime decodable words (list A) —The per-
centage of the running words in a book on a list of the one-
syllable words comprised of the highest utility onsets and
rimes (onsets and rimes with the largest neighborhood sizes
in previous studies). In addition, for this study, rimes that
were themselves among the 100 highest frequency words
were also combined with the highest utility onsets. List A
consisted of such words as bug, hat, and lap.
Percentage of onset-rime decodable words (list B)—The per-
centage of the running words on list A plus one-syllable
words comprised of high- to moderate-utility onsets and
rimes. List B added such words to list A as dill, clock, and
broke.
Percentage of onset-rime decodable words (list C)—This mea-
sure was the percentage of the running words on list B plus
one-syllable words comprised of moderate-utility onsets and
rimes. List C added such words to list B as trail, yam, and
trend.
Most studies of onset-rime decoding have been
based on neighborhood counts (number of words
containing a particular onset or rime) to determine
utility (e.g., Goswami, 1998; Leslie & Calhoon,
1995). The rime utilities used to select the rimes in
these three measures were from the two major studies
of orthographic rime neighborhood sizes (Fry,
Fountoukidis, & Polk, 1985; Wylie & Durrell, 1970).
The onset utilities used in this study were from a re-
cent study conducted as part of a study on first and
second graders’ onset-rime decoding (Cunningham et
al., 1999). The items in the three tests also came from
that study, and the division of the items into three
subtests was based on the results of that study.
Third, because word identification should lead
to word meaning access, we selected one more mea-
sure of word-level demands:
Morphemes per 100 words—The morpheme is the smallest
linguistic unit of meaning. The word, falling, for example,
consists of the free morpheme, fall, and the bound mor-
pheme, ing. This measure was computed by counting the
number of morphemes in each book, dividing by the num-
ber of words, and multiplying the result by 100.
This last measure of word-level demands was
also seen as a decodability measure, albeit of a
grapho-semantic rather than grapho-phonetic kind.
It was included in this study because of renewed in-
terest in the morpheme as another multiletter unit of
decoding besides onsets and rimes (Ehri, 1998;
Minkoff & Raney, 2000).
These nine measures were chosen to indicate
whether leveled texts have word-level demands that
support instruction that teaches students to recog-
nize high-frequency words or that teaches them to
decode unfamiliar words comprised of high-utility
onsets and rimes.
Measures of sentence-level demands. Sentence-
level factors, principally syntax, constitute a major
component of the context within which word identi-
fication occurs (Graesser & Bertus, 1998; Graesser,
Swamer, Baggett, & Sell, 1996). While there are
competing theories for the relationship between word
identification and context during proficient reading,
we could find no advocate of an instructional empha-
sis on high-frequency words or phonics in the litera-
ture who also advocated using only isolated words
rather than texts for students to apply what they were
learning. Indeed, effective teachers who emphasize
word recognition and decoding the most are equally
concerned with teaching their students to apply these
words and skills to reading connected text (Taylor,
Pearson, Clark, & Walpole, 2000).
In this study, two approaches were taken to
generate measures of the sentence-level demands of
leveled texts to indicate whether such texts are sup-
portive of instruction that emphasizes high-frequency
words or phonics. First, we chose two measures of
sentence complexity:
Morphemes per sentence—The number of morphemes in the
book divided by the number of sentences.
Words per sentence—The number of running words in the
book divided by the number of sentences.
According to Miller and Kintsch (1980), average sen-
tence length in a text predicts the number of processing cy-
cles required of a reader by the sentences in a text. In
addition, because we had no reason to believe that either
measure was in the minds of the authors of books leveled
for use in Reading Recovery, they could be expected to pre-
dict the general sentence complexity of those books.
Second, we selected two parallel measures of
syntactic complexity that were based on the T-unit
rather than the sentence. Hunt (1964) proposed di-
viding subjects’ writing samples into repunctuated
sentences he defined as “minimal terminable
units...the shortest grammatically terminable units
into which a connected discourse can be segmented
without leaving any fragments as residue.... For
short, the term T-unit will be used” (p. 34). Each T-
unit is an independent clause with all subordinate
clauses attached to it.
Morphemes per T-unit—The number of morphemes in the
book divided by the number of T-units.
Words per T-unit—The number of running words in the
book divided by the number of T-units.
T-unit length is currently used as an indicator of both re-
ceptive and expressive language syntactic maturity/complex-
ity in research on second-language acquisition (Gass &
Selinker, 2001) and in speech and hearing sciences (Paul,
2001).
These four measures were chosen to indicate
whether leveled texts have sentence-level demands
supporting instruction that emphasize word recogni-
tion or decoding.
Measures of discourse-level demands. Finally, five
measures of discourse-level demands were chosen,
based on previously discussed linguistic units:
Number of morphemes in the book
Number of running words (tokens) in the book
Number of sentences in the book
Number of T-units in the book
Number of unique words (types) in the book
418
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
419
In every case, the measure of discourse-level demands was an
estimate of overall text difficulty in terms of the total amount
of word- or sentence-level processing required to read the
book. Two of the measures, number of sentences in the book
and number of T-units in the book, were estimates of the to-
tal amount of sentence-level processing required by the
book. The other three measures were estimates of the total
amount of word-level processing required by the book.
Each of these measures was chosen for one of two reasons.
The total amount of word-level processing in a book should
indicate the degree of automaticity or fluency of word iden-
tification (Samuels, 2002) required to sustain success
throughout the book. The total amount of sentence-level
processing should indicate the relative ease or difficulty of in-
tegrating word identification with context throughout the
book (Kintsch, 1988/1994).
These 18 measures were chosen as the best
ones available for indicating whether leveled texts
have word-, sentence-, and discourse-level demands
that support instruction that teaches students to rec-
ognize high-frequency words or that teaches them to
decode unfamiliar words comprised of high-utility
onsets and rimes.
Procedure
Scores on all 18 measures of print-based de-
mands were computed for each of the 80 books in
the sample. All the print in a book, exclusive of front
matter, end matter, and illustrations, was analyzed to
yield each score for that book. We planned to exam-
ine the means and standard deviations of these mea-
sures across the 80 books to determine characteristics
of Reading Recovery books as a category of instruc-
tional texts for early readers. We also planned a series
of simple and multiple regression analyses using the
18 measures to predict the levels of the 80 books to
explore Reading Recovery leveling.
Results
Means and standard deviations were computed
for each of the 18 measures of print-based demands.
See Table 1 for these means and standard deviations.
Each of the 18 measures was also correlated
with the Reading Recovery level of the 80 books. We
tested each of the correlations for whether it indicat-
ed that the measure was a significant predictor of in-
creasing curricular demands at higher Reading
Recovery levels. This testing required a two-step
process. The first step was to determine the sign (+
or -) of each correlation that would indicate an in-
creased challenge on that measure as Reading
Recovery level increased. The second step was to test
the statistical significance of any correlation having a
sign that indicated increasing challenge with increas-
ing level.
Before computing the correlations, we assigned
to each of the 18 measures the sign of its correlation
Measure MSD
Word-level demands
Mean U of the words 7918.4 4383.7
Morphemes per 100 words 109.9 19.8
Percentage of high-frequency words (100-word list) 46.1 18.4
Percentage of high-frequency words (500-word list) 58.9 19.1
Percentage of onset-rime decodable words (list A) 9.3 7.2
Percentage of onset-rime decodable words (list B) 12.1 9.0
Percentage of onset-rime decodable words (list C) 16.0 11.4
Type-token ratio 0.4 0.2
U of the word ending the most frequent 75% of the words 171.4 238.6
Sentence-level demands
Morphemes per sentence 10.5 9.0
Morphemes per T-unit 7.0 3.7
Words per sentence 9.6 7.6
Words per T-unit 6.4 3.3
Discourse-level demands
Number of morphemes 227.5 331.3
Number of running words (tokens) 214.2 313.4
Number of sentences 28.4 45.1
Number of T-units 33.5 42.8
Number of unique words (types) 70.1 72.8
TABLE 1
MEANS AND STANDARD DEVIATIONS OF 18 MEASURES OF PRINT-BASED DEMANDS
with Reading Recovery level that would indicate in-
creasing print-based demands as book level in-
creased. Ten of the 18 measures were tagged with a
plus as the sign of the correlation that would indicate
increased challenge on the measure at higher levels.
The remaining eight measures were tagged with a
minus as the sign indicative of such an association.
For example, type-token ratio and words per T-unit
were two of our 18 measures with different signs
tagged to them. Type-token ratio was expected to cor-
relate negatively with book level if Reading Recovery
leveling provides students with more repetition in
identifying the same words at lower levels and less
repetition at higher levels. This was so because
higher type-token ratios indicate more repetition of
the same words. However, words per T-unit was ex-
pected to correlate positively with book level if
Reading Recovery leveling requires students to read
less sophisticated syntax at lower levels and more so-
phisticated syntax at higher levels. This was so be-
cause longer T-units indicate more sophisticated
syntax. (See Table 2 for the signs tagged to each
measure.)
After the first step of the testing procedure was
performed, 10 of the 18 correlations were found to
have the sign indicating increasing print-based
demands as Reading Recovery level increases. In the
second step, the 10 correlations with the tagged sign
were tested for statistical significance. Because the
first step had determined the signs of the correlations
to be tested for statistical significance, these signifi-
cance tests were one-tailed. Seven of the 10 correla-
tions were statistically significant (p < .05).
However, because 10 significance tests had
been performed, there was an increased danger of
making a Type I error. So, to avoid the multiple-
testing fallacy, Bonferroni’s correction of the 10
alpha levels was performed and the 10 correlations
examined again for significance. Bonferronis correc-
tion requires that the alpha level for 10 tests be set
at p < .005 in order to achieve an actual alpha level
of p < .05. The same 7 correlations remained signifi-
cant after Bonferronis correction was performed.
(See Table 2 for the 18 simple correlations, tagged
signs, and statistical significance of correlations with
the tagged sign, both before and after Bonferronis
correction.)
Our plan was to produce two hierarchical multi-
ple regression equations to predict the books’ Reading
Recovery levels in order to (a) evaluate how much of
the increase in Reading Recovery level is determined
by increasing print-based demands and to (b) deter-
420
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Predictor Sign indicating r
increasing challenge
at higher levels
Word-level demands
Mean U of the words - .21
Morphemes per 100 words + .06
Percentage of high-frequency words (100 word list) - .35
Percentage of high-frequency words (500 word list) - .12
Percentage of onset-rime decodable words (list A) - .02
Percentage of onset-rime decodable words (list B) - .09
Percentage of onset-rime decodable words (list C) - .04
Type-token ratio - -.04
U of the word ending the most frequent 75% of the words - -.13
Sentence-level demands
Morphemes per sentence + -.03
Morphemes per T-unit + .53*†
Words per sentence + -.01
Words per T-unit + .54*†
Discourse-level Demands
Number of morphemes + .46*†
Number of running words (tokens) + .47*†
Number of sentences + .40*†
Number of T-units + .36*†
Number of unique words (types) + .69*†
Note. *Correlation with both the expected sign (+ or -) and p < .05, one-tailed test.
†Correlation with both the expected sign (+ or -) and p < .05, one-tailed test, after Bonferroni’s correction.
TABLE 2
CORRELATIONS WITH READING RECOVERY LEVEL OF 18 MEASURES OF PRINT-BASED
DEMANDS
Instructional supportiveness of leveled texts
421
mine which of the print-based demands—word-,
sentence-, or discourse-level—was most associated
with the gradient of difficulty considered important
for texts to have that supports their reading instruc-
tion.
To increase the likelihood of cross-validation of
these equations, we had planned for the first equation
to have only three predictors—the best significant
predictor of increasing demands of each of the three
kinds, in the order of size of simple correlation. We
had planned for the second equation to have only
four predictors—the best significant predictor of in-
creasing demands of each of the three kinds plus the
best significant predictor of increasing word-level de-
mands of the other kind (decodability or word fre-
quency) than was included in the first equation. In
other words, we intended to include both decodability
and word-frequency measures in the second equation
after letting either one of the two serve as the best
measure of increasing word-level demands in the first
equation. Again, in the second equation, the best pre-
dictors of each of the four types would be entered in
the order of size of simple correlation. To produce the
second equation, percentage of onset-rime decodable
words (lists A–C) and morphemes per 100 words were
considered to be word decodability measures. The
rest of the measures of word-level demands were
considered to be word-frequency variables. Each
equation was to be cross-validated by randomly
partitioning the books into two sets of equal size.
Based on the simple correlations, the best sin-
gle predictor overall of increasing curricular demands
as Reading Recovery level increases was number of
unique words (types) in a book. This predictor was
placed into both planned multiple regression equa-
tions as the first predictor. Because this variable is a
measure of increasing discourse-level demands, no
other measure of increasing discourse-level demands
was considered for either planned equation.
The best predictor of increasing challenge of the
other two kinds, word-level demands or sentence-level
demands, and the second best predictor of increasing
demands overall, was words per T-unit. This variable
was placed into both planned multiple regression
equations as the second predictor. Because this vari-
able is a measure of increasing sentence-level de-
mands, no other measure of increasing sentence-level
demands was considered for either planned equation.
Unfortunately, no measure of increasing word-
level demands was significant. As a consequence, no
measure of increasing word-level demands was added
to either planned multiple regression equation. Our
intention to produce two multiple regression equa-
tions was thwarted. Instead, we were left with a sin-
gle equation having two predictors, the first a mea-
sure of increasing discourse-level demands and the
second a measure of increasing sentence-level de-
mands. Using this single set of two predictors, a
multiple regression analysis with Reading Recovery
level as the criterion was performed on the 80 books.
The multiple correlation between the set of
two predictors and Reading Recovery level was .78
(p < .001), accounting for 60% of the variance in
Reading Recovery level. The first predictor account-
ed for 47% of the variance, while the second ac-
counted for 13% (p < .01). A multiple correlation of
.78 lies in the range of .7 to .9 that has been tradi-
tional for most readability formulas predicting the
difficulty of texts with known difficulty (Chall &
Dale, 1995). The multiple regression equation was
Predicted Reading Recovery level = 3.011 + .0465 (number
of unique words in the book) + .663 (words per T-unit)
This equation was cross-validated. The four
books at each of the 20 levels were randomly assigned
to two subsets of 40 books. One of these two sets was
randomly selected and a multiple regression analysis
was performed on those 40 books using the same two
predictors in the same order: number of unique words
(types) and words per T-unit. This analysis produced a
multiple regression equation based on the data from
those 40 books (R = .82). The constant and beta
weights from this analysis were then used to compute
a predicted Reading Recovery level for each of the 40
books in the other subset. Finally, a simple correla-
tion was computed between the predicted and actual
Reading Recovery levels for the 40 books in the sec-
ond subset. This correlation (r = .73) was compared
with the multiple correlation of the first subset of
books as the traditional test of cross-validation. The
decline in the correlation was less than .1 (.08), sup-
porting the probable stability of the multiple regres-
sion equation.
Discussion
Limitations of the study
The 80 books we systematically selected as our
sample may not have been representative of all the
Reading Recovery books available for use with chil-
dren at any one time. Replications of this study with
other samples of books leveled for use in Reading
Recovery would be necessary to determine how rep-
resentative our sample was.
One of our measurement principles was that
each measure in this study should reflect the current
best assessment practices in quantifying the demands
at each level of text structure. Even so, there may be
measures other than the 18 we used that would be
better indicators of the print-based demands at the
word, sentence, and discourse levels of books leveled
for use in Reading Recovery, which indicate how
supportive such texts are for early reading instruction
that emphasizes word recognition or decoding in-
stead of, or in addition to, the emphasis in Reading
Recovery on the three main cueing systems. Or bet-
ter measures may emerge in the literature in the fu-
ture. The findings of this study are limited to the set
of measures we used.
Instructional supportiveness of the
category of Reading Recovery books for
instruction emphasizing word
recognition or decoding
When means and standard deviations of the 18
measures of print-based demands were interpretable
as curricular dimensions, they were examined to
characterize the instructional supportiveness for
word recognition or decoding of the 80 books as a
whole, without regard to how the books are leveled.
To the extent that our sample represented Reading
Recovery books in general, these characterizations
can be said to apply to Reading Recovery books as a
category of instructional texts for early reading.
Word-level demands. We conclude that Reading
Recovery books in general contain a moderate pro-
portion of high-frequency words. In the average
Reading Recovery book in our sample, 46.1% of the
running words were from the 100 most frequent
words in printed English. These highest frequency
words appear slightly less often in Reading Recovery
books than they do in texts written at a third-grade
level where half the running words (tokens) are made
up of the 100 most frequent types (Adams, 1990;
Carroll, Davies, & Richman, 1971). In a similar
manner, in the average Reading Recovery book in
our sample, 58.9% of the running words were from
the 500 most frequent words in printed English.
These high-frequency words appear less often than
would be expected, given that “only 300 different
words and their variants make up 65 percent of all
written material [at third-grade level and beyond]”
(Sakiey & Fry, 1979, p. 2).
The type-token ratio for words in the average
book in our sample was .4 (1 to 2.5). Hiebert (1999)
discussed the instructional role of type-token ratio or
word density ratio of unique to total words” (p.
560). She contended that the type-token ratios of 1:2
or 1:4 in the little books she examined probably do
not provide children with adequate repetition on the
words they contain. Given that standard, the
Reading Recovery books in our sample as a whole do
not provide an adequate amount of repetition on the
high-frequency words.
Unfortunately, there are no guidelines for per-
centage of high-utility onset-rime decodable words in
text that would support instruction in using ortho-
graphic patterns to identify words. Still, when 84.1%
of the books in the sample contain 27.4% (the mean
plus one standard deviation) or fewer onset-rime de-
codable words, and over half the books contain 14%
or fewer, it would seem unreasonable to conclude
that Reading Recovery books as a category support
instruction in using onsets and rimes to decode.
Sentence-level demands. The average Reading
Recovery book in our sample had a mean of 9.6
words per sentence. On the Spache (1953) readability
formula, this mean sentence length is associated with
a predicted grade level of at least 2.2. This mean,
converted to sentences per 100 words, falls on the
central line of the Fry Readability Graph (Fry, 1977)
at the cusp between second- and third-grade level.
Both these estimates are somewhat high given that
Reading Recovery books have a median grade level
of 1.5 or less.
The mean T-unit length of the 80 books was
6.4 words and 7.0 morphemes. To our knowledge
no norms exist for interpreting the T-unit length of
comprehensible input, whether written or oral, at
various ages or grades. However, Loban (1976) com-
puted average number of words per communication
unit (equivalent to Hunt’s T-unit) in expressive oral
language across grades 1–12 and in expressive writ-
ten language across grades 3–12. Using those aver-
ages, the mean T-unit length of 6.4 words in
Reading Recovery books is noticeably less than first
graders’ expressive oral language (7.9) and third
graders’ expressive written language (7.7). Because
the syntactic complexity of young childrens expres-
sive language is generally less than the syntactic com-
plexity of the receptive language they can
comprehend (Paul, 2001), the sentence-level de-
mands of Reading Recovery books appear to be very
moderate.
When contrasting average sentence length and
T-unit length in words, it appears that the estimate
of sentence-level demands employed by readability
formulas such as the Spache may be quite mislead-
422
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
423
ing, at least in books written for first graders to read.
Apparently, Reading Recovery books have longer
sentences than the books of the 1950s, but they still
maintain a moderate syntactic complexity for young
children. Specifically, Reading Recovery books prob-
ably have more compound sentences without having
more syntactic complexity than a readability formula
such as the Spache would predict.
Discourse-level demands. The average Reading
Recovery book in our sample had 214.2 running
words and 70.1 unique words. Children in the
Reading Recovery program, and in regular classroom
programs affiliated with it, are expected to be able to
read about 90% of the running words in a text the
first time they read it (Clay, 1993). Our findings sug-
gest that a student reading a Reading Recovery book
with that degree of challenge would probably en-
counter about 7 unique words that were new or chal-
lenging with about 3 occurrences each. In Reading
Recovery, these encounters presumably provide the
student with the desired number of opportunities to
make oral reading errors or miscues that elicit coach-
ing instruction while using the three main cueing sys-
tems to solve problems. What about classrooms
where the teacher emphasizes word recognition or de-
coding instruction rather than the three main cueing
systems? Using a 90% criterion of oral reading accu-
racy for determining which level of books a student
should be given to read would yield a maximum of
21 opportunities for students to apply what they are
being taught; a 95% criterion would yield a student
about half that many opportunities.
Conclusion. From our analysis of their word-,
sentence-, and discourse-level demands, we conclude
that Reading Recovery books, as a category of texts
for early reading instruction, provide some support
for an instructional emphasis on the recognition of
high-frequency words, but inadequate support for an
instructional emphasis on decoding words using on-
sets and rimes. The strength of the Reading Recovery
books as a category of reading instructional texts ap-
pears to be primarily at the sentence level. If Reading
Recovery books in general contained a small increase
in percentage of high-frequency words and had a
somewhat higher type-token ratio, their average
sentence-level demands could be expected to provide
much better support for students in self-monitoring
their word recognition than traditional instructional
text based on readability formulas ever did.
Likewise, the moderate syntactic complexity of
Reading Recovery books, achieved while avoiding
short unnatural sentences, would seem to provide
ideal support for students in self-monitoring their
onset-rime decoding as well. Unfortunately, Reading
Recovery books as a category contain too few words
comprised of useful onsets and rimes for these
sentence-level advantages to come into play or for us
to interpret the potential usefulness of the average
number of decoding opportunities afforded by this
category of books at the discourse level.
Instructional supportiveness of the
leveling of Reading Recovery books for
instruction emphasizing word
recognition or decoding
The results of the various correlational analyses
between Reading Recovery level and the measures of
print-based demands at three levels of text structure
were interpreted to characterize the instructional
supportiveness for word recognition or decoding of
how Reading Recovery books are leveled.
Word-level demands. The Reading Recovery
books in our sample did not become more challeng-
ing as book level increased along any of the word-
level dimensions assessed by our nine measures.
Sentence-level demands. Sentence length,
whether in words or morphemes, did not significant-
ly increase at higher Reading Recovery levels. This
finding supports the contention that readability for-
mulas using sentence length as a variable cannot pre-
dict the level of Reading Recovery books (Peterson,
1991). However, T-unit length in both words and
morphemes did significantly and positively correlate
with Reading Recovery level, indicating that in-
creased syntactic demands are associated with
Reading Recovery leveling. Although not related to
our research question, this finding lends some sup-
port to the claim that Reading Recovery leveling
supports instruction in the three main cueing sys-
tems. Because the syntax of the sentences in a text is
considered one of the main cueing systems (Clay,
1993), a sequence of texts that supports instruction
in the three main cueing systems should increase in
its syntactic demands as text level increases.
Pertinent to our research question, the increase
in syntactic demands at higher levels of text is consis-
tent with what would support instructional em-
phases on word recognition or decoding. Indeed,
many advocates of teaching high-frequency words or
decoding systematically also advocate teaching stu-
dents to use context to self-monitor their word iden-
tifications during oral and silent reading (e.g.,
Cunningham & Allington, 2003; McCormick,
2003). As students become better readers, it makes
sense that they should learn to use increasingly com-
plex context to perform this self-monitoring act.
Discourse-level demands. All five measures of
book length were significantly and positively corre-
lated with Reading Recovery level, indicating that
discourse-level demands increase as assigned level in-
creases. Hiebert (1999) remarked that, “Within cur-
rent schemes of text selection for beginning readers,
the only concession to word difficulty is attention to
text length” (p. 560). Our findings replicate hers.
The best overall predictor of Reading Recovery level
was the number of unique words (types) in the book.
This single print-based measure accounted for 47%
of the variance in Reading Recovery level across the
80 books in our sample. Although not related to our
research question, this finding lends credence to the
claim that Reading Recovery leveling supports in-
struction in the three main cueing systems. Because
children in Reading Recovery are taught to use the
three main cueing systems to solve problems when
they arise (Hicks & Villaume, 2000), we would ex-
pect harder books in a sequence supporting such in-
struction to have more problems to solve (more
unique words to figure out) as children become
more proficient.
Relevant to our research question, the increase
in discourse-level demands at higher levels of text is
consistent with what would support instructional
emphases on word recognition or decoding. That is,
the better students can recognize or decode words,
the longer they should be able to identify and self-
monitor words before becoming frustrated or fa-
tigued. Moreover, the fact that number of unique
words (types) rather than number of running words
(tokens) was the better predictor of Reading Recovery
level suggests that Reading Recovery leveling would
support decoding instruction better than word-
recognition instruction if the word-level demands
were present to support such instruction.
Conclusion. The increase in sentence- and dis-
course-level demands as Reading Recovery level in-
creases would support an instructional emphasis on
word recognition or onset-rime decoding.
Unfortunately, the Reading Recovery books in our
sample did not become more challenging as book
level increased along any of the nine word-level di-
mensions we measured. If our sample of Reading
Recovery books was representative, this finding
means that the way Reading Recovery books are lev-
eled does not support instruction in recognizing
high-frequency words or decoding words consisting
of high-utility onsets and rimes.
Implications
The instructional supportiveness of
Reading Recovery books for instruction
emphasizing word recognition or
decoding
The sentences in Reading Recovery books
avoid the artificial shortness associated with text
written to achieve a low readability formula score,
yet they still present only moderate syntactic com-
plexity that gradually increases as book level increas-
es. These sentence characteristics are what one would
expect to find from materials designed or selected to
have natural language patterns that support the use
of syntax as one main cueing system for solving
problems while reading.
The best print-based predictor of Reading
Recovery level is the number of unique words in a
book. This discourse characteristic is what one
would expect from books designed or selected and
sequenced to provide students with an optimal num-
ber of opportunities (unfamiliar words) at each point
of their development to use the three main cueing
systems to solve problems (figure out the words).
However, books originally leveled for use in
Reading Recovery are now also widely used in regu-
lar and special classrooms having no affiliation with
Reading Recovery. This led us to pose the research
question for this study: Do books leveled for use in
Reading Recovery support other reading instruction-
al emphases besides the use of the three main cueing
systems?
Reading Recovery books as a category of in-
structional texts for early reading average fewer op-
portunities to recognize high-frequency words than
would be expected from first-grade materials with
the goal of supporting word-study instruction on
such words. The type-token ratio of Reading
Recovery books in general is lower than one would
expect in texts that support instruction in word
recognition. They also average fewer opportunities to
decode words consisting of high-utility onsets and
rimes than one would expect from books chosen or
constructed to support such instruction. Teachers
who attempt to teach their students to recognize
high-frequency words without relying on context, or
424
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
425
to decode words comprised of common onsets and
rimes without relying on context, should find texts
more supportive of that instruction than Reading
Recovery books.
Moreover, the way Reading Recovery books are
leveled provides no support for instruction in recog-
nizing words by their orthography or decoding them
by their phonology. We recommend that advocates
of word recognition or decoding instruction either
seek other kinds of materials or select and relevel a
subset of Reading Recovery books that will provide
increasing word-level demands as the assigned levels
of the books increase.
Reading Recovery lessons incorporate both
study of high-frequency words and phonics instruc-
tion that includes using onset-rime patterns. Yet the
books they select for their program provide little sup-
port for these two instructional components, and the
way they level the books provides none at all. The
Reading Recovery Council of North America may
want to consider revising their leveling approach to
also take into consideration the occurrence of high-
frequency and onset-rime decodable words. If it is
unreasonable to expect a single category of texts to
support word recognition and decoding instruction
as well as instruction in the three main cueing sys-
tems, Reading Recovery may want to consider adding
two additional lists of single-criteria texts for teachers
to use that support their high-frequency word and
onset-rime decoding instruction, respectively.
Books leveled for use in Reading
Recovery and reading assessment
The results of this study warn us about the
practice of schools and school districts where primary-
grade childrens oral reading is assessed using
Reading Recovery books or passages. Because those
assessments typically yield a reading level based on
the Reading Recovery level of the books or passages
used, they may not be valid for students whose
teachers have emphasized word recognition or de-
coding rather than or more than the use of the three
main cueing systems.
Reading educators may be able to select
and relevel subsets of Reading Recovery
books that support the development of
both word recognition and decoding
While the designation of levels [in this chapter] is a useful
way to indicate roughly a progression of difficulty for books
used during Reading Recovery lessons, such a system may
not be suitable for other settings. (Peterson, 1991, p. 128)
This study found that books leveled for use in
Reading Recovery do not consistently increase in
word-level demands as their levels increase. As we
have reasoned elsewhere (Cunningham et al., 2004),
there are so many attractive and interesting books
leveled for use in Reading Recovery that there is rea-
son for optimism about the possibility of finding
subsets of these books with ample percentages of
high-frequency and onset-rime decodable words that
gradually decrease as book difficulty increases.
Perhaps such a new selection and leveling scheme
could permit many of these books to continue to be
used by teachers whose instruction emphasizes word
recognition or decoding more than the use of three
cueing systems.
Implications for further research
Because so little is known about the contribu-
tion of texts to early reading instruction, much re-
search of various kinds is needed. At the least, other
categories of reading instructional texts besides
Reading Recovery books could be examined for evi-
dence that they support particular instructional em-
phases. For example, the phonetically decodable
texts currently mandated in some states could be an-
alyzed to determine the extent to which they also
support the teaching of high-frequency word recog-
nition and decoding by onset-rime patterns. Also,
given that certain kinds of texts have been mandated
without evidence of their effectiveness, intervention
studies of the instructional value of different cate-
gories of texts are needed. Taking an approach simi-
lar to the one in this study to determine the
instructional supportiveness of the texts used in any
intervention will increase the likelihood that the re-
sults of such studies will be interpretable.
The results of our study raise the question of
how important leveling is as a separate factor from
category in supporting early reading instruction. It
has long been assumed in reading education that
texts have levels and that texts should be sequenced
by level when teaching reading. This assumption is
the basis for readability and text leveling, ability and
leveled grouping for reading instruction, and assess-
ment of students’ reading instructional levels with
informal reading inventories or benchmark books.
We share that assumption and have employed it in
this study. However, what if the levels assigned to
books arent as important as their category? That is,
what if the characteristics of the books students read,
more than the order in which they read them, shapes
their reading strategies and supports the reading in-
struction they receive? We could find no previous
research that raised or investigated this question, but
future research could be designed to distinguish the
effects on reading achievement of category from lev-
eling in the materials used to teach reading in the
primary grades. Until such research is done, howev-
er, it seems prudent to follow the consensus of pro-
fessional opinion that books for early reading
instruction should be leveled, and leveled along the
curricular dimensions of the instructional emphasis
the books are expected to support.
JAMES W. CUNNINGHAM is professor emeritus of literacy education
at the University of North Carolina at Chapel Hill where he taught
undergraduate and graduate courses in reading and writing education
for 28 years. His continuing professional interests include improving
elementary classroom writing instruction and investigating first
graders’ sensitivity to orthographic patterns. He can be contacted at
811 Leigh Drive, Gibsonville, NC 27249-2734, USA, or by e-mail at
jwcunnin@email.unc.edu.
STEPHANIE A. SPADORCIA is an assistant professor of language
and literacy at Lesley University in Cambridge, Massachusetts. She
teaches courses in the areas of struggling readers and writers,
assessment, literacy for children with intensive special needs, and
language and literacy development. Her research and writing focus on
improved literacy instruction for struggling readers and writers. She
can be contacted at 29 Everett St., Lesley University, Cambridge, MA
02138, USA, or by e-mail at sspadorc@mail.lesley.edu.
KAREN A. ERICKSON is the director of the center for literacy and
disability studies at the University of North Carolina at Chapel Hill. Her
research focuses on literacy assessment and instruction for children
with significant disabilities including those who use augmentative and
alternative communication. She is the 2004 recipient of the National
Down Syndrome Congress Educator Award and the International
Society for Augmentative and Alternative Communication
Distinguished Literacy Lectureship Award. She can be contacted at
CB# 7335, UNC-Chapel Hill, Chapel Hill, NC 27599-7335, USA, or by
e-mail at Karen_Erickson@med.unc.edu.
DAVID A. KOPPENHAVER is an associate professor in the Language,
Reading, and Exceptionalities Department at Appalachian State
University in Boone, North Carolina. His research focuses on literacy
instruction and assessment of children and adults with developmental
disabilities, including cerebral palsy, autism, and intellectual
disabilities. He can be contacted at Appalachian State University, LRE
Department, 124 Edwin Duncan Hall, Boone, NC 28608, USA, or by e-
mail at koppenhaverd@appstate.edu.
JANET M. STURM is associate professor in the Department of
Communication Disorders at Central Michigan University. She received
her doctorate at the University of Nebraska-Lincoln and completed a
postdoctoral fellowship at the Munroe-Meyer Institute of Genetics and
Rehabilitation in Omaha, Nebraska. Sturm’s professional interests
include computer-supported literacy, tying together literacy
assessment and instructional strategies, classroom communication,
and educational integration of students who use augmentative and
alternative communication. She can be contacted at Department of
Communication Disorders, Health Professions Building, 2167, Central
Michigan University, Mount Pleasant, MI 48859, USA, or by e-mail at
sturm1j@cmich.edu.
DAVID E. YODER is professor emeritus of speech and hearing
sciences at the University of North Carolina at Chapel Hill. His current
research is focused on the literacy needs of persons with severe
speech and physical impairments and assistive technology for adults
with disabilities. Yoder is a member of the research team of the Center
for Literacy and Disabilities Studies at the University of North Carolina,
which is directed by Karen A. Erickson. The team is working to develop
an alternative reading comprehension assessment battery for students
with severe speech and physical impairments. He can be contacted at
CB# 7335, TR# 46, UNC-Chapel Hill, Chapel Hill, NC 27599-7335,
USA, or by e-mail at dyoder@med.unc.edu.
REFERENCES
ADAMS, M.J. (1990). Beginning to read: Thinking and learning about
print. Cambridge, MA: MIT Press.
ALLINGTON, R.L., & WOODSIDE-JIRON, H. (1998, Spring).
Decodable texts in beginning reading: Are mandates and policy based on
research? ERS Spectrum, 16, 3–11.
BARR, R. (1989). The social organization of literacy instruction. In S.
McCormick & J. Zutell (Eds.), Cognitive and social perspectives for literacy
research and instruction (38th yearbook of the National Reading
Conference, pp. 19–33). Chicago: National Reading Conference.
BECK, I.L. (1981). Reading problems and instructional practices. In
G.E. MacKinnon & T.G. Waller (Eds.), Reading research: Advances in the-
ory and practice (Vol. 2, pp. 53–95). New York: Academic Press.
BECK, I.L. (1997, October/November). Response to “Overselling
phonics” [Letter to the editor]. Reading Today, 15(2), p. 17.
BETTS, E.A. (1946). Foundations of reading instruction. New York:
American Book.
CARROLL, J.B. (1971). Statistical analysis of the corpus. In J.B.
Carroll, P. Davies, & B. Richman (Eds.), The American Heritage word
frequency book (pp. xxi–xl). Boston: Houghton Mifflin.
CARROLL, J.B., DAVIES, P., & RICHMAN, B. (1971). The
American Heritage word frequency book. Boston: Houghton Mifflin.
CHALL, J.S., & DALE, E. (1995). Readability revisited: The new
Dale-Chall readability formula. Cambridge, MA: Brookline Books.
CHALL, J.S., CONARD, S.S., & HARRIS-SHARPLES, S. (1991).
Should textbooks challenge students? The case for easier or harder books. New
York: Teachers College Press.
CLAY, M.M. (1985). The early detection of reading difficulties (3rd
ed.). Portsmouth, NH: Heinemann.
CLAY, M.M. (1991). Becoming literate: The construction of inner con-
trol. Portsmouth, NH: Heinemann.
CLAY, M.M. (1993). Reading Recovery: A guidebook for teachers in
training. Portsmouth, NH: Heinemann.
CUNNINGHAM, J.W., ERICKSON, K.A., SPADORCIA, S.A.,
KOPPENHAVER, D.A., CUNNINGHAM, P.M., YODER, D.E., &
MCKENNA, M.C. (1999). Assessing decoding from an onset-rime per-
spective. Journal of Literacy Research, 31, 391–414.
CUNNINGHAM, J.W., KOPPENHAVER, D.A., ERICKSON,
K.A., & SPADORCIA, S.A. (2004). Word identification and text char-
acteristics. In J.V. Hoffman & D.L. Schallert (Eds.), The texts in elemen-
tary classrooms (pp. 21–37). Mahwah, NJ: Erlbaum.
CUNNINGHAM, P.M., & ALLINGTON, R.L. (2003). Classrooms
that work: They can ALL read and write (3rd ed.). New York: Allyn &
Bacon.
EHRI, L.C. (1998). Grapheme-phoneme knowledge is essential for
learning to read words in English. In J.L. Metsala & L.C. Ehri (Eds.),
Word recognition in beginning literacy (pp. 3–40). Mahwah, NJ: Erlbaum.
FOUNTAS, I.C., & PINNELL, G.S. (1996). Guided reading: Good
first teaching for all children. Portsmouth, NH: Heinemann.
FRY, E.B. (1977). Fry’s readability graph: Clarifications, validity and
extension to level 17. Journal of Reading, 21, 242–252.
FRY, E. (2002). Readability versus leveling. The Reading Teacher, 56,
286–291.
FRY, E.B., FOUNTOUKIDIS, D.L., & POLK, J.K. (1985). The
new reading teacher’s book of lists. Englewood Cliffs, NJ: Prentice-Hall.
426
Reading Research Quarterly
OCTOBER/NOVEMBER/DECEMBER 2005 40/4
Instructional supportiveness of leveled texts
427
GASS, S.M., & SELINKER, L. (2001). Second language acquisition:
An introductory course (2nd ed.). Mahwah, NJ: Erlbaum.
GOSWAMI, U. (1998). Rime-based coding in early reading devel-
opment in English: Orthographic analogies and rime neighborhoods. In
C. Hulme & R.M. Joshi (Eds.), Reading and spelling: Development and dis-
orders (pp. 69–86). Mahwah, NJ: Erlbaum.
GRAESSER, A.C., & BERTUS, E.L. (1998). The construction of
causal inferences while reading expository texts on science and technology.
Scientific Studies of Reading, 2, 247–269.
GRAESSER, A.C., SWAMER, S.S., BAGGETT, W.B., & SELL,
M.A. (1996). New models of deep comprehension. In B.K. Britton &
A.C. Graesser (Eds.), Models of understanding text (pp. 1–32). Mahwah,
NJ: Erlbaum.
HABERLANDT, K. (1994). Methods in reading research. In M.A.
Gernsbacher (Ed.), Handbook of psycholinguistics (pp. 1–31). San
Diego, CA: Academic Press.
HICKS, C.P., & VILLAUME, S.K. (2000). Finding our own way:
Critical reflections on the literacy development of two Reading Recovery
children. The Reading Teacher, 54, 398–412.
HIEBERT, E.H. (1999). Text matters in learning to read. The
Reading Teacher, 52, 552–566.
HOFFMAN, J.V. (2002). WORDS (on words in leveled texts for
beginning readers). In D.L. Schallert, C.M. Fairbanks, J. Worthy, B.
Maloch, & J.V. Hoffman (Eds.), 51st yearbook of the National Reading
Conference (pp. 59–81). Oak Creek, WI: National Reading Conference.
HOFFMAN, J.V., ROSER, N.L., SALAS, R., PATTERSON, E., &
PENNINGTON, J. (2001). Text leveling and “little books” in first-grade
reading. Journal of Literacy Research, 33, 507–528.
HOFFMAN, J.V., SAILORS, M., & PATTERSON, E.U. (2002).
Decodable texts for beginning reading instruction: The year 2000 basals.
Journal of Literacy Research, 34, 269–298.
HUNT, K.W. (1964). Differences in grammatical structures written at
three grade levels, the structures to be analyzed by transformational methods
(Cooperative Research Project # 1998). Tallahassee, FL: Florida State
University.
JUEL, C., & MINDEN-CUPP, C. (2000). Learning to read words:
Linguistic units and instructional strategies. Reading Research Quarterly,
35, 458–492.
JUEL, C., & ROPER/SCHNEIDER, D. (1985). The influence of
basal readers on first grade reading. Reading Research Quarterly, 20,
134–152.
KINTSCH, W. (1994). The role of knowledge in discourse compre-
hension: A construction-integration model. In R.B. Ruddell, M.R.
Ruddell, & H. Singer (Eds.), Theoretical models and processes of reading
(4th ed., pp. 951–995). Newark, DE: International Reading Association.
(Reprinted from Psychological Review, 1988, 95, 163–182)
KLARE, G.R. (1984). Readability. In P.D. Pearson, R. Barr, M.L.
Kamil, & P. Mosenthal (Eds.), Handbook of reading research (Vol. 1, pp.
681–744). New York: Longman.
LESLIE, L., & CALHOON, A. (1995). Factors affecting children’s
reading of rimes: Reading ability, word frequency, and rime-neighbor-
hood size. Journal of Educational Psychology, 87, 576–586.
LOBAN, W. (1976). Language development: Kindergarten through
grade twelve (No. 18 in a Series of Research Reports Sponsored by the
NCTE Committee on Research). Urbana, IL: National Council of
Teachers of English.
MARTIN, L.A., & HIEBERT, E.H. (1999). Little books and phonics
texts: An analysis of the new alternative to basals. Ann Arbor, MI: Center for
the Improvement of Early Reading Achievement, University of Michigan.
MCCORMICK, S. (2003). Instructing students who have literacy prob-
lems (4th ed.). Upper Saddle River, NJ: Merrill/Prentice Hall.
MENON, S., & HIEBERT, E.H. (1999). Literature anthologies: The
task for first grade readers. Ann Arbor, MI: Center for the Improvement
of Early Reading Achievement, University of Michigan.
MILLER, J.R., & KINTSCH, W. (1980). Readability and recall of
short prose passages: A theoretical analysis. Journal of Experimental
Psychology: Human Learning and Memory, 6, 335–354.
MINKOFF, S.R.B., & RANEY, G.E. (2000). Letter-detection er-
rors in the word The: Word frequency versus syntactic structure. Scientific
Studies of Reading, 4, 55–76.
PAUL, R. (2001). Language disorders from infancy through adolescence:
Assessment & intervention (2nd ed.). St. Louis, MO: Mosby.
PETERSON, B. (1991). Selecting books for beginning readers. In
D.E. DeFord, C.A. Lyons, & G.S. Pinnell (Eds.), Bridges to literacy:
Learning from Reading Recovery (pp. 119–147). Portsmouth, NH:
Heinemann.
PINNELL, G.S. (1990). Success for low achievers through Reading
Recovery. Educational Leadership, 48, 17–21.
READING RECOVERY COUNCIL OF NORTH AMERICA.
(1997). Reading Recovery comprehensive book list. Columbus, OH: Author.
SAKIEY, E., & FRY, E. (1979). 3000 instant words. Highland Park,
NJ: Dreier Educational Systems.
SAMUELS, S.J. (2002). Reading fluency: Its development and as-
sessment. In A.E. Farstrup & S.J. Samuels (Eds.), What research has to say
about reading instruction (3rd ed., pp. 166–183). Newark, DE:
International Reading Association.
SCHMITT, N. (2000). Vocabulary in language teaching. Cambridge,
MA: Cambridge University Press.
SMITH, N.B. (2002). American reading instruction (Spec. ed.).
Newark, DE: International Reading Association. (Original work pub-
lished 1965)
SPACHE, G.D. (1953). A new readability formula for primary grade
reading materials. Elementary School Journal, 53, 410–413.
TAYLOR, B.M., PEARSON, P.D., CLARK, K.F., & WALPOLE, S.
(2000). Effective schools and accomplished teachers: Lessons about pri-
mary-grade reading instruction in low-income schools. The Elementary
School Journal, 10(2), 121–165.
WRIGHT GROUP. (2004). Leveling correlation chart. Retrieved
April 2, 2004, from, http://www.wrightgroup.com/leveling/leveling
_
cor
relation
_
chart.php
WYLIE, R.E., & DURRELL, D.D. (1970). Teaching vowels through
phonograms. Elementary English, 47, 787–791.
ZENO, S.M., IVENS, S.H., MILLARD, R.T., & DUVVURI, R.
(1995). The educator’s word frequency guide. Brewster, NY: Touchstone
Applied Science Associates.
Received December 15, 2003
Final revision received August 23, 2004
Accepted November 5, 2004
... The most comprehensive study of RR-leveled texts to date was conducted by Cunningham et al. (2005). They examined the predictive strength of 18 variables, including nine word-level variables that addressed word frequency and decodability (e.g., percentage of onset-rime decodable words), four sentence-level measures (e.g., words per sentence and per T-unit), and discourse-level variables (e.g., number of unique and total words). ...
... This study extends previous research on the RR leveling system (Cunningham et al., 2005;Hatcher, 2000;Pitcher & Fang, 2007) through its examination of three programs of texts leveled according to the GRL system. The analysis of the constituents and progression of leveled texts representing three grade bands (kindergarten, the first half of grade one [early first], and the last half of grade one [final first]) is also unique to the present study. ...
... Dominance analysis compares the contribution of each variable to the explanatory power of the full regression model with the contributions of each other variable across all possible models with all possible subsets of variables. Although previous studies of leveled texts have found relationships between a handful of text predictors and text levels (e.g., Cunningham et al., 2005;Hatcher, 2000;Pitcher & Fang, 2007), no studies have determined the relative importance of these predictors (Pearson & Hiebert, 2014). Therefore, a goal of the study was to rank text predictors of GRLs by their importance to better understand how the GRL framework aligns with theoretical and empirical research on text complexity for beginning readers. ...
... The relationship between beliefs and practice is particularly relevant for individuals who provide early literacy professional development (PD) in special education; an area well known for its highly polarized debates about best practices for teaching children to read (Cunningham et al., 2005;Stanovich, 2000;Tunmer & Nicholson, 2011). It seems reasonable to expect that the beliefs that teachers adopt and embrace about how best to teach reading likely have a tremendous influence on their practices. ...
Article
Full-text available
We respond to calls for more research to address whether and how successful professional development (PD) experiences (defined here in terms of student progress) are related to changes in teacher beliefs, specifically about effective literacy instruction for young struggling readers. We developed a measure, a Teacher Belief Score, to identify teacher beliefs present in interview data and we used student achievement data to create two contrasting groups of teachers, those whose students had lower progress and those who had higher. While initially in the fall, lower progress and higher progress teachers differed little in their alignment of beliefs with program features; over time, higher progress teachers trended toward beliefs that were aligned with program features, whereas lower progress teachers trended away. Findings suggest the need for an additional component to Guskey’s model of teacher change: attributing student progress to the new instructional practices learned in PD.
... Effective readers should also be able to parse syntactic structures within a text to help organize main ideas and assign thematic roles where necessary [13] [26]. Two features that allow for quicker syntactic parsing are words or morphemes per t-unit [8] and sentence length [21]. Parsing information in the text helps readers develop larger discourse structures that result in a discourse thread [14]. ...
Conference Paper
In this paper, we introduce the Commonlit Ease of Readability (CLEAR) corpus. The corpus provides researchers within the educational data mining community with a resource from which to develop and test readability metrics and to model text readability. The CLEAR corpus improves on previous readability corpora include size (N = ~5,000 reading excerpts), the breadth of the excerpts available, which cover over 250 years of writing in two different genres, and the readability criterion used (teachers' ratings of text difficulty for their students). This paper discusses the development of the corpus and presents reliability metrics as well as initial analyses of readability.
... A typical text at the level designated as mid-first grade (Level F) has a repetitive structure that describes a group of items or actions (e.g., what it takes to grow a garden). The repetition of words and the patterns within words appear to be given less weight than features such as number of words per page and the nature of the text structure (Cunningham et al., 2005). ...
... A perusal of the word study curriculum for the program that Schwartz [82] identified as most widely used in American beginning classrooms-Leveled Literacy Intervention (LLI) [83]-shows a predominance of lessons on short and long vowels, but little attention to more complex vowel patterns or multisyllabic words. In the texts of the LLI program that students read for practice, however, multisyllabic words are prominent from the beginning levels [39], and a measure of word count but not decodability predicts the level at which texts are placed in leveled text programs [84]. In another widely used word study program, the bulk of attention in kindergarten and first-grade programs addresses short and long vowels as well as high-frequency words [85]. ...
Article
Full-text available
This study describes the features of words known and unknown by first graders of different proficiency levels in six instances of an oral reading fluency assessment: three in winter and three in spring. A sample of 411 students was placed into four groups (very high, high, middle, and low) based on their median correct words per minute in spring. Each word in the assessment was coded on 11 features: numbers of phonemes, letters, syllables, blends, morphemes, percentages of multisyllabic and of morphologically complex words, concreteness, age of acquisition, decodability, and U function. Words were classified as known if more than 50% of the students within a group were able to correctly read those words. Features of known and unknown words were contrasted for all but the highest group, which made no errors, at each point in time. An analysis of the patterns of known words across groups from winter to spring shows that students followed a similar general progression in the number and type of words recognized. The most prominent feature of unknown words in winter and spring for the middle group of students was the presence of multiple syllables. The lowest-performing group of students continued to be limited by word length and frequency in their recognition of words, but on both features, their proficiency increased from winter to spring. The discussion addresses several critical issues, most notably the relationship of words in oral reading assessments to the word recognition curriculum of many beginning reading programs.
... Central to the need for a metric that captures decoding difficulty, readability formulas have also been criticized as not being sensitive to the needs of beginning readers in the primary grades, where fine-grained distinctions (subtler than grade level) are required to match readers to appropriate text (Cunningham et al., 2005;Fry, 2002;Hoffman et al., 2001). For example, Hiebert and Pearson (2010) compared Degrees of Reading Power, Fry, Spache, Lexile, and Coh-metrix formulas across seven types of kindergarten through second grade texts. ...
Article
Full-text available
Quantifying the decoding difficulty (i.e., ‘decodability’) of text is important for accurately matching young readers to appropriate text and scaffolding reading development. Since no easily accessible, quantitative, word-level metric of decodability exists, we developed a decoding measure (DM) that can be calculated via a web-based scoring application that takes into account sub-lexical components (e.g. orthographic complexity), thus measuring decodability at the grapheme-phoneme level, which can be used to judge decodability of individual words or passages. Here we report three experiments using the DM: two predicting children’s word-level errors and one predicting passage reading fluency. Generalized linear mixed effect models showed that metrics from the DM explained unique variance in children’s oral reading miscues after controlling for word frequency in two samples of children (experiments 1 and 2), and that more errors were made on words with higher DM scores for poor readers. Furthermore, the DM metrics predicted children’s number of words read correctly per minute after accounting for estimated Lexile passage scores in a third sample (experiment 3). These results show that after controlling for word frequency (experiments 1 and 2) and estimated Lexile scores (experiment 3) the model including the DM metrics was significantly better in predicting children’s word reading fluency both for individual words and passages. While further refinement of this DM measure is ongoing, it appears to be a promising new measure of decodability at both the word and passage level. The measure also provides the opportunity to enable precision teaching techniques, as grapheme-phoneme correspondence profiles unique to each child could facilitate individualized instruction, and text.
... As words are read, effective readers are able to parse these words into syntactic structures that help organise main ideas and assign thematic roles to arguments (Graesser, Swamer, Baggett, & Sell, 1996;Just & Carpenter, 1980;. In terms of text complexity, a number of text features allow for quicker syntactic parsing including words or morphemes per t-unit 1 (Cunningham et al., 2005) or sentence length (Klare, 1984). As syntactic representations are being developed, readers continuously form larger larger discourse structures to create a discourse thread (Grimes, 1975). ...
Article
Full-text available
Advances in natural language processing (NLP) and computational linguistics have facilitated major improvements on traditional readability formulas that aim at predicting the overall difficulty of a text. Recent studies have identified several types of linguistic features that are theoretically motivated and predictive of human judgments of text readability, which outperform predictions made by traditional readability formulas, such as Flesch–Kincaid. The purpose of this study is to develop new readability models using advanced NLP tools to measure both text comprehension and reading speed. This study used crowdsourcing techniques to collect human judgments of text comprehension and reading speed across a diverse variety of topic domains (science, technology and history). Linguistic features taken from state‐of‐the‐art NLP tools were used to develop models explaining human judgments of text comprehension and reading speed. The accuracy of these models was then compared with classic readability formulas. The results indicated that models employing linguistic features more theoretically related to text comprehension and reading speed outperform classic readability models. This study developed new readability formulas based on advanced NLP tools for both text comprehension and reading speed. These formulas, based on linguistic features that better represent theoretical and behavioural accounts of the reading process, significantly outperformed classic readability formulas.
Article
This paper introduces the CommonLit Ease of Readability (CLEAR) corpus, which provides unique readability scores for ~ 5000 text excerpts along with information about the excerpt’s year of publishing, genre, and other metadata. The CLEAR corpus will provide researchers interested in discourse processing and reading with a resource from which to develop and test readability metrics and to model text readability. The CLEAR corpus includes a number of improvements in comparison to previous readability corpora including size, breadth of the excerpts available, which cover over 250 years of writing in two different genres, and unique readability criterion provided for each text based on teachers’ ratings of text difficulty for student readers. This paper discusses the development of the corpus and presents reliability metrics for the human ratings of readability.
Article
Background: The literature on typically developing children’s literacy acquisition provides a conventional model of development; however, little is available regarding how children with communicative disorders develop literacy abilities over time. This study describes the temporal trajectories of reading development in three children with diagnoses of language disorders and histories of reading difficulty. Method: A qualitative research design employing ethnographic data collection and analysis procedures was used. One male (age 8 years 9 months) and two females (ages 7 years 8 months and 7 years 11 months) identified as having impaired language and a history of reading difficulty were followed for 10 months as they engaged in literacy instruction. Results: Each child demonstrated a unique trajectory of literacy improvement. Trajectories of development were observed across the complexity of the materials read, as well as across variables reflective of comprehension. Variables included formal miscue analysis data, retelling of stories read, and text complexity. Discussion/conclusion: This study suggests that each language-impaired individual should be considered separately in how they progress and in how progression is manifested. If norm-referenced, standardized evaluation methods had been applied, or if only one sampling period had been used, the changes and progress noted in each of these participants may have been missed.
Article
According to interpretations of results from the latest oral reading fluency (ORF) study conducted by the National Assessment of Educational Progress (White et al., 2020), fourth‐grade students who score below the basic standard require interventions that focus on word recognition, phonological decoding, and fluency. Before such mandates for students at the middle grades and beyond are enacted, the available information from ORF assessments should be examined from the view of what students know and what they do not know. In this paper, I examine the performances of students who score below the basic level on oral reading assessments such as those used in the NAEP/ORF study to identify students’ strengths and challenges. Results of analyses of students’ performances are used to address how instruction and interventions can best support students in developing the proficiencies required to perform successfully on silent reading comprehension tasks. Issues that particularly require attention in the design of instruction and interventions for readers in the middle grades who score below the basic level are the mode of reading (silent or oral), text, and reading volume.
Article
Full-text available
Hiebert evaluates several types of texts used for beginning reading instruction by examining the tasks each poses for young readers.
Article
Full-text available
The abstract for this document is available on CSA Illumina.To view the Abstract, click the Abstract button above the document title.
Article
Teachers must constantly adjust tasks and texts so that children's knowledge and their willingness to apply it develop in balanced ways.
Article
In this study, we investigated the reliability of two current approaches for estimating text difficulty at the firstgrade level: the Scale for Text Accessibility and Support (STAS-1) and the Fountas/Pinnell system. We analyzed the performance of 105 first-grade students in texts leveled using these systems in the areas of rate, accuracy, and fluency. Students read these texts under three support conditions: sight reading, read-aloud (modeled), and previewed. The predictive validity of the two rating scales was supported by the performance data. Statistically significant effects were found for the various support conditions. Further, our analysis suggests potential benchmarks for first-grade performance: 95% accuracy; 80 words per minute; and a fluency rating of 3 (on a 1–5 scale).
Article
This study examined the role of various word features (e.g., versatile letter combinations) and basal text factors (e.g., word repetitions) in the developing word recognition skills of 93 first-grade students. The relative influence of the word characteristics in two different types of basal series on the acquisition of word identification skills was evaluated several times during the first-grade year. Results indicate that the text children are exposed to early in first grade may differentially shape their word identification strategies. /// [French] Cette étude a examiné le rôle de diverses caractéristiques de mots (ex: combinaisons de lettres versatiles) et les facteurs de textes de base (ex: répétitions de mots) dans le développement des compétences de reconnaissance de mots parmi 93 élèves de cours préparatoire. On a evalué plusieurs fois durant l'année de cours l'influence relative des caractéristiques de mots dans deux types différents de séries de base sur l'acquisition de compétences d'identification de mots. Les résultats ont indiqué que le texte auquel on expose les enfants tôt en cours préparatoire peut former de manière différentielle leurs stratégies d'identification de mots. /// [Spanish] Este estudio examinó la función de varias características de palabras (v.g.: combinaciones versátiles de letras) y de factores de texto básico (v.g.: repeticiones de palabras) en el desarrollo de las destrezas de reconocimiento de palabras de 93 alumnos de primer grado. La influencia relativa de las características de palabras en 2 diferentes series de textos básicos sobre la adquisición de destrezas de identificación de palabras, fue evaluada varias veces durante el año de primer grado. Los resultados indican que el texto al que son expuestos los alumnos a principios del primer grado, puede diferenciar el desarrollo de las estrategias de identificación de palabras.