The Effect Of Text-To-Speech On Fluency And
Comprehension
Lori Kleinen
Spring 2005
Table of
Contents
Page
Abstract 3
Chapter One:
Introduction
A. Importance of study 4
B. Background on Text-To Speech 6
C. Description of the program 8
D. Statement of the problem 9
E. Research Question 10
F. Definitions of terms 10
G. Summary of research design 11
H. Assumptions 12
Chapter Two: Review of Literature 13-18
Chapter Three: Research Design/Methodology
A. Sample 19
B. Materials 20
C. Research Instruments/Data Collection/Permissions 21
D. Procedure 23
E. Data analysis 25
F. Limitations of study 26
Chapter Four: Results & Conclusions
A. Results 28
B. Conclusion 43
C. Implications for Teaching 44
D. Implications for Further Teaching 45
References 46
Appendices 49-62
Abstract
The purpose of this research was to examine whether
or not synthesized text-to-speech (TTS) might increase my six second graders’
reading fluency and/or comprehension. This type of assistive software
reads text out loud as words are highlighted across a screen. TTS requires
minimal student and teacher training and is easily integrated into the curriculum
in classrooms with computers.
There are two types of TTS – synthesized and
digitized. Past research studies indicated greater increases in comprehension
and/or fluency using digitized TTS, possibly because it sounds more human-like.
While synthesized TTS has improved, no studies using the enhanced software were
found.
Literature reviewed cites repeated and guided
reading, and vocabulary skill development as three factors that increase
fluency and comprehension. It was
reasonable then to hypothesize that both seeing and hearing the word(s) read
repeatedly could be supportive.
Overall findings showed synthetic TTS was
beneficial to my students. Story
selection test scores increased, indicating improvement in both vocabulary
acquisition and comprehension. All participants showed gains in at least one
area of fluency. Reading rates increased for all but one participant, though
more significantly for the weaker readers. Finally, more participants indicated
they liked reading independently after using TTS.
The
Effect Of Text-To-Speech On Fluency And Comprehension
Chapter One
Importance
of Study:
The purpose of this research prospectus was to examine the
value of using text-to-speech (TTS) as a tool to increase my six Title I students’
reading fluency and/or comprehension. This type of assistive software
reads text aloud as words or sentences are highlighted across a screen. A
recently coined term, assistive technology (AT) is defined as any manufactured
or modified device used to improve or maintain an individual’s cognitive and/or
functional capabilities. Assistive devices work toward making the user more
independent in daily living skills, or in the case of students, may enable them
to demonstrate what they know.
In the past, these AT devices were expensive, less effective
and less accessible. However, technology has made some great gains over the
past ten years and the opposite is nearly true, which has resulted in more
students using AT in their classrooms. In fact, the IDEA Act (Individuals with Disabilities
Education Act of 1997) requires that schools provide all students with a
“free and appropriate education” (FAPE), which means students must be educated
in the least restrictive environment (LRE), with the necessary special
education services and modifications needed to make this possible. Under this law, AT must be considered for
students with special needs during the development of an Individualized
Education Plan (IEP). Secondly, Section 504 of the Rehabilitation Act dictates
that schools must provide AT, if needed, as a special education service
to special needs students.
With these mandates in place, the traditional classroom
make-up is changing as we see more special needs children receiving their
education in inclusion classrooms than in substantially separate classrooms.
For example, learning- disabled children with speech disorders come to class
with augmentative voice-output devices that enable them to demonstrate their
knowledge. They work alongside students with reading disabilities who have
portable keyboards with word prediction that offsets their difficulties with
written expression, and regular education students. The classroom teacher must
help all of these students meet their educational goals.
These
teachers must also know how to use their students’ AT devices and find ways to
integrate their use with the curriculum to help these students work toward
their academic goals.
Advances in AT technology have improved TTS software making
it more affordable, and easier to use and implement in the classroom. The
number of students using text-to-speech must be increasing since the
Massachusetts Department of Education now allows 6th, 8th
and 10th graders with learning disabilities to use a TTS program
called Kurzweil 3000 when taking the MCAS test (the Massachusetts Comprehensive
Assessment Test). These students have Individual Education Plans that specify a
need for reading and writing accommodations such as Kurzweil, and the
student(s) must use Kurzweil for regular, routine instruction. If TTS can raise
MCAS scores, why not use it with daily curriculum to help struggling readers
like my Title I students? Statistics indicate some intervention is needed.
The percentage of fourth grade students in the state of
Beginning in the third grade, students are expected
to not only decode, but also comprehend fiction and non-fiction. This task only
becomes more daunting as students move through middle and high school grades
where they encounter longer texts with more difficult vocabulary. We need to
look at other ways to support this population before third grade.
If it can be determined that current synthetic TTS
software effectively increases my students’ comprehension and fluency, it may
encourage similar research using larger population samples to the extent that
findings can be generalized. Should several studies conclude that TTS increases
fluency and comprehension for many readers, TTS would be an easily implemented,
assistive tool that could help struggling readers both in
Background on Text-To-Speech:
Texas Instruments was
one of the first to market text-to-speech, or TTS, patenting Lin et al. and
Goudie’s inventions as early as 1987. Other patents quickly followed. These
early synthetic speech engines converted specified computer text into vowel and
consonant phonemes. Voices were robotic, lacking in prosody (expression), and
predominantly male. The blind were early beneficiaries of this new technology.
In
the 1990’s larger companies like Microsoft entered the assistive technology
market as they realized that TTS could support not only the blind, but also
emergent readers and those with reading disabilities. They created Microsoft
Reader which reads electronic books out loud on-line.
Other
companies, like Living Books, developed CD-ROM storybooks where graphics move
and text is highlighted sentence-by-sentence across the computer screen as
users hear the words read aloud. Computer-aided instruction software with
digital (human) text-to-speech was created as a remedial instruction tool for
teaching reading in the classroom. Finally, synthesized speech became more
expressive and human-like with the arrival of AT&T True female and male
voices that read in multiple languages.
A
number of text-to-speech programs, like Write:OutLoud and Intellitalk, allows
the user to write and then listen to what s/he wrote. Text can be highlighted
and read letter-by-letter, word-by-word, or sentence-by-sentence. These
features encourage good spelling and editing practices too. Software such as
Cast Ereader reads scanned-in text from student textbooks.
Other
TTS software, like Don Johnston’s Start-to-Finish, is used primarily to promote
reading comprehension. Students are able to access classic books that have been
modified at lower reading levels. A digitized, human voice can read text at a
rate of the listener’s choosing. Cloze sentencing and multiple-choice tests are
included on the CD-ROM. However, digital TTS is still more expensive than
synthesized programs, making it difficult for schools to afford.
TTS
could be extremely helpful to learners given what we now know about how people
learn. Brain-imaging technologies reveal that the brain uses isolated parts of
the brain for simple tasks, but accesses a variety of regions of the brain that
network with one another for complex tasks. Recognition networks receive and
analyze information, strategic networks plan and execute actions, and affective
networks evaluate and set priorities. (Dolan & Hall, 2001). It is possible
that text-to-speech may eliminate some of the networking needed to read,
freeing up other networks to construct meaning (i.e., comprehension).
This study seeks to provide
experimental research that focuses on the value of current, affordable,
synthetic TTS software as a useful, assistive tool that may increase my
students’ reading fluency and comprehension.
Description of
the program:
I teach twenty-two
second graders in a private, city school in
I use a number of strategies in my
classroom to encourage and increase my students’ fluency and comprehension. I
read aloud to model fluency. We do echo, choral and partner readings of stories
from a Scott Foreman basal. Before reading each story selection, I activate
their prior knowledge and introduce vocabulary using contextual sentences.
Sometimes students draw pictures of these vocabulary words. I encourage my
students to predict what will happen and guide their reading comprehension by
asking a variety of questions, and encouraging critical thinking. I call
attention to pictures in stories as a comprehension and reading strategy. Our
classroom has a word wall containing high frequency words to assist with
spelling and reading. We listen to CD- ROM readings of the selection stories
and do reader’s theater.
All six Title I students are given a word
ring that contains the current week’s spelling and vocabulary words, along with
a Phonic Take Home Reader that focuses on the spelling skills we are learning
each week. We look for patterns in our spelling words. All 22 students are
tested on spelling and the reading selection weekly.
Title I interventions include use of
Little Big Books that focus on specific spelling patterns and digraphs.
Students work with sentence strips and silly sentences to build sequential
skills and writing conventions. They read out loud. Games such as Bingo are
used to help students learn sight and spelling words. Students participate in a
Reading Derby contest to encourage reading at home. Title I students receive
this instruction in a small group setting.
Statement of the
Problem:
Despite
the interventions and strategies described, my six Title I second graders
continue to struggle with reading. They have weak encoding and decoding skills.
Their selection test scores have not risen. While they may do well on the
spelling tests, they do not necessarily transfer their spelling knowledge when
reading and writing.
Four
of the six Title I students are boys. One boy mostly speaks Arabic at home.
Another is very intelligent according to recent testing, but memorizes words on
a short-term basis; he has not learned phonics rules. The third boy slipped
through the cracks in his large, first-grade classroom in a public school. The
last boy is often tardy and has attention issues that hinder his learning
across the curriculum. He has been tested for but not diagnosed with ADD/ADHD. One
girl’s reading skills seem to fluctuate. The last girl receives speech and
language twice a week. Her most current Individualized Education Plan (IEP)
indicates she has auditory processing issues.
She was the only one of the six on an IEP at the inception of this
study. Two others required IEP’s before this study ended.
These
six students’ weaknesses in reading affect their comprehension across the
curriculum, and transfers over to difficulties with writing. I am concerned
that if they continue at this level with few gains, they will falter
significantly in third grade where they will be reading more difficult and
lengthy content. I am wondering whether or not TTS will give 1) my six struggling
readers a tool to effectively help them meet their learning goals 2) me, as
their teacher, a tool I can implement in my classroom to help all of my
students become successful readers 3) me a tool that could
raise my struggling readers’ standardized testing scores.
Research Question:
This experimental study asks whether or not my six
struggling readers will increase their fluency and comprehension skills using
synthesized TTS software.
Definition of terms:
1. Assistive technology: assistive technology (AT) is defined as any manufactured or modified device used to improve or maintain an individual’s cognitive and/or functional capabilities.
2. Inclusion classrooms – classrooms with typically developing students combined with special needs students.
3. Title I: a grant program wherein students with weaknesses in reading receive additional, individualized or small group reading instruction to reinforce classroom instruction.
4. Scanned in text: An optical character recognition device reads text on paper and scans it into a reader or word processing program.
5. Individualized Education Plan: IDEA mandates students who have been identified with disabilities must have an IEP that outlines student educational goals and benchmarks toward achieving them. The IEP is reviewed and modified, if needed, annually.
6. 504 Plan: strategies documented in a legal document falling under the provision of the Rehabilitation Act of 1973 that outlines strategies to be used in an integrated classroom to accommodate for a student’s disability.
7.
Auditory processing –ability to process
and understand what is heard.
8.
Echo reading – delayed reading after the teacher reads/models fluency.
9.
Choral reading – reading text together.
10.
Partner reading – taking turns reading with a peer.
11.
Reader’s theater – students read a play, but are not required to memorize
their parts or necessarily have props or costumes.
Summary of the Research Design:
My six Title I students are an
availability sample as they are easily accessible to me, however they were also
chosen because they have weak reading comprehension and fluency skills and will
benefit if synthetic TTS is even marginally helpful to them.
They will use TTS in the computer lab for
a half hour three times a week. During this time they will listen to the
current week’s story selection. (The rest of the class will eventually have
access to TTS on a rotating basis after the research period.)
Classroom curriculum-based selection tests developed by Scott Foresman will measure their comprehension before and after TTS use. These selection tests present cloze (fill-in-the-blank) sentences and multiple-choice questions as a means to measure reading comprehension and vocabulary.
Running records will be used to test
participants’ oral fluency before and after TTS use. I will collect running
record data by asking each student to read from a grade-level passage for one
minute. The two grade-level passages I will use before and after TTS use are
from the Scott Foresman’s Individual Reading Inventory and Running Record
publication. Both passages are of equal difficulty and length.
A pre and post-test fluency score
will be determined by counting the number of words read correctly per minute
(WCPM). I will also use a rubric to rate their expression (prosody).
I will compare pre and post-test fluency and
comprehension scores amongst these students. I will measure central tendencies
using mean and mode to identify any gains in fluency and comprehension.
In order to triangulate my data, I will collect qualitative data to provide
additional support to my findings. Participants will complete a survey to
document their attitudes toward reading before and after TTS use.
I will observe these students and
make anecdotal notes as they use TTS, noting such things as: 1) their attention
span and any other behaviors they may exhibit while using TTS and 2) which
highlighting method they choose (word-by-word, sentence-by-sentence or by the
paragraph) throughout the four-week period. I will also note any behaviors
exhibited in the classroom while reading.
Assumptions:
These six students have similar, though not identical weaknesses in reading. I suspect fluency and comprehension test scores will increase over the four-week period TTS is used. Participants do not have any physical conditions that may affect their ability to access the computer or this software. Participants know how to use a mouse. All will be equally motivated to provide their best effort throughout the study. TTS will help these struggling readers to better access their curriculum. Parents of participants will allow their child to be a part of this research. All participants will complete the study and will be present on the days when TTS is used.
Literature Review:
Text-to-speech (TTS) is software that reads text out loud as
words are highlighted across a computer screen. Before looking at research
centered on TTS and its effects on fluency and comprehension, it is important
to understand what fluency and comprehension mean and how they are attained.
Fluency is an important component of reading; it frees the
reader to process meaning, or comprehend. Fluency increases when readers achieve
efficiency in automaticity (word recognition) by being able to accurately and
quickly recognize words with little attention or effort. “Quickly” means there
is a natural, smooth flow to the voice.
Fluent readers are able to read
orally with speed, accuracy, and proper expression. Fluency is the bridge
between decoding and comprehension. In more quantitative terms, fluency is the
fluid combination of accuracy and rate that leads to competent performance
(Binder, 1996). But how do readers achieve fluency?
Current research-based theory agrees that guided, and repeated oral reading increases fluency (Kuhn & Stahl, 2000; The National Reading Panel, 2000). “Guided” means students have the opportunity to listen to a fluent rendering of the same passage while reading, and receive instruction as needed. These renditions provide students with a prosodic model (an example of phrasing and expression that supports an understanding of the text). Repeated reading practice assists the reader with vocabulary recall, leading to a faster, expressive, and more accurate delivery.
Reading fluency is a key skill of effective readers.
Research findings support the following instructional methods for increasing
reading fluency: hearing fluent reading modeled, repeated readings, and
progress monitoring (National Institute for Child Health & Human Development,
2000). Of these three methods, repeated readings are most widely recommended as
a means to increase fluency (Kuhn & Stahl, 2000; The National Reading
Panel, 2000).
Fluency is especially important because the speed and
effortlessness with which readers process text is highly correlated with
comprehension (Deno, Mirkin, & Chiang, 1982; Fuchs, Fuchs, & Maxwell,
1988; Markell & Deno, 1997; Snow, Burns & Griffin, 1998). So if the
reader is not stumbling over decoding words, he or she is then left to
concentrate on comprehension (LaBerge & Samuels, 1974). However, Markell
& Deno (1997) conclude, “Only when increases in oral reading are quite
large can predictions about performance on the comprehension tasks be made with
a high degree of certainty.”(p. 13).
Therefore, achievement in fluency
does not necessarily ensure comprehension.
The National Reading Panel (2000)
identified three areas that are important in developing reading comprehension.
These are 1) vocabulary development 2) text comprehension instruction
and 3) teacher preparation and comprehension strategies instruction.
Vocabulary development (both oral and print) is important
to reading comprehension because the larger the reader’s vocabulary, the easier
it is to make sense of the print. According to the National Reading Panel
(2000), repeated exposure to vocabulary appears to enhance vocabulary
development, as do computers. There are a multitude of software programs
available that can present vocabulary repeatedly, both visually and
orally. For example, most electronic
storybooks offer TTS on demand, where text is highlighted to help the
reader/listener follow along.
Some research has focused
specifically on the value of TTS in teaching vocabulary. Though lacking in
external validity due to the small sample size, Herbert & Murdock’s 1994
study compared digital and synthetic TTS’s effect on student reading and
motivation. They found digital TTS to be most effective for teaching vocabulary
words (as cited in
Text comprehension instruction might include “think alouds”, where
a teacher models aloud the mental processes s/he uses to construct meaning from
texts. Vygotsky (1978) noted the importance of scaffolds to promote
comprehension. Scaffolding strategies might include questions teachers pose
that require students to reflect, summarize, or use other critical thinking
skills.
Comprehension also increases when
students self-question, are provided background knowledge, are taught to draw
from their personal experiences and, when vocabulary is previewed. Companies
are beginning to develop software based on this knowledge. For example,
Scholastic WiggleWorks software allows teachers and/or students to record a
think aloud, prediction, or question related to the text on any given page
displayed.
Other comprehension strategies
include giving students experiences with 1) cooperative learning activities
where students learn reading strategies as a group 2) using graphic organizers
to make sense of the text 3) recalling story structure to answer questions
about what they have read, and 4) summarization.
The
National Reading Panel (2000) concluded that teacher preparation and
comprehension strategies instruction using
these strategies are generally more effective when used collectively, wherein
the teacher can flexibly respond to what works best for his or her students at
any given time. Ideally, they recommended teacher training in these strategies
begin as early as pre-service so that teachers enter the classroom capable and
comfortable with providing a number of comprehension strategies on demand.
However, the National Reading Panel
does not specifically address the importance of a teacher who models reading
with good prosody (expression). Students need to hear how text should be read –
where to inflect emotion, how to end a sentence and begin another, how
punctuation affects reading and comprehension overall.
The importance of prosody in reading
text holds true for TTS as well. Past research has shown that good TTS prosody,
or lack thereof, impacts comprehension.
For example, researchers comparing digital/natural voice with
synthesized TTS concluded synthesized speech “places an increased burden on perceptual
and cognitive resources during the comprehension process.” (Kincaid, 2000)
Silver’s study (1995) tested comprehension of taped passages using low and high-end synthesized speech, versus natural voice. Participants listened to each passage, stopped the tape and responded to ten questions. The dependent measure was the number of correct answers. Comprehension accuracy scores comparisons using DECtalk (high-end synthetic TTS) and natural voice were similar. VOTRAX (a low-intelligibility synthesizer) scores were significantly poorer.
Silver also tested students’ ability
to shadow passages (repeat them word for word). Participants ranging from 4th
grade level readers to college-level readers listened to 33 taped passages on a
cassette with VOTRAX, DECtalk and natural voices. Scoring indicated shadowing
responses using VOTRAX were significantly poorer than those using DECtalk and
natural voice, with natural voice superior to DECtalk when participants
shadowed difficult passages. Similarly, Stine & Wingfield (1987) found when
intonation patterns and contextual clues were removed, recall was more
difficult when irregular sentence structures were used.
Farmer’s et al. (1992)
quasi-experimental research explored the effects of computer-aided, digitized
and synthesized TTS on fluency and comprehension in readers with severe
disabilities. Learning-disabled populations ranging from 2nd to 12th
grade varied in sample sizes (from 10 to 108). Testing included pre- and post
tests, multiple-choice and accuracy and revealed digital TTS (DECtalk) was most
effective; it allowed for self-direction, and students benefited by the onset
rime or sub-syllable level (as cited by Jackson, etal., 1995).
McCullough (1995) investigated the
value of using computers to monitor student progress and to remediate reading.
Using 36 study samples involving learning disabled students with neurological
and low achievement scores in K-9 and 12th grade levels,
participants used digital (DECtalk) and synthetic (RACER) TTS software.
Accuracy and pre/post tests were given to both control and experimental
groups. Results indicated TTS as
effective in providing individualized instruction with onset rime, syllable and
whole word levels using both TTS models. These programs reportedly
increased participant motivation overall (as cited by
Last, are a few studies that examined the value of the multi-modal qualities of TTS. In Clark & Paivio’s 1991 study, visual-text modality was shown to improve recall. Yet Mayer & Moreno’s 1998 study found that participants recalled more with audio-visual versus text-to-speech presentations (as cited in Quirk, 2000).
Emerging from these research studies is the general consensus that the quality of synthesized speech, how closely it emulates prosodic, human speech, impacts fluency and/or reading comprehension scores. The implication is that if TTS is to prove useful for reading, and if a synthetic voice is used, it should closely model natural speech. Providing the listener/reader with prosodic cues (intonation, voice, stress, durational patterns, rhythm, etc.) improves overall intelligibility and limits the demands on working memory. Kincaid’s study recommends further research using TTS software design that has been modified to better emulate prosody, include prosodic cues to promote user acceptability, and contain contextual cues readers can use as support when intelligibility wavers (Kincaid, 2000).
While past research shows synthetic
TTS to be less effective than digital TTS, it is still less expensive than
digitized TTS, and therefore a more viable resource for schools. Also, the quality
of synthetic speech has changed dramatically over the last five years, offering
improved prosody. Clearly synthetic TTS needs to be tested again to determine
whether or not its prosody is good enough to enhance fluency and possibly
comprehension.
With TTS, my students can
independently listen to repeated readings of the stories we read. Vocabulary
words that may otherwise hinder their fluency and understanding of the stories
read in class will be read to them repeatedly as they see the words highlighted,
providing them with a multi-modal presentation of the vocabulary and content. This research will use current synthetic TTS
software to determine whether or not my Title I students will find it a useful,
assistive tool that increases their reading fluency and/or comprehension.
Chapter Three
Sample
This
small-scale research involved studying the effect(s) of text-to-speech on six
Title I students in my second-grade classroom. Specifically, I wanted to know
whether or not text-to-speech would improve this group’s reading comprehension
and/or fluency skills. I recognized results of this study may not be
generalized and that I was using a non-probability sample. Additional studies
need to be conducted and preferably include a control group before results can
be generalized.
These
six students were chosen because they are available, but more importantly,
because they may benefit most since their reading skills are weakest amongst my
twenty-two students. Four of the six Title I students are boys. Two are girls.
“Charles”
speaks Arabic mostly outside of school as he spends much of his time with his
Arabic-speaking grandparents. His mother is Portuguese; this language is also
spoken at home, as is English. He is a young second grader.
“Edward’s”
primary language is English. He has weak
word attack skills and written expression is particularly difficult for him. In
describing how he learns words, Edward says he memorizes them. Recent
psychological testing determined Edward has a high IQ, however his short-term
memory is markedly lower, though still in the average range. He is already receiving
additional tutoring on Saturdays to address his reading difficulties.
“Paul’s”
mother attributes her son’s weak reading and writing skills to his having
slipped through the cracks in his large, first-grade classroom in a public
school last year. Paul tends to ask what a sentence says rather than to read it
first himself. His word attack strategies fluctuate to the extent that he may or
may not be able to sound out beginning and/or middle sounds from one day to the
next. He omits dipthongs when reading and vowels when writing. Paul struggles
with the vocabulary words in our reading selection.
“Alex”
has attention issues that hinder his learning across the curriculum. His
tardiness results in missed instruction. Comprehension fluctuates and is
dependent on the quality of his attention to tasks. Sequencing stories is
difficult for him.
“Kirstin’s”
reading skills seem to fluctuate. She has difficulty with comprehension when
reading independently. While she may do well on weekly spelling tests, word
knowledge is rarely transferred when reading new text or when writing. In terms
of reading fluency, her phrasing, pace and expression are poor. Kirstin does
not like reading out loud in class.
“Mary”
receives speech and language services. I accommodate for her auditory
processing issues by placing her toward the front of the class and checking in
with her individually, especially when directions are given. Mary reads with
some fluency, but needs classroom support for comprehension. She has an
Individualized Education Plan (IEP), but it is not current.
Materials
Participants
worked in a lab setting using a Windows 2000 computer loaded with synthesized
TTS. Originally, Don Johnston’s Write Out:Loud (WOL) text-to-speech software
was going to be used, however it was not compatible with the NTs in the lab. Instead,
a trial version of Intellitalk III, a comparable synthetic program was used.
The following software features were customized according to user preference:
1) text size 2) spacing 4) choice of a variety of male and female voices 7)
voice rate and volume 8) text reads by letter, word, sentence or paragraph 9)
toolbars can read out loud. Headphones were provided. Three stories from Scott
Foresman’s basal reader were input so that participants could listen to/see the
stories read using Intellitalk. Students had access to TTS three times a week
for half-hour intervals.
Scott
Foresman pre and post-test Selection Readings (Appendix A), Tests (Appendix B),
and Foresman Individual
Research
Instruments
1) Scott Foresman Reading basal authors a ten-question
multiple-choice test based on story selection content and vocabulary presented.
(See Appendix B). Test scores prior to TTS use were compared with test scores
after the four weeks of using TTS.
2) Running records were used to test participants’ oral
fluency. Students read from grade-level passages by Scott Foresman (Appendix
C). I marked any miscues or errors made during reading.
a. An oral reading accuracy score was determined by
counting the total number of errors and referring to the Foresman table to
determine a percentage of words the reader read correctly.
b. An oral reading fluency rate was calculated using the
following formula:
Total # of
words read
______________________ x 60 = words per minute
# of seconds needed
c. Prosody was assessed using a rubric that considers
expression and
volume, smoothness, phrasing, and pace. I chose this
rubric because it reflects the ways in which prosody is defined, and is a quick
and efficient means to administer, score and interpret prosody. (See Appendix
D)
3) Descriptive statistics in the form of a pre and post
treatment survey were given to check for student’s attitude
changes about reading, if any, and provided triangulation in my study. (See
Appendix E). I read the survey out loud to be sure participants understood what
was asked.
4) Anecdotal notes were taken during the study to add
credence to my findings. Observations were noted such as specific behaviors
exhibited while using TTS were noted. (See Appendix F).
Data Collection
Equipment
Appendix forms for anecdotal notes, fluency tests and
fluency rubric, in addition to Foresman reading selections and tests.
Permissions
1) Nov.
29th: Administration consent form was submitted and approval was
received to perform study/use school equipment (Appendix H).
2) Dec 8th: IRB application submitted to the professor.
3) After
IRB approval: Parent consent forms were
submitted for approval to execute this project using the participants in my
classroom (See Appendix G).
4) Permission
to use the prosody rubric was received after emailing the
originator (See Appendix D)
Procedure
This study covered four weeks in which participants
continued to receive reading instruction as detailed previously. In addition,
they received lab time for half-hour periods three times a week, as did the
remainder of the class after the baseline data was collected in the first week.
During this lab time, participants listened to/saw the basal reader stories
read to them. The remaining sixteen students worked on language-related
activities on the computer.
Initial Step -- Week One
1. As soon as IRB was approved, participants were tested
for fluency using the first Foresman passage and scoring criteria (Appendix C).
The rubric was also be used to measure prosody (Appendix D).
2. Participants were given the first reading selection
test (Appendix B) based on the story being read in class without TTS use
(Appendix A,). Results from this test were used as the baseline and compared to
the post-test scores (i.e., selection and fluency tests administered at the end
of the study after TTS use).
3. Participants completed the reading attitude survey
(Appendix E).
These steps provided baseline data for my study so
that comparisons could be made after the three-week period of TTS use.
Week Two
1. Participants used TTS in the lab three times for half
hour periods. They saw and listened to the story selection using synthetic TTS.
Anecdotal notes were taken during this time using the form in Appendix F.
2. All students, including participants, took the Week
Two’s Foresman selection test.
Week Three
1. Participants used TTS in the lab three times for half
hour periods. They saw and listened to the story selection using synthetic TTS.
Anecdotal notes were taken during this time using the form in Appendix F.
2. All students, including participants, took Week
Three’s Foresman selection test.
Week Four
1. Participants used TTS in the lab three times for half
hour periods. They saw and listened to the story selection using synthetic TTS.
Anecdotal notes were taken during this time using the form in Appendix F.
2. All students, including participants, read the
Foresman reading selection (See Appendix A) and took Week Four’s Foresman
selection test (Appendix B). Results from this test were compared with the
selection test taken in Week One.
3. Participants were tested for fluency using the second
Foresman passage and scoring criteria (Appendix C). The rubric was also used to
measure prosody (Appendix D).
4. The student survey (Appendix E) was administered
again to see if there were any changes in attitudes toward reading.
Data Analysis
a. Pre and post-test data provided a comparison of
selection test scores before and after TTS use. This data was used to consider
any gains in comprehension, using mean, and mode. Qualitative data (anecdotal
and surveys) was compiled throughout the study. Pre and post TTS survey results were tabulated and
compared by assigning two points for a “yes” answer, and one point for a “no”
answer to each question. Point differences were compared for each question
and as a whole both pre and post use of TTS. Any positive or negative changes
in attitude were noted.
b. This
data was used collectively to determine the causative impact of the treatment
(TTS) on comprehension and fluency (i.e., whether or not any gains were made).
c. Pre- and post-test data was compared relative to
fluency scores using the Foresman Oral Reading Accuracy charts (see Appendix C)
and the Oral Reading Fluency formula:
Total # of
words read
______________________ x 60 = words per minute
# of seconds needed
d. Prosody was
evaluated using the rubric (see Appendix D).Any differences in point scores in
the four areas of prosody were quantified and discussed.
Limitations of this study
1. Participants with unknown additional learning
disabilities other than encoding and decoding may cloud or complicate data
collected.
2. Participants may not be motivated to perform at their optimal level when taking pre-test and post-tests.
3. Participants will become aware that the rest of
the class is not using TTS in the lab. This may cause important motivational
differences that could impact results.
Solution: To address perceived differences and minimize this confounding
variable, the rest of the class will also use TTS software after the four-week
study is complete.
4. Consider the impact computers have on motivating
the user(s) to learn and how that may affect outcomes. In other words, just
being able to use a computer may positively motivate user(s) to the extent that
it positively affects reading fluency and comprehension.
5. Parental permission must be obtained from
parent(s) or guardians of the experimental group. Consider some parents may
decline.
6. Assumes all six students will not be absent during
data collection periods. I have built in flexibility since we’ll only be in the
lab three times a week and I may choose which days since the lab is free on all
five days.
Chapter 4 Results
This research sought to see whether
or not text-to-speech might be a causative factor in increasing my six
struggling students’ reading comprehension and/or fluency.
Comprehension was measured by reviewing story selection pre-test
scores prior to TTS use to post test scores after using TTS for three
successive weeks. (See Appendix B)
Fluency measurements were taken
relative to participants’ 1) reading rate (the number of words per minute a
student reads 2) reading accuracy and 3) prosody (generally also known as
expression). Reading rate and reading accuracy were calculated by listening to
participants read two early 2nd grade level passages, one passage
prior to TTS use and the other after the three-week period of TTS use. A rubric
was used as an instrument to measure prosody. (See Appendices C and D
respectively.)
To add triangulation to this data,
participants completed a survey before and after the research period that
focused on their attitudes toward reading, and themselves as readers. This data
was also compared to see what changes occurred, if any (See Appendix E). Last, I made observational notes as these six
participants used TTS in the lab, focusing primarily on their attention to the
monitor and posture, but also noting any comments they made.
Comprehension Data
The following Figure 1.1 illustrates
pre and post-test data compiled from participants’ scores received on story
selection tests. Prior to TTS use, students collectively scored a mean of 72
and a bimodal score of 70 and 80. Post TTS data from story selections showed a
mean of 95 and a bimodal of 90 and 100. These pre and post tests indicate an
increase in these six students’ comprehension and vocabulary acquisition as a
group. In simple terms, as a group, participants had a B- average prior to TTS
use and an A average after TTS use.
Pre and Post TTS Use Selection Test
Data (Figure 1.1)
Individually, these students’ test
scores also rose if we compare the pre to the post story selection test scores.
Figure 1.1 illustrates this fact. Alex’s test scores rose by 20 points, a 25.0%
increase. Mary’s by 10, a 12.5% increase, and Kirstin’s by 10, an 11.1%
increase. Paul’s scores rose by 20, a 25% increase, Edward’s by 50 – a 125%
increase and Charles’ by 30 – a 42.9% increase.
This data supports the notion that
vocabulary development is important to reading comprehension (National Reading
Panel, 2000). Repeated exposure to vocabulary through listening to these words
in context while using TTS appears to have enhanced these students’ vocabulary
development, making it easier for them to understand the story. The two
remaining areas identified by the National Reading Panel (2000) as being
important components to reading comprehension – 1) text comprehension instruction
and 2) teacher preparation and comprehension strategies instruction remained
the same throughout this study.
Fluency
As mentioned in earlier chapters,
fluency is an important factor to successful reading, and supports
comprehension. Fluency data collected prior to the use of TTS and after the
three-week period of use is shown below in Figures 2.1 through 2.7. This data
is broken down into tested areas relative to fluency - reading rate, reading
accuracy and prosody. Reading rate was calculated using the following formula:
Pre and Post TTS
According to the Scott Foresman
Individual Reading Inventory and Running Record basal, an appropriate reading
rate for a 7-year old is between 60-80 words per
minute (wpm). Figure 2.1 shows three participants – Alex,
Mary and Charles – meet the criteria for an appropriate reading rate for their
age both prior to and after TTS use.
Kirstin, Paul and Edward fell well
below the appropriate reading rate range both prior to TTS use and after. However,
all three of these participants’ reading rates increased after TTS use. Kirstin’s
pre-test score was 28wpm; her post-test score was 34wpm (an increase of 6wpm).
Edward scored 29wpm prior to TTS use and 42wpm on the post-test (an increase of
13). Paul showed the greatest gain of 16 wpm from the pre to post-test.
Curiously, Alex, Mary and Charles’
reading rate increases were less significant. These three are the better
readers of the six participants. Alex gained 4 wpm and Charles gained 5wpm.
Mary’s reading rate decreased by 6 wpm from pre to post TTS use. I can only
conjecture that Mary might have been less motivated the day of the post-test.
My weaker readers benefited more significantly in this area after using TTS.
Reading accuracy percentage scores
were calculated using the following formula:
Accuracy % =
The number
of words in a passage less errors
The number of words in the passage
Participants did not make any
significant gains in oral reading accuracy as seen in Figure 2.2 below. In
fact, Alex and Charles showed no difference while Edward’s accuracy increased
by 3%, Mary’s by 2% and Paul’s by 1%.
Most significant in this data set is Kirsten’s decrease by 5% in accuracy from
pre to post TTS use. Results indicate these participants strive for accuracy
when reading.
Pre and Post TTS Accuracy Percentage
Data (Figure 2.2)
Prosody, the third component used to measure
fluency, was scored using a rubric that considers four areas to rate
prosody: 1) expression and volume 2) phrasing 3) smoothness and 4) pace. Students receive either a 1, 2, 3, or 4 point
scale rating for each category, with 4 being the highest so that the maximum
score an individual could receive is 16 (See Appendix D for a description of
each numbered scale.). According to
Timothy Rasinsky, the rubric’s author, scores of 8 or above indicates the
student is making good progress in fluency. This rubric was used as each
participant read a passage prior to TTS use and after TTS use. These passages
were of equal length and rated by Scott Foresmann authors as early 2nd
grade content. Participants did not read the same passage twice because this
would have positively influenced their fluency. Figure 2.3 compares individual
prosody score totals for all areas prior to and after TTS use.
Figure 2.3 Individual Pre &
Post-Test Prosody Scores
It is clear all participants made
gains in prosody after TTS use. Alex, Kirstin, and Paul increased their overall
prosody by one point. Charles showed a ˝ point gain, while Mary and Edward
earned a two-point increase. It’s interesting that even though Mary’s reading
rate fell, aside from Edward, she showed the greatest prosody gain. Let’s take
a look at individual areas where gains were made.
Figures 2.4 through 2.7 break down
the prosody rubric into the four areas used to rate prosody in this case –
expression and volume, phrasing, smoothness, and pace. Again, these graphs
illustrate pre and post ratings for each area.
Figure 2.4 Individual Pre &
Post-Test Expression & Volume Scores
A reader receiving four points in
this area reads with good expression and enthusiasm throughout the text.
His/her reading sounds natural. S/he is able to vary expression and volume to
match the interpretation of the passage. None of the participants received four
points.
Figure 2.4 (above) indicates Mary’s
and Edward’s expression and volume improved by one point from pre to post TTS
use. This accounts for one of the two points each gained in prosody overall.
Charles’ overall increase in prosody is accounted for here. Alex and Paul made
no gains in this area.
Data increases in volume and
expression are most significant out of all four dimensions of prosody rated
because expressive reading can indicate the reader may also comprehend more of
what s/he reads. Certainly, the fact that Mary’s, Edward’s and Charles’
comprehension scores on the selection tests also rose lends some validity to
this statement. However, it’s harder to find a consistent proportional
correlation between expression and comprehension data if we compare their
scores.
For example, while both Mary and
Edward increased their expression and volume rating by 1 point, Edward’s pre
and post test selection scores increased by 125%. We might expect Mary’s to
have increased by the same amount when in fact she scored a 12.5% increase.
Charles increased his expression and volume by one-half point and his
comprehension score by 42.9%. Still, increases in volume when reading may
indicate Mary, Edward and Charles are becoming more confident readers. In
retrospect, I would revise this rubric by separating expression from volume
because while Edward gained a full point in this area, it had more to do with
increases in volume and less with improved expression.
The following Phrasing Rating chart (Figure
2.5) shows participants’ individual pre and post data. A four-point phrasing
score defines a reader as one whose reading is generally well-phrased, mostly
in clause and sentence units, with adequate attention to expression.
Figure 2.5 Individual Pre &
Post-Test Phrasing Scores
In looking at the data above, both
Alex and Mary attained this goal post TTS use. Their accomplishments account
for the one-third increase in phrasing as a group. The remaining four
participants showed no gains in this area.
Smoothness is another dimension of
prosody. A four-point rating indicates a participant reads smoothly with some
breaks, but word and structure difficulties are resolved quickly, usually
through self-correction.
Figure 2.6 Individual Pre &
Post-Test Smoothness Scores
In looking at Figure 2.6 above, Alex
was the only participant that earned four points, and it was before TTS
use. However, post-TTS use he received a three-point rating. It is possible
Alex’s attentional issues affected his performance, but no notations were made
to this effect during the post reading. Mary, Paul, Charles and Edward showed
no gains in smoothness, but Kirsten went from a one-point rating to two points,
a gain of one. This is the only area in which Kirsten improved. This is
reflected in her improved reading rate score. It took Kirsten 4:17 to read the
first passage and 3:27 to read the second passage. However, the increase in
smoothness was also due to more word substitutions and fewer self-corrections
or sounding outs, resulting in a two-point rating in all prosody areas.
Pace pre and post TTS scores are
shown below (Figure 2.7):
Figure 2.7 Individual Pre &
Post-Test Pace Rating
Pace can be rated as slow and
laborious (1 pt.), moderately slow (2 pts.), an uneven mixture of fast and slow
reading (3 pts.), or consistently conversational (4 pts.). Alex, Mary, Edward and Charles’ pace remained
the same pre and post TTS use. Alex achieved the highest pace score, while Mary
and Charles’ pace remained at 3 points.
Most significant in this data set is
that both Paul and Edward’s pace increased by one point. Paul’s passage reading
prior to TTS use was timed at 4:10, and 2:40 after the three-week period of
using TTS. Edward’s time was similar to Paul’s, at 4:01 and 2:48, respectively.
Their accuracy while reading both passages remained high. Edward scored a 92%
accuracy and Paul 95%.
In summary, at least one participant
showed an increase in each of the four areas rated in prosody. However, more
participants made gains in expression/volume than any other area. Pace (Paul
and Edward gained 1 point each) and phrasing (Alex and Mary gained 1 point
each) gains were equal, while Mary earned an extra point in smoothness and Alex
lost one point. Overall, it appears synthetic TTS positively affected these
students’ prosody.
Of further interest is qualitative
data compiled while the six participants used synthetic TTS. Background, text
and highlight color remained the same throughout the research period.
Participants saw black text against a white background. Words were highlighted
in black, word-by-word. These were the only options for the trial software used
(Intellitalk III). Originally, I intended to use Write:OutLoud, which would
have allowed color customizations of the above, however the software version
available was not compatible with the 2000 NTs in the lab. Intellitalk III is a
comparable, synthetic TTS software program that offers similar features in the
full version to that of Write:OutLoud. Limitations with using the Intellitalk
demo version did not affect the purpose of this study.
The four male participants chose one
of the male voices (Mike) and the two females chose “Mary”. All participants
preferred a reading rate of 2 which scanned text at a natural speaking rate.
Observable differences noted were relative to the posture and attention span
while using synthetic TTS and any comments that were made.
Observational Data - Weeks 1, 2and 3 of TTS Use
Throughout the three-week period of
TTS use, I gave my students very little cuing as to how they should sit or the
amount of attention they should pay to the monitor. I was more interested in
their natural responses to TTS. Of course, just my presence as I neared them
would influence them to some degree. Therefore, I often stood away from them
unless they needed assistance.
During the first week of TTS use, Charles sat
back in his chair and was more concerned about the speed at which the words on
his monitor were moving compared to Paul’s. When I adjusted Charles’ level to
two, he turned to Paul and said, “I’m going to beat you!” Later he asked for
the male voice. In subsequent weeks, Charles’ attention span varied. He would
start off sitting upright with good attention to the monitor and then his
attention and posture would deteriorate after 10 to 15 minutes. In the third
week, Charles stopped the scanning on purpose. Paul started it up again for
him. Charles then asked me, “How many times do we need to listen to the story?”
Paul assumed a lax posture throughout
the three weeks of TTS use, slouching in his chair much of the time. In Week 1
he wanted the scanning to go faster and also asked for the male voice. I
increased scanning to level two and it remained there for Weeks 2 and 3. His
overall attention to the screen varied. He chatted with Charles intermittently.
Occasionally, he assisted Charles with restarting the scanning as Charles was
having some difficulty with the software.
Alex sat upright for most of the
period and was attentive to the monitor. However, he managed to change the
fonts while I wasn’t looking. (The full version has management locks.) During
the second week he played with the mouse cord, talked while “listening” and
needed redirection. He pulled his knees up against the table and leaned against
it. At one point it looked like he was reading along with the computer. (This
distracted behavior is similar to Alex’s regular classroom demeanor.)
Mary sat upright her first week of
TTS use and was extremely attentive to the monitor. Her eyes appeared to be
following the highlighted words. However, after a 15-minute session she asked,
“Can I be done?” Her posture and attentiveness remained in Week 2 as she sat
with her head in her hands, elbows on the table and her eyes on the screen. In
Week 3 she began leaning back in the chair and later asked again if she could
be done after 15 minutes or so of using TTS.
Kirsten’s visual attention to the
monitor was poor throughout the three weeks. During the first week she watched
the monitor initially, and then closed her eyes and assumed a laid back posture
after 15 minutes. The second week revealed equally bad posture as she leaned
back with her knees against the table, however her attention to the screen
improved. Occasionally she would look over at Mary. In Week 3 Kirsten did not take
advantage of the scanning but sat upright; she hardly looked at the screen. She
watched Mary instead.
Edward sat with his elbow under his
chin, then laid back in his chair after ten minutes of watching the screen
attentively. I wondered about adjusting the time, but wanted to adhere to the
three, 30-minute per week sessions proposed in the study. In Week 2 he appeared
to watch the monitor attentively as he sat forward. After ten minutes he sat
back and slouched, but remained fairly attentive. He asked me, “How many times
do I need to listen to this?” Prior to entering the lab in Week 3, Edward
asked, “Are we going to listen to the story in the lab today?” I asked him why
he wanted to know. He said, “Because I want to do the sentences.” The rest of
the students in my class wrote sentences in Microsoft Word while these six
participants listened to TTS. I allowed study participants to do some of this
after the half hour of TTS. During this last session, Edward’s knees came up
(he does this in his regular desk too), however his attention to the monitor
continued.
Survey Data
All six
students completed a survey before and after TTS use. This measuring instrument
was used to determine whether or not reading attitudes would change after using
TTS. Students were encouraged to answer honestly. Students either answered
“yes” or “no” to each question. Two points were assigned for each “yes” answer,
and one point was given for each “no” answer. Tabulated group results are shown
in Table 1 below.
Survey Data (Table 1)
|
Questions |
Pre-Test |
Post-Test |
Difference |
|
1. I like to read out loud. |
8 |
8 |
0 |
|
2. I like to read by myself. |
8 |
11 |
3 |
|
3. I like to read. |
12 |
10 |
2 |
|
4. I like it when someone reads to
me. |
12 |
12 |
0 |
|
5. It is easy for me to sound out
words. |
9 |
9 |
0 |
|
6. I understand what I read. |
10 |
11 |
1 |
|
7. I can figure out new words when I
read. |
9 |
10 |
1 |
|
8. I understand more when someone
reads to me. |
10 |
11 |
1 |
|
9. I read for fun. |
9 |
10 |
1 |
The most positive difference in
reading attitudes was evident in Question 2: “I like to read by myself”. This
was very encouraging since it is much harder to read independently than it is
to be read to.
The data in Table 2 is broken down into “yes” and “no” answers as
well to better calculate how many participants changed their attitudes, whether
they went from positive to negative, or negative to positive, and who was
responsible for those changes.
Survey Data (Table 2)
|
Questions |
Pre-Test |
Post-Test |
Participant |
|
1. I like to read out loud. |
2 yes, 4 no |
2 yes, 4 no |
n/a |
|
2. I like to read by myself. |
2 yes, 4 no |
5 yes, 1 no |
Paul, Charles, Kirsten |
|
3. I like to read. |
6 yes |
4 yes, 2 no |
Paul, Edward |
|
4. I like it when someone reads to
me. |
6 yes |
6 yes |
n/a |
|
5. It is easy for me to sound out
words. |
3 yes 3 no |
3 yes, 3 no |
n/a |
|
6. I understand what I read. |
4 yes, 2 no |
5 yes, 1 no |
Mary |
|
7. I can figure out new words when I
read. |
3 yes, 3 no |
4 yes, 2 no |
Mary |
|
8. I understand more when someone
reads to me. |
4 yes, 2 no |
5 yes, 1 no |
Mary |
|
9. I read for fun. |
3 yes, 3 no |
4 yes, 2 no |
Charles |
Charles, Kirsten and Paul positively
changed their attitudes toward independent reading. However, Paul also decided
he didn’t like to read after using TTS, as did Edward.
Mary showed the most positive number
of attitude changes as shown in Questions 6, 7 and 8. Post TTS use, Mary reported
she can understand what she reads, can figure out new words when reading and
that she understands more when someone reads to her.
Post TTS, Charles was the only participant who changed his
attitude and said he does read for fun.
Though
students did not indicate an attitude change in every area and a few moved from
positive to negative answers, it is still encouraging to see that some of the
students had more optimistic views about reading after TTS use.
The last question in the survey asked
these six students to show how many times a week they read for fun, with
possibilities ranging from 1 to 7, or more. Mary changed the number of times
she reads for fun in a week from 1 or 2 times a week to 4 times a week. Charles
changed his number from a “3” to a “4” times a week. There were no changes in
the remaining participants’ responses.
Out of curiosity, I asked
participants to write a sentence about how they felt about using TTS. I hadn’t
planned to do this, but was curious.. Alex said, “I really liked it.” Paul
reported, “It was okay.” Mary responded, “Good because I like stories. I liked
it!” Charles said, “I really liked it.” Kirsten shared, “I think it is fun. I
got to sit next to Mary. Thank you!” Edward stated, “I did not like it.”
It was interesting that Edward
appeared to like synthetic TTS the least but he appears to have benefited from
it the most. His selection test scores increased 125%, his reading rate
improved by 13 wpm and his prosody score went up a full point post TTS. It may
be that TTS users don’t have to like using TTS to benefit from it.
Conclusions
Available research studies regarding
factors that affect fluency and comprehension revealed that guided and repeated
oral reading assists readers with
vocabulary recall, which leads to better fluency (Kuhn & Stahl, 2000; The
National Reading Panel, 2000) and that reading comprehension relies on
vocabulary skill development 2) text comprehension instruction and 3) teacher
preparation and comprehension strategies instruction. I continued to deliver
the same comprehension instruction and did not acquire or use any new
comprehension strategies during the research period. The sole independent
variable used during the research period was synthetic text-to-speech. It was
hypothesized that by exposing these six students to repeated, guided oral
readings, students might increase their comprehension and fluency.
Overall findings from this research
show synthetic TTS was beneficial to my six struggling readers. Intellitalk’s
synthetic TTS software helped all six participants elevate their selection test
scores, therein indicating increases in both vocabulary acquisition and
comprehension individually, and as a group.
It is possible that pre and post test
story content affected participants’ interest and motivational levels. It is
also reasonable to assume that gains in sight word recognition would be
achieved over time, however it is not certain that four weeks is enough time to
account for these significant gains.
Fluency also improved. Reading rates
rose significantly for my three weakest readers (Kirstin, Paul and Edward),
though much less so for the three more successful readers. Accuracy percentages
remained relatively stagnant for all participants, however these numbers were
already high. Each participant’s prosody improved in at least one of the four
areas, even though the synthetic voices were less expressive (more robotic)
than digitized speech.
Reading attitudes also changed in
mostly positive ways, the biggest change being that more participants decided
they liked reading independently after TTS use. However, it seems that
attitudes are harder to measure since the researcher must rely on the
participant(s)’ honesty, motivation, and self-assessment criteria on any given
day to be consistent.
Implications for Teaching
In a perfect world, every
student in my classroom would have a computer loaded with TTS so that all could
benefit from this assistive technology. I will definitely include synthetic TTS
as a tool to potentially increase my students’ comprehension, vocabulary
acquisition, reading rate and prosody. Though my research model addresses a
small population of struggling readers, the gains made in the above areas
warrant adding TTS to the strategies I currently use to improve these skills.
However, in light of some of the
observational data compiled while students used TTS, I would reduce the time my
students use TTS to two, 15-minute periods per week at most. I would also
experiment with the font, highlight and background colors, if available, to see
if color might be influential.
The fact that synthetic TTS is
relatively inexpensive, easy for students and teachers to use, and is easily
integrated into the curriculum means there is a good possibility that schools
may opt to use it. However, few schools will purchase this software without
further research that might validate my findings. Still, I will recommend that
my colleagues try synthetic TTS and I will offer to train them.
Implications for Further Research
The success of synthetic TTS with my
small student population should encourage more field research in this area. I
recommend the following to provide more conclusive evidence of TTS as a
causative, positive independent variable that impacts dependent variables of
comprehension and fluency so that results might be generalized: 1) Increase the
research time span to provide more conclusive data 2) Increase the participant
population numbers and use random samples 3) Set up a control vs. test group to
increase reliability 4) More research using larger populations who have reading
deficiencies versus those who don’t. 5) More research comparing digitized to
the best synthesized TTS products 6) More research using this assistive tool
along with the color options to determine whether or not color also positively
impacts comprehension and/or fluency.
Further research in this area that
results in positive outcomes would encourage school districts to support the
purchase of TTS software and encourage teachers to integrate this technology
into their curriculum.
In terms of the measurement tools
used in this research, I would recommend taping the participants as they read
to allow the researcher to better compare the pre and post readings when rating
prosody according to the rubric. It is difficult to perform a running record
and rate prosody simultaneously. As mentioned earlier, I would also separate
expression from volume in the rubric, since it’s possible to read with
increased volume but with the same or less expression and still receive a rate
increase.
I also recommend using a Likert scale
of three or more number values for answers to survey questions (as opposed to
yes/no answers) so that standard deviation measurements can be calculated to
better reflect differences in attitudes, if any.
Last, it is clear that funding plays
a major role in research and, without it, research is sometimes limited, as it
was in this case. The trial software came with fewer features than the full
version, but it was free. Comprehension and fluency gains may have been more
significant had participants had access to some better synthesized TTS voices
currently on the market, such as the AT&T True Voices, or digitized speech
(which is more expensive).
References
.
.
Bacon, D. (2004). The contributions of reliability and pretests to effective assessment. Practical Assessment, Research & Evaluation, 9(3). Retrieved July 1, 2004, from PAREonline: http://pareonline.net/getvn.asp?v=3&n=4
Badger, E., & Thomas, B. (1992). Open-ended questions in reading. Practical Assessment Research & Evaluation, 3(4). Retrieved July 1, 2004, from PAREonline: http://pareonline.net/getvn.asp?v=3&n=4
Binder, C. (1996). Behavioral Analyst. Behavior Analyst, 19, 163-197.
Brualdi,
A. C. (1995). Multiple intelligences:
Clark, J. M., & Paivio, A. (1991). Dual coding theory and education. Educational Psychology Review, 3, 149-210.
Deno, S.L., Mirkin, P., & Chaing, B. (1982). Identifying valid measures of reading. Exceptional Children, 52, 219-232.
Dowhower, S. L. (1991). Speaking of prosody: Fluency's unattended bedfellow. Theory Into Practice, 30, 165-174.
Farmer, M. E., Klein, R., & Bryson, S. E. (1992). Computer-assisted reading: Effects of whole-word feedback on fluency and comprehension in readers with severe disabilities. Remedial & Special Education, 13(2), 50-60.
Fuchs, L.S., Fuchs, D. & Maxwell, L. (1988). The validity of informal reading comprehensions measures. Remedial and Special Education, 9, 20-28
Gribbons, B. M., & Herman, J. (1997). True and quasi-experimental designs. Practical Assessment, Research & Evaluation, 5(14). Retrieved July 2, 2004, from PAREonline: http://PAREonline.net/getvn.asp?v=58&n=14
Intellitools, Inc. (2003, August). Intellitools awarded key grant to develop reading program for grades 4-8 students [Product Spotlights]. Closing The Gap, 22(3), 9.
Jackson, J., Harper, K., Ruzic, R., & O'Connell, K. (1995). Text to speech summary source data. Retrieved July 3, 2004, from CAST: http://www.cast.org/ncac/index.cfm?i=1665#s12
Kincaid, J. P. (2000). Linguistic cues and memory for synthetic and natural speech. Human Factors. Retrieved July 1, 2004, from High Beam Research: http://www.highbeam.com
Kuehner, A. V. (1999). The effects of computer instruction on college students' reading skills. Journal of College Reading and Learning. Retrieved July 1, 2004, from High Beam Research: http://www.highbeam.com
Kuhn,
M. R., & Stahl, S. A. (2000). Fluency: A review of developmental and
remedial Practices [CIERA Report#2-008]. Center for the Improvement of Early
LaBerge D. & Samuels, S.J. (1974). Toward a theory of automatic information processing in reading. Cognitive Psychology, 6, 293-323.
McCullough, C. S. (1995). Using computer technology to monitor student progress and
remediate reading problems. School Psychology Review, 24(3), 426-439.
Markell, M. A., & Deno, Stanley L. (1997). Effects of increasing oral reading: Generalization across reading tasks. Journal of Special Education. Retrieved July 2, 2004, from HighBeam Research: http://www.highbeam.com
National
Institute for Child Health & Human Development. (2000, April). Computer
technology and reading instruction. In Report of the National
Quirk, C. (2000). How do Presentation Modality and strategy use influence memory for paired concepts? Journal of Instructional Psychology. Retrieved July 2, 2004, from HighBeam Research: http://www.highbeam.com
Silver, C. N. (1995). Effect of synthetic voice intelligibility on speech comprehension. Human Factors. Retrieved July 2, 2004, from HighBeam Research: http://www.highbeam.com
Snow,
C.E., Burns, M.S., &
Stine, E. L., & Wingfield, A. (1987). Process and strategy in memory for speech among younger and older adults [Special issue]. Psychology and Aging, 2 272-279.
Vygotsky,
L. S. (1978). Mind in society: The development of higher psychological
processes.
Appendices
Appendix A:
Pre TTS
Week 1:
Wanted: Best Friend by A.M.
Monson
Week 2:
Play Ball
by Stephen Krensky
Week 3:
People, People Everywhere! By Nancy Van Laan
Week 4:
The Best Older Sister by Sook Nyui Choi
Appendix B:
Pre TTS
Wanted: Best
Friend
by A.M. Monson
Part I: Vocabulary
Find the word that best fits in each sentence. Mark
the space for your answer.
1. I get to stay up late ___________________.
O sometimes O around O either
2. Josie ________________ that she felt sick.
O dumped O complained O returned
3. Mel rode her bike _______________ me.
O since O across O toward
4. Dad walked ___________ the street.
O since O across O sometimes
5. You can ______________ come with us or stay home.
O best O once O either
Part 2: Comprehension:
Read each question. Mark the space for your answer.
6. Why did
Mouse go home?
O Cat did not
want to play crazy eights.
O Cat had a
new friend.
O Mouse was
tired of games.
1. Cat called The Hollow Log Gazette because he wanted
to ______.
O buy a
newspaper.
O try to find
a friend.
O buy a new
game.
2. Who made the biggest mess in Cat’s house?
O Raccoon
O Mouse
O Mole
3. When Cat and Mouse play crazy eights, Cat will
__________
O try to find
a new friend.
O ask Otter
to come and watch.
O be nicer to
Mouse.
4. What did Cat learn in this story?
O Mouse would
never come back again.
O Mouse was
his best friend after all.
O He could
not beat Mouse at checkers.
The Best Older
Sister
by Sook Nyui Choi
Part I: Vocabulary
Find the word that best fits in each sentence. Mark
the space for your answer.
1. My birthday was a ___________________ day.
O guess O next O special
2. Amy feels it is ________________ to help people.
O across O important O different
3. Babies need lots of _______________.
O attention O creature O thumb
4. We saw a very ___________ show about ducklings.
O ago O interesting O secretly
5. Peggy ______________ a funny joke.
O giggled O lose O told
Part 2: Comprehension:
Read each question. Mark the space for your answer.
6. After
Kiju was born, Sunhi missed her time with _________
O Robin.
O Halmoni.
O Jenny.
7. At first, Sunhi thinks that ______.
O everyone
likes Kiju better than her.
O she is the
best older sister.
O it would be
great to have more babies.
8. Which sentence tells what this story is about?
O Sunhi told
Halmoni about her day at school.
O Kiju always
made such a mess.
O Everything
changed for Sunhi when Kiju was born.
9. Why did Halmoni give Sunhi her surprise
early?
O She could
not keep it a secret.
O She wanted
Sunhi to know that she loved her.
O She was
afraid that Sunhi would find it.
10. In this story, Sunhi learned that _____
O Halmoni is
not her friend.
O it is
better to be an only child.
O everyone is
special.
Appendix C
Passage
1
Kate’s family was going to be in the big parade. Her
sisters were going to march and play music. Her mom and dad would ride on
horses. Everybody but Kate would be in the parade. Kate wondered, “How can I be
in the parade too?”
Kate saw a boy ride by on a bike. At once Kate knew
what to do. Kate worked hard on her bike. She put red and green paper on it.
She made it look pretty and bright. In the parade, many children had bikes.
Kate lined up with the children. When the music started, the parade moved
through the city. People waved to Kate. She had a wonderful time riding in the
parade.
|
Oral Number of Words:
118 |
|
|
Number of Errors |
Percent |
|
1 |
99 |
|
2 |
98 |
|
3-4 |
97 |
|
5 |
96 |
|
6 |
95 |
|
7 |
94 |
|
8 |
93 |
|
9-10 |
92 |
|
11 |
91 |
|
12 |
90 |
Passage 2
Judy’s aunt had a wonderful job. She worked as a
clown. Judy liked spending time with Aunt Barb to learn what a clown does.
Aunt Barb said, “First you must look like a clown.”
Aunt Barb put on a funny costume. She put on white
makeup and a fat, red nose. Then Judy helped Aunt Barb put on a funny wig. They
both laughed. Judy liked Aunt Barb’s clown costume. Judy wanted to look like a
clown too.
Mom gave Judy an old dress and a wig. Then she helped
Judy put on some makeup. Dad gave Judy some of his old shoes. They looked big
and funny on her. Mom and Dad laughed at Judy the clown!
|
Oral Number of Words:
118 |
|
|
Number of Errors |
Percent |
|
1 |
99 |
|
2 |
98 |
|
3-4 |
97 |
|
5 |
96 |
|
6 |
95 |
|
7 |
94 |
|
8 |
93 |
|
9-10 |
92 |
|
11 |
91 |
|
12 |
90 |
Appendix D
Figure 4
Multidimensional Fluency Scale
Use the following scales to rate reader fluency on he
dimensions of expression and volume, phrasing, smoothness, and pace. Scores
range from 4 to 16. Generally, scores below 8 indicate that fluency may be a
concern. Scores of 8 or above indicate that the student is making good progress
in fluency.
|
Dimension |
1 |
2 |
3 |
4 |
|
A. Expression and Volume |
Reads with little expression or
enthusiasm in voice. Reads words as if simply to get them out. Little sense
of trying to make text sound like natural language. Tends to read in a quiet
voice. |
Some expression. Begins to use
voice to make text sound like natural language in some areas of the text, but
not others. Focus remains largely on saying the words. Still reads in a quiet
voice. |
Sounds like natural language
throughout the better part of the passage. Occasionally slips into
expressionless reading. Voice volume is generally appropriate throughout the
text. |
Reads with good expression and enthusiasm
throughout the text. Sounds like natural language. The reader is able to vary
expression and volume to match his/her interpretation of the passage. |
|
B. Phrasing |
Monotonic with little sense of
phrase boundaries, frequent word-by-word reading. |
Frequent two- and three-word
phrases giving the impression of choppy reading; improper stress and
intonation that fail to mark ends of sentences and clauses. |
Mixture of run-ons, mid-sentence
pauses for breath, and possibly some choppiness; reasonable stress/intonation. |
Generally well phrased, mostly in
clause and sentence units, with adequate attention to expression. |
|
C. Smoothness |
Frequent extended pauses,
hesitations, false starts, sound-outs, repetitions, and/or multiple attempts. |
Several “rough spots” in text
where extended pauses, hesitations, etc., are more frequent and disruptive. |
Occasional breaks in smoothness
caused by difficulties with specific words and/or structures. |
Generally smooth reading with some
breaks, but word and structure difficulties are resolved quickly, usually
through self-correction. |
|
D. Pace (during |
Slow and laborious. |
Moderately slow. |
Uneven mixture of fast and slow reading. |
Consistently conversational. |
Source: Adapted from “Training
Teachers to Attend to Their Students’ Oral
(Permission to use this rubric has been received. See next page)
From:
"Tim Rasinski" <trasinsk@kent.edu>
To:
"
Subject:
Re: Permission to use your multidimensional fluency rubric
Date:
Sunday, November 14, 2004 1:47 PM
Lori: You have my permission to use the fluency
scale mentioned below in
the way
described in your email below.
Sincerely,
timothy
rasinski
At
08:26 AM 11/6/2004 -0500, you wrote:
>Hello,
>
>I
found your exceptional prosody rubric scale that rates multidimensional
>fluency
(figure 4 in your article on "Assessing Reading Fluency") and am
>seeking
your permission to use it in a study I'm doing for my thesis at
>
>
>My
study is about sythesized text to speech and I want to see if it will
>help
the struggling readers in my classroom who have comprehension and
>fluency
difficulties.
>
>Please
let me know if I can include and use this rubric in my study. I
>need
to document that I asked and received permission to use it. Thank you.
>
>Lori
Kleinen
Timothy
Rasinski
404
White Hall
330-672-0649
trasinsk@kent.edu
Appendix E:
Reading Survey
Directions:
Circle which answer best fits how you feel.
1. I
like to read out loud. Yes No
2. I
like to read by myself. Yes No
3. I
like to read. Yes No
4. I
like it when someone reads to me. Yes No
5. It is
easy for me to sound out words. Yes No
6.
I understand
what I read. Yes No
7.
I can figure out
new words when I read. Yes No
8.
I understand
more when someone reads to me. Yes No
9.
I read for fun. Yes No
10.
How many times a
week do you read for fun? 1 2
3 4 5
6 7 or more
Appendix
F:
Observational/Anecdotal Data
Participant:
|
Choices: |
Week 1 |
Week 2 |
Week 3 |
Week 4 |
|
Background, Text and Highlight color |
|
|
|
|
|
Voice |
|
|
|
|
|
Speaking rate |
|
|
|
|
|
Read by letter Word, sentence Or paragraph |
|
|
|
|
|
Posture during Use of TTS |
|
|
|
|
|
Attention span |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Other
observations:
Week 1
Week 2:
Week 3:
Week 4:
Effect of TTS 41
Appendix G.
INFORMED CONSENT DOCUMENT
Project Title: The Effect
of Text-to-Speech On Fluency And Comprehension
Investigator: Lori Kleinen
You are being asked to
consent to your child’s participation in a project conducted through the course
of my education at Bridgewater State College. The College requires your signed
agreement for your child to participate in this project.
As the investigator, I will
explain the purpose of this project in detail, the procedures I will use and
any potential benefits and or risks in participating in this project. Below is
a brief outline of the project. After reading it, please feel free to
ask me any questions you may have. If you want your child to participate in
this project, please sign on the last page of this form. You will be given a
copy of this form for your records.
1. Nature and Purpose of the Project: To examine whether or not text-to- speech (TTS) will increase your child's reading fluency and/or comprehension. This type of assistive software reads text aloud as words are highlighted across a computer screen.
2. Explanation of the Procedures: Your child will use TTS in a computer lab three days a week for a half-hour period over four weeks beginning January 3, 2005. They will listen to the current week's story using the software. Data will be collected through fluency tests, selection tests from the stories we read regularly, anecdotal observations, and a survey that questions them about their attitudes about reading before and after text-to-speech use.
3. Discomfort and/or Risks: Risks or discomforts are not anticipated. Your child will not lose any instructional time due to participation in this study.
4. Benefits: Text-to-speech may increase your child's reading fluency and/or comprehension. S/he will both see and hear the stories read as each word or sentence is highlighted on the monitor. Increases in either or both areas may positively affect your child's learning outcomes across the curriculum.
5. Confidentiality: To respect confidentiality, participants’ names will be changed to protect the individual's identity, as will their school name.
6. Refusal/Withdrawal: Refusal or withdrawal from this project will be respected without question. No additional data will be collected from those withdrawing from the project.
I understand also that it is not possible to identify
all potential risks in an experimental procedure, and I believe that reasonable
safeguards have been taken to minimize both the known and potential, but
unknown risks.
Parent/Guardian Signature Date
Witness Signature Date
Appendix H: Administration Letter
November 27, 2004
Dear Principal _____:
I am asking your permission to conduct a study about
the benefits of using text-to-speech. Text to speech is software that
speaks/reads words as they are highlighted across a computer screen. This study
is part of a research project I am doing for my Masters in Instructional
Technology Program through Bridgewater State College. Specifically, I want to
know if text to speech will help six of my students increase their fluency and
reading comprehension.
This study will last four weeks, beginning in January
3, 2005. During this time, my six Title I students will use text to speech
three times a week in the computer lab. They will not lose any instructional
time due to participation in this study. In fact, it is hoped that they will
benefit from it. Results from this study will be forwarded to you.
I would be using the computer lab for this purpose. I
would load text to speech software on six of the computers in the lab. I would
need headphones. The computer technician has indicated these are available. I
would use the lab after two o’clock three days a week. Currently, the lab is
available throughout the week at this time.
Please complete the section below to indicate your
preference. In advance, I thank you for your continued support. Sincerely,
Lori Kleinen
Yes, I will allow you to conduct this study.
Signature:____________________________________________(Date)
_______________
No, I will not allow you to conduct this study.
Signature: ______________________________________(Date)
__________________