the nature of phonological and phonemic … · web viewthe remaining ten (of 37) questions on the...
TRANSCRIPT
1
THE NATURE OF PHONOLOGICAL AND PHONEMIC AWARENESS
IN LITERATE ADULTS
by
Susan Beth Lorenson
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF LINGUISTICS
In Partial Fulfillment of the RequirementsFor the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
2 0 0 4
2
(final examining committee approval form to go here)
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.
SIGNED: ________________________________
4
ACKNOWLEDGEMENTS
5
DEDICATION
6
TABLE OF CONTENTS
1.0 Introduction
1.1 Thesis.................................................................................................p.
1.2 Phonemic awareness: definition and issues
1.3. Phonemic awareness and adults1.3.1 Theory 1: The “no-relationship” relationship1.3.2. Theory 2: Phonemic awareness skills play a negligible role in reading1.3.3. Theory 3: Phonemic awareness plays a crucial role in adult reading
1.4. Thesis and motivation1.4.1. Overview of experiments1.4.2. Theoretical considerations1.4.3. Pragmatic considerations
1.5. Organizational structure of the dissertation
2.0 Previous Research on Phonemic Awareness
2.1 Phonemic awareness research on beginning readers2.1.1. Overview2.1.2. Reading readiness2.1.3. The predictive power of phonemic awareness2.1.4. The value of phonemic awareness training2.1.5. Levels of phonemic awareness2.1.6. Summary & implications
2.2. Phonemic awareness research on special populations2.2.1. Overview2.2.2. Phonemic awareness in illiterates2.2.3. Phonemic awareness in literates with nonalphabetic orthographies2.2.4. Phonemic awareness in dyslexics2.2.5. Phonemic awareness in Children with Down Syndrome2.2.6. Summary & implications
2.3. Phonemic awareness and secondary language activities
2.4. Special considerations: the question of orthographic interference2.4.1. The role of orthography in phonological access2.4.2. Pilot: Delete Segment Task
2.5. Summary3.0. The Experiments
7
3.1. Overview
3.2. Deletion Experiments3.2.1. Experiment One: Delete Segment Task3.2.2. Experiment Two: Delete Syllable Task3.2.2. Comparison of Experiments One and Two
3.3. Substitution Experiments3.3.1. Experiment Three: Substitute Segment Task3.3.2. Experiment Four: Substitute Syllable Task3.3.2. Comparison of Experiments Three and Four
3.4. Reversal Experiments3.4.1. Experiment Five: Reverse Segment Task3.4.2. Experiment Six: Reverse Syllable Task3.4.3. Comparison of Experiments Five and Six
3.5. Summary of Experiments
4.0. Conclusion
LIST OF ILLUSTRATIONS
LIST OF TABLES
ABSTRACT
11
1.0 Introduction
1.1 Thesis
This dissertation is an examination of phonemic awareness in literate adults. Phonemic
awareness, the ability to identify and manipulate the sound segments in a word, has long
been associated with reading ability in children. Myriad studies on children’s phonemic
awareness in the last two decades have consistently shown that good readers possess a
high level of phonemic awareness, whereas poor readers do not. As Mann puts it,
"phoneme awareness bears both a logical and a proven relationship to early reading
success. Its presence is a hallmark of good readers, its deficiency one of the more
consistent characteristics of poor readers" (Mann, 1991, p. 260).
Little is known about the relationship between phonemic awareness and reading ability in
literate adults, however, since no studies have previously been conducted on the subject.
This dissertation reports the results of a study on phonemic awareness in adults, which
was modeled on past phonemic awareness studies with children. The results of the study
indicate that adults, like children, exhibit differing degrees of phonemic awareness, and
that these levels of phonemic awareness are associated with reading ability. The
implications of these findings, both practical and theoretical, are discussed.
An introduction to phonemic awareness is given below, followed by a section on the
possible relationships between phonemic awareness and reading ability in adults, an
expanded thesis statement with an outline of the experiments to be conducted and the
12
motivations for the study, and an overview of the organizational structure of the
dissertation.
1.2 Phonemic awareness: definition and issues
In this section, phonemic awareness is defined, and some of the most salient issues in
previous phonemic awareness research are reviewed.
Phonemic awareness (which shall alternately be referred to in this dissertation by the
abbreviation “PA”) is defined by Stanovich as “the ability to deal explicitly and
segmentally with sound units smaller than the syllable" (Stanovich, 1993, p. 283). That
is, phonemic awareness is both the knowledge that the word cat consists of three distinct
phonemes ([k..æ...t] and not just one inseparable syllable, [kæt]), and the ability to isolate
those phonemes on command (such as stripping the first phoneme off cat and producing
[æt]).
The term "phonemic awareness" has acquired quite a specific meaning in the field of
reading education, referring to various levels of achievement on certain experimental
tasks that are used as diagnostics of reading ability. For the last two decades, the
connection between phonemic awareness and reading ability has been poked, prodded
and otherwise examined by researchers in the fields of Linguistics, Psychology, and
Education.
13
In the last few years, however, phonemic awareness has become something of a hot
topic, and the term “phonemic awareness” has migrated from the jargon of Linguistics
and Educational Psychology into the mass media. It has even become a buzz phrase, a
selling point for a number of products and services. Consider the excerpts on the
following page, which illustrate the marketability of phonemic awareness:
New York Times article on the National Institute of Child Health and Human Development at the NIH, October 27, 1996
Auditory Discrimination in Depth. We offer three to five day Lindamood A.D.D. workshops which develop phonemic awareness - the ability to judge sounds within words. This awareness underlies decoding and spelling . The techniques begin orally with how speech sounds are articulated, and extend into multi-syllable and contextual spelling and reading. Additional training is available and highly recommended: 1 day on the LAC Test; 3 day practicum workshop with students.
Lindamood-Bell web page
Windows to Reading is a series of programs for students and teachers which will focus on early reading and writing instruction for grades K-3. Orientation Telecast on September 30, 1996. [Programs include:]· Phonemic Awareness· Spelling and Phonics Skill Development....
Windows to Reading promotion on the Web
Reading Who? Reading You! can help children achieve these important reading objectives:· Learn letter-sound correspondence· Develop phonemic awareness...
Description of interactive software © 1996, by Software for Success
Phonemic Awareness #61105-3: Phonemic awareness is one of the best predictors of later success in reading. This one day training will enable early childhood educators to become familiar with resources and strategies to facilitate phonemic awareness in young children.
UT 1996 continuing education catalog for early childhood educators
Sesame Street, now in its 27th season, continues its literacy campaign, Let’s Read and Write! The program places additional emphasis on beginning reading skills for preschool children and delves into stories, phonemic awareness, writing sight words, and more.
May ‘96 program description for Sesame Street
How do you know if your child is learning what he must to succeed in school? Hirsch and Holdren say a good kindergarten program will......develop children’s phonemic awareness, which helps them understand that the sounds in a word can be thought of as a string of smaller, individual sounds.
Book review by Lisa Faye Kaplan of the Gannett News Service of What Your Kindergartner Needs to Know
In order to develop an understanding of the alphabetic principle, they must become familiar with letter forms and with the idea that spoken words have identifiable sounds in them -- referred to as the concept of phonemic awareness.
Houghton Mifflin web site, promoting a “a rich literacy environment filled with books” in the classroom
Chicka Chicka Boom Boom is designed to help children build pre-reading skills, said Jan Davidson, president and founder of Davidson & Associates. It’s irresistible music, rhythms, and rhymes build phonemic awareness.
Press release from Davidson & Associates about their new CD-ROM
Web pages, CD-ROMs, software, seminars: phonemic awareness has been embraced by
the mainstream educational community, clearly not just for journal articles anymore.
What explains this phenomenon?
The phonemic awareness hype is due to one simple fact: phonemic awareness skills have
been proven to have an incontestable, unmistakable connection to reading ability in
children.
Previous research on the phonemic awareness/reading ability connection will be
discussed more fully in the next chapter, but a brief overview is presented here.
The majority of work on phonemic awareness has concentrated on a number of claims:
Children must be "phonemically aware” before they are able to read an alphabetic orthography. (Adams, 1990; Ball & Blachman, 1993). The thrust of this claim is that before children are able to decipher a writing system which is based on sound-symbol correspondences, they must be able to break a word down into its individual sounds (phonemes), so that they are able to learn the relationship between these sounds and the letters (graphemes) to which they correspond.
Phonemic awareness tasks can be used to predict a child's later success or failure in reading. (Stanovich, 1993). This claim posits that a child’s ability to manipulate (substitute, delete, etc.) the phonemes in a word predicts his/her future reading ability. A child who knows that “ice” is “nice” without the first sound will be a better reader than a child who does not understand this relationship.
Phonemic awareness training can prove useful in preparing a child for reading instruction. (Lundberg et. al., 1988; Cunningham, 1990; Hatcher, Hume & Ellis, 1994; Bradley & Bryant, 1983). This claim’s proponents see a causal relationship between phonemic awareness and reading ability, and believe that children who are trained in tasks that develop phonemic awareness skills will be better readers than children who do not receive such training.
Reading facilitates phonemic awareness ability. (Tornéus, 1984). This claim is the reverse of claim (1) above, and states that phonemic awareness is not as much a prerequisite of reading or a predictor of reading ability as much as it is a fall-out from reading; those who know how to read an alphabetic orthography can perform phonemic awareness tasks because the task of reading has developed PA skill.
Needless to say, not all researchers on phonemic awareness subscribe to the same theory
of the connection between PA and reading ability. Nevertheless, those who have worked
on elucidating the claims above agree that there is a strong correlation between reading
ability and phonemic awareness. Ball & Blachman (1991) summarize the findings well:
To read or spell phonetically regular words, a child must be aware that words can be broken into phonemes and that each phoneme corresponds to a symbol (or symbols) in our orthography (Ball and Blachman, 1991, p. 63).
That is, reading in an alphabetic system is a process that involves breaking the
"alphabetic code." This decoding includes the application of grapheme-to-phoneme
conversion rules (alternately called graphophonics rules or simply phonics rules), rules
which tell the child that in the word cat, the grapheme “c” corresponds to [k], the
grapheme “a” to [æ], etc.
Although reading involves the application of phonics rules and children must be
phonemically aware to read, it is necessary to understand the important distinction
between phonemic awareness and mere phonics knowledge. The child who is
phonemically aware possesses more than an extensive battery of “x stands for y” rules.
Those [phonics] rules alone are useless unless the child is able to identify and manipulate
the phonemes of a word. The child who has learned his or her phonics rules knows that
“‘c’ stands for [k], ‘a’ stands for [æ], and ‘t’ stands for [t].” The child who is
phonemically aware knows that the isolated phonemes which these rules produce ([k],
[æ], and [t]) yield a word in combination, [kæt], which was once thought of as a single,
impenetrable unit. The child who is phonemically aware is able to effectively use
phonics rules to create and dismantle words.
The phonemic awareness/reading connection is unquestionably strong. The body of
research reviewed in the next chapter has shown that phonemic awareness is more often a
characteristic of good readers than any other measure, including IQ, analytic ability, or
even other linguistic ability (where “linguistic ability” is any language-based skill,
including, but not limited to, phonics knowledge). This last point is particularly striking.
Since phonemic awareness is more strongly associated with reading skill than mere
phonics knowledge, some have suggested that phonemic awareness should be taught,
instead of (or in addition to) teaching phonics. Phonemic awareness training is emerging
as a way to bridge the gap between the often disparate approaches to reading of phonics
instruction and whole language instruction, in which readers are encouraged to read
without relying on explicitly taught rules. It is, perhaps, grounds for a treaty in what have
been dubbed the “reading wars.”
Consider the table below, which illustrates the various qualities of traditional phonics
instruction, 70s-era “whole language” instruction, and phonemic awareness training:
Phonics Instruction
Phonemic Awareness Training
Whole Language Instruction
Graphophonic rules taught explicitlye.g., ‘c’ stands for [k]
Graphophonic rules learned without explicit instruction e.g., students are given texts with juxtaposed rhyming words like “cat” and “hat” and eventually figure out that the ‘c’/’h’ difference corresponds to the [k]/[h] difference
Explicit training in identifying the components of a word (phonemes) e.g., students are taught to “sound out” and “break down” words
Educators in the phonics camp believe that the road to reading is paved with explicit
instruction in sound/symbol correspondence rules (grapheme x stands for phoneme y).
Phonics teachers worry that the whole language approach, with its lack of explicit
instruction, leaves many children in the dark. They claim that though whole language
instruction may work for normal children who have been exposed to books from an early
age and who have already learned to make some connections between the spoken and
printed word, it fails children who do have not have this background or who have another
obstacle (e.g., dyslexia) to making sound/symbol connections on their own. The belief is
that children will best learn to read when they are given the proper tools, that is, the set of
rules which draws connections between the 26 letters of the alphabet and the sounds of
English.
Whole language adherents believe that instruction in phonics is confusing, because there
are so many exceptions to the explicitly taught rules (one can’t help thinking of the
George Bernard Shaw observation that fish could be spelled ghoti, with the gh from
laugh, the o from women, and the ti from caption). They believe that children, given
enough time and exposure to language, will figure out the rules on their own in a natural
way. For instance, if a child is repeatedly exposed to Dr. Seuss’ The Cat in the Hat, in
which the rhyming words “cat” and “hat” are juxtaposed, he or she will eventually
surmise that the letters c and h are responsible for the difference in sound between the
words. Sound/symbol correspondences are not taught, but through the development of
“personal phonics rules...[readers] come to understand the alphabetic system and will
invent ways of relating their own speech to print" (Goodman, p. 108). This “exposure to
print” method of learning allows for the learning of more complicated words. So, when a
reader sees the words face and fake juxtaposed, he or she will use context to figure out
the difference between the two words, and will subsequently create a rule to account for
the difference (k always stands for [k], but c only sometimes does). In theory, the
phonics student would either have to learn a sophisticated rule to tackle face (c stands for
[s] between two vowels), or would be completely thrown off by the c=[k] rule.
Phonemic awareness training shares features with both phonics and whole language
approaches: it is a method of explicit, linguistic instruction, and yet it is not based on the
teaching of inconsistent phonics rules. In phonemic awareness training, children are
taught to play games that foster their phonemic awareness skills. For example, a child
might be given a block for every phoneme in a word (e.g., three blocks for wake, four
blocks for cart). When the first block in a linear series is removed, the child would be
expected to pronounce the remaining word without the first phoneme (ache or art), if the
last block is removed, without the final phoneme (way or car). Note that the
manipulations in this exercise do not rely upon graphophonic rules, and in fact may
“violate” them (wake has four graphemes but three phonemes). Because PA training is
not dependent upon phonics rules, training exercises may also be performed with
preliterate children who have not yet been exposed to phonics rules.
Thus, phonemic awareness training, phonics teaching and whole language instruction are
all quite different approaches to the teaching of reading. With phonemic awareness
training as a novel middle ground, it’s no wonder that educators, parents and software
publishers have begun to jump on the bandwagon.
Despite the extensive research on and recent interest in phonemic awareness, experiments
on the subject have been limited with regard to subject pool. Experiments aimed at
understanding the connection between phonemic awareness and reading ability have been
conducted primarily with children (as in the studies cited above) and special populations,
such as illiterates (Morais et. al.., 1986; Morais et. al., 1991; Koopmans, 1987), literates
with non-alphabetic orthographies (Read, 1986; Mann, 1986; Mann, 1991; Tzeng and
Chang [in press]), dyslexics (Savin, 1972; Lecocq, 1986; Fox and Routh, 1983; Morais,
1983), and children with Down Syndrome (Cossu & Marshall, 1993). There is little, if
any, research on literate adults.1
1.3. Phonemic awareness and adults
Why has no phonemic awareness research been conducted on adults, and just what is the
relationship between phonemic awareness and reading ability in adults? Three theories
explaining the dearth of research in this area are explored in this section.
1.3.1. Theory 1: The “no-relationship” relationship
One explanation for the lack of phonemic awareness research in adults is the following:
(1) Adults, after they have attained a certain level of reading competency, turn into whole word readers. They view all words as single, impervious units and thus no longer use nor require their phonemic awareness skills in reading.
This theory has the advantage of intuitive appeal. Certainly, someone who “sounds out”
words when reading comes across as an immature, unskilled reader. Whole word reading
seems quicker and more efficient than the alternative. However, despite the controversy
in the field of phonemic awareness research (much of it lies in dispute over causal 1 In a 1996 Scientific American article on dyslexia, Dr. Sally Shaywitz (a pediatrician,
neuroscientist, and dyslexia researcher) alludes to some promising results on phonemic awareness and adults from the Connecticut Longitudinal Study. The study began in 1983 with the selection of 445 randomly selected kindergartners, and was originally devised to test, as were many studies at the time, the connection between phonemic awareness skills in preliterate and newly literate children and their future reading ability. Shaywitz was particularly interested in the phonemic awareness skills of dyslexics. When the children were in high school (age 15), she tested the connection between their ability to strip phonemes off the beginning of a word (“Say ‘block’ without the ‘buh’ sound”) and their ability to decode single words. She found that “even in high school students, phonological awareness was the best predictor of reading ability” (Shaywitz, 1996, p. 7). Although overall reading comprehension was not tested (just the ability to identify isolated words), normal readers were Shaywitz’s control group (not the focus of her study), and though one could argue that 15-year-olds are still emerging readers, this is the only study I have found that directly tested the connection between phonemic awareness and some measure of reading ability in literate, experienced readers, and the results are quite promising.
connections, i.e. is phonemic awareness a prerequisite for alphabetic literacy, or does
reading an alphabet foster phonemic awareness?), whether or not adults are phonemically
aware is not up for debate. Even the most dyed-in-the-wool whole language advocate,
who denies the benefits of phonemic awareness training as part of reading instruction,
would concede that all alphabetic literates are phonemically aware, and must remain
phonemically aware.
Moreover, competent, literate adult readers constantly encounter situations that require
the novel implementation of phonemic awareness skills. Consider the following
situations:
Unknown words. Example: A newscaster, seeing the name Boutros Boutros-Ghali for the first time, has no choice but to “sound out” the name, this implementing phonemic awareness skills and the alphabetic code.
Known words unfamiliar in print form. Example: A reader comes across the word imbecile for the first time in print, and realizes upon sounding it out that it is the word he recognizes in spoken form ([imbesel].
Partial words. Example: A crossword puzzle afficionado does the New York Times crossword every week. In doing so, she often has to strip phonemes off words, to see if they meet certain criteria. For example, if the answer is _e_ert and the clue is something meaningless to her, she will only be successful in solving the puzzle if she can think of words that fit that schema (desert, revert, etc.)
The readers in each of these situations are competent readers, and yet their day-to-day
reading experiences must at times invoke phonemic awareness skills.
Goodman (1993), a strong whole language advocate, states that readers have "three
systems of information to bring to any text - graphophonic, syntactic and semantic - and
that each one supports the other two. In the course of making sense of print, we use all
three systems" (Goodman, p. 53). While whole language theorists believe that readers
keep their primary attention focused on comprehension when making sense of a text, they
concede that they “shift their focus to the detail of the text and the cue systems they are
using... when comprehension begins to break down" (p. 83). According to this view,
semantic information provides the most important clues used in reading and drives the
reading process, but the specifics of a particular situation (difficulty, strangeness,
vocabulary, familiarity, nervousness, etc.) may affect which information is used, and may
require the reader to rely on his or her phonemic awareness skills.
Thus, even those most committed to the view of reading as whole-word processing fit
phonemic awareness into the schema, though it plays second fiddle to other processing
mechanisms. This leads us to the whole language view of adult phonemic awareness, and
the second explanation for the dearth of adult phonemic awareness research on adults.
In this section, the theory that adults have no need for phonemic awareness skills was
discussed and quickly refuted. An alternate theory, and one with many adherents,
follows.
1.3.2. Theory 2: Phonemic awareness skills play a minimal, negligible role in adult
reading
Another theory to be considered is this:
(2) Though literate adults are phonemically aware, this awareness plays a minimal role in their reading ability, because it only comes into play when whole word reading and comprehension break down.
This theory holds that competent (adult) readers read differently than beginning readers,
paying less attention to the details involved in reading (like the correspondence between
letters and sounds) and more attention to meaning. Whole language advocates support
this view, and aim to narrow the gap between beginning and experienced readers by
teaching novices that reading is a "psycholinguistic guessing game" (Goodman, 1993) in
which context clues (to a greater degree than graphophonic rules) are used when reading
unfamiliar words. Readers only resort to graphophonic rules as a last-ditch effort; only
when all other avenues (whole word recognition and context clues) have failed do such
rules come into play. Thus, phonemic awareness is not an important part of reading.
Once adults have acquired a baseline (minimal) level of phonemic awareness that can get
them out of sticky reading situations, they turn their attention to comprehension-based
whole word reading.
Possible problems with this theory of reading are discussed in the next section, along
with the third theory of the connection between phonemic awareness and reading ability
in adults.
1.3.3. Theory 3: Phonemic awareness skills play a crucial role in adult reading
There is an alternative to theories (1) and (2) regarding the relationship between
phonemic awareness and adult reading:
(3) Adult reading is an interactive process, in which the ability to apply grapheme-to-phoneme conversion rules is crucial, since it may have to take over at any time when other clues fail (losing your place on a page, etc.). Phonemic awareness skills must remain sharp since they may be called upon at any time, not only as a last resort.
This theory was first advanced by Stanovich (1983). As O'Brien (1988) defines the
difference between whole language (comprehension) theory (2) above and this interactive
(comprehension & phonics) theory, "the [former] holds that readers rely less on graphic
and graphophonic information and more on context as they become more adept at reading
[whereas] interactive positions of reading [postulate] that multiple information sources
are used in creating meaning from texts and that the sources used at any given time may
be controlled by a process that compensates for processing weaknesses" (O'Brien, p.
380).
Now, if the context position holds, adults need not hone their phonemic awareness skills,
since they are essentially whole word readers, and the connection between reading and
PA is minimal, and not likely to vary from adult to adult. On the other hand, if the
interactive position holds, readers often turn to graphophonic rules, and, as such, must be
actively phonemically aware. In this case, the level of phonemic awareness of literate
adults remains an interesting, previously unexplored question.
O’Brien tested the competing comprehension-driven reading vs. interactive reading
theories in a study based on the analysis of reading miscues, and found support for the
interactive theory. Miscues are the mistakes that people make in reading aloud. Not all
mistakes are created equal, though. Some mistakes indicate attention to meaning (saying
“She dunked a doughnut” when the text says, “She dunked a muffin”), whereas others
indicate attention to graphophonic information (“She dunked a muffler”). One of the
assumptions of miscue analysis is that the type of miscue a reader makes indicates the
level of his or her comprehension: "readers who make semantically and syntactically
acceptable miscues that do not result in meaning loss are assumed to be comprehending a
text" (O'Brien, p. 381). In other words, good readers use the semantic and syntactic clues
in a text to make educated guesses about what’s coming up, so miscues that are
semantically connected to the target word should indicate good reading comprehension.
O'Brien conducted an experiment on seventh graders designed to test the relationship
between miscues and comprehension. Subjects were asked to read four passages of
varying difficulty aloud, their miscues were recorded and analyzed, and then those
miscues were compared with the results of post hoc comprehension tests. The results
indicate that the strong context/psycholinguistic position does not hold water. This
position would predict a positive correlation between "good" (context-consistent) miscues
and comprehension, which was found in only one of the four passages used (the most
difficult passage). Moreover, the context position would predict that the best readers
attend to meaning, but O’Brien found that the very best reader did not correct
semantically anomalous miscues as frequently as the worst reader.
O’Brien’s results indicate that readers pay the least attention to context when reading
"easy" passages, which is clearly contra the context position. The whole language
theory/context theory would predict that when reading an easy-to-comprehend passage,
good readers rely heavily on syntactic and semantic information to figure out words.
However, O’Brien found just the opposite. As he puts it, his results indicate that "the
broadest interpretation of such conflicting results across passage content is that oral
reading miscue analysis presents an oversimplification of what readers do to adapt
processing to a variety of texts" (O'Brien, p. 397). The interactive theory, by which
readers - even good readers - often rely on graphophonic converstion (and, by extension,
phonemic awareness) is better able to explain O’Brien’s results.
There are several survey articles and studies on the predictability of text from semantic
and syntactic clues that provide supporting evidence for O’Brien’s conclusions that
reading is an interactive process, and not based predominantly on semantic and syntactic
cuing. For instance:
1. Across a number of studies, the probability of a reader predicting the next word in
a passage from semantic and syntactic information was between .20 and .35,
hardly a safe bet as a main strategy in reading comprehension (Stanovich &
Stanovich, 1995),
2. The words which readers are most able to correctly predict are function words2,
not words bearing semantic content (Gough, 1983), a finding which is difficult to
reconcile with the comprehension-driven theory of reading, and
3. Eye-movement research has indicated that good readers process every letter on a
page, rather than glossing over whole words (Gough, 1983).
O’Brien concludes that the results of his miscue study conflict with the whole language
view of reading, and can be better explained by an interactive view of reading, whereby
readers use different information sources at different times to fill in processing "gaps."
One such information source is graphophonic information. O’Brien’s study then
provides preliminary evidence that a connection between phonemic awareness and
reading may be found in adults, just as it has been with children.
In this section, several possible connections between phonemic awareness and reading in
literate adults have been considered. What has emerged is that this relationship is by no
means dismissable. It is not the case that adults are entirely whole word readers; they
will come across unfamiliar words in their reading, and, according to Goodman, “work
around different possibilities to see if anything might sound right" (p. 50). It is also not
the case that adults rely on their phonemic awareness skills only when comprehension
fails. Though adults are surely better than children at using "context clues" in the
deciphering of unfamiliar words, O’Brien showed that context-appropriate mistakes did
not necessarily indicate better reading ability.
2This may also help explain why in studies in which normal, literate adults are asked to cross out all occurrences of a given letter in a passage, they frequently miss the letters in function words.
It is the case that proficient readers still employ their phonemic awareness skills, the
skills that were so important in learning to read, every day. Adults have a need for
keeping the connections between graphemes and phonemes well-oiled, and yet scant
research has been done on phonemic awareness in adults. Even adults have to deal with
the sticky "details" of reading at times. In this case, those details are graphophonemic
rules, contingent upon phonemic awareness.
1.4. Thesis and motivation
The question explored in this thesis is whether we find the same correlation between
phonemic awareness and reading ability in adults as in children. Goodman suggests that
adults have phonemic awareness skills, but what has not been shown is whether, in
adults, phonemic awareness skill is correlated with reading ability. It has not been shown
that adults, like children, benefit from sharply tuned phonemic awareness skills and that
these skills make them strong and effective readers.
The hypothesis of this study is two-fold:
(1) Although all literate adults are phonemically aware to some degree, they, like children, demonstrate different levels of phonemic awareness (by virtue of talent, inherent linguistic skill, and/or experience).
(2) In adults, as in children, this phonemic awareness is associated with reading ability.
This section is divided into three parts: (1) an overview of the experiments that will test
the above hypothesis, (2) a discussion of the theoretical motivations for these
experiments, and (3) a discussion of the pragmatic motivations for these experiments.
1.4.1. Overview of Experiments
The connection between phonemic awareness and reading ability in adults will be
explored in a series of six experiments. These experiments will be based on previous
experiments conducted with children and special populations, which will be discussed
more fully in Chapter Two. Without delving into great detail in this overview, it is
helpful for the reader to know that previous phonemic awareness experiments have
followed a relatively simple model: gauge a target population’s performance on some
measure of phonemic awareness and on some measure of reading ability, and look for
correlations between the two. For instance, Mann (1986) tested subjects (children) on a
Phoneme Segmentation Test (PST) of phonemic awareness (using the "odd man out"
methodology, in which subjects are asked to identify which word in a series does not
share the same initial, final, or medial phoneme as the others) and on the Woodcock
Reading Mastery Test. She found that scores on the two tests were positively correlated.
It is this basic structure (measuring phonemic awareness and reading ability and looking
for correlations) that is used in the experiments in this study.
A certain challenge lies in constructing experiments that determine the phonemic
awareness of adults, however, because all literate adults, by virtue of their reading
experience, are to a certain degree phonemically aware. Although it is relatively easy to
draw a phonemic awareness line-in-the-sand with children (i.e., Child A says that cat has
three sounds, whereas Child B says cat has one sound; therefore, Child A is phonemically
aware, whereas Child B is not), it is more difficult to distinguish levels of phonemic
awareness in adults.
Adams (1990) identifies five levels of phonemic awareness that can be used to classify
the experimental tasks that have been used in past experiments with children. She points
out that in evaluating the role of phonemic awareness in reading ability,
it is neither the ability to hear the difference between two phonemes nor the ability to distinctly produce them that is significant. What is important is the awareness that they exist as abstractable and manipulable components of the language (Adams, 1990, p.65).
Adams’ levels of phonemic awareness are listed below in increasing order of
sophistication. This order also indicates how well their mastery correlates with
subsequent reading acquisition; the tests described in the next chapter have shown that
mastery of the highest levels of phonemic awareness is most likely to be connected to
reading ability.
Each level is listed with an example of a phonological awareness task that might be used
to assess it:
1. “The most primitive level....involves nothing more than an ear for the sounds of
words.” (Adams, p. 80).
Ex. Nursery rhymes: children (ranging in age from 3 months to 4 years) were tested on their knowledge of five popular nursery rhymes, and it was found that their ability to recite these rhymes “was strongly and specifically
related to development of more abstract phonological skills and of emergent reading abilities” (p. 80)
2. The next level "requires not just sensitivity to similarities and differences in the
overall sounds of words, but the ability to focus attention on the components of their
sounds that make them similar or different" (p.80)
Ex. Oddity tasks, in which children are given a list of words with a phonemic contrast in the beginning, middle or end of the word (e.g. pig, hill, pin) and asked "Which of these doesn't belong?"
3. The third level requires that "the child have a comfortable familiarity with the notion
that words can be subdivided into those small, meaningless sounds corresponding to
phonemes, and second, that she or he be comfortably familiar with the way phonemes
sound when produced 'in isolation'" (p. 80).
Ex. Blending tasks, in which the experimenter gives the child segments in isolation ("/m/.../a/.../p/") and the child is asked to put them together ("map"), and syllable splitting tasks, in which children are asked to split the rime of a syllable from its onset (producing "eel" and "ice" when the experimenter says "feel" and "mice").
4. The fourth level requires that "the child have a thorough understanding that words can
be completely analyzed into a series of phonemes....[and can] so analyze them,
completely and on demand" (p. 80).
Ex. Phonemic segmentation tasks, in which children were given a word and asked to "tap out" how many sounds (phonemes) it contained (so, "map" should get three taps).
5. The fifth and final level, the most sophisticated level of phonemic awareness, requires
that "the child...be able to add, delete, or move any designated phoneme and regenerate a
word (or a nonword) from the result" (p. 80).
Ex. Phoneme manipulation tasks, in which children are asked to leave out an initial, medial, or final phoneme in the pronunciation of a word (e.g. "Say 'hill' without the /h/"; "Say 'monkey' without the /k/," etc.).
Because adults have had years of experience with language, tasks falling into the lowest
levels of Adams' classification will be too compressed; all literate adults are able to
contrast, blend and count the phonemes in a simple word with relative ease. Therefore,
the tasks most useful in the differentiating of adults will be phoneme manipulation tasks
(which fall under the highest level of phonemic awareness).
The subjects, literate university students, will be tested on a variety of phonemic
awareness tasks, and then the results will be compared to their performance on two GRE
tests: (1) a reading comprehension test and vocabulary test, and (2) a test of analytic
ability. The first will be used to determine the subjects’ reading ability (and the
correlation between phonemic awareness and reading ability), whereas the second is a
control measure, to ensure that any correlations between phonemic awareness and
reading ability are not the result of mere cognitive ability, but rather some reading-
specific skill.
The structure of the phonemic awareness portion of the experiment is as follows. Subjects
will be asked to perform three different tasks (a deletion task, a substitution task, and a
reversal task) that require them to manipulate two different types of linguistic units
(syllables and segments), for a total of six tasks in all. This design is illustrated in the
table below:
Segment-based tasks Syllable-based tasksDeletion Delete Segment Task (1) Delete Syllable Task (2)Substitution Substitute Segment Task (3) Substitute Syllable Task (4)Permutation Reverse Segment Task (5) Reverse Syllable Task (6)
Examples of each of these tasks are given below, using the word frantic and some
possible responses (other response possibilities will be discussed in Chapter Three).
Experiment One: Delete Segment TaskInstructions: What would the word ‘frantic’ sound like if the first sound were
taken off?Response: [ræntik]
Experiment Two: Delete Syllable Task Instructions: What would the word ‘frantic’ sound like if the first syllable were
taken off?Response: [tik]
Experiment Three: Substitute Segment Task Instructions: What would the word ‘frantic’ sound like if the first sound were
replaced by the first sound in ‘blasted’?Response: [bræntik]
Experiment Four: Substitute Syllable Task Instructions: What would the word ‘frantic’ sound like if the first syllable were
replaced by the first syllable in ‘blasted’?Response: [blæstik]
Experiment Five: Reverse Segment Task Instructions: What would the word ‘frantic’ sound like if the first sounds in the
two syllables were reversed?Response: [trænfik]
Experiment Six: Reverse Syllable Task Instructions: What would the word ‘frantic’ sound like if the two syllables were
reversed?Response: [tikfræn]
Previous studies on phonemic awareness have shown that tasks which require syllable
manipulation are the easiest for subjects to complete, whereas tasks which require
segment manipulation are the most difficult to complete. Syllabic awareness tasks require
only a low level of phonological skill (probably falling into Adams levels one through
three) found in preliterate children and illiterates. Segmentation tasks like those above
(falling into Adams level five), require a sophisticated level of phonemic awareness that
seems to require alphabetic experience. Therefore, it is predicted that within each task
type, subjects will perform best (on the basis of error rate) on syllable tasks and worst on
segmentation tasks.
The predicted correlation with reading ability should be clear: those subjects who
perform best on the segment manipulation tasks, thus showing the most sophisticated
phonemic awareness, should also be those who perform best on reading tests. Syllable
tasks will serve as a control, since all literate adults should possess the ability to
manipulate syllables.
In section 1.3. above, the issue explored was why we might believe there is a connection
between reading ability and phonemic awareness in literate adults. A related issue is the
importance of such a connection. Why should the Linguistic and Educational
communities care about a correlation between reading and phonemic awareness in literate
adults?
There are both theoretical and pragmatic reasons for exploring the nature of phonemic
awareness in literate adults. Let us explore each of these in turn.
1.4.2. Theoretical considerations
The experiments conducted for this study require literate adults to participate in
phonological experiments. In the experiments, described in further detail below and
presented fully in Chapter Three, subjects are asked to perform a variety of phonological
tasks, including the deletion, reversal, and substitution of syllables and phonemes in
words. The results of these experiments can also contribute to theories of phonological
processing. The following three theories, listed in order of increasing specificity, are
supported by the results of experiments in this dissertation:
10 The syllable, rather than the phoneme, is the basic unit of phonological
processing;
20 Ambisyllabicity exists in English syllables; and
30 [s] is extrametrical in English syllables in certain positions.
Although the phonemic awareness studies in this thesis are not specifically designed to
test any of the above theories, because the experiments conducted require adults to
process and manipulate various phonological units, the results are informative as to the
structure of those units.
First, the experiments support the idea that the syllable is the basic unit of processing, by
showing that adults perform significantly better on tasks in which they are asked to
identify and manipulate syllables than those in which they are asked to identify and
manipulate phonemes. Arguments that the syllable is the basic unit of processing are
grounded in the work of Mehler, Dommergues, and Frauenfelder (1981). In this work,
French subjects were asked to listen to various words, monitoring for a target syllable
(for example, [pa]). Upon hearing the target syllable in an input word, they were to press
a button, indicating whether a match had been detected. The experimenters considered
two alternate theories of processing: either the subjects were processing input words
phoneme-by-phoneme or syllable-by-syllable. If the subjects were using the phoneme as
their basic unit of processing, there would be no difference in the time it took subjects to
detect the target in words beginning with identical phoneme sequences. Consider the
example below:
target sound: p...a... p...a...
input: p...a...l...a...s p...a...l...m...i...e...rresult: match match
In both cases, the match occurs two phonemes into the target word, so detection of [pa] in
pa.las should be no faster or slower than detection of [pa] in pal.mier. This is not what
Mehler et. al. found, however. Subjects were significantly slower in detecting the [pa] in
pal.mier (that is, detecting a CV target in an input word beginning with a CVC syllable).
These results are quite understandable if the syllable is viewed as the basic unit of
processing:
target sound: pa pa
input: pa...las pal...mierresult: match mismatch
Since the target [pa] matches the first syllable in pa.las, but not in pal.mier, the match in
the first is detected significantly faster. Mehler et. al. conclude that the only logical
solution is that subjects process words in terms of syllables and not phonemes.
All previous research on phonemic awareness in children and special populations
suggests that syllabic awareness is an innate ability and phonemic awareness is learned.
If the syllable is the basic unit of processing, we would expect that its detection and
manipulation are basic, innate abilities, and the experiments in this study provide support
for that theory. The experiments require subjects to delete, reverse and substitute the
syllables and phonemes in various words. It will be shown that in all the experiments in
this study, subjects performed significantly better on experiments requiring syllable
manipulation than those requiring phoneme manipulation. Using the word frantic as an
example, subjects found it significantly easier to:
· delete the first syllable in the word (yielding [tik]) than the
first phoneme in the word (yielding [ræntik]),
· substitute the first syllable in the word (yielding [blæstik])
than the first phoneme in the word (yielding [bræntik]), and
· reverse the syllables in the word (yielding [tikfræn]) than
the phonemes in the word (yielding [trænfik]).
These findings show that for adults, like children, syllabic awareness tasks are easier than
phonemic awareness tasks across-the-board; the specific nature of the task is irrelevant.
This bolsters the theory that syllabic awareness is innate and, in turn, that syllables are
the basic units of processing.
Interestingly, when Cutler, Mehler, Norris, and Segui (1986) attempted to replicate the
Mehler et. al. task cited above in English, they ran up against a brick wall, failing to find
evidence for the syllable as the primary unit of speech processing. The data in this study
are consistent both with the theory of the syllable as the basic unit of processing, and with
the Cutler et. al. findings. This is because follow-up studies to the Cutler et. al. study
have suggested that their failure to replicate the original study stems from the fact that
phonemes in certain syllables exhibit ambisyllabicity (membership in more than one
syllable). This is a theory supported by the experiments in this study. This brings us to
the second contribution these experiments have to make to phonological theory: the
experiments conducted provide support for the ambisyllabicity of certain phonemes in
English syllables.
Let us first review some background work on ambisyllabicity in English. Classic theories
of syllabification produce conflicting interpretations of English syllables. The Maximal
Onset Principle (MOP) (Pulgram, 1970) says that when the consonants in a word can be
syllabified in more than one way, the preferred syllabification is that which places as
many consonants as possible in the onset. Following the MOP, then, the syllabification
of the word fresco is fre.sco. The Sonority Dispersion Principle (SDP) (Clements,
1988), however, has a different story to tell. It says that the correct syllabification of a
word results from maximizing sonority dispersion in the onset and minimizing it in the
coda. According to this theory, then, the correct syllabification of fresco is fres.co.
Kahn (1976) was among the first to propose that the syllabification of a word like fresco
might not be as black-and-white as either the MOP or SDP would indicate. Kahn’s
theory contends that in fresco, in which the first syllable is stressed, the [s] sound is
ambisyllabic - a resident of both syllables.
In recent years, a great deal of research has been conducted on the ambisyllabicity of both
medial consonant clusters and single consonants in English. Treiman and Zukowski
(1990) conducted a series of experiments to determine what effect phonotactics, stress,
vowel length, the MOP and sonority have on the syllabification of medial clusters.
Legality was by far the most important factor; subjects syllabified words like atlases and
confetti (in which the medial consonant cluster would be illegal both at the beginning and
the end of a word) between the two consonants (at.lases and con.fetti) between 99% and
100% of the time.
Other factors affecting syllabification were not nearly so definitive. Treiman and
Zukowski found that in cases in which both consonants of a medial consonant cluster
could legally belong to the first syllable, and in which that first syllable was stressed, the
MOP was violated (for the second syllable) and both consonants were placed in the first
syllable between 3% (in a written syllabification task) and 6% (in an oral syllabification
task) of the time. For example, pontiff was sometimes syllabified as pont.iff in a written
syllabification task, and pont.tiff in an oral syllabification task. So, a stressed first
syllable followed by an unstressed syllable attracts consonants.
A final set of experiments considered the role that stress and vowel length play in
syllabification. In words like master and metric (first syllable stressed and containing a
short vowel) the first medial consonant was placed in the first syllable between 78-86%
of the time (in a written task) or ambisyllabic between 43-49% of the time (in an oral
task). This is in apparent violation of the MOP. However, in words like cloister and
apron (first syllable stressed and containing a long vowel) it was more likely that both
consonants would be placed in the second syllable (61-72% of the time, in an oral task).
So, a syllable with a short vowel attracts consonants.
Finally, Treiman and Zukowski considered the effect of second syllable (primary) stress
on syllabification. They found that in words like madrid and estate, the clusters were
placed in the second syllable 41-80% of the time (less in s-clusters than in other clusters).
Again, this is an example of a stressed syllable attracting consonants.
The simplest way to sum up these complicated interactions is that syllables with both
short and stressed vowels try to “grab” as many consonants as they can without violating
phonotactic rules.3 The same factors seem to be at play in the syllabification of single
intervocalic consonants. Treiman & Danis (1988), paying particular attention to the
sonority of the medial consonant, found in an oral task that in words with short and/or
stressed first syllable vowels, medial sonorants were placed in the first syllable about half
the time (between 40-50%) and obstruents were placed in the second syllable about half
the time (49-58%). Since this means that the other half of the time the subjects either
judged the medial consonant ambisyllabic or placed it in the other syllable, these findings
are not very conclusive. Meador and Ohala (1993) found slightly more robust findings
in a written task. They found that medial consonants in a stressed-stressless environment
were judged ambisyllabic 67.9% of the time (as opposed to 32.1% of the time when the
first syllable was stressless). This trend was even more apparent when the vowel in the
first syllable was lax (i.e. palace). In these cases, the medial consonant was judged
ambisyllabic 75.8% of the time. In the syllabification of single consonants, as with
consonant clusters, stress and vowel length play an important role; consonant type is less
important.
To sum up the findings on the syllabification of medial consonants in bisyllabic words,
then
· Syllabication must produce legal syllables (virtually unviolable),
· Syllables with stressed vowels attract consonants (to both their onset and
coda), and
3 To be clear, short syllables with vowels attract consonants to their codas only. Hammond (1993) attributes the affinity medial consonants have for short vowels to a foot-based weighting rule in English. In this rule, unreduced syllables are minimally bimoraic, which can satisfied by resyllabifiying a consonant from the right, resulting in ambisyllabicity.
· Syllables with lax vowels attract consonants to their codas.
The relevance of these findings to the experiments in Chapter Three is as follows: since the experiments
require subjects to undertake syllable manipulation tasks, whether or not a medial consonant is
ambisyllabic determines whether a subject’s response should be deemed correct. Consider the case of the
Delete Syllable Task, in which subjects are asked to strip off the first syllable of a word. The theory of
syllabification adopted greatly affects whether a given response should be judged as correct or incorrect, as
illustrated in the following chart:
TEST ITEM Maximize Onset Sonority Dispersion Ambisyllabicity
fresco fre.sko fres.ko fres.ko or fresk.ko or fres.sko
baby be.bi be.bi be.bi or beb.bi
If ambisyllabicity exists in English syllables, it is difficult to justify one correct
syllabification for fresco and baby. In fresco, the first syllable is stressed and contains a
lax vowel that would seem to suggest the fres.co syllabification. Still, Treiman and
Zukowski found that fresc.co and fres.sco syllabifications were possible (9% and 43% of
the time, respectively). In the case of baby, the first vowel is tense, suggesting the
syllabification ba.by. However, Meador and Ohala found that these cases were
syllabified bab.by over 20% of the time.
add page or two on specifics of theories, giving definitions
Depending on which theory of syllabification is adopted, the correct response to the
Delete Syllable Task for the item fresco is either [sko] (by the MOP), [ko] (by the SDP),
or either of these (by Trieman & Zukowski). It will be shown in Chapter Three that if
there is not ambisyllabicity in English syllables, and an early explain theory of
syllabificaion is adopted, the experimental results for the experiments in this dissertation
fly in the face of everything we know and expect about adult phonemic awareness. If we
are forced to adopt either the Maximize Onset Principle or the Sonority Dispersion
Principle as the correct theory of syllabification in coding subjects’ responses, rather than
allowing ambisyllabicity, adult subjects perform quite poorly on a very simple task: the
deletion of the final syllable in a word. Since school-age children are able to perform this
task with remarkable accuracy, these findings are simply nonsensical.4 On the other
hand, if correct responses are coded taking into consideration ambisyllabicity, adults
perform with the expected competence on the Delete Syllable Task.
The final, and quite specific, contribution of these experiments to linguistic theory is that
the experiments support the characterization of [s] as an extrametrical element in some
environments. The issue surrounding word-initial s-clusters is this: s-clusters violate the
sonority hierarchy ([s] is more sonorous than [t], and yet is further removed from the
syllable nucleus in sticky) and so [s] has been classified by some as extrasyllabic
(Clements and Keyser; 1983). Selkirk (1982) proposed a different explanation for the
anomalous behavior of s-clusters, positing that s-clusters at the beginnings of words were
actually complex onsets. segments? In deleting the first phoneme, subjects may be
deleting the first licensed (non-extrasyllabic) element, or deleting a complex segment,
which can’t be broken down any further. Either way, the status of [s]-consonant
4 Though syllable deletion experiments with children contain stimuli that are easier that those in the experiments here (an issue which will be discussed in a later chapter), this difficulty is presumably offset by the much greater linguistic sophistication of literate adults.
sequences affects the coding of subject responses in Chapter Three. Consider the
following chart: predictions not clear; part of onset vs. extrasyllabic vs. part of complex
single segment
TEST ITEM Response to Delete Segment Task,if [s] is part of cluster
Response to Delete Segment Task,if [s] is extrasyllabic
sticky tiki iki
species pisiz isiz
In the first two experiments described in Chapter Three, subjects are asked to delete the
initial phonemes and initial syllables of a word. It is predicted that subjects will fare
better on the syllable deletion task (because syllable knowledge is innate and taps into a
lower level of phonological awareness) and that subjects’ performance on the delete
phoneme task will be related to their reading ability. If correct responses are coded
assuming that [s], in an initial cluster, behaves like any other initial phoneme (such that
the correct response in the Delete Initial Phoneme task for the item sticky is [tiki]), the
results of the experiments here are unexpected and extremely difficult to explain.
foreshadowing hard to follow If, on the other hand, [s] in initial clusters is treated as an
extrametrical element and responses are coded accordingly (such that the correct
response in the Delete Initial Phoneme task for the item sticky is [iki]), the results, though
still unexpected, can be explained in the confines of a theory of adult phonological
awarenesss. This enigmatic statement will be discussed fully in Chapter Three.
To sum up, the experiments in this dissertation are not designed to directly test any
particular theory of syllable structure or syllable access. However, in coding and
interpreting subjects responses, certain theories about the syllable must be assumed. The
results in this study can best be explained by a theory of the syllable that assumes that the
syllable is the primary unit of processing, that ambisyllabicity exists in English syllables,
and that [s] is extrametrical in certain environments in English syllables. be more
explicit about theory of syllabification assumed
1.4.3. Pragmatic considerations
In the previous subsection, the theoretical motivations for the experiments in this
dissertation were discussed. There are also pragmatic motivations for fleshing out the
exact nature of phonemic awareness and its connection to reading ability in adults.
If the correlation between reading ability and phonemic awareness in adults is shown to
be as strong as in children, there would be great consequences for adult reading
programs. At least three types of programs would benefit from the knowledge that there
is correlation between reading ability and phonemic awareness in adults:
(1) Adult literacy programs,
(2) English-as-a-second-language programs (or any such programs in which the
L2 employs an alphabetic orthography), and
(3) Reading enrichment programs.
First, let us consider the case of adult literacy programs, in which illiterate adults wish to
learn to read and write an alphabetic orthography. Research with children indicates that
phonemic awareness and reading ability are highly correlated, and all effective readers
are phonemically aware. Adams (1991) claims, "toward the goal of efficient and
effective reading instruction, explicit training of phonemic awareness is invaluable" (p.
331). Perhaps, then, phonemic awareness training, rather than mere phonics training,
should be an essential element of adult literacy programs. Ball & Blachman (1991)
conducted a study in which they compared the reading performance of three groups of
children who received various kinds of training. One group was given traditional phonics
(sound-symbol) instruction, a second group received phonics and phonemic awareness
instruction (by being taught phoneme deletion and manipulation "games"), and a third
(control) group received no instruction. They found that although the group which
received both phonics and phonemic awareness instruction outperformed the other
groups, "instruction in letter names and letter sounds alone did not significantly improve
the segmentation skills, the early reading skills, or the spelling skills of the kindergarten
children who participated in the language activities group, as compared with the control
group" (B&B, p. 49).
Are we missing the boat in adult literacy programs? Should we do away with Hooked on
Phonics and teach adults how to play language games? Bagemihl (1988) points out that
segment-reversing language games only appear in languages with alphabetic
orthographies, and concludes that "an alphabetic writing system may be a prerequisite for
a segment reversal ludling [language game] to appear in a language...It seems that the
presence of an alphabetic writing system is necessary for the establishment of some
metalinguistic awareness of the notion of 'segment'; beyond this, however, the
phonological system takes over as the primary basis for reversal" (p. 485). Perhaps, rather
than waiting for segment-based language games to fall out of languages with alphabetic
orthographies, we should use them on the road to teaching such orthographies. An early
theory that there is a critical period for phonemic awareness (Morais, 1986) has come
under fire (Morais, 1991), so there may be nothing preventing adults at any age from
honing their phonemic awareness skills and, in turn, their reading skills. in the wrong
place?
If there is a phonemic awareness/reading connection in adults, this training would also
prove useful for adults learning a language with an alphabetic orthography as a second
language. If the adult’s first language has a non-alphabetic writing system, the benefits
of phonemic awareness training in adults are clear: phonemic awareness in an essential
part of effective reading. However, even if the adult reads and writes an alphabet (and is
thus phonemically aware), phonemic awareness training may be useful. As O’Brien’s
study (discussed earlier in this chapter) showed, effective adult readers read interactively,
relying heavily on graphophonic information. Since the second language is sure to
contain phoneme-grapheme correspondences which are new to the adult learner and thus
more a conscious part of reading, a phonemic awareness “refresher” (that is, training in
the honing or development of phonemic awareness skills) could help in the acquisition of
the second [written] language.
Finally, if phonemic awareness is correlated with reading ability in adults, the
development of phonemic awareness skills could prove a successful part of a reading
enrichment program. Perhaps alongside Ways to Enrich Your Word Power or an Evelyn
Wood-style speed-reading course could sit a phonemic awareness training program,
designed to improve adult reading ability by fostering greater phonemic awareness skills.
1.5. Organizational structure of the dissertation
This chapter began with a definition of phonemic awareness and a brief overview of the
issues raised in phonemic awareness/reading research with children. The motivation for
examining the PA/reading relationship in adults was explored. Finally, an outline of the
experimental design was laid out, followed by a discussion of the theoretical and practical
issues to be raised in the course of this study.
The organization of the remainder of the dissertation is as follows. The next chapter is a
review of previous research relevant to the issue of phonemic awareness in literate adults.
It begins with a review of past phonemic awareness literature on children and special
populations, summarizing previous research in the field, followed by a brief review of
research on orthographic interference in phonological tasks, including an experiment
designed to determine the nature of that interference.
Chapter Three describes the six phonemic awareness experiments outlined above, and the
final chapter is a conclusion, tying together the results of the experiments with previous
research on phonemic awareness, and exploring the theoretical and pragmatic
implications of the findings in this study.
2.0 Previous Research on Phonemic Awareness
This chapter deals with previous research that is relevant to the issue of phonemic
awareness in literate adults. Among the questions that will be explored in this chapter
are:
1. What is the exact nature of the connection between phonemic awareness and reading ability? Is one a prerequisite for the other, or do the two develop simultaneously?
2. Are there levels of phonemic awareness, or is it a developmental on/off switch, in that it is either present or absent?
3. Is phonemic awareness teachable?
4. Why do some people develop phonemic awareness, and others not?
5. Other than explicit instruction in an alphabetic orthography, what types of activities foster phonemic awareness?
6. How might knowledge of a specific orthography affect a subject’s ability to perform a test of phonemic awareness?
This chapter is divided into four sections. The first two sections explore previous
research on phonemic awareness and reading, and the last two sections discuss previous
research on metalinguistic abilities in adults.
Section One is a review of what is by far the most thorough area of research on phonemic
awareness: the relationship between phonemic awareness and reading ability in beginning
readers. In this section, questions (1)-(3) are discussed. The second section is also a
discussion of phonemic awareness research, focusing on special populations (illiterates,
dyslexics, etc.) who struggle with phonemic awareness. Section Two pays particular
attention to question (4), in the context of question (1). In the third section question (5)
is addressed directly, in a discussion of activities that foster phonemic awareness.
Finally, in Section Four, the question of orthographic interference in phonological
processing tasks is discussed (question (6) above).
The chapter concludes with a summary of past findings in phonemic awareness research,
and implications for the experiments in Chapter Three.
2.1 Phonemic awareness research on beginning readers
2.1.1. Overview
In this section, past research on phonemic awareness in children will be reviewed, so as
to frame the six experiments in Chapter Three in their proper context. Though this
dissertation is an examination of phonemic awareness in literate adults, and though
phonemic awareness studies have been conducted on illiterate adults, dyslexics, and other
special populations (see next section), all research on phonemic awareness has grown out
of the original body of research on phonemic awareness and reading ability in children.
The vast majority of past research on phonemic awareness conducted with newly literate
or preliterate children has centered on the following four interrelated claims:
1. Children must be "phonemically aware"5 before they are able to read an alphabetic orthography.
2. PA tasks can be used to predict a child's later success or failure in reading.
3. PA training can prove useful in preparing a child for reading instruction.
4. There are different levels of phonemic awareness.
These are very strong claims with some very strong implications. Researchers have
suggested a number of ways that phonemic awareness tasks might be integrated into the
grade school curriculum: (1) as a screening tool to identify those children most likely to
encounter difficulties in reading, (2) as an accompaniment to traditional phonics
instruction for all students, and (3) as an alternative to traditional phonics instruction
(Mann, 1993).
The degree of consensus on each of these claims is discussed in more detail in the
subsections immediately below. The subsequent sections examine "outside" findings on
PA, touching upon illiterates, literates with non-alphabetic orthographies, dyslexics,
children with Down syndrome, and secondary language activities.
Let us begin with a review of each of the seminal phonemic awareness claims above, in
turn:
2.1.2. Reading readiness
5 Where "phonemically aware," when it is defined, is defined as the ability to perform well on phonemic awareness tasks.
1. Children must be "phonemically aware" before they are able to read an alphabetic orthography.
Ball & Blachman (1991) make a claim which is made by many others, based on the
following intuitively appealing assumption about learning to read:
To read and spell, the beginning reader must make use of the alphabetic code. Thus, the student must come to realize that words can be broken into syllables and phonemes, and that the phoneme is that unit in the speech stream represented by the symbols in an alphabetic script. To a person with well-developed phoneme awareness, our alphabetic system appears to be a reasonable way to represent our language. To those with little or no phoneme awareness, the system probably appears arbitrary. (Ball & Blachman, 1991, p. 51)
They later say that "to read or spell phonetically regular words, a child must be aware that
words can be broken into phonemes and that each phoneme corresponds to a symbol (or
symbols) in our orthography" (Ball & Blachman, 1991, p. 63). Below is a simple
illustration of the stages in a child’s phonemic awareness development:
1. [sn] SUN
2. [sn]
SUN3. [s... ...n]
S...U...N4. [s... ...n]
S...U...N
In stage (1) the child may recognize the aural stimulus [sun] as referring to the big ball of
light, and may recognize the printed word “SUN” as referring to the same object, but he
or she has not necessarily made any direct (phonological) connection between the oral
and written word. In stage (2), the child recognizes that [sun] and SUN are connected,
but is not able to abstract that knowledge - it is a fixed, limited piece of information. In
stage (3), the child has developed phonemic awareness, and can “hear” the word sun as
consisting of three distinct phonemes (and, similarly, has learned the alphabet and sees
“SUN” as three distinct letters). Finally, in stage (4), the child has learned the connection
between the distinct phonemes of [sun] and the distinct graphemes of “SUN.” This
knowledge is abstractable, because it can be applied to other correspondences of [s] and
“S,” for instance.
This oversimplified schema demonstrates how reading is the application of appropriate
grapheme-to-phoneme conversion rules (the "alphabetic code"), which require that the
child is able to identify and manipulate the phonemes in a word.
Breaking the alphabetic code thus requires phonemes to be "dug out of their normal,
subattentional status" (Adams, 1990, p. 294). Children can only figure out the system of
the alphabet when they learn that a word like sun, which they perhaps previously viewed
as a single unit, can be broken down into smaller units (phonemes), and that these smaller
phonological units correspond to letters (graphemes). To Adams and other phonemic
awareness researchers, this requires a step beyond from the normal (innate) way of
perceiving language, which is syllable-based. Children need to be taught that syllables
can indeed be broken down further, and that the resultant units (phonemes) correspond in
a regular fashion to letters.
Whole language advocates view things differently. They do not agree that phoneme-
grapheme rules (which comprise the alphabetic code) need to be taught, nor do they agree
that phonics knowledge is a prerequisite for reading. Goodman (1993) argues that this
alphabetic coding is an "unnatural process" and that "we can liberate children from letter-
by-letter processing of text" by teaching them to view words as whole units and
encouraging them to use semantic clues to figure out visually unfamiliar words.
However, whole language advocates do concede that phonemic awareness is important in
reading; they simply feel that children can figure it out for themselves once they start
reading, rather than being taught a set of rules which are plagued by exceptions. The
phonics teacher tells Johnny that cat begins with [k] and that is spelled with a “c,” and the
whole language teacher gives Johnny a book with many cat/hat rhymes, until he figures
out for himself that the sound and letter similarity between the words isn’t mere
coincidence (and that the sound differences correspond to the “c”/”h” difference). Either
way, he learns that cat can be broken down into three distinct units, and this is the
beginning of phonemic awareness.
2.1.3. The predictive power of phonemic awareness
The second claim about phonemic awareness is:
2. PA tasks can be used to predict a child's later success or failure in reading.
This is far and away the most studied claim about phonemic awareness. Adams (1991)
says that "the familiarity of the letters of the alphabet and awareness of the speech
sounds, or phonemes, to which they correspond, are strong predictors of the ease or
difficulty with which a child learns to read." Stanovich (1993) echoes this sentiment,
pointing out that the phonemic awareness/reading connection is independent of general
intelligence: “[phonemic awareness tasks are] the best predictors of the ease of early
reading acquisition - better than anything else that we know of, including IQ" (p. 331).
An even stronger statement comes from Mann (1993), who argues that either the
presence or absence of strong phonemic awareness skills is significant: "Phoneme
awareness bears both a logical and a proven relationship to early reading success. Its
presence is a hallmark of good readers, its deficiency one of the more consistent
characteristics of poor readers" (p. 260)
Mann's study is outlined in the previous chapter; she found that children who
demonstrated the highest levels of phonemic awareness on the basis of their invented
spellings and segmentation ability turned out to be the best readers. The same
connections have been found with many other researchers (Bradley & Bryant, 1983,
1985; Fox and Routh, 1975, 1980; Stanovich, Cunningham & Cramer, 1984;
Lundberg, Olofsson & Wall, 1980; Mann & Liberman, 1984, among others). In each of
these studies, preliterate children were tested on a variety of phonemic awareness tasks,
including, but not limited to, the following:
Task Type Sample instructions Answer
Phoneme blending Put /m/.../a/.../p/ together and what do you get? [mæp]
Phoneme segmentation What sounds do you hear in the word hot? [h]...[a]...[t]
Phoneme counting Tap out the sounds in map. 3 taps
Phoneme matching Do pipe and pen begin with the same letter? yes
Deleted phoneme What sound do you hear in meat that is missing in eat? [m]
Phoneme oddity Which of these doesn’t belong? pig, hill, pin hill
Task Type Sample instructions Answer
Phoneme deletion What would is left if you take off the first sound in nice? [ays]
Phoneme substitution What would feel sound like if it started with [m]? [mil]
These task descriptions are all drawn from Adams (1990) and Stanovich (1983).
In these studies, children’s performance on phonemic awareness tasks was recorded,
often along with their performance on (1) other phonological tasks, (2) other cognitive
tests, and (3) reading and spelling tests. The other phonological tasks test less
sophisticated levels of children’s linguistic abilities, and may include the tasks above
conducted with syllables as the unit of manipulation (i.e., syllable counting, in which
children are asked to tap out the syllables in a word), or even lower-level tasks (i.e., the
ability to recite nursery rhymes). The other cognitive tests are usually IQ tests, and the
reading tests are standard school measures of reading ability, such as the Peabody Picture
Vocabulary Test and the Metropolitan Reading Readiness Test.6
Each of these classic phonemic awareness studies came to the same conclusion:
preliterate children who perform best on phonemic awareness tests turn out to be the best
readers. Moreover, syllable-based tasks and other tasks tapping into low-level
phonological awareness do not correlate strongly with reading ability. As Adams (1990)
states:
6 The specific tasks and tests used in each study are given in the next chapter, where parallels are drawn in experiments with adults.
It is neither the ability to hear the difference between two phonemes nor the ability to distinctly produce them that is significant. What is important is the awareness that they exist as abstractable and manipulable components of the language (p.65).
So, in the table above, children who perform well on segment deletion tasks are more
likely to be good readers than those who only do well on counting tasks, as the latter
require a lower level of phonemic awareness.
The view of phonemic awareness as a predictor of future reading ability has been
criticized on the grounds that it may be tapping into past reading experience, rather than
future reading ability. Children who score well on PA tests are those who have watched
Sesame Street, been read to by their parents, been encouraged to write at home, etc.
Thus, it is likely that some children who fail to perform well on PA tests do so because of
missed early reading experiences. Does performance on phonemic awareness tasks reflect
potential for future success in reading, or merely gage past reading experience? While
both Adams and Stanovich believe that children who do well on phonemic awareness
tasks will be successful in reading, both also acknowledge that the relationship between
phonemic awareness and reading ability is a murky one. Adams says that the situation is
a bit of a "catch 22...[because] children who have...acquired a solid level of phonemic
awareness before entering school have also begun to read before entering school"
(Adams, p. 8). In other words, it may not be that phonemic awareness per se is a good
predictor of later reading ability, but that early reading experience makes for better
readers. Using her own five-year-old son as an example, Adams notes that between
being read to and watching educational television, John had had 2000+ hours of indirect
reading instruction before even entering a classroom.
No studies have seperated the two? Stanovich (1993) is quick to acknowledge the
correlation between phonological awareness and prior exposure to reading, but does not
think that this renders the results of phonemic awareness tasks in any way invalid. He
says, "phonological awareness...is a good predictor [of success in reading] not just
because it is an incidental correlate of something else, but because phonological
awareness is a foundational ability underlying the learning of spelling-sound
correspondences" (p. 284). Mann also agrees that the arguments about whether
phonemic awareness gauge past reading experience are difficult to sort out, and
ultimately immaterial: "from a practical standpoint, surely it is an important first step to
learn that a child is deficient in phoneme awareness, whatever the basis of the deficiency"
(Mann, p. 268). And how can such a child be helped? This brings us to the next claim:
2.1.4. The value of phonemic awareness training
3. PA training can prove useful in preparing a child for reading instruction.
Adams (1991) claims that "toward the goal of efficient and effective reading instruction,
explicit training of phonemic awareness is invaluable" (p. 331). What follows are
summaries of studies conducted on the value of phonemic awareness training, in
conjunction with other methods of reading instruction:
Phonemic awareness training vs. no training. Tornéus (1984) conducted two early
experiments on phonemic awareness and reading. In the first, she examined the
correlation of performance on four PA tasks (segmentation, blending, positional analysis
and deletion) with reading and spelling ability. She found (as expected from research
cited in the last section) a significant correlation. In her second study, she examined the
role of phonemic awareness training on reading ability by comparing the reading
performance of three differently trained groups: a control, a lower level task, right? not
difference between the last two groups? "sound properties" group (which was trained on
attention to rhyme), and a group which was specifically trained on segmentation ability.
She found that the last two groups outperformed the control, and so concludes
"metaphonological abilities have a causal influence on reading and spelling" (p. 1357).
Phonemic awareness training & phonics instruction vs. phonics instruction alone vs.
no training. One problem with the Tornéus study is that it does not pit phonemic
awareness instruction against traditional reading instruction; it only examines the
benefits of phonemic awareness instruction versus no instruction at all. Ball &
Blachman (1991) conducted a study in which they compared the reading performance of
three groups which received various kinds of training. One group was given traditional
phonics (sound-symbol) instruction, a second group received phonics and phonemic
awareness instruction (by being taught phoneme deletion and manipulation "games"
along with the sound-symbol correspondences), and a third (control) group received no
instruction. They found that although the group which received both phonics and
phonemic awareness instruction outperformed the other groups, "instruction in letter
names and letter sounds alone did not significantly improve the segmentation skills, the
early reading skills, or the spelling skills of the kindergarten children who participated in
the language activities group, as compared with the control group" (B&B, p. 49). That is,
phonics instruction alone did not lead to an improvement in reading skills; it had to be
combined with phonemic awareness instruction to have a significant effect.
Phonemic awareness training vs. no training (again). In contrast to the Tornéus study,
Lundberg et. al. (1988) found that first graders given phonemic awareness training did
not outperform a control group of first graders (given no instruction) on reading tasks.
Second graders in the same study did outperform their counterparts who were not given
phonemic awareness instruction. The authors hypothesize that with the earliest of
readers, phonemic awareness training alone may not be effective in facilitating reading
skills and may have to be combined with traditional reading instruction (as was done in
the Ball & Blachman study). By second grade, most participants in the study would have
already learned sound-symbol correspondences, thus explaining the difference in results
between the two age groups.
So, given the Lundberg et al. and Ball and Blachman findings (that neither reading
instruction alone nor PA awareness alone is the best training approach), what's a reading
teacher to do?
Phonemic awareness training & phonics instruction vs. phonics instruction alone vs.
phonemic awareness training alone. Hatcher, Hulme, and Ellis covered all the bases
in their 1994 study. They tested 7-year-old poor readers given a variety of training:
reading (phonics) with phonemic awareness training, phonics alone, phonemic awareness
training alone, and a control group which received no training. Even though all groups
were given the same amount of training, they found that the phonics with phonemic
awareness training group showed the most progress in reading (though the phonemic
awareness group showed improvement in phonological tasks). Bradley and Bryant
(1983), in a similar study, had found the same results. awk why not this one first?
Hatcher et. al. conclude that "the difference between the group trained in sound
categorization alone and the group who also received training in letter-sound
correspondences is notable... A natural implication of this result is that the integration of
training in phonological skills with letter-sound training (or more broadly with
phonemically based reading instruction) may be particularly effective" (p. 42). They
form the conclusion that "training in phonological skills in isolation from reading and
spelling skills may be much less effective than training that forms explicit links between
children's underlying phonological skills and their experiences in learning to read." (p.
42) This is called the "phonological linkage hypothesis."
In an attempt to test the weight of the phonological linkage hypothesis, Cunningham
(1990) examined the difference between implicit and explicit phonological training on
reading ability. In a 10 week training session, she gave one group of first graders "skill
and drill" (implicit) PA training, in which subjects were trained in how to segment the
phonemes in a word, or blend phonemes together to make a word. The students were
trained with the usual phonemic awareness training methods: phoneme deletion,
phoneme oddity, and phoneme discrimination tests (taken from the Williams program
"The ABDs of Reading"). Another group was given "metalevel" (explicit) PA training in
addition to the implicit training. This training was similar to the implicit phonemic
awareness training, but also emphasized the connection between phonemic awareness
skills (i.e., segmentation and blending), sound-symbol correspondences, and the
development of reading skills. The metalevel group was essentially taught the
phonological linkage hypothesis. Cunningham found that the children in the metalevel
training group performed better on reading tests (though not in phonemic awareness
tasks) why so than those who just received phonemic awareness training: "the children
who reflected upon and discussed the value, application, and utility of phonemic
awareness for the activity of reading at an explicit level performed significantly better on
a transfer measure of reading achievement than the skill and drill experimental group"
(p. 429). Cataldo & Ellis (1988) also found that explicit phonemic awareness was more
highly correlated with reading than implicit phonemic awareness.
Thus, the “phonological linkage hypothesis” is consistent with each of the studies cited
above: phonemic awareness training, when combined with traditional phonics training,
improves children’s reading skills more than no training, or training in phonics alone.
Whereas phonemic awareness training alone is effective is controversial, explicit
phonemic awareness training may be even more valuable.
A final note about phonemic awareness training: in addition to direct phonemic
awareness training programs, Adams also notes that learning to spell can be considered a
form of phonemic awareness training for children. She says,
The process of inventing spelling is essentially a process of phonics. Not surprisingly, then, the phonetic appropriateness of prereaders' invented spellings is found to be predicted by their level of phonemic awareness and to predict their later success in learning to read words....Evidence that invented spelling activity simultaneously develops phonemic awareness and promotes understanding of the alphabetic principle is extremely promising. (Adams, p. 387)
Read (1986) and Treiman (1993) have both done extensive work with children's invented
spellings. Treiman sees the process of learning to spell as reflective of the development
of phonemic competence, as demonstrated in the following chart
(from Treiman, 1993)
Precommunicative (no phonetic or phonological structure)
Semiphonetic (idea that letter=segment; ex. CRPT for "carpet")
Phonetic (no ommissions)
Traditional (mistakes reflect learned correspondences; ex. EIGHTEE for "eighty")
Correct
When children receive training in phonemic awareness, they will succeed in learning to
read (or so the argument goes). Mann (1993) found a correlation between the
sophistication of children’s invented spellings and their reading ability. This correlation
could be because the best invented spellers have, in a sense, received a dose of phonemic
awareness training.
A final conclusion drawn from phonemic awareness studies in children is discussed in the
next section.
2.1.5. Levels of phonemic awareness
4. There are different levels of phonemic awareness.
This is a claim that has already been discussed above, since it is impossible to separate
from other phonemic awareness issues. Children may innately possess a certain level of
phonological (syllabic) awareness, and they eventually work their way up to a level of
phonemic awareness that enables them to decipher the alphabetic code and become
literate. Adams (1990) identifies five levels of phonemic awareness as tested by
phonemic awareness tasks (these were outlined in the introduction) and Treiman (1983)
identifies levels of phonemic awareness which are revealed in children’s spelling errors
The models are related; children at the most primitive level show sensitivity to the sounds
of words, and children at the most sophisticated level are able to manipulate phonemes in
addition, deletion, and movement tasks. The best future readers show the most
sophisticated phonemic awareness.
2.1.6. Summary & implications
Though there is overwhelming agreement that phonemic awareness and reading ability
are connected, the exact nature of the connection is more controversial. One criticism of
the phonemic awareness/reading ability connection is that it is tautological (since
"children who have...acquired a solid level of phonemic awareness before entering school
have also begun to read before entering school" (Adams, p. 8)). Another criticism is that
researchers in the field of phonemic awareness are too committed to a cause-effect
relationship between phonemic awareness and reading ability, and have failed to consider
that some third factor may be affecting these skills (that is, the researchers have a
pedagogical ax to grind in the promotion of phonemic awareness curricula).
The phonological linkage hypothesis merits attention, but it also draws attention to the
fact that PA awareness research done on children attempts to draw conclusions about the
connection between PA and reading ability by examining children's reading ability, thus
muddling itself in a chicken-and-egg argument about the relationship between phonemic
awareness and reading skill. For this reason, nontraditional research on phonemic
awareness (research examining the nature of phonemic awareness in populations other
than preliterate children) also needs to be examined to gain a richer understanding of the
issues involved.
2.2. Phonemic awareness research on special populations
2.2.1. Overview
The claims about phonemic awareness and reading drawn from studies with children may
be questioned on a number of grounds. Is phonemic awareness really a prerequisite for
reading? Is there a causal connection between phonemic awareness and reading? Is
phonemic awareness training valid and/or useful? Are there other, as yet unexplored,
implications of these claims?
Work on phonemic awareness has not been limited to American children in phonics
programs. Moreover, some of the outside work on PA can shed light - or at least raise
questions - about the connection between phonemic awareness and reading ability. We
shall look at work done on PA in illiterates, literates of nonalphabetic orthographies,
dyslexics, and children with Down Syndrome.
2.2.2. Phonemic awareness in illiterates
Morais et. al. (1986) conducted a study on Portuguese illiterates and ex-illiterates (who
learned to read and write as adults). The study compared the two groups with respect to
their performance on a phoneme manipulation (PA) task, a syllable manipulation task,
and a melody segmentation task. The findings were that "the illiterates vs. ex-illiterates
difference was greatest in tasks designed to tap phonemic analysis, irrespective of the
particular operation to be performed, was present but smaller in tasks involving syllables
and absent in a non-linguistic segmentation task (melody segmentation)" (Bertelson
et.al.., p. 114).
The only significant difference was in the phonemic segmentation abilities of the
illiterates (who could not do PA tasks) and the ex-illiterates (who could do PA tasks).
From this finding, the researchers conclude that "while sensitivity to rhyme and analysis
into syllables can develop up to some point in the absence of the experience normally
provided by reading instruction, analysis into phonetic segments requires that experience"
(Morais et. al., 1986, p. 45). So, phonemic awareness is not a maturational development,
but rather a fallout from reading instruction. If this is true, it would seem to violate the
claim that children need to be phonemically aware to begin reading; rather, it appears
that they would become phonemically aware because of reading experience.
Morais (1991) appeals to the idea of "levels" of awareness, but he is talking about
phonological vs. phonemic awareness: "phonological awareness subsumes at least the
following: awareness of phonological strings (a global, nonanalytical level of
awareness); awareness of syllables; awareness of phonemes (also called segmental
awareness); and awareness of phonetic features...it is widely accepted that literacy
instruction is not necessary to elicit all these forms of phonological awareness" (Morais,
1991, p. 6). Since illiterates can manipulate syllables but not phonemes, they do have a
certain level of phonological awareness, an awareness which develops (or more
correctly, is present) without knowledge of an alphabetic orthography. Morais concludes
from the Portuguese study that though phonological awareness may develop in the
absence of reading instruction, phonemic awareness requires that experience.
The Morais studies raise the question, is experience with an alphabet necessary for the
development of phonemic awareness? The study was criticized by Koopmans (1987),
who argues that the "difference between literates and illiterates on the phonetic
segmentation task may be attributable to the lack of formal schooling as such in the latter
group" (Koopmans, P. 110). It is not so clear, according to Koopmans, that illiterates
perform poorly on segmentation tasks simply because they lack phonemic awareness.
Rather, "unschooled subjects, it appears, do not do well on such tasks, not only because
the task contents are often unfamiliar, but even more so because the activity of working
with abstract material is in itself unfamiliar to them" (Koopmans, p. 100). Morais et. al.
disregard this criticism, citing the fact that illiterates did not have impaired performance
on a melody segmentation task, which presumably taps into the same abstract and formal
skills as speech segmentation.
So, the main finding of PA studies with illiterates is that PA follows from (and requires)
experience with an alphabet, rather than developing on its own. What does Morais make
of findings in later studies that illiterates showed significant improvement in
segmentation skills after 20 or 30 training trials? Could Koopmans be right that they did
not understand the task? Morais claims "no," and that there is "no clear indication that
these improvements reflect phonemic awareness...the improvements were due to a
nonsegmental strategy" (Morais, 1991, p, 15). He attributes the improved performance to
the specific training in a task-driven strategy, rather than the speedy acquisition of true
phonemic awareness.
2.2.3. Phonemic awareness in literates with nonalphabetic orthographies
The Morais et. al. study suggests that phonemic awareness absolutely requires alphabetic
experience and will not develop in its absence. On the surface, studies by Read et. al
(1986) and Mann (1986) support this. The study by Read et. al. compared the
performance of two groups of Chinese adults. The first group knew only the traditional
logographic (character) writing system, whereas the second was also familiar with the
pinyin alphabet. Only the latter group (that had received explicit training in the
alphabetic system) performed well on phoneme awareness tasks. Mann's study compared
the performance of American children and Japanese children on syllable and phoneme
manipulation tasks. Though both groups of children performed well on the syllable
manipulation task, only the American children, familiar with an alphabetic writing
system, performed well on a phoneme manipulation task. Japanese children, who knew
only a syllabary, performed poorly on these tasks.
Thus, the Read and Mann studies at first appear to support the idea that PA develops only
in the face of direct alphabetic instruction.
However, the data on non-alphabet readers is not so simple. Mann (1991) cites an
unpublished study of Chinese literates (by Tzeng and Chang) that concludes that there
can be phonemic awareness in the absence of knowledge of pinyin. Mann states that,
“readers of the logography found phoneme deletion an "easy" task, despite being illiterate
in pinyin. Those who were totally illiterate performed less well" (Mann, 1991, p. 59).
So, the literate adults could be easily trained to perform well on a PA task despite lack of
alphabetic experience.
In a similar finding, Mann (1986) found that Japanese fourth graders, despite lacking an
alphabet, performed well on PA tasks (Japanese sixth graders, having learned romanjii, or
the roman alphabet, perform well on these tasks, too, as expected). This suggests that
"with increasing age and educational experience Japanese children may become more and
more capable of manipulating phonemes whether or not they are alphabet-literate"
(Mann, 1986, p. 87).
These results conflict with Morais et. al.'s findings on Portuguese illiterates. Since Tzeng
& Chang’s Taiwanese logographic readers and Mann’s Japanese fourth graders have had
no alphabetic training, why were both groups able to perform well on phonemic
awareness tasks? There are two possible explanations. The first, suggested by Mann, is
akin to a "critical period" for the acquisition of orthography: "perhaps the ability to
manipulate phonemes tends to atrophy unless maintained by appropriate reading
experience" (Mann, 1986, p. 89). "Appropriate reading experience" need not be
alphabetic literacy, but must be some form of literacy. Taiwanese literates can do PA
tasks as long as they receive an adequate training session that enables them to understand
the task at hand. Japanese fourth graders have had enough reading experience to perform
analytical tasks that their first grade counterparts cannot ("it would seem that the age of
the child has an impact on the degree of phoneme awareness" (Mann, 1991, p. 58)).
The Portuguese illiterates, on the other hand, have had no such reading experience, which
explains their inability to perform phonemic awareness tasks at any age.
However, this “critical period” theory would not explain why the older Chinese and
Japanese literates in the Read and Mann studies performed poorly on the phonemic
awareness tasks, as they had clearly had “appropriate reading experience,” in the sense
that they learned to read at a normal age, an age during which phonemic awareness could
have developed.
A second explanation is that subjects who perform well on phonemic awareness tasks
have received a form of alphabetic instruction, though it may be indirect. Morais
suggests that the PA abilities of Japanese fourth graders who had no exposure to an
alphabet
can be understood by taking into account the way kana is taught. Japanese children learn kana by means of a matrix in which all the characters in a row share the same vowel and all the characters in a column share the same consonant, except the first column to the right that consist of isolated vowels. It is possible to obtain a good performance in the deletion task by referring to the matrix (Morais, 1991, p. 17).
He concludes, contra Mann, that PA does not develop with any reading experience, but
requires attention to segments. However, the segmentation instruction does not have to
be as explicit as learning an alphabetic writing system. This explains Mann's results with
Japanese children and may also explain the Tzeng and Chang with Chinese readers
illiterate in pinyin, if some aspect of language pedagogy includes, however subtle,
attention to the initial segments of a word. Fangquie! This could also explain why
phonemic awareness skills were found with younger readers of nonalphabetic writing
systems, but not their older counterparts (if the older subjects were taught kana and
characters through a different method of instruction - or at home - or if the older subjects
were far enough removed from their school-days instruction that their phonemic
awareness skills had atrophied).
However, the “subtle” phonemic awareness instruction idea still fails to account for good
performance on PA tasks by Portuguese illiterates (after an adequate training session).
Commenting on the improved performance of illiterates on a PA task (from 24% correct
responses on the first block of items to 69% on the last block), Morais says that "the
capacity necessary to analyze speech at the segmental level does not seem to atrophy with
age in the absence of reading instruction. There seems to be no critical period for
acquiring segmental awareness" (Morais, 1987, p. 135)
The seeming phonemic awareness of groups who have not had exposure to an alphabet
might also be attributed to secondary language activities (more in the next section).
2.2.4. Phonemic awareness in dyslexics
Dyslexics are an interesting case to examine with regard to the relationship between
phonemic awareness and reading, since they have had the same exposure to print as
normal readers, yet fail to perform well on PA tasks. Though the studies cited in the
previous section seem to suggest a causal relationship between alphabetic experience and
ability to perform PA tasks, "dyslexics have had much experience with alphabetic
material, therefore, experience cannot be the critical factor" (Morais, 1987, p. 135).
Morais (1987) cites a number of studies that demonstrate the relative inability of
dyslexics to carry out PA tasks. In his own 1984 study, Morais found that dyslexics had
only a 14% correct response rate on a consonant deletion task, versus 71-95% correct for
younger normal readers. Savin (1972) observed that young dyslexics can't play Pig
Latin, a game which requires phoneme manipulation. Lecocq (1986), studying groups of
dyslexic and normal 8-9, 10-11, and 12-13 year-olds, found performance on consonant
deletion tasks was significantly worse for dyslexics, who scored between 51% - 78%
correct (vs. 80-100% correct for normal readers). In other words, dyslexics seem to lack
any consistent phonemic awareness.
Somewhat encouraging is the work of Fox and Routh (1983), which showed that after
much training 11-13 year old dyslexics were able to perform a PA task with 96%
accuracy. This is unlike the case of Chinese literates or Portuguese illiterates, in that
"training" here was not merely task repetition after a short and repetitive training session,
but rather years of instruction on phonemic awareness skills. Thus, while the Chinese
and Portuguese cases could be interpreted as evidence that PA develops without explicit
alphabetic instruction, the evidence from dyslexics shows that even with direct alphabetic
instruction, phonemic awareness may fail to develop.
2.2.5. Phonemic awareness in Children with Down Syndrome
Data from children with Down Syndrome are relevant to the discussion of the
relationship between phonemic awareness and reading because "if in some abnormal
group reading is acquired in the absence of ability to perform explicit segmental analysis
tasks, that would seem prima facie evidence that no necessary causal connection holds
between the two skills" (Cossu et. al., p. 130). Cossu looked at the PA skills of Children
with Down Syndrome (mean age 11.4) matched for reading ability with normal children
(mean age 7.3). On four PA tasks (phoneme segmentation (counting), phoneme deletion,
oral spelling, and phonemic synthesis (blending)) the normal children far outperformed
the children with Down Syndrome. Thus, two groups of children with the same level of
reading ability showed different levels of phonemic awareness.
Some might argue that the Down Syndrome subjects performed badly on the PA tasks
merely because they didn't understand what they were supposed to attend to in their
responses. To this, Cossu says that "it would... be tautological to argue that the children
with Down Syndrome merely failed to understand the task demands for the phonological
awareness test: with respect to conscious skills, failure to understand the nature of the
task is failure to be able to perform the task" (Cossu et. al., pp. 134). Since phonemic
awareness is conscious attention to the segmentation of words, the failure to understand
what is meant by "leaving off the first sound of a word" is the failure to be phonemically
aware. O’Connor (1992), et.al., citing their own research with children with Down
Syndrome, conclude that the poor phonemic awareness performance on the part of their
subjects is partially related to their inability to generalize from trained practice items to
test items.
Since children with Down Syndrome perform so much worse on PA tasks than children
they are matched on reading skills with, Cossu concludes that "it is neither the case that
lack of phonological awareness has prevented learning to read, not that learning to read
has developed phonological awareness" (Cossu et. al., p. 134). As with the data from
dyslexics, this data begins to break down presumed strong causal connections between
phonemic awareness and reading ability.
2.2.6. Summary & implications
Research on illiterates, literates with nonalphabetic orthographies, dyslexics and children
with Down Syndrome has interesting implications for phonemic awareness research.
Studies showing that illiterates and literates with nonalphabetic orthographies are unable
to conduct phonemic awareness tasks would seem to indicate that phonemic awareness
results from, and only from, alphabetic instruction. However, dyslexics are exposed to
alphabetic instruction and do not acquire phonemic awareness. Though it could be
argued that dyslexics are also poor readers (so the phonemic awareness/reading
connection holds), Children with Down Syndrome performed worse on phonemic
awareness tasks than their reading-matched counterparts.
Other studies showing that illiterates (with training) and literates with nonalphabetic
orthographies (and indirect exposure to phonemic awareness concepts through method of
language instruction) can perform well on phonemic awareness tasks imply that
phonemic awareness does not have a critical period, nor does it need to be explicitly
taught to be acquired (in keeping with whole language theorists).
This conflicting information can be reconciled with a modified version of Morais’
conclusions from his original study with Portuguese illiterates. Phonemic awareness is
not a developmental milestone; it does not necessarily develop, and it is not subject to a
critical period. It only develops in the context of certain phonemic exposure, though that
exposure may be through direct exposure to an alphabet, indirect exposure to an alphabet
through language pedagogy, or, as discussed in the next section, through secondary
language activities.
2.3. Phonemic awareness and secondary language activities
The studies cited above explore how primary language activity affects phonemic
awareness in special populations, be they readers of other orthographies, illiterate, or
differently abled. What role do secondary language activities play? Secondary language
activities are the games that people play with language as part of everyday life: poems,
songs, and wordplay. They differ from primary language activities (communication-
based activities) in that they involve a particular attention to the structure of language that
primary language activities do not. Though a poet cannot write a haiku without counting
syllables, such counting does not occur in normal discourse. Experience with secondary
language activities may explain why it is that some illiterates can perform phonemic
awareness tasks, or that Chinese unfamiliar with pinyin could break a word up into
segments.
The most extreme view of the studies with readers of nonalphabetic orthographies (Read
et. al; Mann) and illiterates (Morais et. al.) is that the development of phonemic
awareness skills requires alphabetic experience. In addition to the criticisms already
raised, such a statement implies that no one who has not been exposed to an alphabet can
break a word into its component segments. A number of considerations lead us to
conclude that this is far too strong a position.
First, "phoneme awareness can sometimes precede the ability to read an alphabet "
(Mann, 1991, p. 56). Children often enter school able to perform well on PA tasks,
although they have not yet learned to read. How can this be so? Mann says, "factors
other than knowledge of the alphabet may contribute to the awareness of phonemes"
(Mann, 1991, p. 55). So, even without being "taught" the alphabet through reading with
parents or watching educational TV, a child may develop phonemic awareness. What
secondary language activities contribute to phonemic awareness?
Adams cites knowledge of nursery rhymes as an indicator of the most primitive level of
PA, and this is why: rhymes help children learn about spelling categories and develop
phonemic awareness, by learning to break down words into units smaller than the
syllable. In their own study, Bradley and Bryant (1991) found that "preschool
phonological skills play an important part in reading acquisition. rhyme scores...prove to
be reliable predictors of reading ability" (Bradley and Bryant, p. 40). When children
learn to rhyme, they are learning that there are units of words (rimes) that are bigger than
phonemes, but smaller than syllables. This is an intermediate step between syllabic
awareness and phonemic awareness. The role of rhyming games in the development of
phonemic awareness is furthermore a very “whole language” approach to the acquisition
of phonemic awareness, in that it is not explicit. Rhyming as a route to phonemic
awareness is extremely similar to the kana example cited in the previous section. A 3-
year-old American child and a school-age Japanese child might not have been exposed in
any direct way to an alphabet, and yet both may have acquired some phonemic awareness
skills. The 3-year-old has heard a nursery rhyme which groups rhyming words in a
regular fashion (hickory/dickory; dock/clock); the Japanese child has had repeated
exposure to rhyming kana arranged together in a grid (po, mo, do, so...). Both may look
for a way to organize the input; this is found by stripping off the first phoneme.
Secondary language activities may explain why Lundberg (1991) found that 9 out of 51
prereaders gave perfect performance on a segmentation task, even though the children
had participated in similar language activities in and outside the school. Granted, far
more prereaders failed than succeeded on the PA task, but in the absence of alphabetic
instruction why should even 9 students perform so well? Lundberg suggests that
"whereas most researchers now seem to agree that the relationship between phonological
awareness and reading is reciprocal in nature, the question of which affects the other
most and earliest is far from settled" (Lundberg, p. 47). If we adopt Adams' levels of
phonological awareness, we can see where children's secondary language activities play a
role in the development of their PA skill. Yet, at the same time, there does seem to be a
maturational component which cannot be denied: some tasks or secondary language
activities are simply too sophisticated for young children.
If nursery rhymes indicate the most primitive level of phonemic awareness, what comes
next? A number of secondary language activities might hone phonemic awareness skill
(or, rather, hone the analytical aspect of a skill which already exits). For instance,
children love tongue twisters, which require a sensitivity to the fact that all words in a
string begin with the same phoneme (for a child to "get" what makes "Peter Piper picked
a peck of pickled peppers" a tongue twister, he or she must be able to separate/identify
[p]). Language games can tap into various levels of phonemic awareness. In Pig Latin,
the child need only segment and transpose the first onset in a word: latin > atinlay. In
"Op," the child has to be able to segment the onset of every syllable (not just the initial
syllable) away from the rime: segment > sopegmopent. In the "name game," the child
must be able to manipulate both the onset and the rime of the input in constructing the
output: Ken > ken ken bo ben, banana fana fo fen, me my mo men, KEN! Each of these
activities support the idea that "phoneme awareness is facilitated by some other
secondary language experience" (Mann, 1986, p. 88).
Given that secondary language experience can facilitate phonemic awareness, the
seemingly anomalous results of illiterates or non-alphabetic literates on PA tasks can be
explained. Perhaps those subjects did not have explicit instruction in an alphabet, but did
participate in secondary language activities which foster phonemic awareness. Mann
mentions a number of language games in cultures lacking an alphabetic orthography. In
Luganda (Kilbride & Kiblbride, 1974)), a game requires that each syllable be followed by
"z" and the previous vowel: omusajja > omuzusazajjaza. In a Bedouin Hijaze Arabic
game (McCarthy, 1981) single consonants can be metathesized: kaatab > bataak; taakab
> taabak. In one Mandarin game(Yip, 1982), the syllable is split and infixed: ma >
mayka. Though these games occur in languages lacking alphabetic orthographies, the
game-players have certainly developed some level of phonemic awareness.
In this and the previous section, the connection between phonemic awareness and reading
ability has been discussed. Most of the research in this area has been conducted on
populations with little or no exposure to an alphabet (preliterate children, illiterates,
literates with nonalphabetic orthographies) or on populations for whom phonemic
awareness tasks present a special challenge (dyslexics, Children with Down Syndrome).
Before conducting phonemic awareness studies on adults who have had significant
exposure to an alphabet, it is essential to explore the impact alphabetic literacy might
have on such tasks.
2.4. Special considerations: the question of orthographic interference
The previous sections in this chapter contained an overview of past research on phonemic
awareness. As discussed in the previous sections, most of the work in the field has been
done with preliterate children, illiterates, or literates in nonalphabetic orthographies:
those with little or no experience with alphabets. However, there is a wide variety of
evidence that knowledge of orthography, or of a particular orthography, can affect
performance on various phonological tasks. Since the subjects studied in this project
were all literate adults, we must examine the role that orthographic knowledge might play
in a phonemic awareness task.
The work discussed earlier in this chapter makes it clear that literacy (specifically,
alphabetic literacy) affects performance on various phonological tasks. This relationship
manifests itself in the ability to perform phoneme manipulation tasks, and in the encoding
of grapheme-phoneme relationships. But, just what is the nature of this relationship?
How might it affect subjects’ performance on phonemic awareness tasks?
One possibility is that alphabetic literacy might play a role in auditory lexical access.
Since phonemic awareness tasks require a subject to process an auditory stimulus, this
could be problematic.
2.4.1. The role of orthography in phonological access
In fact, there is evidence that orthography interferes in auditory tasks. Surprisingly,
literacy can, in a sense, impede performance. There is a positive correlation between
phonemic awareness and reading ability in studies with children, indicating that
phonemic awareness helps children break the alphabetic code, and go on to become better
readers. Might it be the case that the best adult readers - those most intimate with the
orthography of their language - are hindered by their spelling ability in phonological
tasks? Two studies which have addressed this issue are discussed below.
Tanenhaus, Flanigan & Seidenburg (1980) presented subjects with a series of auditory
word pairs, and asked subjects to ascertain whether or not the pairs rhymed. Some of the
rhyming pairs were orthographically similar (like TURN/BURN) and some were
orthographically distinct (like TURN/LEARN). The main finding of the study was that
orthographically distinct pairs of words took much longer to identify as rhymes than pairs
that were spelled alike. Thus, at some point between the time when subjects are
presented with auditory stimuli and when subjects depress a response key, orthography
interferes.
A second study showing orthographic interference in a phonological task was Taft &
Hambly's 1985 syllable matching study. Subjects were presented with pairs of auditory
stimuli - a syllable and a word - and asked to determine whether the syllable was
contained in the word. Subjects performed particularly poorly on pairs such as
ANK/ANXIOUS ([æk]/[ækes]). These were identified by Taft & Hambly as those
pairs where orthographic interference was likely. Their argument is this: presumably,
though [ækRes] is spelled a-n-x..., [æk] is "spelled"7 a-n-k. Due to an orthographic
mismatch (“x” vs. “k”), subjects would give such items a "no" response, even though, on
the basis of phonology, a "yes" response was expected.
How is the orthographic interference demonstrated in these tasks to be interpreted? As
Taft and Hambly point out, "the interpretation that one favors ultimately depends on how
dominant a role one wishes to ascribe to orthography in spoken word processing" (T&H,
p. 331). This is best illustrated using a lexical access model, such as the simplified
sketch of the Forster model of lexical access (Forster 1976;1990), given below:8
<visual stimulus> <auditory stimulus>
orthographic access code phonological access code
MASTER LEXICONmaster lexicon entry
(orthographic information/phonological information)
According to the strictest interpretation of the Forster model, orthography plays a
minimal role in spoken word processing. Given an auditory stimulus, a purely
phonological code is used in lexical access. Only after access has taken place is
orthographic information about that word (stored in the Master Lexicon entry) available.
7 That is, if subjects are constructing an orthographic representation of /æk/ the one they would construct is a-n-k.
8 Note that Forster's model also includes a semantic access code, which has been omitted here for the sake of simplicity.
An alternative to the Forster model would ascribe a dominant role to orthography in
spoken word processing. Contra the Forster model, such a model of lexical access would
look something like this:
<auditory stimulus>
orthographic access code phonological access code
MASTER LEXICONmaster lexicon entry
with both orthographic and phonological codes involved in the access of an auditory
target. A third alternative lies somewhere in between these two extremes, with
orthographic information neither ignored during auditory word processing nor given the
same importance as phonological information.
Zecker et. al. (1986) lay out these possibilities nicely. In their work, they argue that
phonological and orthographic codes are integrated in the left hemisphere of the brain
(with this finding based on the results of a rhyming judgment task, where they found a
significant interaction between ear of presentation and orthographic similarity of stimulus
items9). Zecker et. al. posit three alternatives - granting various dominance to the role of
orthography in auditory word recognition - to "the relationship between the access of
9 It took subjects considerably longer to judge that orthographically dissimilar pairs like "dirt-hurt" rhymed than orthographically similar pairs like "glue-clue." This is the same finding as that of Tanenhaus, Flanigan, & Seidenberg, cited earlier. However, Zecker et. al. additionally found that this effect was significantly greater when stimulus items were presented in the right ear (i.e. the left hemisphere) than the left ear. They conclude from this that orthographic and phonological information are integrated in the left hemisphere (hence the large orthographic interference).
sound-based codes in visual word recognition and spelling-based codes in auditory word
recognition." We will explore these possibilities in turn, in hopes of discovering what
effect literacy will have on the tasks in the next section.
Zecker et. al. point out that "one possibility is that code integration [that is, orthographic
interference in tasks with auditory stimuli] results from the child learning spelling-sound
mapping rules during beginning reading. Spelling effects in auditory word recognition
are the result of spelling-sound mappings becoming automatized" (Zecker et. al., brackets
mine). In this case, what appears on the surface to be orthographic interference during
access is really the result of "automatized" comparisons of post-access activated spellings
with their pronunciations. To make this more concrete, consider the two tasks discussed
above. In the rhyming recognition task, the stimuli (LEARN/TURN; TURN/BURN) are
accessed via a purely phonological code, but once the words are accessed, their Master
Lexicon entries, orthographic representations and all, are activated. A post-access check
ascertains that l-e-a-r-n and t-u-r-n are not spelled the same, even though "automatized"
sound spelling rules (which would say something like, e-a-r-n sounds like [ern], u-r-n
sounds like [ern]) predict a match. This contradictory information results in a longer
reaction time. The same is true in the ANK/ANXIOUS case. Once ANXIOUS is
accessed, its spelling is available and will thus cause interference in the form of a
mismatch when compared with the speech signal ANK.
Zecker et. al. raise the "second possibility ... that there are sound-spelling rules which are
similar to spelling-sound rules. These rules may be learned by the child in the course of
learning to spell." So, just as a child learning to read learns a rule that the letter 's' sounds
like /s/, that same child, in learning to spell, learns that "long i" (/ay/) is spelled 'i...e.'
This interpretation of orthographic interference as "orthographic recoding" (application of
on-line sound-to-spelling conversion rules) is analogous to the "phonological recoding"
(see Taft, 1991 and others) discussed in the literature on written word recognition. It is a
dual-route access model, whereby both phonological and orthographic codes are used in
auditory word recognition.
However, it is not as clear why an orthographic code might be an automatic part of
spoken word recognition as it is that a phonological code might be an automatic part of
written word recognition. Phonological recoding is hotly debated, even though it has
support in the form of intuitions (the "voice in the head" people hear when they read),
development (children learn to speak before they learn to spell, making phonology the
primary form of linguistic representation) and history (spoken languages existed long
before writing systems were developed for them). In the absence of such support for
orthographic recoding, why should the possibility even be entertained?
Dupoux & Mehler (1992), in their analysis of French phoneme detection tasks, provide
arguments that "not only is an orthographic code available when listening to speech....but
this code is available on-line" (p. 65). They attribute differences in performance on
syllable monitoring tasks by literates and illiterates to the fact that an on-line
orthographic code, which results in orthographic interference among literates, is not
available to the illiterates. Although Dupoux & Mehler are not talking about on-line
orthographic recoding in the sense of sound-spelling conversion rules (in fact, the
conclusions they draw rely on the subjects' awareness of segments), they do not want to
“ tern/bern”
orthographic recoding phonological codingX-u-r-n / X-u-r-n X-ern/X-ern
draw any conclusions about precisely what is contained in the orthographic code: "we do
not have the data to ascertain that the relevant code is really a visual representation of the
letters of the characters, instead of the relevant phonological level that the orthography
captures" (Dupoux & Mehler, p.72). In the tasks that they analyze, the "relevant
phonological level that the orthography captures” is the level of segment vs. syllable.10 It
could well be that the "relevant phonological level" in a different task is the level of
phoneme-to-grapheme conversion rules. Again, a consideration of the previously
discussed tasks serves to make this more concrete. If it is the case that orthographic
recoding happens on-line, this is what happens with the rhyming task:
TURN/BURN case:
In the case where two words are pronounced alike, the sound-spelling conversion rules
will yield the same spelling for both words. If the words are in fact spelled alike, this will
not be a problem once the words are accessed. On the other hand, consider the case in
which two words pronounced alike are spelled differently:
TURN/LEARN case:
10 Dupoux et. al. examine the results from syllable monitoring tasks. They cite orthographic interference as the reason that subjects will sometimes say (erroneously) that the syllable "pal" is contained in "palace." Though the two words are mismatched on syllabic grounds (the first syllable of "palace" is "pa"), on orthographic grounds (that is, at the level of the segment), the two stimuli are matched, in that they both consist of the sequence p-a-l.
“tern/lern”
orthographic recoding phonological codingX-u-r-n / X-u-r-n X-ern/X-ern
OR
Again, since the words are pronounced alike, orthographic recoding would spell them in
the same way . The outputs of two possible spelling-sound conversion rules are listed
above. No matter which rule is chosen, such recoding would cause interference (since
*lurn and *tearn would not be listed in the lexicon).
Thus, on-line orthographic recoding is a second possible explanation for orthographic
interference effects in tasks with auditory stimuli.
In addition to the two possible explanations for orthographic interference listed above,
Zecker et. al. propose a third possibility for the role of orthography in auditory word
recognition, which lies somewhere in between the dominant and minimal roles discussed
thus far: "the child learns to integrate spelling and sound information as a learning
heuristic." Thus, the access route for auditory words is not purely phonological (latter
tapping into orthographic information), nor orthographic and phonological in parallel, but
would rather involve some integrated code. This account is favored by Taft and Hambly,
who conclude that "the phonological task is performed in a phonological code, though
this code is influenced by orthography" (T&H). Lexical access taps into what Taft and
Hambly call "orthographically influenced phonological representations." These might
include something like the “spelling pronunciations” that Lorenson (1993) found:
subjects have trouble classifying [izlænd] (for island) and [kolonel] (for colonel) as
nonwords, apparently because these pronunciations warrant a pseudo-lexical entry (an
"orthographically influenced phonological representation”) in the brain.
So, orthographic knowledge does interfere with performance on phonological tasks, be it
through phonologically accessed orthographic representations, orthographic recoding, or
orthographically influenced phonological representations. This must be taken into
consideration in the six experiments in the next chapter, which are designed to assess the
connection between phonemic awareness and reading ability in literate adults. The tasks
in these experiments are based on tasks used in previous phonemic awareness studies,
most of which involve children. In each of the experiments with children, task
performance is gauged by subject error rate on phonemic awareness tasks. Since adults
have greater cognitive capabilities than children, and since moreover they have greater
linguistic capabilities than children, the tasks used in experiments with children are not
challenging enough to adults to yield any interesting results. For instance, all literate
adults can, with nearly 100% accuracy “count” the number of sounds in a word in which
there is a one-to-one correspondence between the number of phonemes and the number
of graphemes; adults are helped in this task type by their orthographic knowledge.
However, the orthographic knowledge that adults possess might be enough of an
impediment that it can be used against them, e.g. by generating higher error rates on
phonemic awareness tasks. The pilot task described below tested this assumption.
2.4.2. Pilot: Delete Segment Task
A task used in many phonemic awareness studies with children is the Delete Segment
Task (see Bradley and Bryant, 1983, among others). In this task, children are asked to
delete the initial phoneme in a word. A pilot experiment was conducted to determine if a
segment deletion task would suggest that adults possess differing levels of phonemic
awareness. The purpose of the experiment was to alter the original methodology
minimally, but enough to generate errors among the adult subjects.
METHOD
Subjects
Subjects were thirteen University of Arizona undergraduates, all of whom were native
speakers of English.
Tasks and Procedure
In segment deletion tests with children, the instructions often are something like, “Listen
to the word task. If you take away the /t/ sound, what word is left?” (Stanovich et. al.,
1984; Cunningham, 1990) and the stimuli are such that when the first sound is stripped
off, a real word of English remains (for example, pink, told, man ink, old, an). The
stimuli are words which begin with a consonant-vowel sequence, not a cluster. Literate
adults find this task extremely easy, especially since the real-word status of the correct
response serves as a “check” to their answers. It was determined that replicating the
Stanovich/Cunningham segment deletion task with adults would fail to elicit enough
errors to be at all illuminating.
However, it was important to remain faithful to the original methodology, since the
experiments in this study attempt to draw the same kinds of conclusions about phonemic
awareness and reading as experiments with children. The least drastic means of adapting
the experiment to adults was to retain the original task and procedure while replacing the
stimuli with much more challenging items. The instruction to subjects was, “I’m going to
give you a word, and I want you to tell me what that word would sound like if the first
sound were taken off.” They then heard the task modeled on ten practice items, which
were followed by thirty test items. Subjects heard instructions through a set of
headphones and spoke responses into a hand-held microphone.
The stimuli in the Cunningham and Stanovich experiments were monosyllabic words
beginning with a CV sequence which were still words of English when the first
consonant was deleted. In this experiment, the stimuli were bisyllabic words beginning
with a variety of sequences (more below) which are nonwords when the first consonant is
deleted.
Based on the studies on orthographic interference and phonological access cited in the
previous section, it was determined that one way to make the stimuli more difficult was
to include stimuli which lack a one-to-one relationship between the phonemes and
graphemes of their initial segment (the segment to be deleted). Because aural stimuli
would call up “orthographically influenced phonological representations,” orthography
could be used as a tool to make stimuli more challenging.
Each stimulus word fell into one of the following four categories:
1. a single phoneme spelled with a single consonant, like baby [bebi] (a one-
to-one correspondence of phonemes and graphemes)
2. a single phoneme spelled with a digraph, like thesis [iss] (a one-to-two
correspondence of phonemes and graphemes)
3. a consonant cluster, spelled with a single grapheme, like music [myuzk]
(a two-to-one correspondence of phonemes and graphemes)
4. a consonant cluster, spelled with more than one grapheme, like climax
[klymæks] (a two-to-two correspondence of phonemes and graphemes)
It should be the easiest for subjects to delete an initial segment from a stimulus in
category (1), since doing so does not require extraction of a phoneme from a cluster or
involve a mismatch of phonological and orthographic representations. Conditions (2) and
(3) should present subjects with some difficulty if they are using orthographic
representations in their task strategy (that is, they may give [hiss] as a response to thesis
in the segment deletion task, or [uzk] as a response to music). Finally, conditions (3)
and (4) should be difficult because they begin with clusters, so segment deletion requires
subjects to extract the initial segment from the rest of the onset.
Results
Responses were considered correct if subjects deleted the first (and only the first)
phoneme in a word. The test items and correct responses are given in the table below:
Stimulus Stimulus (IPA) Correct Response
theory iri Iri
beauty byui Yuri
scarcely skersli kersli
pupil pyupl yupl
humor hyumr yumr
parish per er
champagne æmpen æmpen
tiny tyni yni
storage stord tord
colleague klig lig
thermal rml rml
frantic fræntk ræntk
sentry sentri entri
shaky eki eki
quarter kwrtr wrtr
climax klymæks lymæks
gentile dntyl ntyl
fifty ffti fti
puny pyuni yuni
chimney tmni mni
penny pni ni
cubic kyubk yubk
captive kæptv æptv
feudal fyudl yudl
music myuzk yuzk
shadow ædo Ædo
pony poni Oni
thesis iss iss
temper tmpr mpr
baby bebi Ebi
Subjects made the fewest errors on words beginning with, and spelled with, a single
consonant. Performance on the other three conditions was compared to that condition, as
shown in the table below:
Condition Example T-test comparison Significance(1) one-to-one baby --- ---(2) one-to-two thesis 2.23 .05 > p > .01(3) two-to-one music 6.87 p < .001(4) two-to-two climax 10.94 p < .001
As predicted, conditions (2)-(4) were all significantly more difficult than condition (1).
However, the difference in subjects’ performance on conditions and (1) and (2) was only mildly
significant, whereas the differences in performance between both (1) and (3) and (1) and (4)
were strongly signifcant. Since (3) and (4) are the conditions in which subjects are asked to
extract an initial segment from a cluster, it may be concluded that the cluster condition requires a
more sophisticated level of phonemic awareness than extraction of an onset from its rime (as
required in condition (1)). This finding is consistent with Adams’ levels of phonological
awareness, in that onset/rime knowledge is at a lower level of phonological awareness than
segment manipulation. Performance on condition (2) vs. (1) indicates mild orthographic
interference in this phonological task. However, subjects found it harder to delete a segment that
was part of a cluster (regardless of how it was spelled) than a segment which was spelled like a
cluster.
There are two conclusions to be drawn from this study. First, orthography does play a
part in phonological tasks (involving auditory stimuli), but orthographic interference
(orthographic difficulty) takes a back seat to phonological difficulty. The second
conclusion is more practical: if phonemic awareness tasks based on experiments with
preliterate children are to be made more difficult for literate adults by using different
stimuli, stimuli which are chosen on the basis of a difficult phonological pattern will
generate more errors than stimuli chosen because of a confusing
phonological/orthographic mismatch.
2.5. Summary
At the beginning of this chapter, a number of questions about the connection between
phonemic awareness and reading ability were raised:
What is the exact nature of the relationship between phonemic awareness and reading ability? Is one a prerequisite for the other, or do the two develop simultaneously?
Are there levels of phonemic awareness, or is it a developmental on/off switch, in that it is either present or absent?
Is phonemic awareness teachable? Why do some people develop phonemic awareness, and others not? Other than explicit instruction in an alphabetic orthography, what types of
activities foster phonemic awareness? How might knowledge of a specific orthography affect a subject’s ability to
perform a test of phonemic awareness?
The research cited in this chapter leads us to the following view of phonemic awareness:
The relationship between phonological awareness and reading ability is strong; the two
measures are highly correlated in studies with preliterate children and young readers. The
question has arisen as to whether phonological awareness fosters reading ability or
whether reading experience breeds phonemic awareness. Many researchers who work on
emergent literacy in children have viewed the PA/reading relationship as causal in the
sense that phonemic awareness tasks can “predict” later success or failure in reading; the
children who are most phonemically aware become the best readers. However, this has
been criticized as a "catch-22" claim, in that children who perform well on PA tasks are
perhaps those who have already had some reading-related experience, however indirect it
might be.
Some argue that phonological awareness is a prerequisite for learning to read an
alphabet. This is difficult to reconcile with the findings that most illiterates and readers
of non-alphabetic orthographies lack phonemic awareness. Research with these
populations has shown that phonemic awareness is a consequence of reading (alphabetic)
experience; it does not develop in a vacuum. If phonemic awareness is a prerequisite for
reading an alphabet, and yet does not exist in many who lack alphabetic experience, how
does it develop?
Morais (1987) argues the point that phonemic awareness develops only from alphabetic
literacy:
It is important, in my view, to distinguish between the ability of segmental analysis and the cognitive capacity that underlies this ability ... the
alphabetic system of writing is a cultural product, used by only a minority of the more recent humans. And alphabetic literacy is probably the only function of segmental awareness. When, then, should people be born with a specific capacity for segmental awareness that would be only waiting for a particular experience to develop? (Morais, 1987, p. 134)
Morais’ theory has support in research on populations who do not develop phonemic
awareness. Phonemic awareness fails to develop in certain people either because they
(1) exhibit a neurological condition which prevents certain connections from being made
(e.g., dyslexia) or, more commonly, because they (2) lack sufficient alphabetic
experience. People falling into the latter category may be school-age children who lack a
sufficiently rich reading background upon entering school (and thus have trouble
catching up when reading instruction begins), illiterates (who never receive reading
instruction), or literates in nonalphabetic orthographies (because they, too, lack explicit
instruct instruction or experience with an alphabetic orthography).
The theory the phonemic awareness is a consequence of alphabetic reading (rather than a
prerequisite) also has problems, however. Some prereaders, illiterates, and readers of
syllabaries and logographies - populations who have had no direct experience with
alphabetic literacy - do exhibit varying degrees of success on phonemic awareness tasks.
This is because explicit instruction in an alphabetic orthography is not the only route to
phonemic awareness, as other types of activities foster phonemic awareness.. Secondary
language activities include language games, poetry, kana charts, etc. and explain why
preliterate children exhibit different levels of phonemic awareness before they have
begun phonics instruction, and why some literates in nonalphabetic orthographies are
able to perform phonemic awareness tasks. Any of these experiences might develop
phonemic awareness.
So, rather than measuring past reading experience or future reading ability in any
concrete way, phonemic awareness tasks, at their core, measure the ability to abstract
and manipulate phonemes. The only logical connection between PA and reading ability is
the following: phonemic awareness and reading ability are in a symbiotic relationship, in
which the two skills develop in tandem. All successful alphabetic readers exhibit a
certain degree of phonemic awareness, and those children who exhibit the most
sophisticated phonemic awareness will go on to be the best readers.
Phonemic awareness does appear to be teachable through phonemic awareness training,
though it is most effective when combined with phonics instruction. The combination of
fact-based learning (i.e., graphophonic rules about sound-symbol correspondences)
combined with skill-based learning (i.e., phonemic awareness skills, such as
segmentation and blending), has proved more effective than either on its own (or
nothing). This is encapsulated in the phonological linkage hypothesis, which states that
the most effective form of reading instruction aids children in forming explicit links
between their phonemic awareness skills and their reading-based skills; explicit
instruction in the symbiotic nature of the relationship.
As phonemic awareness develops (though primary or secondary language activities),
people demonstrate different levels of phonemic awareness, from the most basic
(knowledge of nursery rhymes) to the most sophisticated (the ability to manipulate
phonemes on command, by deleting or substituting a phoneme in a word). Phonemic
awareness is not a developmental on/off switch, and appears to require some type of
alphabetic experience to grow and develop to its fullest extent (which explains why most
illiterates and literates in nonalphabetic orthographies cannot perform phonemic
awareness tasks).
Adult phonemic awareness is difficult to assess since adults, by virtue of their years of
experience with an alphabetic orthography, are conflated at the highest levels of
phonemic awareness. Moreover, it is clear that literate adults may experience
phonological interference in performing phonemic awareness tasks, because of
orthographically influenced phonological representations or orthographic recoding of
phonologically processed material. Fortunately, the results of a pilot task indicate that
orthographic complexity of a word is less important in processing than the phonological
complexity of a word when performing phonological tasks, so adults’ phonological and
phonemic awareness can be tested in isolation.
The experiments described in the next chapter did just that.
3.0. The Experiments
3.1. Overview
The experiments discussed in this chapter were designed to test the phonemic awareness
skills of literate adults, and the relationship of those skills to reading ability. Six tests of
phonological awareness were conducted in all. The six experiments are related, but
distinct, and all have varying degrees of difficulty. Performance on each test of
phonemic awareness was compared to performance on a test of reading ability.
In the experiments discussed below, the phonological awareness of subjects is assessed
by having them perform three different phonological awareness tasks (deletion,
substitution, and permutation of phonological units) on two different phonological units
(syllables and segments), resulting in six experiments total (3x2=6). The table below
illustrates this design:
Segment-based tasks Syllable-based tasksDeletion Delete Segment Task (1) Delete Syllable Task (2)Substitution Substitute Segment Task (3) Substitute Syllable Task (4)Permutation Reverse Segment Task (5) Reverse Syllable Task (6)
The numbers in parentheses refer to the experiment numbers.
As the table illustrates, for each phonological manipulation (operation) that subjects were
asked to perform, a pair of sister experiments was constructed: one in which syllables are
the unit of manipulation, and one in which segments are the unit of manipulation. This
design allows for three distinct analyses of the data:
1. An analysis of the relationship between subjects’ reading ability and their performance on the phonological awareness tasks outlined above,
2. A comparison of subjects’ performance on sister experiments, that is, the two experiments in which subjects are asked to perform the same type of operation on two different linguistic units (segments and syllables), and
3. An analysis of subjects’ performance on the three different types of manipulations of the same phonological unit
This three-part analysis will explore several aspects of phonemic awareness in literate
adults. Analysis (1) above will test for a relationship between phonemic awareness and
reading ability, while (2) and (3) will test whether there are differing levels of phonemic
and phonological awareness in literate adults. The predictions are as follows: With
regard to (1), it is expected, following the vast body of research done on children’s
phonemic awareness and reading ability, that there will be a correlation between subjects’
performance on reading and phonemic awareness tests. However, since all literate adults
exhibit some level of phonemic awareness, and since the ability to manipulate syllables is
present before the ability to manipulate phonemes and prior to reading experience (as shown in
preliterate children and illiterates), it is predicted that reading ability will correlate with performance on the
three phonemic awareness tests (the segment-based tasks) and not the remaining phonological awareness
tests (the syllable-based tasks). This is illustrated below:
Segment-based tasks Syllable-based tasks
Deletion Delete Segment Task (1) Delete Syllable Task (2)
Substitution Substitute Segment Task (3) Substitute Syllable Task (4)
Permutation Reverse Segment Task (5) Reverse Syllable Task (6)
Prediction: As (1), (3), & (5) tap into phonemic awareness, they will be correlated with reading ability
Prediction: As (2), (4), & (6) tap into lower-level phonological awareness only, they will not be correlated with reading ability
Because adults have achieved a baseline level of phonological awareness, their ability to
manipulate syllables is unrelated to their reading ability (since all adults are able to manipulate
syllables with relative ease). This brings us to the second set of predictions.
Because syllables are accessible phonological units to illiterates and preliterates, it is assumed
(Adams, 1990, and many others) that syllabic awareness is a precursor to phonemic awareness.
Therefore, it is expected that when the comparisons of the sister experiments discussed in (2)
above are carried out, it will be found that subjects perform significantly better on the syllabic-
based tasks than the segment-based tasks for each operation. This is illustrated below:
Performance on Syllable-based tasks > Performance on Segment-based tasks
Deletion Delete Syllable Task (2) > Delete Segment Task (1)
Substitution Substitute Syllable Task (4) > Substitute Segment Task (3)
Permutation Reverse Syllable Task (6) > Reverse Segment Task (5)
Finally, it is predicted that when performances on different types of manipulations are
compared, it will be shown that adults, like children, perform best on deletion tasks and
worst on permutation tasks (Adams, 1990). This prediction, based upon findings with
children, is an expected fall-out from the experimental design, but is not central to the
thesis here. It will be touched upon the in discussion of all six experiments.
The tasks in the experiments below are all based on previous studies. A question that
arises in these types of studies is the degree to which the experimental design tests true
natural language ability, versus quirky task-oriented skills. After all, the results of such a
study are only interesting insofar as they reveal something about a subject’s natural
language ability. Why, then, have researchers in phonemic and phonological awareness
assumed that the seemingly unnatural feats subjects are asked to perform in an
experimental setting somehow reveal their linguistic competence? English speakers
don’t reverse segments in their everyday speech, and segment reversal is not employed in
English morphology. How, then, can we judge a subject’s phonemic awareness ability
by performance on an arguably unnatural task? Cross-linguistic evidence shows that
such tasks are not necessarily unnatural.
The tasks in phonemic awareness experiments usually take the form of invented language
games (invented by linguists, educators, psychologists, and speech pathologists for
training and diagnostic purposes), but naturally occurring language games (invented by
speakers for no greater reason than language play) take the same forms. Anderson (1992)
argues that language games may tell us more about linguistic abilities than naturally
occurring morphology:
Some [morphological] rule types are unattested not because they are beyond the bounds of human linguistic capacity, but rather because there is not coherent sequence of possible historical change that would give rise...In the phenomenon of ‘secret languages’ or ‘language games’ [there are] a wide range of process types which are not found in natural languages. Since they are not constrained by the limits on historical change, they are freer to exploit the limits of human linguistic capacities. (Anderson, 1992, p.63).
Anderson’s point is that “natural” is not synonymous with “broadly occurring”; due to
historical change or other factors, perfectly “natural” types of syllable and phoneme
manipulations may become rare. There are certainly parallels with the phonemic
inventory of a language; though [x] is a perfectly natural sound, it is not present in
modern-day English. Historical change assures that a given sound or phonological
process may be present in one language but absent in another.
Each of the “invented” tasks in the experimental design laid out above occurs in some
form in a naturally occurring language game. A version of segment deletion occurs in
that old standby, Pig Latin, in which the player has to delete (extract) the initial segment
and move it to the end of the word (game: [gem] [emge]). A similar game in Cuna, a
Chíbchan language of Panama, invokes syllable deletion: the first syllable of a word is
stripped off and then moved to the end of the word ([uwaya] [wayau]) (Bagemihl,
1989). Phoneme substitution can be seen in a Luganda game, in which each syllable
beginning with a consonant is reduplicated, with [z] in that consonant’s place ([omusajja]
[o-mu-zu-sa-za-jja-za]) (Kilbride & Kilbride, 1972). Syllable substitution is attested
in an informal language game played by my Swarthmore College peers, in which words
ending in -ter had their final syllable supplanted by [tri]: water: [war] [watri],
bitter: [br] [btri]. Substitution games are further attested in examples of
children’s spontaneous language play. Two examples that illustrate this are deanut dutter
dandwich (“peanut butter sandwich”) as segment substitution (Kleeck & Bryant, 1984)
and Mommy, is it an a-dult or a nuh-dult? as syllable substitution (Gleitman, Gleitman, &
Shipley, 1972). Even reversal tasks, which are less likely to occur as morphological
processes, make their way into language games. Bagemihl cites Zande, a
Niger-Congo/Adamawa-Ubangian language of Zaire, as an example of a language with a
syllable reversing language game ([tikpo] [kpoti]), and Chasu as an example of a
language with segment reversing games ([sano] [naso]). So, the tasks in Experiments
One through Six below are attested in natural language processes; the operations which
subjects are asked to perform all are found in naturally occurring language games.
The fact that even the most difficult tasks in the experiments below are attested in natural
language games should not be interpreted to mean that all human language speakers
possess the phonemic awareness necessary to play the games. On the contrary:
It appears that an alphabetic writing system may be a prerequisite for a segment reversal ludling [language game] to appear in a language...however, this is not a sufficient criterion, since many languages with alphabetic systems have only syllable reversing ludlings. Moreover, in languages with segment reversing ludlings, the reversal is clearly not based on the orthographic representations, as the English examples...illustrate....It seems that the presence of an alphabetic writing system is necessary for the establishment of some metalinguistic awareness of the notion of 'segment'; beyond this, however, the phonological system takes over as the primary basis for reversal. (Bagemihl, 1989, p. 485)
So, citing language games to justify task naturalness in no way implies that phonemic
awareness is an innate ability; it only confirms that the abilities to be tested in these
experiments are within the realm of human linguistic capacity. After all, not everyone
has a talent for Pig Latin!
The six experiments in this section have the advantage of clinical control; the task
instructions, task familiarity, learning period, and stimulus pattern are the same in each.
Because of this control, the data from different tasks can be compared with each other
and to the parallel measure of reading ability without concern that the results are
confounded by some task-specific factor.11
3.2. Deletion Experiments
3.2.1. Experiment One: Delete Segment Task
Experiment One is the Delete Segment Task. The experiment has two parts: an oral test
of phonemic awareness and a written task that measures reading ability.
The purpose of the experiment is two-fold. In and of itself, it is designed to test whether
adults exhibit the same connection between phonemic awareness and reading as children.
11 ? Still, as Mann suggests, it would be intriguing to study data from naturally occurring language games: “One would like to see research that investigates the phoneme awareness of children and adults who speak such secret languages [who play language games]. It remains to be seen whether these speakers will achieve high levels of performance on the same materials used in previous research on phoneme awareness.” (Mann, 1991, p. 61)
In comparison with its sister test, the Delete Syllable Task, it is designed to test whether
adults exhibit different capabilities in the deletion of segments versus syllables.
The predictions for the outcome of the experiment are as follows. It is expected that, in
keeping with results from similar experiments performed on children, the adults who
perform best on this measure of phonemic awareness are those adults who also perform
best on the reading test. A secondary prediction is that though subjects will perform
relatively well on this test, as deletion is a simple operation, they will fare better on the
Delete Syllable Task, which asks them to perform the same operation (deletion), but with
a more accessible unit of deletion (the syllable, rather than the segment).
METHOD
Subjects
Forty-two University of Arizona students participated in the experiment in exchange for
extra credit in a class. Nine students were excluded from the study. Of these one was not
a native or near-native speaker of English12, one did not complete the oral part of the
experiment, and one failed to follow directions. Six subjects were unable to complete
portions of the experiment due to a tape malfunction. Excluding these, the subject pool
consisted of thirty-three subjects who were unaware of the purpose of the experiment.
The subjects were arbitrarily divided into two groups, henceforth referred to as Group A
and Group B. There were 18 subjects in Group A and 15 in Group B.
12 Subjects who learned English while preliterate (for purposes of this experiment, before the age of three) were included.
Tasks and Procedure: Oral Test of Phonemic Awareness
For Experiment One, the test of phonemic awareness administered was the Delete
Segment Task. This is a classic phonemic awareness test which has been employed
routinely in phonemic awareness tests with children (Bryant, et. al., 1990; Bradley and
Bryant, 1983; Cunningham, 1990; Lundberg, et. al., 1988; Mann, 1986; Stanovich, et.
al., 1984; Tornéus, 1984; among others). In these tasks, subjects are given a list of
words and asked to delete the initial phoneme. The basic methodology employed here
differs little from that used in children’s tasks, though there are some minor adjustments.
These are detailed below.
Since the subjects in this case are adults, it was determined that the task had to be made
more difficult than parallel tasks which have been performed on children, while
remaining true to the original methodology. As such, the instructions, stimuli, and
response time were all slightly adapted. The instructions to the subjects were similar to
those given children, except that the real word/nonword status of the response was not
made explicit (that is, children were told that the correct response would be a real word of
English). The stimuli in the adult experiment were made considerably more difficult.
First, all correct responses were nonwords (as opposed to the children’s experiments, in
which all correct responses were words). Secondly, since the results of a pilot (see
Chapter Two) showed that literate adults have little trouble stripping off a single
consonant from a word (even when that consonant is not in a one-to-one relationship with
the first grapheme of the word), all test items in this experiment began with consonant
clusters.13 Finally, the adult subjects were given a limited amount of time in which to
respond to a stimulus, unlike previous studies with children. All of these changes were
instituted to make the experiment more challenging to adults, with the goal of eliciting
higher error rates.
Subjects were played taped instructions through headphones. Each subject gave
responses into a microphone, which could be held in one’s hand or placed on a desk.
Subjects heard the following instructions:
We’re going to play some games with words. In each game, you will be given a set of instructions and some examples showing how the game is played. You will have a limited amount of time to respond to each item, and there is no stopping or going back. There are six games in all.
The instructions specific to Experiment One were as follows: “I’m going to give you a
word, and I want you to tell me what that word would be if the first sound were taken off.
Here are some examples.” The subjects then heard five practice items, followed by the
correct responses to those items. Upon completion of the practice items, subjects heard
the instruction, “Now you try,” which was followed by twelve test items, presented
approximately four seconds apart (as timed by an electric metronome).14 Following these
instructions, the expected response to the stimulus beauty is [yui].
13 Or, pehaps it should be said, what appear to be consonant clusters. Whether the [ky] sequence that begins a word like [kyut] is a cluster is a matter of some debate. This will be discussed further in the results section.
14 Recording equipment and timing were the same for all experiments.
Each subject was presented stimuli from one of two lists; Group A subjects heard List 1
stimuli and Group B subjects heard List 2 stimuli. The stimuli on the two lists were
matched for frequency (using PHONDIC15) and phonological pattern. The stimuli shared
the following characteristics:
All test items were two syllable words in which the second syllable contained an unreduced, stressless vowel;
All test items began with what could be characterized as a CCV sequence, which was either a [Cy] word (i.e., [byuri]) a [Cw] word (i.e., [kwazi]), or a “straight” cluster (i.e., [fresko])
These three types of initial clusters were chosen based on the results of the Delete
Segment Pilot discussed in Chapter Two.
The practice items on Lists 1 and 2 modeled the Delete Segment Task on both [CV...]
words and [CCV] words, so that there would be no confusion about the unit subjects were
being asked to delete. After hearing both [bebi...ebi] and [trbyut...rbyut] modeled,
subjects knew that this was a truly a delete initial segment task, and not a delete onset
task.
The practice items and stimuli for Experiment One are given below:
15 ? PHONDIC is an ASCII file of 20,000+ English words. For each word, the file contains its transcription, spelling, part of speech, and frequency.
List 1 Stimuli List 2 StimuliPractice Items: Practice Items:
baby moodytribute translateplenty glory
charcoal checkmategraphite precinct
Test Items: Test Items:beauty bureautwenty treatyquarry quasifrantic frenzysticky stuccocubic fury
franchise climaxscarcely speciesmusic punyclergy proxyfresco credo
swallow statue
Tasks and Procedure: Written Test of Reading Ability
Subjects were administered a test of reading ability compiled from tests previously used
in General Record Exams (GREs). The test consisted of 37 questions, 27 of which were
drawn from the exam’s verbal ability section, consisting of both vocabulary and reading
comprehension questions. This test was deemed appropriate for adults because of its
similarity to the tests of reading ability often used in tests of the relationship between
phonemic awareness and reading ability in children. Although the Peabody Picture
Vocabulary Test - Revised (PPVT-R) is a popular test with very young subjects (Ball &
Blachman, 1991; Bowey & Francis, 1991), the Metropolitan tests are frequently used
with older children (as seen in Stanovich, Cunningham & Cramer, 1984; Cunningham,
1990; Hatcher et. al., 1994; Mann, 1993). The Metropolitan Achievement Tests, Primer
and Primary levels, are one such test and are used to “measure sound-symbol
correspondence, word recognition, and reading comprehension” (Cunningham, 1990, p.
432). Obviously, the written test administered in this experiment does not offer a direct
test of sound-symbol correspondence, but since the subjects in this experiment are literate
adults rather than preliterate children, this is an expected (and ultimately irrelevant)
difference.
The remaining ten (of 37) questions on the written test were drawn from the exam’s
analytical ability section, and served as a control: it was not expected that there would be
any correlation at all with subjects’ performance on phonemic awareness tasks, as no
correlations were found in control test with children. (Control tests which have been
used in phonemic awareness experiments with children include IQ tests (Stanovich, et.
al., 1984), music segmentation tests (Morais, et. al, 1986), and nonphonological
linguistic tests (i.e., syntax tests) (Lundberg, 1991)). If, in fact, a correlation was found
between the analytical questions and phonemic awareness ability, it would indicate that
subjects’ performance on phonemic awareness tasks was based more on analytical
strategy than linguistic skill.
Subjects were not informed as to the origin of the questions they received. They had 30
minutes to complete the test. All subjects were given the same test, and the test was used
as the measure of reading ability in Experiments 1-6.
RESULTS AND DISCUSSION
Results: Oral Test of Phonemic Awareness
The results for the Delete Segment Task were coded as correct or incorrect. For all six
experiments in this study, any items for which all subjects in a group gave correct or
incorrect responses were excluded. This exclusion serves to minimize any item effect
(i.e., the ease or difficulty of a particular stimulus, rather than the ease or difficult of the
task) on the correlations being tested.
The method of coding for the Delete Segment Task was rather conservative; that is to
say, the only responses scored as “correct” were those in which the first consonant in a
cluster (but not the second) was deleted. Acceptable correct responses for both lists are
shown in the chart below:
List 1 Stimuli
correct response List 2 Stimuli correct response
beauty yui bureau yuro
twenty wnti treaty rii
quarry wri quasi wzi
frantic ræntk frenzy rnzi
sticky tki stucco tko
cubic yubik fury yurifranchise ræntyz climax lymæks
scarcely kersli species piiz
clergy lrdi puny yuni
music yuzk proxy rksi
fresco rsko credo rido
swallow wlo statue tætu
The coding of most of the clusters is clear; when the second consonant is anything but
[w] or [y], there is no question that it is a consonantal part of the onset. Performing the
Delete Segment Task correctly requires subjects to do more than separate onset and rime.
They must break the onset in two and extract the first of these consonants away from the
onset.16 Doing so exhibits a relatively sophisticated level of phonemic awareness:
subjects must be able to explicitly deal with the onset of a syllable as comprised of
separate units.
Maybe this should go under tasks/procedure? The [Cy] cases and [Cw] cases are a little
trickier. Are the [y] and [w] in these cases really the second consonants of consonant
clusters, or rather on-glides to the following vowel? Davis and Hammond (1994) argue
16 The special status of s-clusters will be discussed later in this chapter.
that [w] and [y] should be thought of differently: [w] is part of the onset (that is, the
second consonant of a consonant cluster in the cases cited here), but [y] is co-moraic with
the following vowel, and thus part of the coda. They cite the following evidence for
their claim:
· Cw has far fewer phonotactic constraints on it: Cw occurs before all vowels but [], [aw], and [oy], whereas [Cy] occurs only before [u] only (indicating that y+u sequences are in actuality a diphthong, /u/)
· In the language game Pig Latin, [CwV] is unambiguously treated as part of the onset (for example, [unswe] is always the response to “swoon”), while [CyV] sequences are subject to variation in which some respondents treat it as part of the vowel (so, “cute” might yield either [yutke] or [utke]), and
· In the “Name Game,” the [w] is deleted in [Cw] names like “Gwen” (gwn, gwn, bo bn, benæn fæn fo fn...), but the [y] is sometimes retained (that is, treated like part of the vowel) in [Cy] names like “Beulah” (byul, byul, bo byul,benæn fæn fo fyul...).
Thus, there are many reasons for thinking that the [Cy] words in this experiment are not
truly words that begin with consonant clusters, whereas the “Cw” words are. Why, then,
include them at all? First of all, Davis & Hammond’s evidence from Pig Latin and the
“Name Game” is not without variation; there are dialects of each game in which the [y]
of [Cy] sequences is treated as part of the onset. Secondly, in the pilot experiment,
subjects made many more errors on words beginning with [Cy] sequences than they did
words beginning with a single consonant, indicating that the [y] in [Cy] words is not an
indisputable part of the vowel. So, for the purposes of this experiment, all clusters ([Cy],
[Cw], and [CC]) will be treated in the same fashion, as sequences of two consonants.17
17 Davis and Hammond, in citing the Pig Latin response of [utke] for “cute,” suggest an
There is reason to think, however, that s-clusters (included among the “straight is [ki],
not [tki]. S-clusters violate the sonority hierarchy ([s] is more sonorous than [t], and yet
is further removed from the syllable nucleus in sticky) and so have been classified by
some as extrasyllabic (Clements and Keyser; 1983). Selkirk (1982) proposed a different
explanation for the anomalous behavior of s-clusters, positing that s-clusters at the
beginnings of words were actually complex onsets. In deleting the first phoneme, subjects
may be deleting the first licensed (non-extrasyllabic) element, or deleting a complex
segment, which can’t be broken down any further. Thus, responses to s-cluster stimuli
were marked correct if both the [s] and the following consonant were stripped off,
yielded some interesting results. ” clusters) should be treated differently, and that a
correct response to sticky in the Delete Segment Task
There were no items in the Delete Segment Task which all subjects got correct or
incorrect, so all items were included in the final analysis. The results of the Delete
Segment Task were as follows: Group A subjects had a mean correct response rate of
7.94 (out of 12), with a standard deviation of 3.22, and Group B subjects had a mean
correct response rate of 9 (out of 12), with a standard deviation of 2.17.
Results: Written Test of Reading Ability
intermediate stage of [utkye]. If, indeed, [y] is co-moraic with the following vowel in [Cy] words, it may be the case that when subjects are asked to delete the first segment in [myuzk] and respond with [uzk], there is an intermediate (unvoiced) [yuzk] response. In the current study, however, there is no way to determine if such an intermediate stage exists, nor is there any clear reason to assume that the proposed [iu] diphthong is treated differently from other diphthongs.
The written tests of reading ability and analytic ability were scored according to the
answer sheet in the Graduate Record Exam instruction book.
Results: Segment Deletion and Reading
There was no significant correlation between reading ability and ability to delete an
initial phoneme for either Group A or Group B subjects (Group A: r(18)= .464; Group
B: r(15)= -.040). This was an unexpected result given that in similar experiments with
children, the correlation between phonemic awareness and reading ability was
statistically significant. This result will be discussed further in the sections below.
Analytical ability, which was used as a control, was also not correlated with performance
on this task (Group A: r(18)= .149; Group B: r(15)= -.032). This was the expected
result, as no past phonemic awareness studies have shown a connection between PA and
intelligence, mathematical, or analytical ability.
3.2.2. Experiment Two: Delete Syllable Task
In the Delete Syllable Task, subjects are asked to strip off the initial syllable in a word.
In comparison with its sister test, the Delete Segment Task, it is designed to test whether
adults exhibit different capabilities in the deletion of segments versus syllables.
It is expected that performance on the Delete Syllable Task will exceed that on any of the
other tasks, because (1) deletion is the simplest phonological operation of the three
employed in this set of experiments, and (2) syllables are more accessible than phonemes.
In fact, it is not expected that the best readers will do significantly better on this measure
of phonemic awareness than the poorest readers, since syllable deletion is a task requiring
a very low level of phonemic awareness, and should therefore be unrelated to reading
ability.
METHOD
Subjects
There were thirty-three subjects; the same subjects as in Experiment One.
Tasks and Procedure: Oral Test of Phonological Awareness
For Experiment Two, the test of phonological awareness administered was the Delete
Syllable Task. This phonemic awareness test has been used in tests with children, as it is
here, as a means of explicitly gauging the effect the unit of manipulation has on a task
(Lundberg, 1988). In this task, subjects are given a list of words and asked to delete the
initial syllable. The methodology is exactly the same as that in Experiment One, except
in this experiment subjects are told, “I’m going to give you a word, and I want you to tell
me what that word would be if the first syllable were taken off.” As in Experiment One,
the subjects heard five practice items (with their correct responses), followed by twelve
test items.
Again, each subject was presented with stimuli from one of two lists. The stimuli were
the same as those in Experiment One. Group A subjects (who received List One stimuli
in Experiment One) received List Two stimuli in Experiment Two, and Group B subjects
heard List One stimuli in Experiment Two. This list-balancing was incorporated into the
design so that any conclusions drawn from a comparison of these sister experiments was
sure to be the result of differences in phonemic awareness rather than differences in the
difficulty of stimuli.
Tasks and Procedure: Written Test of Reading Ability
The test of reading ability and the control test of analytical ability were the same as in
Experiment One.
RESULTS AND DISCUSSION
Results: Oral Test of Phonological Awareness
The coding of correct and incorrect answers for the Delete Syllable Task was
considerably more complicated than that for the Delete Segment Task. In one of life’s
great linguistic ironies, even though the basis of phonemic awareness research is that
syllabic awareness is innate and phonemic awareness is a more complicated ability only
learned through exposure to an alphabet, it is much easier to determine what comprises a
segment than what comprises a syllable for this task. Since phonemic awareness research
on children has traditionally employed stimuli that are either monosyllabic or disyllabic
and obviously bimorphemic, it is of little help in determining how the answers to this
Delete Syllable Task should be coded.
It is widely accepted that the syllable is a bona fide phonological unit (Anderson, 1969;
Fudge, 1969; Kahn, 1976). There are syllable-based writing systems, syllable-based
language games, and phonological rules which depend on syllable boundaries. So, it is
not the existence of the syllable which is in question here, it is the composition of the
syllable.
The stimuli in the Delete Syllable Task are all bisyllabic words with primary stress on the
first syllable a nonreduced vowel in the second syllable. Some of the stimuli contain a
medial consonant cluster (i.e. fresco), while others contain a single medial consonant (i.e.
baby). How should these words be syllabified? Consider the cases of fresco and baby
If responses that violate phonotactic rules of English are marked as incorrect, there are
still several possible responses to the question, “What does X sound like when the first
syllable is removed?” For fresco, three possibilities present themselves: [sko], [ko], and
[o]. For baby, there are two: [bi] and [i]. How are we to determine which response(s)
should be considered appropriate?
As discussed in Chapter One, theories of syllabification confuse, rather than illuminate,
the matter. The Maximal Onset Principle (Pulgram, 1970), the Sonority Dispersion
Principle (Clements, 1988) and theories of ambisyllabicity (Kahn, 1976) all lead to
different coding of answers in the Delete Syllable Task, as demonstrated in the chart
below:
TEST ITEM Maximize Onset Sonority Dispersion Ambisyllabicity
frsco fr.sko frs.ko frsk.ko or frs.sko
baby be.bi be.bi be.bi or beb.bi
The Maximize Onset and Sonority Dispersion principles both force a strict interpretation
of correct answers in the Delete Syllable Task. If either principle is used as the basis of
coding, only one correct answer presents itself for each test item. Theories of
amibsyllabiicity, on the other hand, allow for several different responses, since medial
segments may belong to both syllables.
Recall that Treiman & Zukowski (1990), Treiman & Danis (1988), and Meador & Ohala
(1993) all found support for ambisyllabicity in syllabification tasks. This would support
a liberal coding of responses in the Delete Syllable Task, by which frsk.ko or frs.sko
could be the correct syllabification of fresco, and therefore [ko] or [sko] could be the
correct response.
Two other considerations support a more liberal coding of responses in the Delete
Syllable Task. The first consideration is a linguistic one: it is not clear from the three
studies cited above how the nonreduced vowels in the stimuli affect syllabification. Are
the second syllables in fresco and baby more likely to attract consonants than the second
syllables of master and palace? The second consideration is strategy-oriented. Recall
that the instructions to subjects in this experiment are, “I’m going to give you a word, and
I want you to tell me what that word would be if the first syllable were taken off.” In
cases of ambisyllabicity, it is possible that these instructions could elicit two different
responses. Let’s say two subjects hear the word baby and judge the medial consonant
ambisyllabic. When prompted for a response, Subject 1 might view the task as
identifying the first syllable [beb] and spitting out the remainder [i]. Treiman and Danis
found that subjects tended to “force” singly spelled consonants into one syllable or
another, even in the most stereotypically ambisyllabic environments. Subject 2, on the
other hand, might view the task not as stripping off the first syllable, but identifying the
second, and so would say [bi].
Given all these factors, responses in the Delete Syllable Task were only marked incorrect
if they violated rules of English phonotactics (for example, the response [ksi] to the
stimulus proxy). Acceptable responses are given in the chart below:
List 1 Stimuli Exp. 2 response List 2 Stimuli
Exp. 2 responseTest Items: Test Items:
Beauty di, ti, i bureau ro, o
Twenty ti, i treaty di, ti, i
Quarry ri, i quasi zi, si, i
Frantic tk, k frenzy zi, i
Sticky ki, i stucco ko, o
Cubic bk, k fury ri, i
Franchise tyz, yz climax mæks, æks
Scarcely li, sli species iz, siz, iz
Clergy di, i puny ni, i
Music zk, k proxy si, i
Fresco ko, o, sko credo do, o
Swallow lo, o statue tu, u
Subjects performed extremely well on the Delete Syllable Task. When all responses
(both Group A and Group B) were considered, there was a mean correct response rate of
10.15 (out of 12), with a standard deviation of 3.09.
The chart below offers support for the decision to code subjects’ responses to the delete
syllable task as argued above.
Delete Segment Delete Syllable
Maximize Onset Sonority Dispersion Ambisyllabicity
Mean for All Subjects = 8.46 Mean = 7.26 Mean = 7.69 Mean = 10.15
If the Maximize Onset or Sonority Dispersion Principles were used to code responses,
subjects would actually have performed worse on the Delete Syllable Task than the
Delete Segment Task, an unexpected result that cannot be reconciled with previous
phonemic awareness experiments. A coding according to a theory of English
ambisyllabicity, however, is consistent with earlier findings that the syllable is easier to
manipulate than the segment.
There were no items on either list that all subjects got incorrect. There were no items
which all subjects in Group A got correct, however, there were seven items which all
subjects in Group B got correct. The items were: quarry, franchise, scarcely, music,
clergy, fresco, and swallow. There appears to be no difference in phonological pattern
between the items that all subjects got correct vs. those that all subjects did not get
correct (beauty, twenty, frantic, sticky, and cubic). This may be explained by the fact that
subjects, on the whole, performed extremely well on the Delete Syllable Task, and made
few errors. Group A subjects had a mean correct response rate of 9.33 (out of 12), with a
standard deviation of 3.51, and Group B subjects had a mean correct response rate of 4.33
(out of 5), with a standard deviation of .82.
Results: Written Test of Reading Ability
The scoring of the written tests of reading ability and analytic ability was exactly the
same as in Experiment One, as it was the same test.
Results: Segment Deletion and Reading
There was no significant correlation between reading ability and ability to delete an
initial syllable for either Group A or Group B subjects (Group A: r(18) = .295; Group B:
r(15)= -.100). This was the expected result for the syllable deletion task, particularly
given the high mean correct response rate. Syllable deletion is an easy enough task that
all subjects performed well on it. Since syllabic awareness is a low-level, arguably innate
phonological skill, it is not at all surprising that it is unrelated to reading ability.
What is surprising is that analytical ability, the control measure, was correlated with
ability to delete an initial syllable in Group B subjects: Group B: r(15) = .546,
significant at the .05 level. (There was no such correlation in Group A subjects: r(18)
= .379, p=.061). We have seen that syllabic awareness is not a product of reading ability,
and apparently not facilitated by it. Perhaps the reason that the best analyzers were the
same subjects who performed best on the Delete Syllable Task is that they took an
analytical approach to the task. Viewing syllable deletion as an analytic task (a
possibility since it does not require any special linguistic skill or sophisticated
phonological awareness), they were able to strip the first syllable from a word easily.
Another explanation may be found in a task-order effect, a consequence of the fact that
subjects performed the delete syllable task immediately after the delete segment task. At
least two subjects switched to a delete-segment strategy in the middle of the delete
syllable task and consequently made errors on several consecutive test items, rendering a
lower score on the Delete Syllable Task than expected.
3.2.2. Comparison of Experiments One and Two
It was predicted that there would be a significant difference between performance on the
Delete Segment Task on the Delete Syllable Task. Results were mixed. A t-test showed
that Group B subjects did perform significantly better on the Delete Syllable Task than
the Delete Segment Task (t=-3.79628, p<.001). However, there was no parallel
difference in performance by the Group A subjects (t= -1.636739, not significant). That
is, Group A subjects did not perform significantly better on the Delete Syllable Task than
the Delete Segment Task.
The two charts below show subjects’ performance on the deletion experiments as
correlated with reading ability:
Performance on Delete Segment and Delete Syllable Tasks as Correlated with Reading Ability (Group A):
Performance on Delete Segment and Delete Syllable Tasks as Correlated with Reading Ability (Group B):
To recap the findings thus far (as illustrated on the charts):
(1) Group A subjects performed no better on the syllable-based task (an unexpected
result), and although there appears to be a correlation of both tasks with reading
ability, this correlation was not significant (another unexpected result).
(2) Group B subjects performed significantly better on the syllable-based task than
the segment-based task (an expected result), but there was no correlation with
reading ability for either the syllable-based or segment-based task (an unexpected
result).
These mixed results are confusing. The difference found in Group B between segment
deletion and syllable deletion abilities was expected because segment deletion requires a
sophisticated level of phonemic awareness. If such a difference exists (as it appears to)
why wasn’t it connected to subjects’ reading ability?
Group A subjects performed no differently on the Delete Segment and Delete Syllable
Tasks, and there was no correlation between Segment Deletion ability and reading ability.
These are two unexpected results, and yet they are easier to explain than the one
unexpected result for Group B subjects. It was expected that subjects’ would fare better
on syllable deletion than segment deletion, because the latter requires a more
sophisticated level of phonemic awareness. The data show that Group A subjects found
it no more difficult to strip off an initial segment than an initial syllable. A likely
explanation is that literate adults have achieved a high enough level of phonemic
awareness than segment deletion is not a challenge. The only difference between
segment deletion and syllable deletion, skill-wise, is that one is acquired while the other
is innate. The lack of a correlation between segment deletion and reading only supports
this. Even the worst readers must possess some level of phonemic awareness to be
literate, and that level includes the ability to delete an initial phoneme from a word.
With this as a working theory, the question that remains is why Group B subjects
performed significantly better on the syllable based task.
3.3. Substitution Experiments
3.3.1. Experiment Three: Substitute Segment Task
Experiment Three is the Substitute Segment Task, in which subjects are asked to
substitute the initial segment in a word with another segment. Like Experiment One, it is
designed to test whether a correlation exists between reading ability and phonemic
awareness.
Experiment Three, like Experiment One, is a segment manipulation task. However, it is
predicted that subjects will not perform as well as they did in Experiment One.
Substitution tasks are considered more difficult than deletion tasks because subjects must
in a sense perform two operations: the must delete an initial segment (as in Experiment
One), but then replace it with a new segment. This is considerably more difficult than
deletion alone.
The results of Experiment Three will also be compared with the results of its sister
experiment, the Delete Syllable Task (Experiment Four). It is predicted that subjects will
perform significantly better on Experiment Four, since it requires only syllable
manipulation.
METHOD
Subjects
The subjects were the same as in previous experiments.
Tasks and Procedure: Oral Test of Phonemic Awareness
For Experiment Three, the test of phonemic awareness administered was the Substitute
Segment Task. This measure of phonemic awareness test has been used in tests with
children (Stanovich, et. al., 1984; Treiman, 1985), but as in Experiment One, the
methodology needs to be adapted slightly with adults. An example of instructions in a
children’s substitution task (Stanovich, et. al., 1984) is the following: “If I say the word
go and then change the first sound by changing it to /n/, the new word will be no.” As in
delete segment tasks with children, both the input and output in this case are real words.
Moreover, the segment to be substituted is explicitly pronounced.
The methodology here was adapted for two reasons. First, the task had to be made more
challenging for adults. Second, it seemed preferable to avoid pronouncing the substituted
segment in isolation as (1) the taped stimuli would make perception of isolated obstruents
extremely difficult, and (2) using the consonant + schwa pronunciation (i.e. [d] for [d]) -
a syllabic pronunciation - would cloud the distinction between the Delete Segment and
Delete Syllable Tasks.
To solve these problems, the instructions to subjects were as follows: “I’m going to give
you two words, and I want you to tell me what would happen if you replaced the first
sound in the first word with the first sound in the second word.” As in previous
experiments, the subjects then heard five examples which modeled the task (“What
would happen if you replaced the first sound in frozen with the first sound in credo?...
[pause]... [krozn].”). Following these instructions, the expected response to the stimulus
pair brisket/fluster is [frskt]. By altering the instructions in this way, the task was
made more difficult (subjects were not “given” the substituted phoneme, but had to
isolate it themselves), but less confusing (there is nothing to indicate that the unit to be
substituted is the minimal syllable [k]).
For each of the twelve test items, there were two stimuli: the stimulus that had its first
segment replaced, and the stimulus containing the replacement segment. There were two
lists of stimuli (sets), matched for frequency and phonetic pattern between sets using
PHONDIC. All test stimuli were two syllable words beginning with a consonant cluster
and with primary stress on the first syllable. This pattern was chosen for two reasons:
(1) This is the same basic phonetic pattern as the words in Experiments One & Two,
and
(2) As with Experiments One and Two, it was felt that stimuli with word-initial
clusters would generate higher error rates among subjects (thus eliciting more
dramatic differences in performance).
The practice items and stimuli for Experiment Three are given below:
List 1 Stimuli List 2 StimuliPractice Items: Practice Items:profit/blarney frozen/credobrutal/clipper sweater/trickybandage/suction mandate/pistoltravel/pliant treason/freewaysecret/bundle tandem/dismalTest Items: Test Items:blister/presto slander/grizzlyprincess/blunder triplet/bleachersblanket/crumple problem/clobbercrimson/twinkle brandish/tractorquisling/spastic drastic/sloganflounder/clumsy crumpet/slendertrinket/cluster quagmire/skepticgremlin/plastic crinkle/flankingprosper/glisten brisket/flusterfrosting/blasted proxy/plentyplaintiff/scarlet swelter/crystalcrafty/plasma traction/clanging
Tasks and Procedure: Written Test of Reading Ability
The written test of reading ability and control test of analytical ability were the same as in
previous experiments.
RESULTS AND DISCUSSION
Results: Oral Test of Phonemic Awareness
Answers that were coded as correct in the Substitute Segment Task are listed in the table
below:
List 1 Stimuli correct response List 2 Stimuli correct responseblister/presto plstr slander/grizzly gændr
princess/blunder brnss triplet/bleachers brplt
blanket/crumple klækt problem/clobber krblm
crimson/twinkle trmzn brandish/tractor trænd
quisling/spastic spsl drastic/slogan slæstk
flounder/clumsy klwndr crumpet/slender slmpt
trinket/cluster krkt quagmire/skeptic skægmayrgremlin/plastic prmln crinkle/flanking frkl
prosper/glisten grspr brisket/fluster frskt
frosting/blasted brst proxy/plenty prksi
plaintiff/scarlet skentf swelter/crystal kltr
crafty/plasma præfti traction/clanging kræken
The coding of correct answers in this experiment follows the reasoning of the coding
from Experiment One, in that the only answers coded as correct were those in which only
the first segment of a consonant cluster was replaced. As in Experiment One, responses
that treated s-clusters as complex segments were marked correct. Phonotactics were also
taken into account in the interpretation of some responses. For example, the expected
response to the quisling/spastic pair is *[spwsl] if s-clusters are treated as complex
segments. Since *[spw] is an illegal onset in English, responses beginning with the legal
cluster [sp] were accepted.
Subjects performed extremely poorly on the Substitute Segment Task. There were no
stimuli on either list that all subjects got correct. For Group A, there were also no
stimuli which all subjects got incorrect, but the mean correct response rate was 3 (out of
12), with a standard deviation of 2.57. There were three sets of stimuli on List Two that
were excluded from the results because they elicited incorrect responses from all subjects
in Group B. Excluding these stimuli, Group B subjects had a mean correct response rate
of 1.67 (out of 9), with a standard deviation of 1.71.
The three excluded pairs of stimuli were (1) brandish/tractor, (2) drastic/slogan, and (3)
traction/clanging. It is interesting that brandish/tractor gave subjects difficulty, because
it was the only pair (on either list) for which transposing the entire onset of the second
word [tr] would yield the same results as transposing only the first phoneme [t]. It seems
that subjects may have been thrown off by this, and may have second-guessed their
responses.
Drastic/slogan is one of the pairs that involves an s-cluster. However, there were other
pairs on the same list with s-clusters (e.g., crumpet/slender) to which some subjects gave
correct responses, so no generalizations can be made. There is nothing distinct about the
traction/clanging pair. In fact, subjects made so many errors on this task (with only 3 or
4 subjects getting any given response correct), it is difficult to categorize the excluded
items.
Results: Written Test of Reading Ability
The written tests of reading ability and analytic ability were the same as in the two
previous experiments.
Results: Segment Substitution and Reading
Results were mixed. It was predicted that there would be a correlation between reading
ability and performance on the segment substitution task. For Group A subjects, there
was no significant correlation between these two measures (r(18) = .186), but for Group
B there was a significant correlation (Group B: r(15) = .642, p<.01).
Analytical ability, which was used as a control, was not correlated with either group’s
performance on this task (Group A: r(18) = .049; Group B: r(15) = .232)).
3.3.2. Experiment Four: Substitute Syllable Task
Experiment Four is the Substitute Syllable Task, in which subjects are asked to replace
the initial syllable in a word with a syllable from another word. In comparison with its
sister test, the Substitute Segment Task, it is designed to test whether adults exhibit
different capabilities in the substitution of segments versus syllables.
It is expected that performance on the Substitute Syllable Task will exceed performance
on the Substitute Segment Task because syllable manipulation requires a lower level of
phonological awareness than segment manipulation. It is also expected that performance
on this task will not be as good as performance on the Delete Syllable Task, since
substitution is a more complicated operation than deletion. As with the Delete Syllable
Task, it is not expected that the best readers will do significantly better on this measure of
phonemic awareness than the poorest readers.
METHOD
Subjects
Subjects were the same as those in Experiment Three.
Tasks and Procedure: Oral Test of Phonological Awareness
For Experiment Four, the test of phonological awareness administered was the Substitute
Syllable Task. The methodology and stimuli are the same as those used in Experiment
Three, except that the unit of manipulation is the syllable. Subjects were told “I’m going
to give you two words, and I want you to tell me what would happen if you replaced the
first syllable in the first word with the first syllable in the second word.” As in previous
experiments, subjects heard five practice items (with their correct responses), followed by
twelve test items. Following these instructions, the expected response to the stimulus
brisket/fluster is [flskt].
Lists were counterbalanced as in Experiments One and Two; subjects received stimuli
from whatever list they did not receive in the Substitute Segment Task.
Tasks and Procedure: Written Test of Reading Ability
The test of reading ability and the control test of reading ability were the same as in
previous experiments.
RESULTS AND DISCUSSION
Results: Oral Test of Phonological Awareness
The coding of correct and incorrect answers for the Substitute Syllable Task was based
on conclusions drawn in Experiment Two. The most common correct answers are given
in the table below:
List 1 Stimuli common response
List 2 Stimuli common response
blister/presto prstr slander/grizzly grzdr
princess/blunder blnss triplet/bleachers blitlt
blanket/crumple krmkt problem/clobber klblm
crimson/twinkle twksn brandish/tractor trækd
quisling/spastic spæsl drastic/slogan slo(s)tk
flounder/clumsy klmdr crumpet/slender slnpt
trinket/cluster klskt quagmire/skeptic
skpmyr
gremlin/plastic plæsln crinkle/flanking flækl
prosper/glisten glspr brisket/fluster flskt
frosting/blasted blæst proxy/plenty plnsi
plaintiff/scarlet skrtf swelter/crystal krstr
crafty/plasma plæsti traction/clanging klæn
This listing is not exhaustive, however. Consider the blanket/crumple example.
Subjects are asked to replace the first syllable of blanket with the first syllable in
crumple. The research cited in Experiment Two tells us that these words yield more than
one syllabification possibility. Both words have stressed, short vowels in their first
syllables, which may create an ambisyllabic environment for the second consonant in the
medial cluster. So, though the preferred syllabifications for blanket and crumple are
blan.ket and crum.ple (in which the word is split between the consonants in the cluster),
blank.ket and crump.ple are also possible. Thus, responses to the Substitute Syllable
Task for this item could be [krmkt], [krmpkt], [krmpt], and [krmt].18 Even
though all of these responses would have marked correct, the latter two never presented
themselves.
18 These last two are based on the assumption that when subjects delete the initial syllable of blanket prior to substitution, they delete the ambisyllabic consonant along with it, leaving only [t]. This strategy was addressed in the discussion of Experiment Two.
Subjects exhibited a fair performance on the Substitute Syllable Task. There were no
items on either list that all subjects got correct or incorrect. For Group A subjects, there
was a mean correct response rate of 5.17 (out of 12), with a standard deviation of 3.79.
For Group B subjects, there was a mean correct response rate of 8.33 (out of 12), with a
standard deviation of 3.58. This performance was worse than that on the Delete Syllable
Task, as predicted.
Results: Written Test of Reading Ability
The scoring of the written tests of reading ability and analytic ability was the same as in
previous experiments.
Results: Syllable Substitution and Reading
For both Groups A and B, there was a significant correlation between reading ability and
the ability to substitute syllables (task (Group A: r(18) = .580, p<.05; Group B: r(15)
= .536, p<.05). This was an unexpected result, since this experiment involved syllable
manipulation, and syllabic awareness is a skill that is thought to be unrelated to reading
experience.
There was no significant correlation between analytical ability and performance on this
task, as expected (Group A: r(18) = .180; Group B: r(15) = .064).
3.3.2. Comparison of Experiments Three and Four
It was predicted that there would be a significant difference between performance on the
Substitute Segment Task and the Substitute Syllable Task, with subjects performing
better on the syllable task. A t-test confirmed that this was the case for both Group A
and Group B subjects. (Group A: t = 2.33, p<.05; Group B: t = 7.70; p<.05).
It was also correctly predicted that there would be no correlation of analytical ability with
ability to substitute a segment or syllable. This prediction was borne out.
However, the charts below illustrate some of the unexpected results in Experiments Three
and Four:
Performance on Substitute Segment and Substitute Syllable Tasks as Correlated with Reading Ability (Group A):
Performance on Substitute Segment and Substitute Syllable Tasks as Correlated with Reading Ability (Group B):
To review, it was predicted that there would be a significant correlation between reading
ability and performance on the Substitute Segment Task, but this was only found in
Group B subjects (the .185 correlation in Group A subjects was not significant). It was
predicted that there would not be a correlation between reading ability and performance
on the Substitute Syllable Task, but a significant correlation was found for both groups of
subjects.
How should these results be interpreted? There were unexpected results in Experiments
One and Two, to be sure, but they made sense upon closer examination. A missing
connection between reading ability and segment deletion ability was not unexpected
when subjects’ fine performance on the phonemic awareness task (differing not at all
from performance on the syllable deletion task) was taken into account. Here, though,
the findings are more difficult to explain. Performance on the segment and syllable tasks
does seem to differ, indicating that subjects exhibit syllabic awareness abilities superior
to their phonemic awareness abilities. This is not problematic until we consider that
subjects’ reading ability, which we expect to effect segmentation ability (as found in
studies with children), effects syllabification ability (at which we expect all adults to
express similar competence).
A closer examination of the results suggest that the data from the substitution
experiments are better dismissed as anomalies than explained. As mentioned above,
subjects’ unexpectedly poor performance on the Substitute Segment Task and
inconsistent performance on the Substitute Syllable Task both hint at a problem with the
task. Scrutiny of responses shows that the task demands were too strenuous; many
subjects were confused by the directions and simply did not perform up to par. With
such high error rates on both experiments, it is difficult to draw conclusions about the
data.. Had only Experiment Three performance been flawed, the results could still be
attributed to a difference in syllabification and segmentation abilities. With performance
on both tasks so low, it is more likely the result of task confusion.
A few examples illustrate the confusion subjects’ experienced in the substitution tasks.
There were many subjects who transposed the stimuli in the “Replace the first
segment/syllable of A with the first segment/syllable of B” instructions. When instructed
to replace the first syllable or segment of triplet with the first syllable or segment of
bleachers, they responded not with [blit_let] (in the syllable substitution task) or [br_plet]
(in the segment substitution case), but with [tr_ptRerz] or [tlitRerz]. Other incorrect
responses show that subjects were far removed from the task at hand. One subject gave
[kr_kflæk] as her response to the substitute segment task for the stimulus
crinkle/flanking (correct response: [fr_kel]), which is a compound of the first syllables
of both words. A different subject gave [klæk_k_t] as her response in the substitute
syllable task for the same stimulus (correct response: [flækel]), which is difficult to
make heads or tails of.
Unlike errors in the other experiments, many of the errors in Experiments Three and Four
are impossible to characterize. In Experiment One (the Delete Segment Task), nearly all
errors involved the deletion of an onset, rather than a segment. In Experiment Five (the
Reverse Segment Task, to be discussed below, in which subjects were asked to reverse
the segments beginning the syllables in a word) the majority of errors occurred when
subjects reversed the syllables in a word, rather than the segments. In both of these cases,
the errors fit a pattern. Moreover, the errors are an attempt on the part of the subject to
make the unit of manipulation larger and more accessible (an onset or a syllable as
opposed to a segment), thus making the task easier. Thus, these errors support the theory
that phonemic awareness tasks are difficult because the manipulation of phonemes is an
unnatural undertaking. By making the unit of manipulation larger, subjects avoid having
to perform a true phonemic awareness task.
In the substitution experiments, the errors are not always part of a pattern (as shown by
the crinkle/flanking examples), and even when they are (i.e. the cases in which subjects
transposed the directions), they do not make the task any easier. Thus, the errors in
Experiments Three and Four are task-oriented, not linguistic, in nature. This, taken with
subjects’ poor performance on the experiments, indicates that the data from these
experiments should be excluded from any larger analysis of adults’ phonemic awareness
abilities. The methodology in the substitution experiments was flawed; the instructions
were obviously too complicated for most subjects to process.
3.4. Reversal Experiments
3.4.1. Experiment Five: Reverse Segment Task
Experiment Five is the Reverse Segment Task, in which subjects were asked to reverse
the initial segments of a given word’s syllables. As in the other segment-based
experiments, the results of the Reverse Segment Task will be compared with those of a
reading test to test the correlation between the measures. It is predicted that reading
ability will be correlated with segment reversal abilities, since those abilities require a
sophisticated level of phonemic awareness.
Other predictions for the outcome of Experiment Five are parallel to predictions made for
previous segment-based experiments. It is predicted that performance on the segment
reversal task will be worse than that on the segment deletion task, since reversal is a more
difficult operation. When reversing the segments in a word, subjects essentially need to
conduct two segmental abstractions (versus one in a deletion task): they must extract the
first segment in each syllable of a word from the rest of the syllable, and then reposition
those segments in each other’s place. Moreover, previous experiments have shown that
the initial and final phonemes in a word are easier to manipulate than the medial
phonemes in a word (Byrne & Fielding-Barnsley, 1989; Mann & Brady, 1988). Since the
reversal task requires subjects to access and manipulate medial phonemes, it is expected
to be more difficult.
Performance on the Reverse Segment Task should be inferior to that on the Reverse
Syllable Task, for the reason cited many times in the last few sections: segment
operations require phonemic awareness, whereas syllabic operations only require a lower
level of phonological awareness.
METHOD
Subjects
The subjects were the same as in previous experiments in this chapter.
Tasks and Procedure: Oral Test of Phonemic Awareness
Experiment Five’s test of phonemic awareness was the Reverse Segment Task. This is
the one measure of phonemic awareness that has not been used in tests with children.
Segment Reversal is a difficult task, requiring a phonological sophistication that
preliterate and newly literate children simply do not have. The methodology here is
based on that used in syllable reversal tasks with children (see below), with the only
difference being that the unit of manipulation is the segment.
Subjects were told, “I’m going to give you a word, and I want you to tell me what that
word would be if the first sounds in each syllable were reversed.” As in previous
experiments, the subjects then heard five practice items that modeled the task, followed
by twelve test items. Following these instructions, the expected response to the stimulus
baptize is [tæpbyz].
The stimuli in Experiment Five were all two syllable words with primary stress on the
first syllable and unreduced vowels in the second syllable. They all had the phonetic
pattern CVCCVC(C) and were matched for frequency between lists using PHONDIC. In
the practice items, the correct responses which were modeled were those in which the
syllable boundary occurred between the CC sequence in the middle of the word (e.g.
hectic [tkhk]; sucrose [ruksoz]. The practice items and stimuli for Experiment
Five are given below:
List 1 Stimuli List 2 StimuliPractice Items: Practice Items:hectic sequingarnish nitratebackwash tadpolechastise pepticrustic sucroseTest Items: Test Items:baptize cashmeretoxic septichormone hydratecombat cosmicmastiff sentencedictate fabricmystic mascotpublish tarnish
costume cyclonegarlic formatphosphate vibratemustang darling
Tasks and Procedure: Written Test of Reading Ability
The written test of reading ability and control test of analytical ability were the same as in
previous experiments.
RESULTS AND DISCUSSION
Results: Oral Test of Phonemic Awareness
Answers that were coded as correct in the Reverse Segment Task are given in the table
below:
List 1 Stimuli correct response
List 2 Stimuli correct response
baptize tæpbyz cashmere mækir
toxic saktk septic tpsk
hormone morhon hydrate rydhet*
combat bmkæt cosmic mazkk
mastiff (s)tæ(s)mf* sentence tnsns
dictate tkdet fabric ræbfk*
mystic (s)t(s)mk* mascot (s)kæ(s)mat*
publish lub(p)* tarnish nrti
costume (s)t(s)kum* cyclone lykson*
garlic largk format morfæt
phosphate fsfet vibrate rybvet*
mustang (s)te(s)mæ* darling lardi
The coding of correct answers in this experiment was very straightforward. The only
answers that were coded as correct were those in which the first and third consonants in
the CVCCVC(C) word were reversed. As in Experiments One and Three, s-clusters
were treated somewhat differently, so that answers that treated s-clusters as a complex
segment (i.e. [stæmf] as a response to the Reverse Segment Task for mastiff) were also
marked as correct.
Stimuli marked above with an asterisk (*) contain a medial cluster that could be
syllabified in more than one way. Though it was expected that subjects would place a
syllable break in between the two consonants of the stimulus (since Treiman and
Zukowski found subjects preferred this syllabification over 90% of the time), responses
that indicated a different syllabification were also marked correct. Since the word
cyclone, for example, could be syllabified as cyc.lone or cy.clone, both [lykson] and
[kyslon] are possible right answers. With some stimuli, like vibrate, the alternate
syllabification (vi.brate, instead of vib.rate) produces a phonotactically illegal, but not
unpronounceable, response to the Reverse Segment Task. For instance, the
syllabification vi.brate yields the response [bryveyt] to this task. This is a
pronounceable and legal string of letters in English on the surface, but if subjects are
performing the task correctly, the syllabification of this string should be [by.vreyt],
which has an illegal cluster as the onset to its second syllable. As it happens, no subject
ever gave a response to the Reverse Segment Task that indicated a CV.CCVC
syllabification, even when the resulting onset cluster was perfectly legal. Thus, the only
relevant correct responses are those listed in the chart above.
Performance was poor on the Reverse Segment Task. There were no items for which all
subjects in either group gave correct responses. Group A subjects had a mean correct
response rate of 2.42 (out of 12), with a standard deviation of 2.83. There were three
items on List 2 for which all Group B subjects gave incorrect responses; these were
excluded from the analysis. On the remaining items, Group B subjects had a mean
correct response rate of 2.2 (out of 9), with a standard deviation of 2.40.
Interestingly, the three items which all Group B subjects got incorrect were hydrate,
cyclone, and vibrate. These were the only items on either list with tense (long) vowels in
the first syllable, which might lead subjects to prefer a CV.CCVC (vs. CVC.CVC)
syllabification. This could create two problems for subjects. First, in the cases of
hydrate [hy.dreyt] and vibrate [vy.breyt], the segment reversal for a CV.CCVC
syllabification would result in the responses of [dy.hreyt] and [by.vreyt], both of
which contain syllable-initial clusters [hr] and [vr] that violate the phonotactics of
English. Second, a CV.CCVC syllabification of a stimulus item requires that subjects
extract the first consonant of the second syllable from a consonant cluster (vs. simply
reversing the whole onsets in a CVC.CVC word like fabric [fb.rk]). Because this puts
an extra burden on the subject (he/she must break up an onset and then reverse two
segments), it is in a sense combining aspects of the Delete Segment and Reverse Segment
tasks, and it is not surprising that subjects did not do well.
In retrospect, the three problematic items in List Two (hydrate, cyclone, and vibrate)
should have been excluded from the study or at the very least balanced with
phonologically similar items in List One. This was an oversight on the part of the
researcher. However, given Group A’s poor performance on the Reverse Segment Task
(a mean correct response rate of 2.42 vs. Group B’s 2.2), it is not clear that the Group B
subjects would have performed much better on this task had they been tested on different
stimuli.
As predicted, subjects performed much worse on this measure of phonemic awareness
than they did on the Delete Segment Task.
Results: Written Test of Reading Ability
The written tests of reading ability and analytic ability were the same as in the two
previous experiments.
Results: Segment Reversal and Reading
There was a strong, statistically significant correlation between reading ability and the
ability to reverse an initial phoneme in Group A subjects (r(18) = .831, p<.05). There
was no significant correlation between reading and segment reversal in Group B subjects
(r(15) = .387). Since it was expected that these two measures would be correlated, these
were mixed results.
There was no correlation in either group between analytical ability and performance on
this task (Group A: r(18) = .367; Group B: r(15) = -.226). This was the expected
result.
3.4.2. Experiment Six: Reverse Syllable Task
Experiment Six is the Reverse Syllable Task, in which subjects are asked to switch the
two syllables in a word.
It is expected that performance on the Reverse Syllable Task will be superior to
performance on its sister test, the Reverse Segment Task. Because syllabic awareness is
subsumed by phonemic awareness, subjects are expected to perform quite well on the
syllable-based task, but have difficulty with the segment-based task.
Performance on the Reverse Syllable Task is expected to be worse than on the Delete
Syllable Task, since reversal is a more complicated operation than deletion.
Since all adults have achieved a reasonably sophisticated level of phonological awareness
that includes and surpasses syllabic awareness, it is predicted that reading ability will
have no effect on syllable reversal ability.
METHOD
Subjects
Subjects were the same as in the previous Experiment.
Tasks and Procedure: Oral Test of Phonological Awareness
The test of phonological awareness used in Experiment Six was the Reverse Syllable
Task, in which subjects are asked to swap the two syllables of a word. The methodology
in this experiment was modeled after Treiman & Danis (1988), who used it to test the
syllabification of intervocalic consonants. They told their subjects, “We’re going to play
a game with words. I’ll show you how it goes by giving you some examples. When I
say grandfather, you say father...grand. When I say catfood, you say food..cat” and
then gave them bisyllabic (noncompound) words as stimuli. The instructions in this
experiment were virtually the same (and parallel those in Experiment Five). Subjects
were told, “I’m going to give you a word, and I want you to tell me what that word would
be if the two syllables were reversed.” They then heard the practice items (with their
correct responses) and twelve stimuli.
Stimuli were the same as in Experiment Five, with any given subject receiving whatever
list he or she did not receive in that experiment.
Tasks and Procedure: Written Test of Reading Ability
The test of reading ability and the control test of reading ability were the same as in
previous experiments.
RESULTS AND DISCUSSION
Results: Oral Test of Phonological Awareness
As in Experiment Five, which used the same CVCCVC(C) words, it was presumed that
subjects would syllabify stimuli between the medial consonants. However, responses
consistent with an alternate syllabification (CV.CCVC) were also deemed acceptable
when the response did not violate any rules of English phonotactics (alternate responses
are indicated by parentheses in the table below). A correct syllable reversal for hydrate
could then be [rethyd] or [drethy].
Acceptable responses are as follows:
List 1 Stimuli correct response
List 2 Stimuli correct response
baptize tyzbæp cashmere mirkæ
toxic sktk septic tksp
hormone monhor hydrate (d)rethy(d)
combat bætkm cosmic mk(k)z
mastiff tfmæs sentence tns(s)n
dictate tetdk fabric (b)rkfæ(b)
mystic (s)tkm(s) mascot (s)ktmæ(s)
publish lpb tarnish ntr
costume (s)tumk (s) cyclone (k)lonsy(k)
garlic lkgr format mætfor
phosphate fetfs vibrate (b)retvy(b)
mustang (s)tæm(s) darling ldr
Subjects performed extremely well on the Reverse Syllable Task. There was one item
which all Group A subjects got correct; the mean correct response rate on the remaining
items was 9.39 (out of 11) with a standard deviation of 1.79. There were three items that
all Group B subjects got correct; the mean correct response rate on the remaining items
was 7.53 (out of 9), with a standard deviation of 1.30.
The items that all subjects got correct are not distinguished from other list items by any
notable phonological characteristics. They were baptize, combat and garlic in List One,
and cyclone in List Two.
Why did subjects do so well on the Reverse Syllable Task, when it was predicted that
they would do worse than on the Delete Syllable Task? A simple answer is that reversal
is not a more difficult operation than deletion. This is contradicted, however, by both
common sense and empirical results. If deletion and reversal are equally complex, how
can we explain the vast difference between performances on the Delete and Reverse
Segment Tasks? The methodology and stimuli for those two experiments were the same
as in their sister experiments, so the type of operation must be the culprit. The only
solution is that task performance is based on a more complicated interaction of unit of
manipulation and type of manipulation than previously thought. Reversal is a more
difficult operation than deletion, but only when the linguistic unit is partially inaccessible.
Since syllables are so accessible to literate adults, differences in operation difficult are
overshadowed. Differences between these operations surface in segment-based tasks,
since phonemes are more unattainable units. This will be discussed further below.
Results: Written Test of Reading Ability
The reading and analytical tests were the same as in previous experiments.
Results: Syllable Reversal and Reading
There was no significant correlation between reading ability and ability to reverse the
syllables in a word for either Group A or Group B subjects (Group A: r(18) = .204;
Group B: r(15)= .083). This was the expected result.
There was also no significant correlation between reading ability and ability to reverse
the syllables in a word for either Group A or Group B subjects (Group A: r(18) = .232;
Group B: r(15)= .007, also the expected result.
3.4.3. Comparison of Experiments Five and Six
It was predicted that there would be a significant difference between performance on the
Reverse Segment Task and the Reverse Syllable Task. A t-test confirmed that this was
the case for both groups (Group A: t = -10.65, p<.001; Group B: t = -14.72, p<.001).
Subjects found it much easier to reverse two syllables in a word than to reverse two
segments. This was the predicted result. Segment reversal, which requires both the
manipulation of an inaccessible unit (the segment) and a difficult operation (reversal),
apparently requires a level of phonemic awareness that most literate adults have not yet
achieved. Also, as predicted, there was no correlation between subjects’ ability to
perform reversal tasks (of either segments or syllables) and their analytical ability.
Performance on Reverse Segment and Reverse Syllable Tasks as Correlated with Reading Ability (Group A):
Performance on Reverse Segment and Reverse Syllable Tasks as Correlated with Reading Ability (Group B):
A mixed result, illustrated on the charts above, is that segment reversal ability was only
correlated with reading ability in Group A subjects (the correlation in Group B
subjects, .387, was not significant). As illustrated by the flat trend lines in the charts
above, the ability to reverse syllables was not correlated with reading ability in either
group of subjects (the predicted result).
3.5. Summary of Experiments
In the concluding section below, the results of all the experiments are analyzed more
comprehensively in the context of past and future phonemic awareness research.
Implications for further study and applications of these findings are also discussed.
4.0. Conclusion
This dissertation has examined the nature of phonological and phonemic awareness in literate adults. In this final section, it will be shown that
1. Adults demonstrate differing levels of phonemic and phonological awareness, even though all are phonemically aware to some degree,
2. There is a correlation between reading ability and phonemic/phonological awareness in adults, and that
3. These conclusions about phonemic awareness and reading ability in adults have both linguistic and educational implications.
Let us explore each of these conclusions in turn.
In Chapter Two, past research on phonemic awareness was discussed. A common claim
about phonemic awareness is that it exists alongside alphabetic experience. Although
there has been some evidence that phonemic awareness can develop in the absence of
alphabetic experience (Mann, 1991), and some evidence that those who possess
alphabetic literacy may be phonemically unaware (Cossu, et. al., 1993), there is much
more evidence that phonemic awareness is found only in those with alphabetic experience
(Read, et. al, 1986; Morais, et.al., 1986). Efforts to determine whether phonemic
awareness is a prerequisite to learning an alphabet or if phonemic awareness is a fallout
from alphabetic literacy often yield confusing results (Bryant, et. al, 1990); the strongest,
least uncontroversial conclusion that can be drawn is that the two abilities develop in
tandem.
Literate, English-speaking college students, are all, to some degree, phonemically aware.
Recall that both Adams (1990) and Treiman (1993) identify levels of phonemic
awareness, which are judged on the basis of phonemic awareness task performance
(Adams) and spelling errors (Treiman). In both the Adams and Treiman models, literate
adults have achieved the highest level of awareness; they are able to manipulate the
phonemes in a word, as demonstrated by Delete Segment Task (Adams), and they make
spelling mistakes that reflect learned phoneme-grapheme correspondences, like s-u-p-r-i-
s-e (Treiman). While learning how to read and write, they pass through many stages of
phonological awareness. They learn that words can be broken down into syllables,
syllables into rimes and onsets, rimes and onsets into segments. Ultimately, they break
the alphabetic code, learning that written words are composed of segments 19 and those
segments correspond in relatively regular ways with the letters of the alphabet. These
skills enable them to perform well on phonemic awareness tasks that elude preliterate
children and illiterate adults.
4.1 Recap of results from Chapter Three
However, not all adults reach the same level of phonemic awareness. Adams only divides
phonemic awareness into five levels, and these levels are useful in comparing literate
adults with children or illiterates. They do not suffice, though, for comparing adults to
each other. Just as phoneme counting serves to distinguish literates and illiterates, the
tasks discussed in Chapter Three serve to distinguish very phonemically aware from
19 An interesting area of future research might be a study of emergent phonemic awareness in the blind. Pring (1994) found that the blind never go through a logographic (whole-word) stage in reading, since they are unable to see words and word breaks on the printed page. From the very beginning, blind readers must conceive of their writing system alphabetically.
moderately phonemically aware adults. Recall the six tasks (with corresponding
experiment numbers in parentheses):
Segment-based tasks Syllable-based tasksDeletion Delete Segment Task (1) Delete Syllable Task (2)Substitution Substitute Segment Task (3) Substitute Syllable Task (4)Reversal Reverse Segment Task (5) Reverse Syllable Task (6)
Although Adams would classify phoneme deletion, substitution and reversal tasks as
tapping into the highest level of phonemic awareness, she also discusses the fact that
reversal tasks are more difficult than deletion tasks. So, since the tasks in this experiment
require the manipulation of different phonological units (segment and syllable), in
different ways (deletion, substitution, and reversal), we would expect that subjects might
perform better on some tasks than others, even though all adults are to some degree
phonemically aware. The man correct response rates for the six tasks (taking into
account all responses) are listed in the table below:
Mean Correct Response Rates on the Six Tasks
Del. Seg. Del. Syll. Sub. Seg. Sub. Syll. Rev. Seg. Rev. Syll.List A subjects 66% 78% 25% 43% 21% 87%List B subjects 75% 94% 14% 70% 18% 88%
As is clear from the correct response rates, subjects performed well on the Delete
Segment, Delete Syllable, and Reverse Syllable Tasks, had a mid-range performance on
the Substitute Syllable Task, and performed poorly on the Substitute Segment and
Reverse Segment Tasks. In short, some tasks were more difficult than others.
4.1.1. Recap and analysis of Segment-Based Task results
Moving on to more specific predictions, it was expected that subjects’ phonemic
awareness (as measured by their ability to perform well on tasks which tap into phonemic
awareness) would be correlated with reading ability. The charts that follow summarize
the findings from the experiments in Chapter Three.
First, it was predicted that subjects’ performance on segment-based tasks (which tap into
the highest levels of phonemic awareness) would be correlated with reading ability, but
not with analytic ability. The two tables below summarize the predictions and findings
(predicted findings are shaded):
Correlation of Segment-Based Tasks with Reading Ability
Deletion Tasks Substitution Tasks Reversal TasksGroup A Group B Group A Group B Group A Group B
Prediction Yes Yes Yes Yes Yes YesResult No No No Yes Yes Nor-value .464 -.040 .186 .642 .831 .387
Correlation of Segment-Based Tasks with Analytical Ability
Deletion Tasks Substitution Tasks Reversal TasksGroup A Group B Group A Group B Group A Group B
Prediction No No No No No NoResult No No No No No Nor-value .149 -.032 .049 .232 .367 -.226
The results are mixed. In no case was performance on a segment-based task correlated
with analytical ability, which was the expected result. However, performance on
segment-based tasks did not correlate with reading ability across-the-board, as predicted.
Though Group B’s performance on the Substitute Segment Task and Group A’s
performance on the Reverse Segment Task were correlated with reading ability, neither
group’s performance on the Delete Segment Task was correlated with reading ability.
In light of these mixed results, a new measure called “Total Segment Score” was
calculated for each group, combining performance on the three segment based tasks.
This score was then measured against performance on the analytical and reading
measures. The results are illustrated on the graphs below:
Group A Total Segment Score as Correlated with Reading and Analytical Score
Group B Total Segment Score as Correlated with Reading and Analytical Score
In both groups, the total segment score is significantly correlated with reading ability at
the p=.05 level (A r(18)=.678; B r(15)=.568), but not with analytical ability (A
r(18)=.254; B r(15)=-.013). This is exactly what we would predict: adults who are the
most phonemically aware (as demonstrated by their overall performance on segment-
based tasks) are those who are the best readers. Moreover, their phonemic awareness is
not correlated with their analytical ability. So, although the results of the individual
segment-based tasks are less clear-cut, overall performance on segment tasks supports the
idea that adult PA and reading ability are correlated.
4.1.2. Recap and analysis of Syllable-Based Task results
With regard to the syllable-based tasks, it was predicted that subjects’ performance on
these tasks would not be correlated with reading ability, as syllable-based tasks are
expected to tap into a lower level of phonological awareness than the segment-based
tasks, a level at which all literate adults would be comparable with regard to phonological
ability. The results are summarized in the following tables (with predicted outcomes
shaded):
Correlation of Syllable-Based Tasks with Reading Ability
Deletion Tasks Substitution Tasks Reversal TasksGroup A Group B Group A Group B Group A Group B
Prediction No No No No No NoResult No No Yes Yes No Nor-value .295 -.100 .580 .536 .204 .083
Correlation of Syllable-Based Tasks with Analytical Ability
Deletion Tasks Substitution Tasks Reversal TasksGroup A Group B Group A Group B Group A Group B
Prediction No No No No No NoResult No Yes No No No Nor-value .379 -.546 .180 .064 .232 .007
As with the results of the segment-based tasks, the results here are mixed, and not
entirely as expected. In neither Group A nor Group B did subjects’ performance on
syllable deletion or reversal correlate with reading ability, but in both groups,
performance on syllable substitution was (unexpectedly) correlated with reading ability.
Strangely, Group B’s performance on the Delete Syllable Task was correlated with
analytical ability (though it was predicted that no measure of phonological awareness
would be correlated with analytical ability). Again, an overall measure of performance
on syllable-based tasks (“total syllable score”) was calculated to shed light on the matter.
The results are illustrated in the graphs below:
Group A Total Syllable Score as Correlated with Reading and Analytical Score
Group B Total Syllable Score as Correlated with Reading and Analytical Score
In Group A subjects, the total syllable score is correlated with reading ability at the p=.05
level (r(18)=.585), but not with analytical ability (r(18)=.395). In Group B subjects, the
correlation between total syllable score and reading ability is not statistically significant
(B r(15)=.422, although it can be seen in the trend line that it is much stronger than the
correlation between total syllable score and analytical ability for the same subjects
(r(15)=.079).
While it was predicted that performance on syllable-based tasks would not be correlated
with reading ability in any case, it is telling that total syllable score is more strongly
correlated with reading ability than analytical ability for both sets of subjects, and is
significant in the case of Group A subjects. Subjects did perform at various levels of
ability on the syllable-based tasks; could it be that these differences reflect differences in
levels of phonological awareness? Is it the case that adults not only demonstrate differing
levels of phonemic awareness (as demonstrated by performance on segment-based tasks),
but also differing levels of phonological awareness (as demonstrated by performance on
syllable-based tasks)? And how might these different levels of awareness correlate with
reading ability?
4.1.3. Total Phonological Awareness Score
In an attempt to tackle some of these questions, a third aggregate score was computed:
“Total Phonological Awareness Score.” This score was computed by calculating
subjects’ performance on all six experimental tasks, with segment-based tasks and
syllable-based tasks combined. The results were then checked for correlations with
reading ability and analytical ability. The results are listed in the tables below, along
with a recap of the Total Segment and Total Syllable Score findings (predicted results are
shaded):
Correlation of Total Phonological Awareness Performance with Reading Ability
Total Segment Score Total Syllable Score Total PA Score
Group A Group B Group A Group B Group A Group BPrediction Yes Yes Yes Yes Yes Yes
Result Yes Yes Yes No Yes Nor-value .668 .568 .585 .422 .673 .505
Correlation of Total Phonological Awareness Performance with Analytical Ability
Total Segment Score Total Syllable Score Total PA Score
Group A Group B Group A Group B Group A Group BPrediction No No No No No No
Result No No No No No Nor-value .254 -.013 .395 .079 .345 .061
In Group A subjects, the total phonological awareness score is correlated with reading
ability at the p=.05 level (r(18)=.673), but not with analytical ability (r(18)=.345). In
Group B subjects, the correlation between total syllable score and reading ability is not
statistically significant at the p=.05 level (r(15)=.505) (statistical significance at the p=.05
level would require that r be greater than or equal to .515). Still, the correlation is much
stronger than that between total syllable score and analytical ability for the same subjects
(r(15)=.061). The graphs below provide a visual illustration of these results:
So, the three aggregate measures require some unpacking. Total Segment Score was
correlated with reading ability and not with analytical ability (both as predicted). Neither
Total Syllable Score nor Total Phonological Awareness score were correlated with
analytical ability (as predicted). However, both Total Syllable Score and Total
Phonological Score were significantly correlated with reading ability for Group A
subjects, and there was a near-significant correlation of these measures with reading
ability in Group B subjects.
4.2. The Degree of Difficulty Model
What conception of adult phonemic and phonological awareness can help to explain these
results? The “degree of difficulty” model, outlined below, sheds some light on the
matter. It will be argued that although all adults possess some degree of phonemic
awareness, when a task is sufficiently difficult, differences in phonemic and phonological
differences emerge and are reflected in task performance, and those differences are
correlated with reading ability.
Phonemic awareness theory holds that segment-based tasks will be more difficult that
syllable-based tasks. For each set of “sister” experiments (the same task performed on
segments vs. syllables), a t-test was run to determine whether subjects’ performance on
segment-based tasks differed significantly from their performance on syllable-based
tasks. The results are summarized in the chart below:
Difference in Subjects’ Performance on Segment vs. Syllable Tasks
Deletion Tasks Substitution Tasks Reversal TasksGroup A Group B Group A Group B Group A Group B
Prediction Yes Yes Yes Yes Yes YesResult* No Yes Yes Yes Yes Yes
t Stat -1.64 -3.80 -2.34 -7.70 -10.65 -14.72P(T<=t) one-tail 0.06 0.00 0.02 0.00 0.00 0.00*p < .05
In all cases, subjects had lower correct mean response rates for the segment-based tasks
than the paired syllable-based tasks. Excepting Group A’s performance on the deletion
task, these differences were statistically significant at the p=.05 level.
So, it can be safely claimed that the syllable-based tasks in this study were more difficult
than the segment-based tasks. This is consistent with past phonemic awareness research
and theory. Returning to the original six tasks, let’s assign a degree of difficulty (d.o.d.)
to each task. These are given in the following table:
Segment-based tasks (2) Syllable-based tasks (1)Deletion (1) Delete Segment; d.o.d.=2 Delete Syllable, d.o.d.=1Reversal (2) Reverse Segment, d.o.d.=4 Reverse Syllable, d.o.d.=2Substitution (3) Substitute Segment, d.o.d.=6 Substitute Syllable, d.o.d.=3
These numbers are admittedly arbitrary, but serve a conceptual purpose. They were
assigned in the following fashion: all syllable-based tasks have a d.o.d. of 1, while
segment-based tasks have a d.o.d. of 2. This distinction reflects the “subattentional
status” (to borrow a phrase from Adams) usually given to phonemes. Similarly, deletion
tasks, the easiest tasks, get a d.o.d. of 1, while reversal and substitution and permutation
are assigned (2) and (3) respectively (overall correct response rate on substitution tasks
was worse than that on reversal tasks). With these assignments, let us return to a revised
version of the chart from the beginning of this chapter that summarized mean correct
response rates by task:
Mean Correct Response Rates By Task (Percentages shaded and in bold are those that were correlated with reading ability)
Del Syll Rev Syll Del Seg Sub Syll Rev Seg Sub Segd.o.d. 1 2 2 3 4 6
List A subjects 78% 87% 66% 43% 21% 25%List B subjects 94% 88% 75% 70% 18% 14%
The chart now lists response rates sorted by degree of difficulty. The double line in the
middle of the table separates tasks in which no performance was correlated with reading
ability from those in which at least one group’s performance was correlated with reading
ability. It also separates a d.o.d. of 2 or less from a d.o.d. of 3 or greater.
It is postulated that the all the literate adults in this study have achieved a level of
phonemic awareness up to and including degree of difficulty 2, thus explaining why the
performance on the first three tasks in the table above was generally good. These are the
tasks to the left of the double line; correct response rate for these tasks was 66-94%, so
subjects correctly responded at least two-thirds of the time. None of these performances
were correlated with reading ability. In more concrete terms, this implies the following:
all literate adults, by virtue of their alphabetic experience, are expected to have some
degree of phonemic awareness. In this study, that level of awareness includes the ability
to delete the initial phoneme of a word, and to delete and reverse the syllables in a word.
When tasks tap into this level of phonemic awareness, the d.o.d. of the tasks will be low
(2 or less), the task performance will be uniformly good (better than 66% correct
response rate), and the slight differences in subjects’ performances will not be correlated
with reading ability.
The tasks to the right side of the double line in the chart above are those with a degree of
difficulty of 3 or higher. These include the Substitute Syllable, Reverse Segment and
Substitute Segment tasks. Because of the difficult of these tasks, performance was poor.
Subjects answered correctly only 14%-70% of the time (except for Group B’s
performance on the Substitute Syllable task, subjects gave correct answers less than half
the time). For all three of these more difficult tasks, performance by at least one group
was significantly correlated with reading ability. These tasks are presumably, because of
their difficulty, tapping into higher-level phonemic and phonological awareness abilities,
which some adults possess and some do not. Syllables are more available to adults than
segments, and reversal and substitution tasks are more difficult than deletion tasks.
When these differences present themselves, they are correlated with reading ability.
4.4. Summary and Conclusions
In this chapter, we have made the following observations about phonemic and
phonological awareness in literate adults:
Overall performance on segment-based tasks (“Total Segment Score”) was significantly correlated with reading ability in both groups of subjects.
Overall performance on syllable-based tasks (“Total Syllable Score”) was significantly correlated with reading ability in Group A subjects.
Overall task performance (“Total Phonological Awareness Score”) was significantly correlated with reading ability in Group A subjects (and was a strong trend in Group B subjects).
There were no correlations of any aggregate measure (Total Segment Score, Total Syllable Score or Total Phonological Awareness Score) with analytical ability, the control measure.
Recall that in phonemic awareness studies with children, the main conclusion is that
reading ability and performance on phonemic awareness tasks are positively correlated.
An even more interesting finding from children’s studies is that performance on more
difficult measures of phonemic and phonological awareness (say, an Adams level four
counting task) is more likely to be correlated with reading ability than a lower level
measure (for instance, an Adams level two oddity task). An examination of results from
adults in this study reached the same conclusion:
Task difficulty was the factor that determined whether performance on the six individual tasks was correlated with reading ability. Subjects performed better on syllable-based tasks than segment-based tasks (a linguistic measure of difficulty), and better on deletion tasks than on substitution or reversal tasks (a non-linguistic measure of difficulty). When a task was sufficiently difficult, differences in phonological and phonemic awareness levels of subjects emerged, and these differences were correlated with reading ability.
To recap the conclusions thus far, it has been shown that adults, though all phonemically
aware to some degree, exhibit different levels of phonemic and phonological awareness.
These differences are most apparent when the tasks are sufficiently difficult. Overall
correlations of task performance with reading ability are strong, and inconsistencies are
explained by differences in task difficuly. What are the implications of these findings?
Contrary to popular belief, adults to not use wholly different strategies in reading than
children, relying on whole word knowledge or morphology to the exclusion of
phonological analysis. Though Forster, in his work on prefix stripping, showed that
adults exhibit morphological interference in the recognition of words, grapheme-
phoneme relationships play a part as well. As indicated in the introduction, this is
probably particularly true in the identification of unfamiliar words (which may be
familiar after orthographic recoding).
Segments, even for literate adults, are less available units of phonology than syllables.
This gives credence to theories of the syllable as the basic unit of perception. Segments
are unnatural, learned constructs. One oft-cited argument for the existence of the
syllable is the presence of alphabetic writing systems. Yet, an examination of the history
of writing systems indicates that the invention of the alphabet was a historical accident of
sorts; an attempt by Greeks to adopt a Semitic syllabary ill-suited to their richer syllabic
inventory (Olson, 1993).
4.5. Implications
This research also has implications for the educational and applied linguistics
communities. Studies with children have shown that a combination of phonemic
awareness training and phonics may be the best way to facilitate learning to read. PA
training for adults might then be incorporated into (1) Adult literacy programs, (2)
English-as-a-second-language programs (or any such programs in which the L2 employs
an alphabetic orthography), and (3) Reading enrichment programs.
Phonemic awareness research need not be limited to children and special populations.
Just because literate adults have had significant experience with an alphabetic writing
system does not mean that their relationship with phonemic awareness is over;
preliminary evidence from this study indicates that it continues to develop, and to serve
them throughout adulthood.
Summary of Correlations (Correlations in bold are significant at the p = .05 level)
Group A Reading Score Analytical ScoreReading Score 1Analytical Score 0.439786 1Delete Segment Task 0.464077 0.149008Reverse Segment Task 0.831242 0.366641Substitute Segment Task 0.185878 0.048682Delete Syllable Task 0.294728 0.379228Reverse Syllable Task 0.203664 0.232132Substitute Syllable Task 0.580289 0.179847Total PA Score 0.673095 0.34571Total Segment Task Score 0.66774 0.253796Total Syllable Task Score 0.585082 0.394829
Group B Reading Score Analytical ScoreReading Score 1Analytical Score 0.056152 1Delete Segment Task -0.03967 -0.0324Reverse Segment Task 0.387164 -0.22604Substitute Segment Task 0.642144 0.232003Delete Syllable Task -0.10047 0.545671Reverse Syllable Task 0.082538 0.007204Substitute Syllable Task 0.535983 0.064104Total PA Score 0.50542 0.061202Total Segment Task Score 0.567699 -0.01333Total Syllable Task Score 0.42181 0.079216
REFERENCES
Adams, Marilyn Jager. (1990). Beginning to Read: Thinking and Learning about Print Cambridge, MA: MIT Press.
Anderson, Stephen R. (1969). West Scandanavian Vowel Systems and the Ordering of Phonological Rules. PhD Dissertation, MIT.
Anderson, Stephen R. (1992). A-Morphus Morphology. Cambridge: Cambridge University Press.
Bagemihl, Bruce. (1989). The Crossing Constraint and 'Backwards languages.' NLLT 7, pp. 481-549.
Bao, Z. (1990) "Fanqie Languages and Reduplication" Linguistic Inquiry, 27.3 307-350.
Ball, E. and B. Blachman. (1991). Does phoneme awareness training in kindergarten make a difference in early word recognition and developmental spelling? Reading
Research Quarterly 26, pp. 49-66.
Bertelson, P., Morais, J., Cary, L. and J. Algeria. (1987). Interpreting data from illiterates: Reply to Koopmans. Cognition 27, pp. 113-115.
Bowey, J. and Francis, J. (1991). Phonological analysis as a function of age and exposure to reading instruction. Applied Psycholinguistics, 12, pp. 91-121.
Bradley, L. and P. Bryant. (1983). Categorizing Sounds and Learning to Read: A causal connection. Nature 271, pp. 419-421.
Bradley, L. and P. Bryant. (1985). Rhyme and Reason in Reading and Spelling. Ann Arbor: University of Michigan Press.
Bradley, L. and P. Bryant. (1991). Phonological Skills Before and After Learning To Read. In S. Brady and D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 37-46). Hillsdale, NJ: Lawrence Erlbaum Associates.
Bryant, P. MacLean, M., Bradley, Linguistics., and Crossland, T. (1990). Rhyme and Alliteration, Phoneme Detection, and Learning to Read. Developemental Psychology 26:3, pp. 429-438.
Byrne, B. & Fielding-Barnsley, R. (1989). Phonemic awareness and letter knowledge in the child’s acquisition of the alphabetic principle. Journal of Educational Psychology 81 pp. 313-321.
Cataldo, S. and Ellis, N. (1988). Interaction in the development of spelling, reading and phonological skills. Journal of Reading Research 11:2, pp. 86-109.
Chaney, C. (1992). Language development, metalinguistic skills, and print awareness in 3-year-old children. Applied Psycholinguistics 13, pp. 485-514.
Clements, George N. (1988). The role of the sonority cycle in core syllabification. Working papers of the Cornell Phonetics Laboratory. Ithaca, NY: Cornell University.
Clements, G.N. and Keyser, S.J. (1983). CV Phonology: A generative theory of the syllable. Cambridge, MA: MIT Press.
Cossu, G., F. Rossini, and J.C. Marshall. (1993). When reading is acquired but phonemic awareness is not: A study of literacy in Down's syndrome. Cognition 46, pp. 129-138.
Cunningham, Anne. (1990). Explicit versus Implicit Instruction in Phonemic Awareness. Journal of Experimental Child Psychology 50, pp.429-444.
Cutler, A., Mehler, J., Norris, D., and Segui, J. (1986). The syllable’s differing role in the segmentation of French and English. Journal of Memory and Language 25, pp. 385-400.
Davis, S. and M. Hammond. (1994). On the Status of On-glides in English Syllable Structure. University of Indiana and University of Arizona, ms.
Durgunonglu, A., Nagy, W., and Hancin-Bhatt, B. (1993). Cross-Language Transfer of Phonological Awareness. Journal of Educational Psychology 85:3, pp. 453-465.
Dupoux, E. and Mehler, J. (1992). Unifying Awareness and On-Line Studies of Speech: A tentative Framework. In Analytic Approaches to Human Cognition., J. Algeria,
D. Hoender, J. Jenca de Morais, and M. Radeau (eds.).
Forster, K. (1976). Accessing the Mental Lexicon. In E.C.J. Walker & R.F. Wales (eds.), New Approaches to Language Mechanisms. Amsterdam: North Holland Publishing.
Forster, K. (1990). Lexical Processing. In D.N. Osherson and H. Lasnik, eds., Language: An Invitation to Cognitive Science, Vol. 1. Cambridge: MIT Press.
Fox, B. and Routh, D.K. (1975). Analyzing spoken language into words, syllables, and phonemes: A developmental study. Journal of Psycholinguistic Research, 4, pp. 331-342.
Fox, B. and Routh, D.K. (1980). Phonemic analysis and severe reading disability in children. Journal of Psycholinguistic Research, 9, pp. 115-119.
Fudge, E. C. (1969). Syllables. Journal of Linguistics 5, pp. 253-286.
Goodman, Ken. (1993). Phonics Phacts. Portsmouth, NH: Heinemann Press.
Gough, P. (1983). Context, form, and interaction. In K. Rayner (Ed.), Eye movements in reading, pp. 331-358. Cambridge, MA: MIT Press.
Goyen, J. (1989). Reading Methods in Spain: The effect of regular orthography. The Reading Teacher, February, 1989.
Hammond, M. (1993) Resyllabification in English and heavy trochees, presented at FLSM.
Hammond, M. (1997) "Vowel quantity and syllabification in English", Language 73, 1-17.
Hanna, P.R., Hodges, R.E., & Hanna, J.S. (1971). Spelling: Structure and strategies. Boston: Houghton Mifflin.
Hatcher, P. Hulme, C. And Ellis, A. (1994) . Ameliorating Early Reading Failure by Integrating the Teaching of Reading and Phonological Skills: The Phonological Linkage Hypothesis. Child Development 65, pp. 41-57.
Hooper, J. B. (1972). The syllable in phonological theory. Language 48, pp. 525-540.
Kahn, D. (1976). Syllable-based generalizations in English phonology. Bloomington, IN: Indiana University Linguistics Club.
Koopmans, M. (1987). Formal schooling and task familiarity: A reply to Morais et.al. Cognition 27, pp.109-11.
Lorenson, S. (1993). The role of orthography in auditory processing. University of Arizona, ms.
Lundberg, Ingvar. (1991). Phonemic Awareness Can be Developed Without Reading Instruction. In S. Brady and D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 47-54). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lundberg, I., Frost, J. and Petersen, O. (1988). Effects of an extensive program for stimulating phonological awareness in preschool children. Reading Research Quarterly 23, pp. 263-283.
Lundberg, I., Olofsson, A. and Wall, S. (1980). Reading and spelling skill in the first school years predicted from phonemic awareness skills in kindergarten. Scandanavian Journal of Psychology, 21, pp. 159-173.
Manis, F., Cusodio, R. and Szeszulski, P. (1993). Development of Phonological and Orthographic Skill: A 2-Year Longitudinal Study of Dyslexic Children. Journal Of Experimental Child Psychology, 56, pp. 64-86.
Mann, Virginia. (1986). Phonological awareness: The role of reading experience. Cognition 24, pp. 65-92.
Mann, Virginia. (1991). Are We Taking Too Narrow a View of the Conditions for Development of Phonological Awareness? In S. Brady and D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 55-64).Hillsdale, NJ: Lawrence Erlbaum Associates.
Mann, Viriginia. (1993). Phoneme Awareness and Future Reading Ability. Journal of Learning Disabilities 26 (4), pp. 259-269.
Mann, V.A. & Brady, S. (1988). Reading disability: The role of language deficiencies. Journal of Consulting and Clinical Psychology 56(6), pp. 811-816.
Meador, D. and Ohala, D. (1993). Ambisyllabicity in English. University of Arizona, ms.
Mehler, J. Dommergues, J.Y., and Frauenfelder, U. (1981). The syllable’s role in speech segmentation. Journal of Verbal Learning and Verbal Behavior 20, pp.
298-305.
Morais, J., Bertelson, P., Cary, L. and Alegria, J. (1986). Literacy training and speech segmentation. Cognition 24, pp.45-64.
Morais, Jose. (1987). Segmental Analysis of Speech and its Relation to Reading Ability. Annals of Dyslexia 37, pp. 126-141.
Morais, Jose. (1991). Constraints on the Development of Phonemic Awareness. In S. Brady and D. Shankweiler (Eds.), Phonological Processes in Literacy (pp. 5-28). Hillsdale, NJ: Lawrence Erlbaum Associates.
O'Brien, David. (1988). The Relation Between Oral Reading Miscue Patterns and Comprehension: A Test of the Relative Explanatory Power of Pyscholinguistics and Interactive Views of Reading. Journal of Psycholinguistic Research Vol.
17(5), pp.379-401.
Pring, L. (1994), Touch and Go: Learning to Read Braille. Reading Research Quarterly, 29:1, pp. 67-74.
Pulgram, E. (1970). Syllable, Word, Nexus, Cursus. The Hague: Mouton.
Read, Charles. (1986) Children's Invented Spellings. London: Routledge & Kegan Paul.
Read, C. Yung-Fei, X., Hong-Yin, N. and D. Bao-Qing. (1986). The ability to manipulate speech sounds depends on knowing alphabetic writing. Cognition 24, pp.31-44.
Segui, J., Dupoux, E., and Mehler, J. (1985). The role of the syllable in speech segmentation, phoneme identification, and lexical access. Ms.
Selkirk , E. O. (1982). The Syllable. In Harry Van der Hulst and Norval Smith (Eds.) The Structure of Phonological Representations (Part II). Cinnaminson: Foris, pp. 337-384).
Shaywitz. Sally E. (1996). Dyslexia. Scientific American, November 1996.
Stanovich, Keith. (1993). Romance and Reality. The Reading Teacher, Vol. 47, No. 4, pp. 280-291.
Stanovich, K. and Cunningham, A. (1992). Cognitive Correlates of Print Exposure. Journal of Memory and Cognition, 20 (1), pp. 51-68.
Stanovich, K.E., Cunningham, A. & Cramer, B. (1984). Assessing phonological awareness in kindergarten children: Issues of task comparability. Journal of Experimental Child Psychology 38, pp. 175-190.
Stanovich, K. & Stanovich, P. (1995). How research might inform the debate about early reading acquisition. Journal of Research in Reading 18 (2), pp. 87-105.
Taft, M. and Hambly, G. (1985). The Influence of Orthography on Phonological Representations in the Lexicon. Journal of Memory and Language, 24, pp. 320-335.
Taft, M. (1991). Phonological Recoding. In Reading and the Mental Lexicon. Erlbaum.
Tanenhaus, M.K., Flanigan, H.P., & Seidenberg, M.S. (1980). Orthographic and Phonological activation in Auditory and visual word recognition. Memory and Cognition, 8, pp. 513-520.
Tornéus, M. 1984. Phonological Awareness and Reading: A Chicken and Egg Problem? Jounral of Educational Psychology 76, pp. 1346-1358.
Tranel, B. (1981). Concreteness in Generative Phonology: Evidence from French. Berkeley: University of California Press.
Treiman, Rebecca. (1993) Beginning to Spell. Oxford: Oxford University Press.
Treiman, R. & Danis, C. (1988). Syllabification of Intervocalic Consonants. Journal of Memory and Language 27, pp. 87-104.
Treiman, R. & Weatherston, S. (1992). Effects of Linguistic Structure on Children’s Ability to Isolate Initial Consonants. Journal of Educational Psychology, 84:2, pp. 174-181.
Treiman, R. & Zukowski, A. (1990). Toward an Understanding of English Syllabification. Journal of Memory and Language 29, 66-85.
Zecker, S. M. Tanenhaus, Alderman, L. and Siqueland, L. (1986). Lateralization of Lexical codes in Auditory Word Recognition. Brain and Language, 29, pp. 372-389.